Extraction and Analysis of EEG signals for the detection of ADHD … · Extraction and Analysis of...

VISVESVARAYA TECHNOLOGICAL UNIVERSITY“JNANA SANGAMA”, BELAGAVI-590018.

2015-2016

AProject Report

On

“Extraction and Analysis of EEG signals for the detection ofADHD syndrome – A novel approach”

Submitted in partial fulfillment for the award of the degree ofBACHELOR OF ENGINEERING

IN

Submitted By

Chandana S [1JB12EC024]Karthik B Balaraj [1JB12EC041]Sushmitha Pol [1JB12EC112]Usha P [1JB10EC115]

UNDER THE GUIDANCE OF

||Jai Sri Gurudev||Sri Adichunchanagiri Shikshana Trust ®

DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING

SJB INSTITUTE OF TECHNOLOGYB G S HEALTH AND EDUCATION CITY

Kengeri, Bangalore-560060.

Dr. K. V. Mahendra PrashanthProfessor, Dept. of ECE

Chief Co-ordinator, R&D and PG studies

ACKNOWLEDGEMENT

We would like to express our profound grateful to His Divine Soul Jagadguru Padmabhushan Sri Sri Sri Dr.

Balagangadharanatha Mahaswamiji and His Holiness Jagadguru Sri Sri Sri Dr. Nirmalanandanatha Swamiji

for providing us an opportunity to complete our academics in this esteemed institution.

We would also like to express our profound thanks to Revered Sri Sri Prakashnath Swamiji, Managing Director,

SJB Institute of Technology, for his continuous support in providing amenities to carry out this project in this

admired institution.

We express our gratitude to Dr. Puttaraju, Principal, SJB Institute of Technology, for providing us an excellent

facilities and academic ambience; which have helped us in satisfactory completion of project work.

We extend our sincere thanks to Dr. K. R. Nataraj, Head of the Department, Electronics & Communication

Engineering for providing us an invaluable support throughout the period of our project work.

We wish to express our heartfelt gratitude to our guide Dr. K.V. Mahendra Prashanth, for his valuable guidance,

suggestions and cheerful encouragement during the entire period of our project work.

We wish to extend our heartfelt gratitude to Mrs. Vijayalakshmi. K (BMSCE) and her assistants, for their

valuable guidance, suggestions and cheerful encouragement during the entire period of our project work.

We would extend our sincere thanks to Srushti Special Academy for their kind co-operation and providing valuable

inputs during this project work.

We would also like to express our thanks to the Subjects (Children), Parents and Clarity Medical System for all

their concern and support towards data collection for our project work.

We express our truthful thanks to Dr. K. Mahantesh, Project Coordinator, Department of ECE, for his valuable

support.

Finally, we take this opportunity to extend our earnest gratitude and respect to our parents, Teaching & Non teaching

staffs of the department, the library staff and all our friends, who have directly or indirectly supported us during the

period of our project work.

Regards,

Chandana S

Karthik B Balaraj

Sushmitha Pol

Usha P

iii

ABSTRACT

Attention Deficit Hyperactivity disorder (ADHD) is a common mental disorder that

begins in childhood and can continue through adolescence and adulthood. EEG is most often

used to diagnose epilepsy, sleep disorders, coma, encephalopathy, and brain death. EEG is used

to be a first-line method of diagnosis for tumors, stroke and other focal brain disorders.

The present project work is mainly designed to predict the probable region of brain that

shows abnormality due to ADHD syndrome. EEG data of seven normal and ADHD subjects

(children) of age group 4-17 years has been collected following a protocol which contains 4

events- Eyes close, Eyes open, Visual Cue and Motor activity. The segmented data is subjected

to extract features like, energy that corresponds to Alpha, Beta, Delta and Theta frequency

ranges; absolute power and amplitude levels at different electrode positions. Single map analysis

and Frequency map analysis is also performed and comparative analysis is done between the

normal and ADHD subjects. Neural network algorithm is implemented to distinguish between

ADHD and normal subjects. Finally, 3-D plotting is done for the ease of visualization and

diagnosis purpose.

This project work draws the conclusion that, posterior region of the brain, covering

electrode positions O1, O2, P3 and P4 could be considered to predict the abnormality during

eyes open and eyes close; Anterior portion of the brain, covering pre-frontal regions and frontal

region of the brain, with electrode positions FP1 and FP2 could be used to predict the

abnormality. It also proves that, neural networks, if well-trained can be used to distinguish

between ADHD and normal subjects, given the pattern of EEG as input.

iv

LIST OF FIGURES

Sl. No. Figure No. Name of Figure Page No.

1. 1.1 Brain 4

2. 1.2 Electroencephalogram 7

3. 1.3 An EEG setup 7

4. 3.1 Flowchart 17

5. 3.2 International 10-20 electrode system 21

6. 3.3 Braintech traveller portable EEG device 22

7. 3.4 Longitudinal Bipolar montage 26

8. 3.5 Protocol used 28

9. 3.6 Block diagram of FFT algorithm 30

10. 3.7 Frequency Maps 35

11. 3.8 Frequency Tables 35

12. 3.9 Training and verification 37

13. 3.10 Single Layer feed-forward 39

14. 3.11 Multi-Layer feed-forward 39

15. 3.12 Recurrent Network 40

16. 3.13 The Neural Network Diagram 40

17. 3.14 Performance plot 41

18. 3.15 Error Histogram 42

v

19. 3.16 Regression Plot 43

20. 3.17 Confusion Plot 43

21. 4.1 Frequency map of ADHD subject in eyes close event 53

22. 4.2 Frequency map of normal subject in eyes close event 54

23. 4.3 Frequency map of ADHD subject in eyes open event 54

24. 4.4 Frequency map of normal subject in eyes open event 55

25. 4.5 Frequency map of ADHD subject in visual cue event 55

26. 4.6 Frequency map of normal subject in visual cue event 56

27. 4.7 Frequency map of ADHD subject in motor activity event 56

28. 4.8 Frequency map of normal subject in motor activity event 57

29. 4.9 Performance Plot (Case-1) 58

30. 4.10 Regression Plot (Case-1) 59

31. 4.11 Training State Plot (Case-1) 60

32. 4.12 Confusion Matrix - Train State (Case-1) 61

33. 4.13 Confusion Matrix - Test State (Case-1) 61

34. 4.14 Error Histogram 1 (Case-1) 62


36. 4.16 Performance Plot (Case-2) 62

37. 4.17 Regression Plot (Case-2) 63

38. 4.18 Training State Plot (Case-2) 63

39. 4.19 Confusion Matrix - Train State (Case-2) 64

vi

40. 4.20 Confusion Matrix - Test State (Case-2) 64



LIST OF TABLES

Sl. No. Table No. Name of Table Page No.

1. 1.1 Activation Patterns in Brain 5

2. 1.2 Classification of EEG Signals 9

3. 3.1 Details of ADHD Subjects 20

4. 3.2 Details of Normal Subjects 21

5. 3.3 Operating Environment of Braintech Traveller 23

6. 3.4 General Parameters of Braintech Traveller Device 23

7. 3.5 Channels of Longitudinal Bipolar Montage 26

8. 4.1 Single Map Analysis 44

9. 4.2 Standard Deviation Table for Eyes Close Event 45

10. 4.3 Standard Deviation Table for Eyes Open Event 45

11. 4.4 Amplitude Values for ADHD Subjects 46

12. 4.5 Amplitude Values for Normal Subjects 47

13. 4.6 Standard Deviation Table for Visual Cue 48

14. 4.7 Relative Power Percentage Table for ADHD Subjects in

Eyes Close Event

48

vii

15. 4.8 Relative Power Percentage Table for Normal Subjects in

Eyes Close Event

49


Eyes Open Event

49


Eyes Open Event

49


Visual Cue Event (Animal)

50


Visual Cue Event (Color)

50


Visual Cue Event (Number)

51


Visual Cue Event (Animal)

51


Visual Cue Event (Color)

51


Visual Cue Event (Number)

52


Motor Activity

52


Motor Activity

53

viii

TABLE OF CONTENTS

Abstract iii

List of Figures iv

List of Tables vi

CHAPTER 1: INTRODUCTION 1

1.1 Attention Deficit Hyperactivity Disorder (ADHD) 1

1.1.1 Types of ADHD 1

1.1.2 Symptoms of ADHD 2

1.1.3 Causes for ADHD 2

1.1.4 Brain Functioning in ADHD 3

1.1.4.1 Structural Abnormality 3

1.1.4.2 Functional Abnormality 4

1.1.4.3 Functional circuits involved in the pathophysiology of ADHD 4

as identified in a review of the neurobiology of ADHD

1.1.4.4 Electrical activity 5

1.1.4.5 Chemical Imbalance 6

1.2 Electroencephalography 6

1.2.1 EEG Signal Classification 8

1.2.2 Advantages 8

1.2.3 Disadvantages 11

1.2.4 EEG vs. fMRI, fNIRS and PET 11

1.2.5 EEG vs.MEG 12

CHAPTER 2: LITERATURE SURVEY 13

2.1 Motivation 15

2.2 Objectives 16

2.3 Challenges 16

CHAPTER 3: METHODOLOGY 17

3.1 Subjects 18

3.1.1 Questionnaire 18

ix

3.1.2 Subject Details 20

3.2 EEG Data Acquisition 21

3.2.1 Braintech Traveler Portable EEG Device 22

3.2.1.1 Configurations 22

3.2.1.2 General Parameters 23

3.2.1.3 More Features 24

3.2.2 Electrode Positions 25

3.2.3 Protocol 27

3.3 Data Segmentation 28

3.4 Feature Extraction 30

3.4.1 Energy Calculation 30

3.4.2 Statistical Parameters Calculation 31

3.5 Analysis 33

3.5.1 Single Map Analysis 34

3.5.2 Frequency Map Analysis 34

3.5.3 Frequency Tables 35

3.6 Neural Network Algorithm and its Importance 36

3.6.1 Why Neural Networks 37

3.6.2 Feed forward back propagation 37

3.6.3 Network architectures 38

3.6.3.1 Single Layer Feed-forward 38

3.6.3.2 Multi-layer feed-forward 39

3.6.3.3 Recurrent network 40

3.6.4 The Neuron 40

3.6.5 Performance plot and Training state plot 41

3.6.6 Error Histogram 42

3.6.7 Regression Plot 42

3.6.8 Confusion Plot 43

CHAPTER 4: RESULTS 44

4.1 Single Map Analysis 44

4.2 Power Analysis 48

x

4.2.1 Eyes Close ADHD Kids 48

4.2.2 Eyes Close Normal Kids 49

4.2.3 Eyes Open ADHD Kids 49

4.2.4 Eyes Open Normal Kids 49

4.2.5 Visual Cue ADHD Kids (Animal Slide) 50

4.2.6 Visual Cue ADHD Kids (Color Slide) 50

4.2.7 Visual Cue ADHD Kids (Number Slide) 50

4.2.8 Visual Cue Normal Kids (Animal Slide) 51

4.2.9 Visual Cue Normal Kids (Color Slide) 51

4.2.10 Visual Cue Normal Kids (Number Slide) 52

4.2.11 Activity for ADHD Kids 52

4.2.12 Activity for Normal Kids 53

4.3 Frequency Maps 53

4.4 Neural Network Algorithm 57

CHAPTER 5: CONCLUSION 66

CHAPTER 6: FUTURE SCOPE 68

REFERENCES 69

PUBLICATIONS : Extraction and Analysis of EEG Signals for the Detection 71of ADHD Syndrome - A Novel Approach

AWARDS & RECOGNITION 81

Sponsored Project from KSCST 81

Best Project Award 83

Extraction and Analysis of EEG Signals for the Detection of ADHD Syndrome- A Novel Approach

Department of ECE, SJB Institute of Technology, Bangalore Page 1

CHAPTER – 1

INTRODUCTION

1.1 ATTENTION DEFICIT HYPERACTIVITY DISORDER (ADHD)

Attention Deficit Hyperactivity Disorder (ADHD) is a common mental disorder that

begins in childhood and can continue through adolescence and adulthood. It makes it hard for a

child to focus and pay attention. Some children may be hyperactive or have trouble being patient.

For children with ADHD, levels of inattention, hyperactivity and impulsive behaviors are greater

than for other children in their age group. ADHD can make it hard for a child to do well in

school or behave at home or in the community.

Children of all backgrounds can have ADHD. Teens and adults can have ADHD too. It is

estimated that about 3-5 percent of all kids, have this disorder. Also the most unfortunate thing is

that in many cases this disorder goes totally undiagnosed .The child is often mistaken as hyper-

energetic kid or as just being more playful, but the reality is that many such kids suffer from a

medical disorder known as Attention Deficit Hyperactive Disorder or ADHD.

1.1.1 Types of ADHD

Kids having ADHD usually depict either of three patterns of behavior.

1. Inattentive type: having difficulty in concentrating and easily getting distracted.

2. Hyperactive type: means that they are more physically hyperactive (constantly moving

jumping running fidgeting squirming, having difficulty sitting in one place etc).

3. Combined type: having symptoms of hyperactivity and inattention both.

Nobody exactly knows what causes ADHD. Recent researches point out the causes could

be either neurobiological or genetically. Certain chemical levels in the brain have been found

low (dopamine -a chemical that sends messages to the part of the brain involved in attention,

movement and motivation). Increased levels of heavy metals (lead) in the body is also seen in

some kids .The kind of parenting given to the child has no role in the causation of this disorder,

although bad parenting can influence the severity of this disorder. Implication of ADHD can

have far reaching effects, with the schooling of the child taking the biggest hit. Poor performance



in school due to ADHD can lead to a hampered self esteem in the child .ADHD can also lead to

learning disabilities, defiant behavior, conduct disorder (anti-social behavior, stealing ,fighting),

anxiety disorders and depression .

Severe forms of ADHD are difficult to differentiate from milder forms of autism, a

careful approach is necessary in diagnosing and treating such forms. Such severe forms affect the

child’s normal mental development. For example speech and cognitive skills can be markedly

affected in such cases.

1.1.2 Symptoms of ADHD

ADHD has many symptoms. Some symptoms at first may look like normal behaviors for

a child, but ADHD makes them much worse and occur more often.

1. Inattentive Type (Category-A)

Difficulty sustaining attention and gets easily distracted.

Doesn’t seem to listen when spoken to directly, often forgetful.

Avoids taking tasks that require close attention.

Careless mistakes, difficulty in organizing tasks or activities.

Not able to follow instructions and fails to follow school or home work.

2. Hyperactivity and Impulsivity Type (Category-B)

Constantly running around jumping, always on the move as if driven by motor.

Fidgeting with hands or feet or constantly moving in seat.

Often has difficulty playing quietly, talks excessively.

Has difficulty waiting for turn. Is not able to sit in one place, has to wander away.

Difficulty in sleeping. Constantly squirming and moving in bed.

1.1.3 Causes for ADHD

ADHD probably stems from interactions between genes and environmental or non-

genetic factors. ADHD often runs in families. Researchers have found that much of the risk of

having ADHD has to do with genes. Many genes are linked to ADHD, and each gene plays a



small role in the disorder. ADHD is very complex and a genetic test for diagnosing the disorder

is not yet available.

Among the non-genetic factors that may increase a child’s risk for developing ADHD

are:

Smoking or drinking during pregnancy.

Birth complications or very low birth weight.

Exposure to lead or other toxic substances.

Extreme neglect, abuse or social deprivation.

Food additives like artificial food colors, which might make hyperactivity worse.

Some additives and dyes may worsen hyperactivity and inattention, but these effects are

small and do not account for most cases of ADHD.

Besides DNA, other factors can influence who develops ADHD. These risk factors

include:

Environmental exposure: Exposure to lead may increase a child’s risk for ADHD.

Brain injury: A small number of children who suffer a traumatic brain injury may

develop ADHD.

A mother’s alcohol and tobacco using during pregnancy: A study by the Washington

University School of Medicine found that mothers who smoke while pregnant increase

their child’s risk for developing ADHD. Women who drink alcohol and use drugs during

pregnancy also put their child at risk for the disorder.

Low birth weight or premature delivery: Babies born before their due date are more

likely to have ADHD when they are older.

1.1.4 Brain Functioning in ADHD

Neuro-imaging studies have associated various structural, functional, electrical activity

and chemical correlates with attention-deficit hyperactivity disorder (ADHD) in children,

adolescents and adults.

1.1.4.1 Structural Abnormality:

Brain imaging studies have associated structural abnormalities with ADHD in children

and adolescents, including:



Delayed cortical development

Cortical thinning, and reductions in the volume of grey and white matter

Reductions in the volume of several regions of the brain, including: the posterior inferior

vermis, splenium of the corpus callosum, total and right cerebral volume, right caudate,

right global pallidus, right anterior frontal region, cerebellum, temporal lobe, and

pulvinar.

Brain abnormalities associated with ADHD in children and adolescents may persist into

adulthood. MRI studies in adults have yielded similar evidence as described above with regards

to reduction in the volume of several regions of the brain, cortical thickness, and grey matter;

particularly in the frontal cortex of the brain, compared with controls; as well as structural

abnormalities in connecting brain cells within networks that regulate attention and emotion.

1.1.4.2 Functional Abnormality:

Brain structures implicated in ADHD correspond to brain networks, including some

involving frontal regions, and some that support executive function and attention.

1.1.4.3 Functional circuits involved in the pathophysiology of ADHD as

identified in a review of the neurobiology of ADHD:

Figure 1.1: Brain

Activation abnormalities are associated with ADHD in children, adolescents and adults,

with meta-analyses demonstrating significant activation reductions in various frontal regions of

the brain including the anterior cingulate, dorsolateral prefrontal and inferior prefrontal cortices,

and related regions including the basal ganglia, thalamus, and areas of parietal cortex.



Furthermore, a typical functional network connectivity in the default mode network. In

addition, there is some evidence that patterns of under and over activation of certain regions of

the brain differ between children and adolescents versus adults, as indicated by a meta-analysis

of 55 fMRI studies which compared children, adolescents and adults with ADHD with healthy

controls. This is explained in figure 1.1 and table 1.1.

Activation patterns in the brain as indicated by a meta-analysis of 55 fMRI studies

Age group Regions associated with over-activation Regions associated with under-

activation

Children and

adolescents

right angular gyrus

middle occipital gyrus

posterior cingulate cortex

midcingulate cortex

frontal regions

(bilaterally)

putamen (bilaterally)

right parietal region

right temporal region

Adults right angular gyrus

middle occipital gyrus

right central sulcus

precentral gyrus

Table 1.1: Activation patterns in brain

1.1.4.4 Electrical activity:

A meta-analysis of worldwide studies reported that quantitative electroencephalography

(EEG; the recording of electrical activity along the scalp) may be used to identify changes in

brain electrical activity. An increase in the theta/beta (two EEG frequency bands) ratio was

observed in all studies included in the review. Further research is required to substantiate EEG

findings for use as a biomarker in ADHD diagnosis. Individual EEG patterns associated with

ADHD are under early investigation for utility in personalizing neurofeedback protocols,

computer-assisted training to self-regulate brainwave activity, as non-pharmacological

treatment options for ADHD.



1.1.4.5 Chemical Imbalance:

Delayed maturation of certain dopaminergic neural pathways has been observed in

children and adolescents with ADHD, as well as an imbalance in the levels of both dopamine and

noradrenaline in the brains of children, adolescents and adults with ADHD compared with

healthy controls. Dopamine and noradrenaline have been implicated in influencing impulsivity,

and dopamine in influencing inattention.

Emerging evidence also suggests possible roles for other signaling systems in the

neurobiology of ADHD. Deficiencies in glutamate signaling in some regions of the brain may

have a modulatory role in adults with ADHD. Furthermore, polymorphisms in the serotonin

transporter gene have been associated with differential response to ADHD treatment, and the

presence of comorbid conduct disorder in children and adolescents with hyperkinetic disorder

(an alternative description of ADHD).

Current research indicates the frontal lobe, basal ganglia, caudate nucleus, cerebellum, as

well as other areas of the brain, play a significant role in ADHD because they are involved in

complex processes that regulate behavior (Teeter, 1998). These higher order processes are

referred to as executive functions. Executive functions include such processes as inhibition,

working memory, planning, self-monitoring, verbal regulation, motor control, maintaining and

changing mental set and emotional regulation. According to a current model of ADHD

developed by Dr. Russell Barkley, problems in response inhibition are the core deficit in ADHD.

This has a cascading effect on the other executive functions listed above (Barkley, 1997).

1.2 ELECTROENCEPHALOGRAPHY

Electroencephalography (EEG) is an electrophysiological monitoring method to record

electrical activity of the brain. It is typically non-invasive, with the electrodes placed along the

scalp, although invasive electrodes are sometimes used in specific applications. EEG measures

voltage fluctuations resulting from ionic current within the neurons of the brain. In clinical

contexts, EEG refers to the recording of the brain's spontaneous electrical activity over a period

of time, as recorded from multiple electrodes placed on the scalp. Diagnostic applications

generally focus on the spectral content of EEG, that is, the type of neural oscillations (popularly

called "brain waves") that can be observed in EEG signals. (Refer figure 1.2 and 1.3).



Figure 1.2: Electroencephalogram

EEG is most often used to diagnose epilepsy, which causes abnormalities in EEG

readings. It is also used to diagnose sleep disorders, coma, encephalopathy, and brain death. EEG

used to be a first-line method of diagnosis for tumors, stroke and other focal brain disorders, but

this use has decreased with the advent of high-resolution anatomical imaging techniques such

as magnetic resonance imaging (MRI) and computed tomography(CT). Despite limited spatial

resolution, EEG continues to be a valuable tool for research and diagnosis, especially when

millisecond-range temporal resolution (not possible with CT or MRI) is required.

Figure 1.3: An EEG setup

A routine clinical EEG recording typically lasts 20–30 minutes (plus preparation time)

and usually involves recording from scalp electrodes. Routine EEG is typically used in the

following clinical circumstances:

To distinguish epileptic seizures from other types of spells, such as psychogenic non-

epileptic seizures, syncope, sub-cortical movement disorders and migraine variants.



To differentiate "organic" encephalopathy or delirium from primary psychiatric syndromes

such as catatonia.

To serve as an adjunct test of brain death.

To prognosticate, in certain instances, in patients with coma.

To determine whether to wean anti-epileptic medications.

Additionally, EEG may be used to monitor certain procedures:

To monitor the depth of anesthesia.

As an indirect indicator of cerebral perfusion in carotid endarterectomy.

To monitor am barbital effect during the Wada test.

EEG can also be used in intensive care units for brain function monitoring:

To monitor for non-convulsive seizures/non-convulsive status epileptics

To monitor the effect of sedative/anesthesia in patients in medically induced coma (for

treatment of refractory seizures or increased intracranial pressure)

To monitor for secondary brain damage in conditions such as subarachnoid

hemorrhage (currently a research method)

1.2.1 EEG Signal Classification

EEG signals are basically classified into 4 types: Delta, Theta, Alpha and Beta. Recently,

two more types have been included – Mu and Gamma. Table 1.2 shows the classification,

frequency ranges and importance of different types of EEG signals.

1.2.2 Advantages

Several other methods to study brain function exist, including functional magnetic

resonance imaging (fMRI), positron emission tomography, magneto encephalography (MEG),

Nuclear magnetic resonance spectroscopy, Electrocorticography, Single-photon emission

computed tomography, Near-infrared spectroscopy (NIRS), and Event related optical signal

(EROS).



Type Frequency

(Hz)

Location Normally

Delta Up to 4 Frontally in adults,

posterior in children,

high amplitude waves

adults slow sleep wave

in babies

has been found during some continuous

attention tasks

Theta 4 - 8 Found in locations not

related to task at hand

young children

drowsiness or arousal in older children and

adults

idling

associated with inhibition of elicited responses

(has been found to spike in situations where a

person is actively trying to repress a response

or action).

Alpha 8 - 13 Posterior regions of

head, both sides, higher

in amplitude on non-

dominant side. Central

sites (C3-C4) at rest.

relaxed/reflecting

closing the eyes

Also associated with inhibition control,

seemingly with the purpose timing inhibitory

activity in different locations across the brain.

Beta >13 - 30 Both sides,

symmetrical

distribution, most

evident frontally, low

amplitude waves

alert/working

active, busy or anxious thinking, active

concentration

Gama 30 - 100+ Somatosensory cortex Displays during cross-modal sensory

processing (perception that combines two

different senses, such as sound and sight)

Also is shown during short term memory

matching of recognized objects, sounds, or

tactile sensations

Mu 8 - 13 Sensorimotor cortex Shows rest state motor neurons.

Table 1.2: Classification of EEG signals



Despite the relatively poor spatial sensitivity of EEG, it possesses multiple advantages over some

of these techniques:

Hardware costs are significantly lower than those of most other techniques

EEG prevents limited availability of technologists to provide immediate care in high traffic

hospitals.

EEG sensors can be used in more places than fMRI, SPECT, PET, MRS, or MEG, as these

techniques require bulky and immobile equipment. For example, MEG requires equipment

consisting of liquid helium-cooled detectors that can be used only in magnetically shielded

rooms, altogether costing upwards of several million dollars; and fMRI requires the use of a

1-ton magnet in, again, a shielded room.

EEG has very high temporal resolution, on the order of milliseconds rather than seconds.

EEG is commonly recorded at sampling rates between 250 and 2000 Hz in clinical and

research settings, but modern EEG data collection systems are capable of recording at

sampling rates above 20,000 Hz if desired. MEG and EROS are the only other noninvasive

cognitive neuroscience techniques that acquire data at this level of temporal resolution.

EEG is relatively tolerant of subject movement, unlike most other neuroimaging techniques.

There even exist methods for minimizing, and even eliminating movement artifacts in EEG

data.

EEG is silent, which allows for better study of the responses to auditory stimuli.

EEG does not aggravate claustrophobia, unlike fMRI, PET, MRS, SPECT, and sometimes

MEG.

EEG does not involve exposure to high-intensity (>1 tesla) magnetic fields, as in some of the

other techniques, especially MRI and MRS. These can cause a variety of undesirable issues

with the data, and also prohibit use of these techniques with participants that have metal

implants in their body, such as metal-containing pacemakers.

EEG does not involve exposure to radioligands, unlike positron emission tomography.

ERP studies can be conducted with relatively simple paradigms, compared with IE block-

design fMRI studies.

Extremely un-invasive, unlike Electro-corticography, which actually requires electrodes to

be placed on the surface of the brain.



EEG also has some characteristics that compare favorably with behavioral testing:

EEG can detect covert processing (i.e., processing that does not require a response).

EEG can be used in subjects who are incapable of making a motor response.

Some ERP components can be detected even when the subject is not attending to the stimuli

Unlike other means of studying reaction time, ERPs can elucidate stages of processing

(rather than just the final end result).

EEG is a powerful tool for tracking brain changes during different phases of life. EEG sleep

analysis can indicate significant aspects of the timing of brain development, including

evaluating adolescent brain maturation. Brain activity can also be monitored by Computed

tomography (CT scan).

In EEG there is a better understanding of what signal is measured as compared to other

research techniques, i.e. the BOLD response in MRI.

1.2.3 Disadvantages

Low spatial resolution on the scalp. fMRI, for example, can directly display areas of the

brain that are active, while EEG requires intense interpretation just to hypothesize what areas

are activated by a particular response.

EEG poorly measures neural activity that occurs below the upper layers of the brain (the

cortex).

Often takes a long time to connect a subject to EEG, as it requires precise placement of

dozens of electrodes around the head and the use of various gels, saline solutions, and/or

pastes to keep them in place. While the length of time differs dependent on the specific EEG

device used, as a general rule it takes considerably less time to prepare a subject for MEG,

fMRI, MRS, and SPECT.

Signal-to-noise ratio is poor, so sophisticated data analysis and relatively large numbers of

subjects are needed to extract useful information from EEG.

1.2.4 EEG vs. fMRI, fNIRS and PET

EEG has several strong points as a tool for exploring brain activity. EEGs can detect

changes over milliseconds, which is excellent considering an action potential takes



approximately 0.5-130 milliseconds to propagate across a single neuron, depending on the type

of neuron. Other methods of looking at brain activity, such as PET and fMRI have time

resolution between seconds and minutes. EEG measures the brain's electrical activity directly,

while other methods record changes in blood flow (e.g., SPECT, fMRI) or metabolic activity

(e.g., PET, NIRS), which are indirect markers of brain electrical activity. EEG can be used

simultaneously with fMRI so that high-temporal-resolution data can be recorded at the same time

as high-spatial-resolution data, however, since the data derived from each occurs over a different

time course, the data sets do not necessarily represent exactly the same brain activity. There are

technical difficulties associated with combining these two modalities, including the need to

remove the MRI gradient artifact present during MRI acquisition and the ballistocardiographic

artifact (resulting from the pulsatile motion of blood and tissue) from the EEG. Furthermore,

currents can be induced in moving EEG electrode wires due to the magnetic field of the MRI.

EEG can be used simultaneously with NIRS without major technical difficulties. There is

no influence of these modalities on each other and a combined measurement can give useful

information about electrical activity as well as local hemodynamic.

1.2.5 EEG vs. MEG

EEG reflects correlated synaptic activity caused by post-synaptic potentials of

cortical neurons. The ionic currents involved in the generation of fast action potentials may not

contribute greatly to the averaged field potentials representing the EEG . More specifically, the

scalp electrical potentials that produce EEG are generally thought to be caused by the

extracellular ionic currents caused by dendritic electrical activity, whereas the fields

producing magneto encephalographic signals are associated with intracellular ionic currents.

EEG can be recorded at the same time as MEG so that data from these complementary

high-time-resolution techniques can be combined.



CHAPTER - 2

LITERATURE SURVEY

Early reviews of studies of electrophysiological measures collected on hyperactive or ADHD

children concluded that the disorder was most likely associated with problems of under

reactivity to stimulation and task demands with less evidence supporting resting under

arousal in the disorder.[1-3]

In the largest EEG study of ADHD to date (with a sample of over 400 children), found that

children with ADHD displayed increased theta power, slight elevations in frontal alpha

power, and diffuse decreases in beta mean frequency.[5,6]

Studies have found mixed results for alpha power, with some studies showing it to be

increased, others showing it to be decreased, and still other studies finding it to be similar

between children with ADHD and normal controls.[4-6]

The findings for beta (indicative of heightened cortical arousal) activity have been less

consistent, with several studies finding decreased beta activity in frontal and central regions

and others not. There also appears to be a small group of ADHD children (~15%) who show

an excess of beta activity.[4-8,15]

Researchers suggested further research be conducted in which the tests elicit changes in the

higher frequency gamma wave range, which have been "intimately related to selective

attentional processes."

EEG abnormalities in male adolescents with ADHD during an eyes-closed resting condition

and found absolute dominance in delta and alpha activity and a higher theta/beta ratio

compared to the control subjects.[8,15]

Most of the studies concern the ADHD children and summarize that they have reduced

power in alpha and beta bands and an increased power in delta and theta bands in comparison

with the healthy control groups.[13]



Furthermore, Theta/Beta ratio has been diagnostic marker in ADHD children. The children

with ADHD have a higher theta/beta ratio than normal children.[14,15]

Electrophysiological measures were among the first to be used to study brain processes in

children with attention deficit hyperactivity disorder (ADHD). Early reviews of studies of

electrophysiological measures collected on hyperactive or ADHD children concluded that the

disorder was most likely associated with problems of under reactivity to stimulation and task

demands with less evidence supporting resting under arousal in the disorder. Collectively, the

EEG findings in children, adolescents, and adults with ADHD are increased slow-wave

activity in frontal regions, suggesting cortical hypo arousal, especially in the ADHD

Combined subtype. Several researchers have reported that EEG measures discriminate well

between children with and without ADHD.[16]

Attention deficit disorder with or without hyperactivity (ADHD) has the prevalence rate of 6-

9% among school age children within United States and persists in adolescence in 70% and

in adulthood in 40-60%. In the largest EEG study of ADHD to date (with a sample of over

400 children), found that children with ADHD displayed increased theta power, slight

elevations in frontal alpha power, and diffuse decreases in beta mean frequency.[17]

EEG studies of children with ADHD have typically found increased theta activity, increased

posterior delta, decreased alpha and beta and an increase in the theta/beta ratio compared to

normal children. This study investigated the presence of EEG subtypes of children with the

combined type of ADHD. Results indicated the presence of 3 distinct EEG-defined clusters

of children, characterized by increased slow wave activity and deficiencies of fast wave,

increased high amplitude theta with deficiencies of beta activity, and a beta-excess

group.[18]

The findings for beta (indicative of heightened cortical arousal) activity have been less

consistent, with several studies finding decreased beta activity in frontal and central regions

and others not. There also appears to be a small group of ADHD children (~15%) who show

an excess of beta activity .Studies have found mixed results for alpha power, with some

studies showing it to be increased, others showing it to be decreased, and still other studies

finding it to be similar between children with ADHD and normal controls.[19]



This study investigated age-related changes and sex differences in the EEGs of two groups of

children with attention-deficit/hyperactivity disorder (ADHD) combined type and ADHD

predominantly inattentive type, in comparison with a control group of normal children. With

increasing age, the EEG of the ADHD inattentive group was found to change at a similar rate

to the changes found in the normal group, with the differences in power levels remaining

constant. In the ADHD combined group, the power was found to change at a greater rate than

in the ADHD inattentive group, with power levels of the two ADHD groups becoming

similar with age.[20]

Quantitative electroencephalogram (QEEG) findings indicated significant maturational

effects in cortical arousal in the prefrontal cortex as well as evidence of cortical slowing in

both ADHD groups, regardless of age or sex. Sensitivity of the QEEG-derived attentional

index was 86%; specificity was 98%. These findings constituted a positive initial test of a

QEEG-based neurometric test for use in the assessment of ADHD. Furthermore, Theta/Beta

ratio has been diagnostic marker in ADHD children. The children with ADHD have a higher

theta/beta ratio than normal children.[21]

The authors have considered the fMRI data as an example to illustrate the utility of fully

connected cascade (FCC) artificial neural networks (ANN) architecture for performing

classification. They employed various directional and non directional brain connectivity-

based methods to extract discriminative features which gave better classification accuracy

compare to raw data. Their accuracy for distinguishing ADHD from healthy subjects was

close to 90% and between the subtypes was close to 95%. Further, they showed that, If

properly used FCC ANN performs very well compare to other classifiers.

2.1 MOTIVATION

According to the literature survey and surveys done by the recognized organizations,

many facts were found out. The prime motivations of this project are as follows:

Over 5 million children in this country are diagnosed with attention deficit hyperactivity

disorder (ADHD) and of those children, about 3 million are medicated each year.

It is a common problem seen in about 3 to 10 per cent of school age children. It is 4-6 times

commoner in boys as compared to girls (l, 4).



According to an ASSOCHAM survey, the number of such cases in the country has gone up

from 4% in 2005 to 11% in 2011.

In a survey undertaken about 6 years ago that covered 2,300 school children, it was found

that 4-5% of city children in the age group of 6-9 years were suffering from ADHD.

Several studies estimated the prevalence of ADHD, in USA 4-8%, Korea 7.6% to 9.5%,

India 20%, and Emirates 29.7% in United Arab8-11.

In many cases this disorder goes totally undiagnosed.

2.2 OBJECTIVES

In relation with EEG signals, a lot of works has been done to find the significant

difference between the neural activity of ADHD and normal subjects by using different

methodologies. This project is mainly designed to predict the probable region of brain that shows

abnormality due to ADHD syndrome. In this process, frequency characteristics of ADHD and

normal subjects are extracted and compared to differentiate between the each. Several analyses

are conducted to find out the probable region of abnormality and finally, 3-D plotting is

performed for ease of visualization.

2.3 CHALLENGES

Following are the challenges which we faced during this project work:

Finding the special schools which train kids with ADHD and taking the consent of

parents and the organization. This is because, medical data is purely confidential and

it is unavailable even in the public websites and government organizations. Also, the

parents of such kids mostly had come from a poor background and were panic when

they had to expose their wards for such studies.

Finding purely ADHD children since many of them were suffering from multiple

disorders like mental retardation, intellectual disability etc.

Acquisition of EEG data.

Data Segmentation.

Calculation of Statistical Parameters

Prediction of probable region of brain which is responsible for ADHD.



CHAPTER - 3

METHODOLOGY

The process involved in this project work is depicted in the flowchart shown in figure

3.1.

Figure 3.1: Flowchart

The complete information about the steps followed is explained further.

PARAMETER CALCULATION

COMPARATIVE ANALYSIS TODETECT ABNORMALITY

CONVERSION OF 2- D TO 3-DPLOTTING

PREDICTION OF PROBABLE REGIONOF ABNORMALITY



CHAPTER - 3

METHODOLOGY


3.1.



SUBJECT

EEG DATA ACQUISITION

DATA SEGMENTATION

FEATURE EXTRACTION


COMPARATIVE ANALYSIS TODETECT ABNORMALITY





CHAPTER - 3

METHODOLOGY


3.1.





3.1 SUBJECTS

Subject means the child who has been considered for the studies in this project. So, here 7

normal and 7 ADHD children were identified and considered for the research. An interview was

conducted to the parents/ guardians of ADHD subjects as a part of this project work. The

questionnaire and answers obtained are as follows:

3.1.1 Questionnaire

1. When did you notice the syndrome at first?

Majority of the parents said, they noticed it between the age 7 months and 1 and ½ years,

but many were not educated about it and couldn’t recognize the symptoms. Many said,

they thought it’s normal for kids to be hyper-active and that’s the reason, it took time for

them to notice and bring into the notice of doctor.

2. How did you notice it?

Five out of 8 kids witnessed high fever accompanied by epilepsy which appears to be the

root cause for the syndrome. 2 out of 8 said, their kid had slow growth and had no clarity

in speech, even at the age of 3. 4 out of 8 said, that the kids were hyper-active (like

running, screaming, not listening to anybody, complaints from schools etc).

3. What sort of behaviors do the children with this syndrome have?

All 8 parents mentioned that the kids were –

Uncontrollable

Screaming

Running

Cannot stay in a single place for more than few seconds

Poor-concentration and less understanding capability

Unable to identify the person, until he/she becomes very familiar and is a

regular visitor. But in some cases, the kids weren’t able to do so, even after

many visits.

Extremely high energy

Takes time to remember things



Gets too angry, sometimes, hurts himself or others

No observation ability

Likes noise

Problem in speech

Cannot differentiate between good and bad

Slow-learner

Stubborn

While, some of the parents said, their ward talks indefinitely and repeats whatever was

told by others.

4. What interests the kid with ADHD?

This question gave us variety of answers from parents as each of them had a variety of

interests. Some of them are, listening to music, painting, riding bikes, computer, talking

over a phone call, cooking, keeping things neat, paper cutting and scribbling.

5. Is she/he under the medication? If yes, what is the outcome of it?

Yes, all the kids were under medication. The common answer about the outcome was

that, the kid kept calm and slept when he was under medication (i.e., when he/she

consumed tablet), but remained to be hyperactive, once the medicines lost its power.

6. Does anybody in the family have this disorder?

Here, we wanted to know, if the disease was genetic. But, the answer was “No”. None of

them in the family had previously had this disorder.

7. Is the kid social?

6 out of 8 said that their kid wasn’t social. And 2 out of 8 kids were scared to mingle with

people, but could make an attempt after several visits.

8. Are these kids independent at their works?

6 out of 8 were dependent on others for their day-to-day activities.

9. Are the kids eligible for day-to-day transactions- like shopping, billing etc?

No. The kids were not able to do the daily transaction by themselves, since they weren’t

able to count things or cash and analyze about it by themselves.



10. Is the syndrome curable? What did the doctors say after diagnosis?

No, there wasn’t any permanent solution for the problem and it can only be controlled for

some amount of time with the help of pills. The kids will carry the same to their

adulthood too.

3.1.2 Subject details

Sl.

No.

Name of the

subject

Age( in

years)

Whether under

medication(Yes/No)

Disabilities Period of

Training

1 Chetan Naik 11 Yes ADHD 3 years

2 Jayesh 9 Yes ADHD 6 months

3 Vinay 10 Yes ADHD 6 months

4 Jyothi 14 Yes ADHD + Mild MR 5 years

5 Siddhanth Patil 9 No ADHD + Mild MR 2 years

6 Dileep gowda 9 Yes ADHD + MR 4 years

7 Chandan 9 No ADHD + MR 5 years

Table 3.1: Details of ADHD subjects

Subjects of age group 4-17 were chosen for this study. Normal subjects were those who

weren’t suffering from any disorders and who were not under any medications, i.e., children who

were healthy. ADHD subjects are those who were affected by ADHD. Here, it became difficult

to identify a pure ADHD case, therefore kids with multiple disabilities like Mental retardation,

intellectual disability and so on were also considered. All these details along with their age is

given in tables 3.1 and 3.2.



Sl. No. Name of the subject Age(in years) Whether under Medication

(Yes/No)

1 Dhruthi.N.R 12 No

2 Dhruthi.L 8 No

3 Darshan.L 13 No

4 Vishwas Gowda.C 11 No

5 Vikas Gowda.C 11 No

6 Likitha.N 13 No

7 Sumanth Gowda.N 10 No

Table 3.2: Details of Normal Subjects

3.2 EEG DATA ACQUISITION

EEG signals of 7 normal and 7 ADHD subjects were obtained according to International

standard 10-20 electrode system using BrainTech traveller portable EEG device with 24 channel

digital EEG machine having and A/D conversion of 16bits with the sampling frequency of 256

Hz. Here, 10-20 denotes the placement of electrodes in a particular manner as shown in figure

3.2.

Figure 3.2: International 10-20 electrode system



3.2.1 Braintech Traveler portable EEG device

Braintech traveler portable EEG device is the one, which:

Operates with USB Power.

Most user-friendly tool bar.

Online/Offline reformatting of Gain, Filter, Sweep Speed.

EEG Flash with High intensity white LED's with built in rechargeable Li-ion battery.

Facility of editing and marking of events and comments Online/Offline.

Highly Advanced Database management with different searching options.

Figure 3.3 shows the portable braintech traveller EEG device

Figure 3.3: Braintech traveller portable EEG device

3.2.1.1 Configurations

1. General:

It includes Com port (port compatible with this software), sensitivity (or gain), Low pass

filter (LPF); High pass filter (HPF), Notch, Muscle filter, Deflection, system frequency

(Computer) and Post HV time. User can also set the calibration frequency, calibration level,

calibration type. Montage sequencing (automatic selection of montages as per the user defined

‘Set’ during recording) can be ‘On/Off’ and sequencing set can be selected from the list (User

defined set of montages). The last option ‘Channel for Email Compression’ is to use Email

Utility. User can select one of the three options whichever is suitable, for which he wants to send

the Email, it reduces the size and weight of the data.



2. Display:

User can change the display configuration as per its own requirement. If user wants to

show major grid, minor grid, negative up etc, he can tick the check box. Also user can change the

colors of various options. User can change the position of the Montage bar, but left position is

preferable. Also, can On/Off Recorded mode parameter.

3. User Definable:

In this user can fill the Clinic details and set the sensitivity, low pass filter and high pass

filter as per the test requirements.

4. Printer:

Here, user can set its own printer settings like ‘Print major, minor grid, montage pictorial,

control bar, montage on right, patient photo and footnote. User can print the page in black &

white also.

Technical specifications of the device are as follows:

SL.NO. PARAMETERS

1. Operating Temperature 5 to 45 0C

2. Operating Humidity 5 to 95% RH, non-condensing

3. Storing Temperature 0 to 55 0C

Table 3.3: Operating environment of Braintech traveller

3.2.1.2 General parameters

SL.NO. PARAMETERS

1. No. of Input Channels 24

2. No. of Display Channels 24

3. Frequency Band Range 0.1 to 100 Hz

4. Input Impédance > 10 M

5. Noise < 1 µV P-P

6. Input Range ± 8 mV

7. CMRR > 95 dB



8. Sampling Rate 1024 Hz at ADC, 256 Hz for internal storage

9. A/D Conversion 16 bit

10. Impedance Measurement Automatic, Built-in Yes

11. High Cut Filter (Digital) 0.5 – 100 Hz

12. Low Cut Filter (Digital) 0.2– 7 Hz

13. Notch Filter (Digital) 50 / 60 Hz

14. Time Resolution 7.5, 15, 30, 60 mm/s

15. Power Supply USB Power (With built-in rechargeable Li-ion

Battery)

16. Photo Stimulation High intensity white LEDs

17. Flash Rate 1 – 60 Hz

Table 3.4: General parameters of Braintech traveller device

3.2.1.3 More features

There are more features provided to the user for the convenience of the user and for easy

use.

1. Video On/Off: In EEG Acquisition mode, if the user use the video facility of the EEG

machine to record the video recording of the patient, then the video of the patient is

shown in Analysis mode (By default). The Video of the patient is shown in two ways:

a) Video window On/Off: User can 'On' the video screen by clicking the Video shortcut,

if he wants to study the patient’s behavior by watching the video recording.

b) Splitted screen: By using this option screen is divided into two parts one is of

waveform and other shows the video recording of the patient.

2. Set Transparency: To view the recorded screen fully user can set the transparency

level as per its own need. The result of this, Montage pictorial, Control Bar and Montage

Bar will appear transparent.

3. Mark Page: User can mark the specific pages for further use. Marked pages can be

print/Saved as per the requirement to review the recorded data.



4. Print Options: If User wants the hard copy of the Report/Marked Pages/Current

page, User can print the data. User can also save the data.

5. Auto Speed: Auto speed is the option to set the speed at which the recorded screen

moves in forward (Continuous) or backward direction (Backward Scroll). User can set

the speed as per its own need by using the screen shortcut 'Auto Speed'.

6. Patient Report: Patient Report is the summary of the patient’s records written by the

technical person/Doctor for further reference. User can edit the reports in the 'Patient

Report(s)' form and also can print/Save the report as per its own need. Report name can

be changed. Also the selected report can be attached to the current patient. Attached

report can be printed.

3.2.2 Electrode positions

Electrode positions have been named according to the brain region below that area of the

scalp: frontal, central (sulcus), parietal, temporal, and occipital. In the bipolar method, the

EEG is measured from a pair of scalp electrodes. The pair of electrodes measures the difference

in electrical potential (voltage) between their two positions above the brain. A third electrode is

put on the earlobe as a point of reference, 'ground', of the body's baseline voltage due to other

electrical activities within the body. There are different types of montages available. They are:

Bipolar montage: Each channel (i.e., waveform) represents the difference between two

adjacent electrodes. The entire montage consists of series of these channels.

Referential montage: Each channel represents the difference between a certain electrode

and a designated reference electrode.

Average reference montage: The outputs of all the amplifiers are summed and averaged,

and this averaged signal is used as the common reference for each channel.

Laplacian montage: Each channel represents the difference between an electrode and a

weighted average of the surrounding electrodes.



In this project work, Longitudinal bipolar montage is used which is shown in figure 3.4.

Figure 3.4: Longitudinal Bipolar montage

Channel number Longitudinal bipolar double banana

1 FP1-P3

2 F3-C3

3 C3-P3

4 P3-O1

5 FP2-F4

6 F4-C4

7 C4-P4

8 P4-O2

9 FP1-F7

10 F7-T7

11 T7-P7

12 P7-O1

13 FP2-F8



14 F8-T8

15 T8-P8

16 P8-O2

17 FZ-CZ

18 CZ-PZ

19 EKG

Table 3.5: Channels of Longitudinal bipolar montage

During the data acquisition, many noises get added to the actual data. They are:

Subject Dependent noises- like Scalp impedance, Physical movement of electrodes,

chewing, Blinking etc.

Environmental noises – like AC line current, video monitors, AC light, Pagers etc.

In order to minimize this noise, several measures were taken and they are as follows:

1. Hair, Hair products, sweat and other scalp debris are the sources of impedance. To

minimize this, scalp abrasions were used and conductive gel was applied.

2. In order to reduce the environmental noises, acoustically and electrically shielded cables

were used. Notch filter is used to remove the power line interference and to reduce the

distortion of signal of interest.

3.2.3 Protocol

In educational research many studies share the common goal of assessing learning or

performance, and in this chapter provide information on methods for collecting and analyzing

learning and performance data. Even though learning and performance are conceptually

different, many of the data collection and analysis techniques can be used to assess both.

Selection of ADHD affected subject is the most crucial task. Once we have the subjects,

EEG data is collected by the help of EEG data acquisition device.

To collect data is easy but without a common protocol analysis of same is

difficult.

Since we cannot predict the thinking of a person, about what and when he thinks,

we need to maintain a protocol in order to favor the analysis based on the thought



processing of different persons or subjects on the same object or task.

Here we used common protocol of eyes close and eyes open on the subjects.

We requested the subjects to open or close their eyes for few seconds in alternating sets of two

with varying time record length. This helps us to study the alpha wave activity.

The second step in protocol was visual cue. In this visual cue we made the subjects

whose data we are collecting, to watch a slide show. This slide show consisted of some common

aspects such as colors, numbers, and animals which are identifiable by both normal and

abnormal subjects. There were total 23 slides with three distinguishing slides between the

transitions of subject of the slides. Each of the slides had the duration of 2 seconds and making

total of 46 seconds of EEG data. We studied the delta, theta, alpha, and beta waves behavior.

The next step in data collection protocol was motor activity, where all the subjects in the

end were asked to arrange the spoons and hand it over to their care taker. In this step the subjects

took their own time, but we were interested in the thought processing.

These are the protocols followed when we collected the data from both type of subjects,

normal and abnormal.

The figure 3.5 shows the pictorial representation of protocol used.

Figure 3.5: Protocol used

3.3 DATA SEGMENTATION

Data segmentation is the process of taking the data i.e., EEG signal and segmenting it so

that it can be easy to analyze the brain activities.

As mentioned earlier that, EEG signals of ADHD and normal children are collected for

analysis. The data is acquired on 24 channels. We selected data manually with minimum

redundancy. The electrode positions were also selected manually based on the interest of our

study. We found that there were many artifacts. During data acquisition, there was a lot of noise

induced. The noises are mentioned in earlier section.

Eyes close Eyes open Visual Cue Motor Activity



Therefore, to eliminate these artifacts and the selection of electrode positions, the data

segmentation is done manually using markers during acquisition of data. The markers were

inserted based on events and timings during the data acquisition.

As mentioned in the protocol, we had conducted following events:

Eyes Close

Eyes Open

Visual Cue

Motor activity

The markers were inserted by noting down the timings. Like, in visual cue we had given

the children to identify few things, such as, Animals, Numbers and Colours. While the children

were identifying these things the timings were noted down and accordingly the markers were

inserted. The timings for each event in visual cue are 2 seconds. Similarly, for eyes close and

eyes open around 3-5 seconds and for motor activity about 1-2 minutes. The electrode positions

which are selected based on our requirement are:

1. FP1

2. FP2

3. P3

4. P4

5. O1

6. O2

7. T3

8. T4

The reason for selecting these electrode positions is:

Frontal- FP1and FP2 contains voluntary motor functions and areas for planning, mood,

smell and social judgment.

Parietal- P3 and P4 contain the area for sensory reception and integration of sensory

information.

Occipital- O1 and O2is visual centre of brain.

Temporal- T3 and T4 contains areas of hearing, smell, learning, memory, emotional

behavior.



Therefore, these data were selected and segmented according to the area of our interest,

which made the analysis easier and simpler. Hence, the requirement of data segmentation is

specified.

3.4 FEATURE EXTRACTION

Different features of EEG signals such as frequency bands (alpha, beta, delta and theta) ,

energy corresponding to the same, power and statistical parameters such as mean, SD and

variance were calculated.

3.4.1 Energy calculation

Energy is calculated using FFT algorithm. A Fast Fourier Transform (FFT)

algorithm computes the Discrete Fourier Transform (DFT) of a sequence, or its inverse. Fourier

analysis converts a signal from its original domain (often time or space) to a representation in

the frequency domain and vice versa. An FFT rapidly computes such transformations

by factorizing the DFT matrix into a product of sparse (mostly zero) factors.

Fast Fourier Transform (FFT) algorithms have computational complexity O(n log n)

instead of O(n2). When using FFT algorithms, a distinction is made between the window

length and the transform length. The window length is the length of the input data vector. It is

determined by, for example, the size of an external buffer. The transform length is the length of

the output, the computed DFT. An FFT algorithm pads or chops the input to achieve the desired

transform length. The following figure 3.6 illustrates the two lengths.

x y

Window Transform Transformlength length

m ny = fft ( x , n )

Figure 3.6: Block diagram of FFT algorithm

The execution time of an FFT algorithm depends on the transform length.

Let x0, ...., xN-1 be complex numbers. The DFT is defined by the formula

FFT

BUFFER PAD/CHOP EFFICIENT DFT



X = x e π k = 0, . . . . , N − 1Evaluating this definition directly requires O(N2) operations: there are N outputs Xk, and

each output requires a sum of N terms. An FFT is any method to compute the same results in

O(N log N) operations.

Steps to compute energy calculations are as follows:

1. The raw EEG signal should be in time domain.

2. EEG is nothing but a time series data. Hence, FFT (EEG data) will give the frequency

domain representation of the time domain EEG signal.

3. We use a specific band pass filter belonging to the frequency range of Alpha, Beta, Delta

and Theta bands.

4. The frequency ranges are:

i. Delta band : 0.0 – 4.0 Hz

ii. Theta band : 4.0 – 8.0 Hz

iii. Alpha band : 8.0 – 12.0 Hz

iv. Beta band : 12.0 – 25.0 Hz

5. After specifying the frequency ranges and filtering the data by using FFT algorithm we

compute Energy for corresponding amplitude values for each sample of each event and of

each electrode position.

3.4.2 Statistical parameters calculation

Statistical parameter is an important component of any statistical analysis. In simple

words, a parameter is any numerical quantity that characterizes a given set of data or some aspect

of it. This means the parameter tells us something about the whole data.

The different types of parameters that we have calculated are:

a) Mean



b) Variance

c) Standard deviation

d) Absolute power

e) Relative power (%)

f) Amplitude levels

g) Energy

Mean: The mean is the arithmetic average of a set of data and the geometric mean. The mean

gives someone an idea where the centre lies for a set of data. The arithmetic mean is obtained

by taking the sum of all the numbers in the set and dividing by the total number of samples in

the set of data.

= ∑Variance: Variance is the square of the standard deviation. It has an intimate mathematical

relationship with standard deviation. Variance is defined as the average of the square of the

deviations of a set of data from their mean. =Where,

σ is the standard deviation

Standard deviation: The standard deviation is a veritable bulwark in the sea of statistics.

Along with the mean, the standard deviation is a theoretical cornerstone in inferential

statistics. The standard deviation gives an approximate picture of the average amount each

number in a set varies from the centre value.

= ∑( − )Energy: Energy is mathematically defined as the multiplication of time with power. It

indicates the strength of the signal as it gives the area under the curve at any interval of time.

Where,

= the mean∑ = the sum of all the scores in set

N = the number of scores or observation in the set



= | ( )|Algorithm used to compute these parameters are:

1. Importing raw data from excel

2. Selection of electrodes - transpose of matrix (For ease of observation)

3. Data segmentation according to events - Feature extraction

4. Calculation of mean, variance, standard deviation, absolute power, relative power

percentage and exporting the same to excel.

5. FFT algorithm is used for the calculation of Energy corresponding to Delta, Theta, Alpha

and Beta.

3.5 ANALYSIS

A systematic examination and evaluation of data or information, by breaking it into

its component parts to uncover their interrelationships (or) an examination of data and facts to

uncover and understand cause-effect relationships, thus providing basis for problem

solving and decision making. This is called Analysis.

The analysis is done based on brain maps. This advanced neuro-diagnostic technology

measures the different frequencies of electrical activity that are generated from the brain. These

frequencies called delta, theta, alpha and beta provide information concerning the active state of

various regions of the brain, effectively producing a map of the brain’s activity.

There are various types of brain maps to study, like:

Single Map

Tri Maps

Frequency Maps

Frequency Spectrum

Progressive Amp Maps

Progressive Freq. Maps

Frequency Tables

User Frequency Bands



In our project we have analyzed the segmented data in different ways. The different types of

analysis are:

1. Single map analysis

2. Frequency map analysis

3. Frequency tables

3.5.1 Single Map Analysis

In single map analysis the amplitude levels of the EEG data is obtained. The amplitude

levels are measured for each event as per the protocol used in the project. The unit for the

measured amplitude levels is in micro volts (µV).These amplitude levels indicate the brain

activities. The amplitude varies from person to person. It may be positive or negative values.

3.5.2 Frequency Map Analysis

In frequency map analysis, obtained data is converted from 2D to 3D plot. The power

percentage is also analyzed here. It shows the lightning or shooting of signal in the region where

the brain activity is taking place. The range of amplitude levels is given in the form of colours.

As shown in figure 3.7, the colour band at the right hand side represents the power levels. The

lower power is shown wherever the brain activities are suppressed and higher power represents

that the brain is active in those regions. It is measured in µV2/Hz.

The below figure represents brain activities for:

delta (1-4 Hz)

theta (4- 8 Hz)

alpha (8-12 Hz)

beta (12-25 Hz)

high beta (18-25 Hz)

mid beta (15-18 Hz)

Low beta (12-15 Hz).



Figure 3.7: Frequency Maps

3.5.3 Frequency Tables

Frequency table represents the absolute power (µV2) and relative power percentage.

These values are calculated for different frequency ranges as mentioned earlier. The values are

computed for all the electrodes chosen as our area of interest. The powers are calculated for all

the events mentioned in the protocol of the project. Figure 3.8 represents sample of Frequency

tables.

Figure 3.8: Frequency tables



3.6 NEURAL NETWORK ALGORITHM AND ITS IMPORTANCE

A neural network algorithm is a machine learning approach inspired by the way in which

the brain performs a particular learning task. Knowledge about the learning task is given in the

form of examples. Inter neuron connection strengths (weights) are used to store the acquired

information (the training examples).During the learning process the weights are modified in

order to model the particular learning task correctly on the training examples.Neural network are

they used for:

Classification: Pattern recognition, feature extraction, image matching

Noise Reduction: Recognize patterns in the inputs and produce noiseless outputs

Prediction: Extrapolation based on historical data

Ability to learn: Neural Network's (NN) figure out how to perform their function on their

own.

Determine their function based only upon sample inputs.

Ability to generalize: i.e. produce reasonable outputs for inputs it has not been taught

how to deal with. The weights in a neural network are the most important factor in

determining it function. Training is the act of presenting the network with some sample

data and modifying the weights to better approximate the desired function.

There are two main types of training

Supervised Training: Supplies the neural network with inputs and the desired outputs.

Response of the network to the inputs is measured. The weights are modified to reduce

the difference between the actual and desired outputs.

Unsupervised Training: Only supplies inputs. The neural network adjusts its own

weights so that similar inputs cause similar outputs. The network identifies the patterns

and differences in the inputs without any external assistance.

Epoch: One iteration through the process of providing the network with an input and updating

the network's weights. Typically many epochs are required to train the neural network

Hidden Layers and Neurons: For most problems, one layer is sufficient. Two layers are

required when the function is discontinuous. The number of neurons is very important: Too few

Under fit the data – NN can’t learn the details. Too many Over fit the data – NN learns the

insignificant details.



Training and Verification: There are two different ways in which training can be implemented:

incremental mode and batch mode. In incremental mode, the gradient is computed and the

weights are updated after each input is applied to the network. In batch mode, all the inputs in the

training set are applied to the network before the weights are updated. For most problems, when

using the Neural Network Toolbox™ software, batch training is significantly faster and produces

smaller errors than incremental training

The set of all known samples is broken into two orthogonal sets as shown in figure 3.9:

Training set: A group of samples used to train the neural network

Testing set: A group of samples used to test the performance of the neural network

Figure 3.9: Training and Verification

3.6.1 Why Neural Networks?

Neural networks are used to distinguishing ADHD from healthy subjects. Several

researches have been conducted with this and artificial neural networks show the best

performance in classification of ADHD and healthy subjects.

3.6.2 Feed forward back propagation:

The back propagation learning algorithm can be divided into two phases: propagation and

weight update.

Phase 1: Propagation

Each propagation involves the following steps:

Forward propagation of a training pattern's input through the neural network in order to

generate the propagation's output activations.

Known Samples

Trainingset

Testingset



Backward propagation of the propagation's output activations through the neural

network using the training pattern target in order to generate the deltas (the difference

between the targeted and actual output values) of all output and hidden neurons.

Phase 2: Weight update

For each weight-synapse follow the following steps:

1. Multiply its output delta and input activation to get the gradient of the weight.

2. Subtract a ratio (percentage) of the gradient from the weight.

This ratio (percentage) influences the speed and quality of learning; it is called the

learning rate. The greater the ratio, the faster the neuron trains; the lower the ratio, the more

accurate the training is. The sign of the gradient of a weight indicates where the error is

increasing; this is why the weight must be updated in the opposite direction. Repeat phase 1 and

2 until the performance of the network is satisfactory.

Most common measure of error is the mean square error:

E = (target – output)^2

3.6.3 Network architectures:

The architecture of a neural network is linked with the learning algorithm used to train.

Three different classes of network architectures

Single-layer feed-forward

Multi-layer feed-forward

Recurrent

In these three class neurons are organized in acyclic layers

3.6.3.1 Single Layer Feed-forward:

The simplest kind of neural network is a single-layer perceptron network, which consists

of a single layer of output nodes; the inputs are fed directly to the outputs via a series of weights.

In this way it can be considered the simplest kind of feed-forward network. The sum of the

products of the weights and the inputs is calculated in each node, and if the value is above some



threshold (typically 0) the neuron fires and takes the activated value (typically 1); otherwise it

takes the deactivated value (typically -1). It is shown in figure 3.10.

Figure 3.10: Single Layer Feed-forward

3.6.3.2 Multi-layer feed-forward:

The Multi-layer feed forward is divided into three layers: the input layer, the hidden layer

and the output layer, where each layer in this order gives the input to the next. The extra layers

give the structure needed to recognize non-linearly separable classes. Refer figure 3.11.

X

X

Y=SOM

X

X

Input Layer Hidden Layer Output Layer

Figure 3.11: Multi-layer feed-forward



3.6.3.3 Recurrent network:

A recurrent neural network (RNN) is a class of artificial neural network where

connections between units form a directed cycle. This creates an internal state of the network

which allows it to exhibit dynamic temporal behavior. Unlike feed forward neural networks,

RNNs can use their internal memory to process arbitrary sequences of inputs. It is shown in

figure 3.12.

Elemental wise

addition

Activation function

Routes information

can propagate

Involved in modifying

information flow

and values

Figure 3.12: Recurrent network

3.6.4 The Neuron:

The neuron is the basic information processing unit of a NN. It consists of:

1) A set of synapses or connecting links, each link characterized by a weight:

W1, W2, …, Wm.

2) An adder function (linear combiner) which computes the weighted sum of the

inputs: u=∑ Wj kj3) Activation function (squashing function) for limiting the amplitude of the

output of the neuron. Y= ∂(u+b).

The Neural Network Diagram is shown in figure 3.13.

Figure 3.13: The Neural Network Diagram

hi-1 hi hi+1

xi xi+1



A neuron receives input from many other neurons, changes its internal state (activation)

based on the current input, sends one output signal to many other neurons, possibly including its

input neurons (NN is recurrent network). Back-propagation is a type of supervised learning, used

at each layer to minimize the error between the layer’s response and the actual data.

3.6.5 Performance plot and Training state plot:

From the training window, we can access four plots: performance, training state, error

histogram, and regression. The performance plot shows the value of the performance function

versus the iteration number. It plots training, validation, and test performances. The training state

plot shows the progress of other training variables, such as the gradient magnitude, the number

of validation checks, etc. The error histogram plot shows the distribution of the network errors.

The performance plot is shown in figure 3.14.

Figure 3.14: Performance plot



3.6.6 Error Histogram:

The histogram can give you an indication of outliers, which are data points where the fit

is significantly worse than the majority of data. The blue bars represent training data, the green

bars represent validation data, and the red bars represent testing data. Refer figure 3.15 for error

histogram.

Figure 3.15: Error Histogram

3.6.7 Regression plot:

The following regression plots display the network outputs with respect to targets for

training, validation, and test sets. For a perfect fit, the data should fall along a 45 degree line,

where the network outputs are equal to the targets. For this problem, the fit is reasonably good

for all data sets, with R values in each case of 0.93 or above. If even more accurate results were

required, we could retrain the network by trying using these approaches:

Reset the initial network weights and biases to new values with init and train again.

Increase the number of hidden neurons.

Increase the number of training vectors.

Increase the number of input values, if more relevant information is available.

Try a different training algorithm.

This will change the initial weights and biases of the network, and may produce an

improved network after retraining. In this case, the network response is satisfactory, and we

can put the network to use on new input. It is as shown in figure 3.16.



Figure 3.16: Regression plot Figure 3.17: Confusion plot

3.6.8 Confusion plot:

On the confusion matrix plot, the rows correspond to the predicted class (Output Class),

and the columns show the true class (Target Class). The diagonal cells show for how many (and

what percentage) of the examples the trained network correctly estimates the classes of

observations. That is, it shows what percentage of the true and predicted classes match. The off

diagonal cells show where the classifier has made mistakes. The column on the far right of the

plot shows the accuracy for each predicted class, while the row at the bottom of the plot shows

the accuracy for each true class. The cell in the bottom right of the plot shows the overall

accuracy.

Note: Increasing the learning rate decreased the training time. Increasing the epoch size, the

number of samples per epoch, decreased the number of epochs required and seemed to aid in

convergence (reduced fluctuations). Increasing the number of hidden neurons decreased the

number of epochs required. Each time a neural network is trained, can result in a different

solution due to different initial weight and bias values and different divisions of data into

training, validation, and test sets. As a result, different neural networks trained on the same

problem can give different outputs for the same input. To ensure that a neural network of good

accuracy has been found, retrain several times. Refer figure 3.17 for confusion plot.



CHAPTER - 4

RESULTS

4.1 SINGLE MAP ANALYSIS

Comparative analysis of EEG signals between the ADHD and Normal subjects yields

the following results as shown in table 4.1.

EVENTS NORMAL SUBJECTS ADHD SUBJECTS

EYES CLOSE

Lower amplitude levels.

Lower SD when compared to

the ADHD subjects.

Higher SD in both FP2 (50%)

and FP1 (50%).

Higher Amplitude levels.

Higher SD for those who

are extremely hyper-

active, but relatively less

for the trained subjects.

EYES OPEN

Higher amplitude levels.

SD is relatively lower than the

ADHD subjects.

SD is higher in FP1 position.

Lower amplitude levels.

Higher SD for those who

are extremely hyper-

active, but relatively less

for the trained subjects.

VISUAL CUE

Higher amplitude in frontal

region and relatively lower

amplitude levels in occipital

region.

Higher amplitude in

occipital region when they

couldn’t identify.

Higher amplitude in

frontal region when they

could identify.

MOTOR ACTIVITY

Higher amplitudes are observed

in almost all the electrodes, but

highest in electrode positions

FP2 and O2.

Higher amplitude is

observed only in the

frontal region.

Table 4.1: Single map analysis



1. Eyes close:

The table 4.2 provides the range of Standard deviation for ADHD and normal subjects for

electrode positions FP1, FP2, O1, O2, P3, P4, T3 & T4 for eyes close position. As observed

in the table, lower SD values are observed for normal subjects when compare to ADHD. This

is one such parameters which helps in differentiating between normal and ADHD subjects.

Electrode positions Normal subject ADHD subject

FP1 4-15 9-130

FP2 5-11 7-116

O1 5-17 6-130

O2 5-10 6-62

P3 3-7 5-31

P4 4-8 5-33

Table 4.2: Standard deviation table for eyes close event

From the table we infer that the amplitude varied from -30 to 9µV for normal subjects,

whereas, in ADHD subjects we observed that the amplitude levels were very high and it

varied from -30 to 70µV. The higher amplitudes correspond to higher neuronal activities.

In eyes close position, usually the EEG amplitudes gets suppressed to lower positive

values or negative values for a normal subject, whereas, here there is a significant

increase in the amplitude level for an ADHD subject.

2. Eyes open:


FP1 5-42 15-57

FP2 5-30 10-203

O1 5-14 6-107

O2 4-11 6-123

P3 3-7 6-44

P4 3-8 5-57

Table 4.3: Standard deviation table for eyes open event



From table 4.3, we observed that the amplitude varied from -20 to 30µV for normal

subjects, whereas, in ADHD subjects we observed that the amplitude levels were low and

it varied from -70 to 30µV.

Normally, when eyes are open, neuronal activities increase and it will have more positive

values.

Here, there is increased amplitude in normal subjects, whereas for ADHD the activity has

reduced.

The following tables (4.4 and 4.5) show the amplitude values of normal and ADHD

subjects in different electrode position and for different events.

Chetan Naik -ADHD subject1Electrode /

EventEyesclose

Eyesopen

Eyesclose

Eyesopen Blue Green Yellow Orange

FP1 0 6 -6 -39 -6 -10 -3 -2FP2 0 5 15 -32 -13 5 4 -7F7 4 -2 5 -12 -15 2 18 -11F3 4 0 8 -9 -15 6 17 -10FZ 4 -3 15 -15 -18 3 22 -7F4 2 2 9 -12 -17 1 22 -7F8 -1 0 7 -17 -17 0 18 -7T3 8 -6 4 -12 -15 5 17 -8C3 8 -3 6 -11 -17 4 19 -7CZ 7 -3 10 -12 -17 5 21 -6C4 5 -1 7 -9 -18 2 23 -4T4 -1 -4 9 -13 -17 -2 18 -6T5 9 -7 9 -12 -13 5 17 -7P3 6 -7 5 -12 -17 8 17 -7PZ 3 -8 5 -13 -19 7 17 -8P4 6 -3 5 -11 -20 6 17 -6T6 0 -6 7 -12 -18 4 16 -5O1 4 -7 8 -9 -2 3 14 1O2 4 -4 9 -13 -9 6 10 -3

Averagevalues: 3.7894 -2.6842 7.2105 -14.4736 -14.8947 3.1578 16 -6.1578

Table 4.4: Amplitude values for ADHD subjects



Similarly, for a normal subject the values are:

Likitha .N-normal subject5Electrode /

EventEyesclose

Eyesopen

Eyesclose

Eyesopen Blue Green Yellow Orange

FP1 -9 -5 8 1 4 -7 14 5FP2 -2 8 -24 -11 -16 -1 3 6F7 0 -7 1 3 -4 -2 -5 3F3 -2 -1 -20 -4 -7 -3 1 4FZ 2 7 -6 1 -5 1 -5 -5F4 4 5 -5 1 -8 0 2 -4F8 5 3 -1 -4 -3 2 1 -6T3 0 -7 2 4 0 -1 -7 10C3 1 1 -2 1 -6 -12 -13 -12CZ 15 5 -3 1 -2 6 -5 -2C4 4 2 2 2 2 4 1 -2T4 3 -5 -1 -3 18 4 11 1T5 -4 -4 9 2 0 -1 -5 2P3 -1 4 3 2 1 4 -7 2PZ 4 0 4 -1 3 6 -1 3P4 1 -5 6 0 6 5 1 2T6 -5 0 -4 0 4 -1 1 -20O1 -9 12 8 -2 -10 3 -2 6O2 -10 -6 8 -5 7 -3 2 -2

Averagevalues: -0.158 0.368 -0.789 -0.632 -0.842 0.211 -0.684 -0.474

Table 4.5: Amplitude values for normal subjects

3. Visual Cue:

The table 4.6 shows the SD for visual cue. Improved neuronal activities are seen here

since visual processing takes place.

In the protocol that we had 23 events where the subjects had to identify colours, animals

and numbers.

This whole event was conducted for 46 seconds. The normal subjects were able to

identify all the events given to them and hence had higher amplitude levels in frontal and

occipital region compared to other regions.

Whereas, ADHD affected subjects could identify few events and few were left

unidentified. Therefore, we can observe that when the things were identified by them the



higher amplitude levels were found in frontal region and when they couldn’t identify the

higher amplitude levels were found in occipital region.

The standard deviation is lower in normal subjects than in ADHD subjects.


FP1 6-32 6-41

FP2 8-28 5-36

O1 6-18 5-16

O2 5-16 5-19

P3 4-26 4-20

P4 4-9 4-21

Table 4.6: Standard deviation table for visual cue

4. Motor activity:

During this activity we observed that higher amplitudes were present in almost all the

electrodes, but highest in electrode positions FP2 and O2 for normal subjects. Whereas, for

ADHD affected subject it was observed that higher amplitude was present in the frontal

region.

4.2 POWER ANALYSIS

Here we calculate the absolute power and find the relative power percentages for the

analysis. This is performed for every event and for 8 selected electrode positions. 1st row is for

delta band, 2nd for theta, 3rd for alpha and fourth for beta.

4.2.1 Eyes close ADHD kids:

frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 97-99% 96-98% 94-97% 89-93% 87-95% 89-94% 92-97% 91-95%

Theta 2- 4% 1- 2% 1- 2% 2- 3% 3- 8% 5- 6% 13-14% 2- 3%

Alpha 0- 1% 0- 1% 1- 3% 2- 5% 2- 4% 2- 4% 2- 5% 1- 1%

Beta 0- 1% 0- 1% 1- 2% 1- 3% 2- 4% 1- 3% 1- 4% 2- 4%

Table 4.7: Relative power percentage table for ADHD subjects in eyes close event



Relative power percentage table foe ADHD in eyes close event is shown in table 4.7. For

eyes close in normal person it should show beta and alpha activity in P3 and P4 positions which

are for cognitive processing. But in ADHD subjects we can see that they have higher delta band

activity at all electrode positions.

4.2.2 Eyes close normal kids:

frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 95-98% 92-97% 68-73% 73-82% 65-70% 78-82% 81-88% 58-67%

Theta 1- 2% 1- 4% 4- 7% 3- 4% 7- 8% 6- 9% 4- 7% 13-16%

Alpha 1- 2% 1- 2% 17-23% 4- 7% 42-46% 8-17% 4- 8% 13-20%

Beta 1- 3% 1- 2% 7-10% 4- 6% 8- 9% 9-10% 13-17% 7-12%

Table 4.8: Relative power percentage table for normal subjects in eyes close event

In normal subjects, they have higher alpha activity in P3 position, i.e. right side of the

brain which is responsible for cognitive processing. A considerably good alpha activity is seen in

the T4 position which indicates that normal subject think or memorize during eyes close event.

4.2.3 Eyes open ADHD kids:

frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 91-96% 96-98% 94-96% 72-75% 88-96% 90-94% 92-97% 75-89%

Theta 1- 5% 11-13% 3- 4% 13-17% 19-21% 2- 3% 4- 9% 20-23%

Alpha 2- 2% 1- 2% 1- 1% 4- 4% 1- 2% 1- 3% 1- 4% 1- 3%

Beta 1- 2% 1- 2% 4- 5% 5- 6% 1- 5% 2- 5% 2- 4% 2- 6%

Table 4.9: Relative power percentage table for ADHD subjects in eyes open event

During eyes open event, ADHD subjects showed delta activity in most of the electrodes.

They even showed theta activity at electrode position FP2, O2, P3, and T4. They have very less

alpha and beta activity, which should have been more if they were normal kids.

4.2.4 Eyes open normal kids:

frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 94-97% 94-97% 59-66% 78-84% 80-87% 57-69% 69-75% 75-78%

Theta 13-15% 1- 3% 9-14% 10-15% 6-10% 4- 7% 13-13% 10-12%

Alpha 0- 2% 1- 2% 3- 5% 4- 5% 5- 9% 6-15% 6- 7% 5-11%

Beta 1- 2% 1- 2% 4- 7% 2- 4% 6- 6% 3- 4% 24-26% 13-14%

Table 4.10: Relative power percentage table for normal subjects in eyes open event



In normal kids, we see at different electrode positions, different bands are showing

activity. Like O1 & O2 are showing theta activity, these positions are responsible for visual

processing. T3 & T4 are showing beta activity, indicating that the memorization processes is

seen at this event, and where as P3 & P4 show the alpha activity.

4.2.5 Visual cue ADHD kids (animal slide):

frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 82-92% 83-93% 84-93% 82-92% 83-92% 78-87% 83-93% 85-95%

Theta 7- 8% 1- 3% 1- 4% 1- 4% 1- 4% 15-17% 1- 3% 1- 4%

Alpha 0- 1% 1- 2% 1- 3% 1- 4% 2- 7% 3- 8% 1- 4% 1- 2%

Beta 1- 4% 2- 6% 1- 3% 1- 4% 2- 7% 1- 6% 1- 3% 2- 6%

Table 4.11: Relative power percentage table for ADHD subjects in visual cue event (Animal)

ADHD subjects showed high delta activity at all electrode positions in this visual cue

event. They were able to recognize the subject in the slide (i.e. Parrot, bird) with a delay. This

delay can be cause of less logical attention and cognitive process. Logical attention and cognitive

processing are related to FP1, FP2, P3 and P4 positions. In this positions delta activity is more

and theta activity is slightly higher than the alpha and beta bands.

4.2.6 Visual cue ADHD kids (color slide):

frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 85-95% 85-95% 84-94% 84-94% 85-95% 64-84% 84-94% 85-95%

Theta 0- 1% 1- 4% 2- 6% 1- 2% 1- 3% 18-22% 1- 2% 1- 2%

Alpha 0- 2% 1- 2% 1- 3% 1- 3% 1- 2% 1- 2% 1- 3% 1- 3%

Beta 4-11% 2- 7% 1- 2% 1- 3% 1- 4% 1- 3% 1- 4% 1- 4%

Table 4.12: Relative power percentage table for ADHD subjects in visual cue event (Color)

The analysis of this table is similar to that of table: Visual Cue ADHD kids (Parrot slide),

except that in FP1 and FP2 position beta activity is more. This indicates that different colors can

catch the attention of ADHD subjects.

4.2.7 Visual cue ADHD kids (number slide):

The ADHD subjects have shown theta alpha and beta activities in some electrode

positions. The subjects were trained in identifying the characters, numbers, objects, etc., due to



frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 84-94% 83-93% 78-87% 79-89% 55-71% 80-89% 83-93% 75-95%

Theta 1- 3% 1- 4% 3- 7% 1- 6% 2-11% 18-20% 5-12% 5- 9%

Alpha 0- 1% 0- 1% 11-12% 1- 3% 3- 6% 1- 3% 1- 2% 10-11%

Beta 1- 2% 0- 2% 4- 9% 3- 6% 12-18% 2- 5% 1- 4% 5-11%

Table 4.13: Relative power percentage table for ADHD subjects in visual cue event (Number)

this they were able to identify the only one digit in a number. So they are showing thought

processing in right part of the brain and memorization. P3 position shows us the beta activity,

which indicates that it is processing and identifying the character. Even the position T4 and O1

shows a good alpha and beta activity which are for memorizing and visual processing

respectively.

4.2.8 Visual cue normal kids (animal slide):

frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 92-94% 90-92% 70-77% 70-77% 72-78% 86-88% 73-84% 81-88%

Theta 5- 8% 6- 7% 5- 8% 9-11% 12-16% 5- 6% 10-11% 4- 8%

Alpha 1- 2% 1- 2% 3- 4% 7- 8% 11-12% 10-11% 6- 7% 2- 3%

Beta 1- 2% 1- 3% 3- 4% 6-10% 3- 4% 3- 4% 9-17% 4- 6%

Table 4.14: Relative power percentage table for normal subjects in visual cue event (Animal)

The normal kids have shown good alpha beta and theta activity during the identification

of the subject compared to ADHD. The visual processing, cognitive processing and memorizing

is good since the respective electrode positions display a considerably high theta , alpha and beta

power levels.

4.2.9 Visual cue normal kids (color slide):

frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 88-93% 85-91% 72-74% 76-85% 80-84% 78-81% 56-65% 68-78%

Theta 25-28% 4- 6% 11-17% 9-16% 8-12% 15-19% 9-13% 5- 8%

Alpha 2- 3% 1- 2% 3- 4% 3- 4% 9-11% 3- 5% 5- 6% 2- 5%

Beta 1- 3% 2- 6% 7-10% 4- 5% 3- 4% 5- 7% 3- 6% 14-15%Table 4.15: Relative power percentage table for normal subjects in visual cue event (Color)

The explanation for table 4.14 holds good even for this table.



4.2.10 Visual cue normal kids (number slide):

frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 90-92% 90-95% 82-89% 81-86% 82-88% 75-88% 57-70% 59-72%

Theta 5- 6% 11-15% 11-12% 10-11% 14-17% 11-16% 11-18% 15-17%

Alpha 4- 6% 1- 2% 3- 4% 2- 4% 4-14% 25-28% 9-11% 20-23%

Beta 1- 2% 2- 4% 4- 6% 2- 5% 6- 8% 5- 7% 7-16% 18-24%

Table 4.16: Relative power percentage table for normal subjects in visual cue event (Number)

In this event normal kids have shown very high theta, alpha and beta power levels in cognitive

processing (P3&P4 positions), memorizing (T3&T4 positions) and visual processing (O1&O2

positions).

Comparing normal and ADHD subjects, we see that ADHD subjects show very low

cognitive and memorizing process, which is directly related to thinking and concentration level

of them. In normal subjects we see a good cognitive and memorizing process. With validation

from Neurologist we may be able to classify the ADHD and normal subjects based on this

relative power analysis.

4.2.11 Activity for ADHD kids:

frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 84-94% 83-93% 82-92% 79-89% 73-93% 83-92% 82-92% 74-94%

Theta 1- 2% 5- 7% 1- 3% 1- 5% 15-17% 1- 3% 6- 9% 9-10%

Alpha 3- 3% 1- 2% 0- 1% 1- 4% 0- 1% 0- 1% 1- 2% 3- 4%

Beta 0- 1% 0- 1% 0- 2% 3- 5% 1- 2% 0- 1% 0- 1% 1- 2%

Table 4.17: Relative power percentage table for ADHD subjects in motor activity

We can observe that for activity maximum relative power level or power percentage of

all electrode positions lies in delta band.

P3 position in 10/20 system’s principal function is cognitive processing in right half of

brain. From the table we can see that the cognitive processing in the ADHD subjects lies in theta

band.

T3 and T4 position’s principal function is verbal and non verbal memory formation and

storage. From the table, memory formation is taking place in delta and theta band in ADHD

subjects.



4.2.12 Activity for normal kids:

frequency/electrode

FP1-AVG

FP2-AVG

O1-AVG

O2-AVG

P3-AVG

P4-AVG

T3-AVG

T4-AVG

Delta 81-86% 80-85% 76-78% 47-52% 67-75% 67-73% 61-69% 75-82%

Theta 7- 9% 11-14% 11-11% 11-15% 24-27% 15-16% 11-16% 9-12%

Alpha 1- 3% 6- 6% 5- 5% 6- 9% 9-10% 6- 9% 3- 4% 4- 5%

Beta 1- 3% 1- 2% 8- 9% 29-31% 2- 5% 6- 8% 3- 5% 9-12%

Table 4.18: Relative power percentage table for normal subjects in motor activity

In comparison with ADHD subjects, normal subjects have shown the activity in all

bands. The visual processing can be seen in the normal kids at beta band level.

The cognitive processing is more in theta alpha and beta compared with ADHD kids,

where they had less cognitive processing in left, but the normal kids have good cognitive

processing in both right and left part of the brain. The memory formation and storage during

activity is the significant information for differentiating between the ADHD and normal kids, i.e.

normal subjects learn easily because they will focused on one particular thing.

4.3 FREQUENCY MAPS

In frequency maps, we plot the power levels of EEG signals corresponding to different

frequency ranges like delta, theta, alpha, beta, high beta, mid beta and low between, just in order

to find out the neuronal activities of ADHD and healthy subjects during different events and it

also helps specialists in detecting the abnormality.

1. Eyes closed position –ADHD

Figure 4.1: Frequency map of ADHD subject in eyes close event



Inference for Eyes closed position: Refer figures 4.1 and 4.2.

Different colour levels indicate different power values (in the RHS), whereas in the LHS,

each of the electrodes are given with different colour coding: Orange for Delta, Yellow

for theta, Green for alpha and Blue for beta.

3-D plots show increased amplitude levels corresponding to delta for ADHD subjects in

all the regions, whereas, for Normal subjects, we can observe the lower amplitude levels.

Difference in levels of amplitude indicates the difference in calmness levels. Higher the

amplitude, higher is the neuronal activities and the person is considered to be hyper –

active.

2. Eyes closed position – Normal

Figure 4.2: Frequency map of normal subject in eyes close event

3. Eyes open position – ADHD

Figure 4.3: Frequency map of ADHD subject in eyes open event



4. Eyes open –Normal

Figure 4.4: Frequency map of normal subject in eyes open event

Inference for Eyes open position: Refer figures 4.3 and 4.4.

3-D plots show increased amplitude levels corresponding to delta for ADHD subjects in

all the regions, whereas, for Normal subjects, we can observe the lower amplitude levels.

Difference in levels of amplitude indicates the difference in calmness levels. Higher the

amplitude, higher is the neuronal activities and the person is considered to be hyper –

active.

Here, ADHD subjects do not show any change in activity, whereas, normal subjects still

remain to be calm, since we had instructed them to not think upon anything, except

concentrating on one thing.

5. Visual Cue – ADHD

Figure 4.5: Frequency map of ADHD subject in visual cue event



6. Visual cue – Normal

Figure 4.6: Frequency map of normal subject in visual cue event

Inference for Visual Cue: Refer figure 4.5 and 4.6.

Higher neuronal activities were found when the ADHD subjects were trying to identify

the objects, this shows that their neuronal activities shoot up just because they are unable

to identify it and thinking over it or may be even diverted.

For the normal subjects, since they could identify the objects, only Frontal region which

relates to identifying shapes, sizes, colours and objects shows the higher amplitude.

This is an evident parameter to clearly distinguish between ADHD and normal subjects.

7. Motor activity – ADHD

Figure 4.7: Frequency map of ADHD subject in motor activity event



Inference for motor activity: Refer figures 4.7 and 4.8.

Since, the activity which we gave involves counting of objects and arranging them in a

order, it requires visual processing, memory and cognitive ability.

The task seems simpler for the normal kids and they could do without thinking much,

thus , there is less neuronal activities, whereas for the ADHD subjects, since, the visual

processing and cognitive ability is lesser, the neuronal activities tend to be more.

8. Motor Activity – Normal

Figure 4.8: Frequency map of normal subject in motor activity event

4.4 NEURAL NETWORK ALGORITHM

Case 1:

1. Performance plot

This is drawn to know the performance of neural network during the training i.e.,

whether the neural network is able to fit the target data within the test data or not.

Here, at 52nd iteration, the validation performance has reached minimum i.e.,

0.078113. The training is continued for 6 more iterations (epochs) before the training

stopped.



Performance plot is shown in figure 4.9.

Figure 4.9: Performance Plot (Case-1)

Generally, the error reduces after more epochs of training, but might start to increase

on the validation data set as the network starts over fitting the training data. In the

default setup, the training stops after six consecutive increases in validation error,

and the best performance is taken from the epoch with the lowest validation error.

This figure does not indicate any problems with the training. The validation and test

curves are very similar.

If the test curve had increased significantly before validation curve increased, then it

is possible that some over fitting of data might have occurred.

2. Regression plot:

The next step in validating the network is to create a regression plot, which shows the

relationship between outputs of the networks and the targets. It is shown in figure 4.10.

If the training were perfect, the network outputs and the targets would be exactly equal,

but the relationship is rarely perfect in practice.

Here, we have considered 4200 samples of data keeping 8 electrode positions i.e., FP1, FP2, O1,

O2, P3, P4 T3 & T4 as attributes.

The first 3 plots represent training, validation and testing. The dashed line in each plot

represents the perfect results- outputs= targets.



The solid line represents the best fit linear regression line between the outputs and

targets.

Figure 4.10: Regression Plot (Case-1)

The R value is an indication of the relationship between outputs and targets. If R=1, this

indicates there is an exact linear relationship between the outputs and targets. If R is close

to zero, then there is no linear relationship between outputs and targets.

Here, the R value is greater than 0.84 which shows a good fit of data and the linear

relationship between the input and outputs. Further, this can be improved by changing the

number of hidden layers and by considering different attributes.

3. Training state plot

Plottrainstate(tr) plots the training state from a training record tr returned by train. This is

shown in figure 4.11 below.

Generally, there are two different modes: Incremental mode and Batch mode, with

batch mode being the fastest. But, Incremental modes are used in case of

discontinuous signals.

Here, since discontinuous mode of operation is used, gradient and Mu are calculated.



This is plotted to check for the training state validations i.e., whether the network has

been trained properly or not.

Figure 4.11: Training State Plot (Case-1)

As seen in the last plot of the above figure, validation check has been done for 58

epochs and of which, 52 epochs show zero failure in training. Whereas, after this, 6

continuous validation checks shows increase in the error of detection.

Therefore, for this network, 52 epochs gives the best results of distinguishing

between ADHD and normal subjects.

4. Confusion matrix

Plotconfusion(targets,outputs) returns a confusion matrix plot for the target and output

data in targets and outputs, respectively.

On the confusion matrix plot, the rows correspond to the predicted class (Output Class),

and the columns show the true class (Target Class).

The diagonal cells show for how many (and what percentage) of the examples the trained

network correctly estimates the classes of observations. That is, it shows what percentage



of the true and predicted classes match. The off diagonal cells show where the classifier has

made mistakes.

The column on the far right of the plot shows the accuracy for each predicted class,

while the row at the bottom of the plot shows the accuracy for each true class.

The cell in the bottom right of the plot shows the overall accuracy.

To train the neural network, 4200 samples were considered with 8 attributes.

The matrix shows that, the network could correctly learn for 4197 samples (99.9%) and

network has got confused for the 3 samples (0.1%). Overall correct predictions are

99.9% and the failure is 0.1% as shown in figure 4.12.

Figure 4.12: Confusion Matrix-Train State Figure 4.13: Confusion Matrix-Test State

(Case-1) (Case-1)

Once the network is trained, it is checked with the test samples to validate the

training. Here we have used 1260 samples. This is shown in figure 4.13.

The network provides 99.9% of accuracy in distinguishing ADHD from healthy

subjects. There is 0.1% of error while distinguishing between them. This shows more

accuracy (99.9%) in distinguishing ADHD from normal subjects when compared to

the other works mentioned in literature survey which shows the accuracy of 90%.

5. Error Histogram

Error histogram is plotted to fit the test, validation and training state errors into the same

graph and it is shown in the figure that, zero error is found when the difference between targets

and outputs is 0.06124.



Error histogram is shown in figures 4.14 and 4.15. In the first plot, all training, test and

validation data is fit in the same graph and in the second, only training data is taken into

consideration.

Figure 4.14: Error Histogram 1 (Case-1) Figure 4.15: Error Histogram 2 (Case-1)

Case 2:

Similar work has been done having 4 attributes FP1, FP2, O1 and O2 for the events eyes

open and eyes close and the results are as follows: For training 3600 samples are used. For

testing 1075 are used.

1. Performance plot

Figure 4.16: Performance Plot (Case-2)



Here, at 24th iteration, the validation performance has reached minimum i.e.,

0.097072. The training is continued for 6 more iterations (epochs) before the training

stopped. It is shown in figure 4.16.

2. Regression plot

Figure 4.17: Regression Plot (Case-2)

Here, the R value is greater than 0.80 which shows a good fit of data and the linear

relationship between the input and outputs. Refer figure 4.17.

3. Training state plot

Figure 4.18: Training State Plot (Case-2)



As seen in the last plot of figure 4.18, validation check has been done for 30 epochs

and of which, 24 epochs show zero failure in training. Whereas, after this, 6

continuous validation checks shows increase in the error of detection.

Therefore, for this network, 24 epochs gives the best results of distinguishing

between ADHD and normal subjects.

4. Confusion Matrix

Figure 4.19: Confusion matrix- Train state

Figure 4.20: Confusion Matrix–Test State (Case-2)

The network provides 100% of accuracy in distinguishing ADHD from healthy

subjects. There is no error found while distinguishing between them.



This shows more accuracy in distinguishing ADHD from normal subjects when

compared to the other works mentioned in literature survey which shows the

accuracy of 90% and also, when considered in previous attributes.

Confusion matrices for training and testing data are shown in figures 4.19 and 4.20

respectively.

5. Error Histogram

Figure 4.21: Error Histogram 1 (Case-2) Figure 4.22: Error Histogram 2 (Case-2)

Error histogram is plotted to fit the test, validation and training state errors into the same

graph and it is shown in the figure that, zero error is found when the difference between

targets and outputs is 0.01955.

The figures 4.21 and 4.22 depict the error histogram plot for the current case.



CHAPTER - 5

CONCLUSION

This project work included seven ADHD and seven normal subjects. We incorporated

two types of evaluation: 1. Behavioural evaluation 2. Educational evaluation. In the behavioural

evaluation, subjects’ parents and teachers were interviewed to obtain the child’s previous

medical history, social behaviour, symptoms of disorder and so on. Here, we found that, the

disease wasn’t hereditary, the ADHD child was under medication, had improved in day-to-day

activities after training, and mainly, the root cause for the syndrome was quoted to be due to the

severe epilepsy and fever within the age 1 year (as informed by doctors) and partially due to the

improper growth of brain.

In educational evaluation, standard protocol (mentioned earlier) was incorporated to

study the nature of subjects like – class room observations, academic skills, problems of

hyperactivity (such as blurting out answers to the teachers questions or interrupting the teacher or

other students in the class), inattention (such as becoming easily distracted, making careless

mistakes, or failing to finish assignments on time) and impulsivity (such as blurting out answers

to the teachers questions or interrupting the teacher or other students in the class).

Several analysis performed indicate that, ADHD subjects show increased neuronal

activities in the eyes closed position, whereas normally, healthy subjects tend to suppress their

neuronal activity. Generally, alpha and beta waves are high in case of a healthy subject during

eyes close event in posterior region of brain i.e., in electrode positions O1, O2, P3 and P4. Here,

the ADHD cases show very low alpha and beta power levels in posterior region of the brain.

When eyes is open, the alpha and beta activities gets suppressed in a healthy subject, but for the

ADHD subject, alpha activities increase significantly in the posterior region, whereas in anterior,

right and left parts of the brain, it remains to be low. Extremely high standard deviations are

found in the ADHD subjects in eyes close, eyes open and in motor activity, whereas in the

normal subjects, all parameters show significantly lower values. This indicates the deviation

from the mean value and implies about the hyper-active nature of ADHD subjects.



Delta waves are usually the dominant one of all the waves in both normal and abnormal

subject. Severity of delta waves (as in amplitude and power levels) will indicate the abnormality.

The frequency maps show that delta waves have the highest amplitude levels in ADHD when

compared with that of normal subjects. In the motor activity, theta waves are dominantly seen in

ADHD cases, which indicates the abnormality because, theta waves if found dominantly can be

considered as abnormal. Neural networks used to classify ADHD and normal subjects gave the

accuracy of 99.9%, whereas, literature survey says, previous research could attain the accuracy

of 90% in distinguishing between ADHD and normal; and 95% accuracy in distinguishing

between the subtypes.

This project work draws the conclusion that, posterior region of the brain, covering

electrode positions O1, O2, P3 and P4 could be considered to predict the abnormality during

eyes open and eyes close; Anterior portion of the brain, covering pre-frontal regions and frontal

region of the brain, with electrode positions FP1 and FP2 could be used to predict the

abnormality. It also proves that, neural networks, if well-trained can be used to distinguish

between ADHD and normal subjects, given the pattern of EEG as input.



CHAPTER – 6

FUTURE SCOPE

This project work has its own social importance and if further proceeded for research would:

Help in proper detection of the abnormality and assists for further treatment plan. Since,

several studies estimated the prevalence of ADHD, in USA 4-8%, Korea 7.6% to 9.5%,

India 20%, and Emirates 29.7% in United Arab8-11 and every 3 out of 5 children in the

world are affected by this syndrome.

Supports cognitive trainers to improvise the behaviour of children with ADHD syndrome.

This is done by periodically evaluating the improvements in kid’s academic performance

and day-to-day activities.

The cost involved to detect ADHD syndrome as per the literature is significantly high,

whereas, detection of the same as proposed in the project is cost-effective. Therefore, this

could be brought out as a product.

EEG readings might be used to diagnose ADHD syndrome and can also be used by brain-

trainers to measure the effect of cognitive training.

We have predicted the probable region of abnormality and electrode positions that could

be considered for detecting the abnormality. The researchers and specialists in medical field can

take this further and accurately detect the exact region of abnormality which might help in

diagnosing the syndrome.



REFERENCES

[1] Barkley, R. A. (1992). Is EEG biofeedback treatment effective for ADHD children?

Proceed with much caution. Attention Deficit Disorder Advocacy Group newsletter.

[2] Barkley, R. A. (1998). Attention-deficit hyperactivity disorder.Scientific American, 279(3),

66–71.

[3] Barkley, R. A., DuPaul, G. J.,&McMurray,M. B. (1991). Attention deficit disorder with

and without hyperactivity: clinical response to three dose levels of methylphenidate.

Pediatrics, 87, 519–531.

[4] Barry, R. J., Clarke, A. R., & Johnstone, S. J. (2003). A review of electrophysiology in

attention-deficit/hyperactivity disorder: I.Qualitative and quantitative

electroencephalography. Clinical Neurophysiology, 114, 171–183.

[5] Chabot, R. J., Merkin, H., Wood, L. M., Davenport, T. L., & Serfontein, G. (1996).

Sensitivity and specificity of QEEG in children with attention deficit or specific

developmental learning disorders. Clinical Electroencephalography, 27, 26–34.

[6] Chabot, R. J., & Serfontein, G. (1996). Quantitative electroencephalographic profiles of

children with attention deficit disorder. Biological Psychiatry, 40, 951–963.

[7] Kuperman, S., Johnson, B., Arndt, S., Lindgren, S., &Wolraich, M.(1996). Quantitative

EEG differences in a nonclinical sample of children with ADHD and undifferentiated ADD

[see comments]. Journal of the American Academy of Child & Adolescent Psychiatry, 35,

1009–1017.

[8] Lazzaro I, Gordon E, Li W, Lim CL, Plahn M, Whitmont S, Clarke S, Barry RJ, Dosen A,

Meares R. Simultaneous EEG and EDA measures in adolescent attention deficit

hyperactivity disorder. Int J Psychophysiol. 1999;34:123–134.

[9] Clarke AR, Barry RJ, McCarthy R, Selikowitz M. EEG analysis in attention-deficit

hyperactivity disorder: a comparative study of two subtypes. Psychiatry Research. 1998.

[10] Clarke AR, Barry RJ, McCarthy R, Selikowitz M. Age and sex effects in the EEG:

differences in two subtypes of attention-deficit hyperactivity disorder. 2001; 112:815–826.

[11] Clarke AR, Barry RJ, McCarthy R, Selikowitz M. Electro-physiological differences in two

subtypes of attention deficit hyperactivity disorder. Psychophysiology. 2001; 38:212–221.



[12] Barry R, Clarke A, Johnstone S. A review of electrophysiology in attention

deficit/hyperactivity disorder: Qualitative and quantitative electroencephalography. Clin

Neurophysiol. 2003;114:171–183.

[13] Lubar JF. Discourse on the development of EEG diagnostics and biofeedback for attention-

deficit/hyperactivity disorders.Biofeedback Self Regulation. 1991; 16:201–225.

[14] Monastra V, Lubar J, Linden M, VanDeusen P, Green G, Wing W, Phillips A, Fenger T.

Assessing attention deficit hyperactivity disorder via quantitative electroencephalography:

an initial validation study. Neuropsychology. 1999; 13:424–433.

[15] Lubar JF. Discourse on the development of EEG diagnostics and biofeedback for attention-

deficit/hyperactivity disorders.Biofeedback Self Regulation. 1991; 16:201–225.

[16] “Clinical Utility of EEG in Attention Deficit Hyperactivity Disorder” Sandra K. Loo and

Russell A. Barkley, Applied Neuropsychology (2005).

[17] “Sensitivity and specificity of QEEG in children with attention deficit or specific

developmental learning disorders”. Clinical Electroencephalography, Chabot, R. J.,

Merkin, H., Wood, L. M., Davenport, T. L., & Serfontein, G. (1996).

[18] “Electro-physiological differences in two subtypes of attention deficit hyperactivity

disorder”. Psychophysiology. 2001 Clarke AR, Barry RJ, McCarthy R, Selikowitz .M.

[19] “A review of electrophysiology in attention deficit/hyperactivity disorder: Qualitative and

quantitative electroencephalography”. Clinical Neurophysiology disorder: Barry R,

Clarke A, Johnstone S.

[20] "Age and sex effects in the EEG: differences in two subtypes of attention-

deficit/hyperactivity disorder." Clarke AR1, Barry RJ, McCarthy R, Selikowitz.M.

[21] “Assessing attention deficit hyperactivity disorder via quantitative electroencephalography:

an initial validation study.” Neuropsychology. 1999; Monastra V, Lubar J, Linden M,

VanDeusen P, Green G, Wing W, Phillips A, Fenger T.

[22] “Fully connected cascade Artificial Neural network architecture for Attention deficit

hyperactivity disorder classification from fMRI data”. Gopikrishna Deshpande; Dept. of

Elns and Comm. Eng., Auburn Univ., Auburn, AL, USA; Peng Wang; Rangaprakash;

Bogdon Wilamowski.

EXTRACTION AND ANALYSIS OF EEG SIGNALS FOR THEDETECTION OF ADHD SYNDROME- A NOVEL APPROACH

Chandana.SDept of ECE, SJBIT, Bangalore-560060, India

Email: [email protected] B Balaraj

Dept of ECE, SJBIT, Bangalore-560060, IndiaEmail: [email protected]

Usha PDept of ECE, SJBIT, Bangalore-560060, India

Email: [email protected] Pol

Dept of ECE, SJBIT,Bangalore-560060, IndiaEmail: [email protected]

Abstract – Attention Deficit Hyperactivity disorder(ADHD) is a mental disorder, which makes it hardfor the affected subjects to focus and payattention. EEG is most often used to diagnoseepilepsy, sleep disorders, coma, encephalopathy,and brain death. Analysis of EEG signals isusually be used as a first-line method of diagnosisfor tumors, stroke and other focal brain disorders.

In the present work attempt has been made todetect probable region of brain that showsabnormality due to ADHD syndrome. EEG dataof seven normal and ADHD subjects (children) ofage group 4-17 years were collected on following aprotocol which contains 4 events- Eyes close, Eyesopen, Visual Cue and Motor activity. Thesegmented data is subjected to extract featureslike, energy that corresponds to Alpha, Beta, Deltaand Theta frequency ranges; absolute power andamplitude levels at different electrode positions.Single map analysis and Frequency map analysiswere also performed and comparative analysiswas done between the normal and ADHDsubjects. Finally, 3-D plotting of the EEG signalswas done for ease of visualization and neuralnetwork algorithm is used to distinguish betweennormal and ADHD affected subjects.

From the results it is observed that the ADHDaffected subjects show increased neuronalactivities during eyes closed condition; whereasnormally, healthy subjects tend to suppress theirneuronal activity. Higher power and higherstandard deviation were observed in the

ADHD subjects; when eyes closed, eyes open andin motor activity; which indicates that they arehyper active. However, in the normal subjects, all

Dr. K.V. Mahendra PrashanthProfessor, Dept of Electronics and

communication, Chief Coordinator, R&D and PGstudies, SJBIT, Bangalore-560060, India

Email: [email protected]

Mrs. Vijayalakshmi. KAssociate Professor, Dept of Medical Electronics,BMS College of Engineering, Bangalore-560019,

IndiaEmail: [email protected]

the parameters showed significant lower values.Usage of feed-forward back propagationtechnique in Neural network algorithm providesthe accuracy of 99.9% when observed in theconfusion matrix, which implies that, the appliedmethod can clearly classify between ADHDaffected subjects and normal (healthy) subjects;which would help neurologists for effectiveanalysis and treatment.

Further, posterior region of the brain, coveringelectrode positions O1, O2, P3 and P4 could beconsidered to predict the abnormality during eyesopen and eyes close; Anterior portion of the brain,covering pre-frontal regions and frontal region ofthe brain, with electrode positions FP1 and FP2could also be used to predict the abnormality. Italso proves that, neural networks can be used todistinguish between ADHD and normal subjects,on given the pattern of EEG signals as input.

Keywords: Attention Deficit Hyperactivity Disorder(ADHD), Electroencephalography (EEG), Neuralnetwork algorithm, 3-D plots.

I. INTRODUCTIONAttention Deficit Hyperactivity Disorder (ADHD) isa common mental disorder that begins in childhoodand can continue through adolescence and adulthood.It is hard for a child to focus and pay attention, whois diagnosed with ADHD. Some children may behyperactive or have trouble being patient. Forchildren with ADHD, levels of inattention,hyperactivity and impulsive behaviors are greaterthan for other children in their age group. ADHD canmake it hard for a child to do well in school orbehave at home or in the community.

It is estimated that about 6-9 percent of all kids, havethis disorder. In many cases this disorder goes totallyundiagnosed. [17]

Types of ADHDChildren having ADHD usually depict either of threepatterns of behavior: Hyperactive type; Inattentivetype; and Combined type.[20]

It has not yet been studied and research is being doneto detect causes of ADHD syndrome. Recentresearches point out the causes could be eitherneurobiological or genetically. Certain chemicallevels in the brain have been found low (dopamine -achemical that sends messages to the part of the braininvolved in attention, movement and motivation).Increased levels of heavy metals (lead) in the bodyare also seen in some children. The kind of parentinggiven to the child has no role in the causation of thisdisorder, although bad parenting can influence theseverity of this disorder. ADHD can also may lead tolearning disabilities, defiant behaviour, conductdisorder (anti-social behaviour, stealing, fighting),anxiety disorders and depression.

Symptoms of ADHD

ADHD has many symptoms. Some of the symptomsat first may look like normal behaviours for a child,but ADHD makes them much worse and occur moreoften.

Inattentive Type (Category-A) : Difficulty sustainingattention and gets easily distracted; Doesn’t seem tolisten when spoken to directly, often forgetful;Careless mistakes and not able to follow instructions.

Hyperactivity and Impulsivity Type (Category-B) :Constantly running around jumping, always on themove as if driven by motor; Fidgeting with hands orfeet or constantly moving in seat; Not able to sit inone place, has to wander away; Difficulty in sleeping

Causes for ADHD

ADHD probably stems from interactions betweengenes and environmental or non-genetic factors.Researchers have found that much of the risk ofhaving ADHD has to do with genes. Many genes arelinked to ADHD, and each gene plays a small role inthe disorder. ADHD is very complex and a genetictest for diagnosing the disorder is not yet available.

Some additives and dyes may worsen hyperactivityand inattention, but these effects are small and do notaccount for most cases of ADHD.

Besides DNA, other factors can influence whodevelops ADHD. These risk factors include:

Environmental exposure: Exposure to lead mayincrease a child’s risk for ADHD.

Brain injury: A small number of children who suffera traumatic brain injury may develop ADHD.

A mother’s alcohol and tobacco using duringpregnancy: A study by the Washington UniversitySchool of Medicine found that mothers who smokewhile pregnant increase their child’s risk fordeveloping ADHD. Women who drink alcohol anduse drugs during pregnancy also put their child at riskfor the disorder.

Low birth weight or premature delivery: Babies bornbefore their due date are more likely to have ADHDwhen they are older.

In the largest EEG study of ADHD to date (with asample of over 400 children), found that childrenwith ADHD displayed increased theta power, slightelevations in frontal alpha power, and diffusedecreases in beta mean frequency.[5,6]

Studies have found mixed results for alpha power,with some studies showing it to be increased, othersshowing it to be decreased, and still other studiesfinding it to be similar between children with ADHDand normal controls.[4-6]

The findings for beta (indicative of heightenedcortical arousal) activity have been less consistent,with several studies finding decreased beta activity infrontal and central regions and others not. There alsoappears to be a small group of ADHD children(~15%) who show an excess of beta activity.[4-8,15]

EEG abnormalities in male adolescents with ADHDduring an eyes-closed resting condition and foundabsolute dominance in delta and alpha activity and ahigher theta/beta ratio compared to the controlsubjects.[8,15]

Most of the studies concern the ADHD children andsummarize that they have reduced power in alpha andbeta bands and an increased power in delta and thetabands in comparison with the healthy controlgroups.[13]

Attention deficit disorder with or withouthyperactivity (ADHD) has the prevalence rate of 6-9% among school age children within United Statesand persists in adolescence in 70% And in adulthoodin 40-60% .In the largest EEG study of ADHD todate (with a sample of over 400 children), found thatchildren with ADHD displayed increased thetapower, slight elevations in frontal alpha power, anddiffuse decreases in beta mean frequency.[17]

EEG studies of children with ADHD have typicallyfound increased theta activity, increased posteriordelta, decreased alpha and beta and an increase in thetheta/beta ratio compared to normal children. Thisstudy investigated the presence of EEG subtypes ofchildren with the combined type of ADHD. Resultsindicated the presence of 3 distinct EEG-definedclusters of children, characterized by increased slowwave activity and deficiencies of fast wave, increasedhigh amplitude theta with deficiencies of betaactivity, and a beta-excess group.[18]

The findings for beta (indicative of heightenedcortical arousal) activity have been less consistent,with several studies finding decreased beta activity infrontal and central regions and others not. There alsoappears to be a small group of ADHD children(~15%) who show an excess of beta activity .Studieshave found mixed results for alpha power, with somestudies showing it to be increased, others showing itto be decreased, and still other studies finding it to besimilar between children with ADHD and normalcontrols.[19]

The authors have considered the fMRI data as anexample to illustrate the utility of fully connectedcascade (FCC) artificial neural networks (ANN)architecture for performing classification. Theyemployed various directional and non directionalbrain connectivity- based methods to extractdiscriminative features which gave betterclassification accuracy compare to raw data. Theiraccuracy for distinguishing ADHD from healthysubjects was close to 90% and between the subtypeswas close to 95%. Further, they showed that, Ifproperly used FCC ANN performs very well compareto other classifiers.

OBJECTIVES

This work is mainly designed to predict the probableregion of brain that shows abnormality due to ADHDsyndrome. Frequency characteristics of ADHD andnormal subjects are extracted and compared todifferentiate between them. Several analyses wereconducted to find out the probable region ofabnormality and 3-D plotting is performed for ease ofvisualization. Further, neural network algorithm wasadopted to differentiate between ADHD affected andnormal subjects.

II. METHODOLOGY

SUBJECTS

We have considered 7 normal and 7 ADHD children

for the research. An interview was conducted to theparents/ guardians of ADHD subjects as a part of thiswork.Subjects of age group 4-17 were chosen for thisstudy. Normal subjects are the healthy children.ADHD subjects are those who were affected byADHD. Here, it became difficult to identify a pureADHD case, therefore kids with multiple disabilitieslike Mental retardation, intellectual disability wereconsidered.


EEG signals of 7 normal and 7 ADHD subjects wereobtained according to International standard 10-20electrode system using BrainTech traveller portableEEG device which is a 24 channel digital EEGmachine having the A/D conversion of 16bits withthe sampling frequency of 256 Hz.

Figure 4 depicts the flowchart of the proposedmethodology

Figure4. Methodology flowchart

Here, 10-20 denotes the placement of electrodes in aparticular manner as shown in figure 5.

SUBJECT


DATA SEGMENTATION

FEATURE EXTRACTION


COMPARATIVE ANALYSIS TO DETECTABNORMALITY



Figure5. International 10-20 electrode system

Longitudinal bipolar montage is also is being used.See figure 6.

Figure 6: Longitudinal Bipolar montage

During the data acquisition, many noises get addedto the actual data like Scalp impedance, Physicalmovement of electrodes, chewing, Blinking, AC linecurrent, video monitors, AC light, Pagers etc.

In order to minimize this noise, several measureswere also taken like: Hair, Hair products, sweat andother scalp debris are the sources of impedance. Tominimize this, scalp abrasions were used andconductive gel was applied; In order to reduce theenvironmental noises, acoustically and electricallyshielded cables were used; Notch filter is used toremove the power line interference and to reduce thedistortion of signal of interest.

PROTOCOL USED

A common protocol of eyes close and eyes open isused which was done in alternating sets of two withvarying time record length which helps in studyingthe alpha wave activity.

The second step in protocol was visual cue whichconsisted of some common aspects such as colors,numbers, and animals which are identifiable by bothnormal and abnormal subjects. Each of the slides had

the duration of 2 seconds and making total of 46seconds of EEG data.

The next step was motor activity, where all thesubjects were asked to arrange the spoons and hand itover to their care taker.

The figure7 shows the pictorial representation ofprotocol used.

Figure 7: Protocol used

DATA SEGMENTATION

The data was selected manually with minimumredundancy. The electrode positions were selectedbased on the interest of our study. The datasegmentation is done manually using markers duringacquisition of data. The markers were inserted basedon events and timings during the data acquisition.

Feature extraction

Different features of EEG signals such as frequencybands (alpha, beta, delta and theta), energy, powerand statistical parameters such as mean, SD andvariance were calculated and following procedurewas adopted.

The EEG signal should be in time domain. FFT ofEEG data provides the frequency domainrepresentation of the time domain EEG signal. Aspecific band pass filter belonging to the frequencyrange of Alpha, Beta, Delta and Theta bands wereused. The frequency ranges are: Delta band: 0.0 – 4.0Hz; Theta band: 4.0 – 8.0 Hz; Alpha band: 8.0 – 12.0Hz; Beta band: 12.0 – 25.0 Hz. After specifying thefrequency ranges and filtering the data by using FFTalgorithm energy for corresponding amplitude valuesfor each sample of each event and of each electrodeposition were computed.

ANALYSIS

The frequencies ranges delta, theta, alpha and betaprovide information concerning the active state of

Eyes close Eyes open VisualCue

MotorActivity

various regions of the brain.

Analyses adopted were: Single map analysis- Insingle map analysis the amplitude levels of the EEGdata is obtained; Frequency map analysis- obtaineddata is converted from 2D to 3D plot; Frequencytables- Frequency table represents the absolute powerand relative power percentage.

The probable affected region can be predicted usingneural network algorithm.

III. RESULT

A. SINGLE MAP ANALYSIS

Comparative analysis of EEG signals between theADHD and Normal subjects yields the followingresults-

Table1. Comparison analysis of single maps

B. POWER ANALYSIS

The absolute power was calculated and the relativepower percentages were determined for the analysis.This is performed for every event and for eightselected electrode positions. First row is for deltaband, second for theta, third for alpha and fourth forbeta.

Table2. Power percentage of ADHD subjects

Table3. Power percentage of Normal subjects

The above tables (table 2, 3) are shown only for oneevent. Similarly, the relative power percentage iscalculated for all the events mentioned in theprotocol.

Eyes closed: Relative power percentage of ADHDsubjects in eyes close event is shown in table2.Normally, Beta and alpha activity in P3 and P4positions are observed. But in ADHD subjects we cansee that they have higher delta band activity at allelectrode positions and in normal subjects, they havehigher alpha activity in P3 position. A considerablygood alpha activity is seen in the T4 position.

Eyes open: The ADHD affected subjects showedtheta activity at electrode position FP2, O2, P3, andT4. They have very less alpha and beta activity. Thenormal subjects have shown the activity in all bands.There is more theta, alpha and beta activity.

C. FREQUENCY MAPS

In frequency maps, the power levels of EEG signalscorresponding to different frequency ranges likedelta, theta, alpha, beta, high beta, mid beta and lowbeta were plotted just in order to find out theneuronal activities of ADHD and healthy subjectsduring different events and it also helps specialists indetecting the abnormality.

Different colour levels indicate different powervalues (in the RHS), whereas in the LHS, each of theelectrodes are given with different colour coding:Orange for Delta, Yellow for theta, Green for alphaand Blue for beta.

Eyes closed position: 3-D plots show increasedamplitude levels corresponding to delta for ADHDsubjects in all the regions, whereas, for Normalsubjects, lower amplitude levels can be observed.

Eyes open position: 3-D plots show increasedamplitude levels corresponding to delta for ADHDsubjects in all the regions, whereas, for Normal

Figure9. Frequency maps for ADHD affectedsubjects.

Figure9. Frequency maps for Normal subjects

subjects, we can observe the lower amplitude levels.

Visual Cue: For the normal subjects, since theycould identify the objects, only Frontal region whichrelates to identifying shapes, sizes, colours andobjects shows the higher amplitude. Whereas, ADHDaffected subjects were unable to identify.

Motor activity:

Since, the activity which we gave involves countingof objects and arranging them in a order, it requires

visual processing, memory and cognitive ability.

The task seems simpler for the normal kids and theycould do without thinking much, thus , there is lessneuronal activities, whereas for the ADHD subjects,since, the visual processing and cognitive ability islesser, the neuronal activities tend to be more.

IV. CONCLUSION

Several analysis performed indicate that, ADHDsubjects show increased neuronal activities in theeyes closed position, whereas normally, healthysubjects tend to suppress their neuronal activity.Generally, alpha and beta waves are high in case of ahealthy subject during eyes close event in posteriorregion of brain i.e., in electrode positions O1, O2, P3and P4. Here, the ADHD cases show very low alphaand beta power levels in posterior region of the brain.When eyes is open, the alpha and beta activities getssuppressed in a healthy subject, but for the ADHDsubject, alpha activities increase significantly in theposterior region, whereas in anterior, right and leftparts of the brain, it remains to be low. Extremelyhigh standard deviations are found in the ADHDsubjects in eyes close, eyes open and in motoractivity, whereas in the normal subjects, allparameters show significantly lower values. Thisindicates the deviation from the mean value andimplies about the hyper-active nature of ADHDsubjects.

Delta waves are usually the dominant one of all thewaves in both normal and abnormal subject. Severityof delta waves (as in amplitude and power levels)will indicate the abnormality. The frequency mapsshow that delta waves have the highest amplitudelevels in ADHD when compared with that of normalsubjects. In the motor activity, theta waves aredominantly seen in ADHD cases, which indicates theabnormality because, theta waves if founddominantly can be considered as abnormal. Neuralnetworks used to classify ADHD and normal subjectsgave the accuracy of 99.9%, whereas, literaturesurvey says, previous research could attain theaccuracy of 90% in distinguishing between ADHDand normal; and 95% accuracy in distinguishingbetween the subtypes.

This work draws the conclusion that, posterior regionof the brain, covering electrode positions O1, O2, P3and P4 could be considered to predict theabnormality during eyes open and eyes close;Anterior portion of the brain, covering pre-frontalregions and frontal region of the brain, with electrodepositions FP1 and FP2 could be used to predict theabnormality. It also proves that, neural networks, if

well-trained can be used to distinguish betweenADHD and normal subjects, given the pattern ofEEG as input.

V. FUTURE SCOPE

This work has its own social importance and iffurther proceeded for research would: Help in properdetection of the abnormality and assists for furthertreatment plan; Supports cognitive trainers toimprovise the behaviour of children with ADHDsyndrome;The cost involved to detect ADHDsyndrome as per the literature is significantly high,whereas, detection of the same as proposed in theproject is cost-effective. Therefore, this could bebrought out as a product.

The probable region of abnormality has beenpredicted and electrode positions that could beconsidered for detecting the abnormality are found.The researchers and specialists in medical field cantake this further and accurately detect the exactregion of abnormality which might help indiagnosing the syndrome.

ACKNOWLEDGEMENTWe would like to acknowledge Srushti specialacademy for providing their valuable time andsupport towards our research. We would also like tothank the subjects involved in this work and theirparents for their kind cooperation, also the scholars ofBMSCE who have helped us in all the stages. We arealso grateful to the management, principal and all thefaculties of SJB Institute of Technology fortremendous support.

REFERENCES[1] Barkley, R. A. (1992). Is EEG biofeedbacktreatment effective for ADHD children? Proceed withmuch caution. Attention Deficit Disorder AdvocacyGroup newsletter.

[2] Barkley, R. A. (1998). Attention-deficithyperactivity disorder.Scientific American, 279(3),66–71.

[3] Barkley, R. A., DuPaul, G. J.,&McMurray,M. B.(1991). Attention deficit disorder with and withouthyperactivity: clinical response to three dose levels ofmethylphenidate. Pediatrics, 87, 519–531.

[4] Barry, R. J., Clarke, A. R., & Johnstone, S. J.(2003). A review of electrophysiology in attention-deficit/hyperactivity disorder: I.Qualitative andquantitative electroencephalography. ClinicalNeurophysiology, 114, 171–183.

[5] Chabot, R. J., Merkin, H., Wood, L. M.,Davenport, T. L., & Serfontein, G. (1996). Sensitivityand specificity of QEEG in children with attentiondeficit or specific developmental learning disorders.Clinical Electroencephalography, 27, 26–34.

[6] Chabot, R. J., & Serfontein, G. (1996).Quantitative electroencephalographic profiles ofchildren with attention deficit disorder. BiologicalPsychiatry, 40, 951–963.

[7] Kuperman, S., Johnson, B., Arndt, S., Lindgren,S., &Wolraich, M.(1996). Quantitative EEGdifferences in a nonclinical sample of children withADHD and undifferentiated ADD [see comments].Journal of the American Academy of Child &Adolescent Psychiatry, 35, 1009–1017.

[8] Lazzaro I, Gordon E, Li W, Lim CL, Plahn M,Whitmont S, Clarke S, Barry RJ, Dosen A, Meares R.Simultaneous EEG and EDA measures in adolescentattention deficit hyperactivity disorder. Int JPsychophysiol. 1999;34:123–134.

[9] Clarke AR, Barry RJ, McCarthy R, Selikowitz M.EEG analysis in attention-deficit hyperactivitydisorder: a comparative study of two subtypes.Psychiatry Research. 1998; 81:19–29.

[10] Clarke AR, Barry RJ, McCarthy R, SelikowitzM. Age and sex effects in the EEG: differences intwo subtypes of attention-deficity hyperactivitydisorder. Clinical Neurophysiology. 2001; 112:815–826.

[11] Clarke AR, Barry RJ, McCarthy R, SelikowitzM. Electro-physiological differences in two subtypesof attention deficit hyperactivity disorder.Psychophysiology. 2001; 38:212–221.

[12] Barry R, Clarke A, Johnstone S. A review ofelectrophysiology in attention deficit/hyperactivitydisorder: Qualitative and quantitativeelectroencephalography. Clin Neurophysiol.2003;114:171–183.

[13] Lubar JF. Discourse on the development of EEGdiagnostics and biofeedback for attention-deficit/hyperactivity disorders.Biofeedback SelfRegulation. 1991; 16:201–225.

[14] Monastra V, Lubar J, Linden M, VanDeusen P,Green G, Wing W, Phillips A, Fenger T. Assessingattention deficit hyperactivity disorder viaquantitative electroencephalography: an initialvalidation study. Neuropsychology. 1999; 13:424–433.

[15] Lubar JF. Discourse on the development ofEEG diagnostics and biofeedback for attention-deficit/hyperactivity disorders.Biofeedback SelfRegulation. 1991; 16:201–225.

[16] “Clinical Utility of EEG in Attention DeficitHyperactivity Disorder” Sandra K. Loo and RussellA. Barkley, Applied Neuropsychology

(2005).

[17] “Sensitivity and specificity of QEEG in childrenwith attention deficit or specific developmentallearning disorders”. Clinical Electroencephalography,Chabot, R. J., Merkin, H., Wood, L. M., Davenport,T. L., & Serfontein, G. (1996).

[18] “Electro-physiological differences in twosubtypes of attention deficit hyperactivity disorder”.Psychophysiology. 2001 Clarke AR, Barry RJ,McCarthy R, Selikowitz .M.

[19] “A review of electrophysiology in attentiondeficit/hyperactivity disorder: Qualitative andquantitative electroencephalography”. ClinicalNeurophysiology disorder: Barry R, Clarke A,Johnstone S.

[20]” Age and sex effects in the EEG: differences intwo subtypes of attention-deficit/hyperactivitydisorder.” Clarke AR1, Barry RJ, McCarthy R,Selikowitz.M.[21] “Assessing attention deficit hyperactivitydisorder via quantitative electroencephalography: aninitial validation study.” Neuropsychology. 1999;Monastra V, Lubar J, Linden M, VanDeusen P,Green G, Wing W, Phillips A, Fenger T.

[22] “Fully connected cascade Artificial Neuralnetwork architecture for Attention deficithyperactivity disorder classification from fMRIdata”. Gopikrishna Deshpande; Dept. of Elns andComm. Eng., Auburn Univ., Auburn, AL, USA; PengWang; Rangaprakash; Bogdon Wilamowski.

Extraction and Analysis of EEG signals for the detection of ADHD … · Extraction and Analysis of...

Documents

Transcript of Extraction and Analysis of EEG signals for the detection of ADHD … · Extraction and Analysis of...