Subtraction Method for Removing Powerline Interference from ECG

Subtraction Method for Removing Powerline Interference from ECG Chavdar LevkovChavdar Levkov11, Georgy Mihov, Georgy Mihov22, Ratcho Ivanov, Ratcho Ivanov22, Ivan K. Daskalov , Ivan K. Daskalov 33

Ivaylo ChristovIvaylo Christov33, Ivan Dotsinsky, Ivan Dotsinsky33

11SIGNACOR laboratory, Lyubliana str. No 46, office 706, 1618 Sofia, Bulgaria, SIGNACOR laboratory, Lyubliana str. No 46, office 706, 1618 Sofia, Bulgaria, [email protected]@signacor.com

22Faculty of Electronic Engineering and Technologies (FETT), Technical Faculty of Electronic Engineering and Technologies (FETT), Technical University – Sofia, Kl. Ohridsky str. No 8, block 1, 1797 Sofia, Bulgaria, University – Sofia, Kl. Ohridsky str. No 8, block 1, 1797 Sofia, Bulgaria,

[email protected]@tu-sofia.bg

33Centre of Biomedical Engineering, Bulgarian Academy of Sciences, Acad. G. Centre of Biomedical Engineering, Bulgarian Academy of Sciences, Acad. G. Bonchev Str., bl. 105, 1113 Sofia, Bulgaria, [email protected] Str., bl. 105, 1113 Sofia, Bulgaria, [email protected]

The Thirteen International Scientific and Applied Science Conference ELECTRONICS - ET’2004

Bulgaria, Sozopol, September 22 - 24, 20041

In Memoriam of Professor Ivan Konstantinov Daskalov



1. Introduction

- Contract No 1238-23/79 named “Microprocessor based ECG - Contract No 1238-23/79 named “Microprocessor based ECG selector for Screening Analysis”selector for Screening Analysis”- Contractor - Institute for Medical Equipment, Sofia;- Contractor - Institute for Medical Equipment, Sofia;- Developer - Research Sector of Higher Mechanical and - Developer - Research Sector of Higher Mechanical and Electrotechnical Institute - now Technical University – Sofia with Electrotechnical Institute - now Technical University – Sofia with director of the project Prof. V. Zlatarov, technical research manager director of the project Prof. V. Zlatarov, technical research manager R. Ivanov and technical assistant G. Mihov;R. Ivanov and technical assistant G. Mihov;- Collaborator - Central Laboratory for Electromedical - Collaborator - Central Laboratory for Electromedical Instrumentation (CLEMA) at the Medical Academy Sofia with Instrumentation (CLEMA) at the Medical Academy Sofia with director Prof. I. K. Daskalov and research assistants Ch. Levkov director Prof. I. K. Daskalov and research assistants Ch. Levkov and M. Matveev.and M. Matveev.



The background - beginning in 1979

3

The problemRemoving the Powerline interference from ECG.Removing the Powerline interference from ECG.



Powerline interference can be Powerline interference can be removed with classical linear filters. removed with classical linear filters. However, important spectral However, important spectral components of the ECG are components of the ECG are affected. After the filtration the low affected. After the filtration the low frequency components are frequency components are preserved. P, T waves and S-T preserved. P, T waves and S-T segments are undistorted. But segments are undistorted. But important frequency components in important frequency components in ECG are affected which leads to ECG are affected which leads to unacceptable distortions of the unacceptable distortions of the shape of the QRS complex. shape of the QRS complex.

Analysis

4

The solution



Linear digital filtration is Linear digital filtration is applied only in these applied only in these segments where the ECG segments where the ECG frequency components are frequency components are not affected (in so called not affected (in so called linear segments). At the linear segments). At the same time the powerline same time the powerline interference signal is interference signal is extracted and saved. Later extracted and saved. Later on this interference signal on this interference signal is subtracted from the is subtracted from the other non-linear segments other non-linear segments of the ECG (e.g in the of the ECG (e.g in the QRS complex).QRS complex).

5

Bulgarian Patent - 1980



[1] Levkov C., Mihov G., Ivanov [1] Levkov C., Mihov G., Ivanov R., and Daskalov. I.R., and Daskalov. I. Method and Method and device for elimination of power-device for elimination of power-line frequency disturbance during line frequency disturbance during analysis of biological signalsanalysis of biological signals. . Bulgarian Patent No 30792 Bulgarian Patent No 30792 30.06.1980, 30.06.1980, ((in Bulgarian).in Bulgarian).

6

First Publication - 1981



[2] Daskalov I., Ivanov R., [2] Daskalov I., Ivanov R., Levkov C. and Mihov G. Levkov C. and Mihov G. Elimination of power-line Elimination of power-line frequency disturbance by frequency disturbance by microprocessormicroprocessor. Proceedings of . Proceedings of the XVI Scientific Session “Radio the XVI Scientific Session “Radio day ‘81”, May 1981, Sofia, 97-100 day ‘81”, May 1981, Sofia, 97-100

((in Bulgarian).in Bulgarian).

7

First Device - ECG Selector - 1981



[3] Daskalov I., Ivanov R. and [3] Daskalov I., Ivanov R. and Mihov G. Mihov G. Microprocessor Microprocessor controlled ECG slector for controlled ECG slector for screening analysisscreening analysis. Proceedings . Proceedings of the XVI Scientific Session of the XVI Scientific Session “Radio day ‘81”, May 1981, Sofia, “Radio day ‘81”, May 1981, Sofia, 93-96 93-96 ((in Bulgarian).in Bulgarian).

8

First Ph.D Thesis - 1983



[4] Mihov G. [4] Mihov G. Software automated Software automated electronic devi-ces for processing electronic devi-ces for processing and visualization of electrocar-and visualization of electrocar-diosignalsdiosignals. Ph.D. thesis, TU-. Ph.D. thesis, TU-Sofia, 1983 Sofia, 1983 ((in Bulgarian).in Bulgarian).

9

First Publication Abroad - 1984



[5] Levkov, C., Mihov, G., Ivanov, [5] Levkov, C., Mihov, G., Ivanov, R., and Daskalov. I., 1984, R., and Daskalov. I., 1984, Subtraction of 50 Hz interference Subtraction of 50 Hz interference from the electrocardiogramfrom the electrocardiogram. . Medical & Biological Engineering Medical & Biological Engineering

& Computing, 22, 371-373& Computing, 22, 371-373..

10

New Investigations - 1988



[6] Christov, I. and Dotsinsky, I., [6] Christov, I. and Dotsinsky, I., 1988, 1988, New approach to the digital New approach to the digital elimination of 50 Hz interference elimination of 50 Hz interference from the electrocardiogramfrom the electrocardiogram. Medical . Medical & Biological Engineering & & Biological Engineering & Computing, 26, 431-434.Computing, 26, 431-434.

11

2. The Subtraction Method



Basal Structure of the Subtraction Method

12

Main Stages



– – Linear segment detectionLinear segment detection. Every ECG sample is tested whether . Every ECG sample is tested whether it belongs to a linear segment. A special criteria for linearity is used it belongs to a linear segment. A special criteria for linearity is used for this purpose;for this purpose;– – Interference extractingInterference extracting. If the criteria for linearity is fulfilled the . If the criteria for linearity is fulfilled the powerline interference is extracted with appropriate digital filter. At powerline interference is extracted with appropriate digital filter. At the same time the interference is removed from these linear ECG the same time the interference is removed from these linear ECG segments;segments;– – Interference restoringInterference restoring. The extracted values of the interference . The extracted values of the interference are saved in FIFO buffer. After appropriate amplitude and phase are saved in FIFO buffer. After appropriate amplitude and phase processing they are used for restoring the interference;processing they are used for restoring the interference;– – Interference subtractingInterference subtracting. The restored values of the interference . The restored values of the interference are subtracted from the original signal in non-linear segments.are subtracted from the original signal in non-linear segments.

13

2.1. Linear segment detection



a: Typical linear and non-linear segments of a real ECG signal

Criteria for linearity is based on the second difference (signed Criteria for linearity is based on the second difference (signed DDii) of the ) of the

ECG. ECG. DDii is compared with a threshold is compared with a threshold MM which is obtained experimentally which is obtained experimentally:: MDi

b: Aproximate frequency spectrum of a linear segment

14



Getting the second difference from linear segments of the ECG. Maximal and minimal length for analysing

Sampling rate Ф Powerline frequency FSampling period N = Φ/F

In even multiple sampling N is integer and = 2n In odd multiple sampling N is integer and = 2n+1In non-multiple sampling N is rationale.

a: for odd multiplicity; b: for even multiplicity

15



Different Criteria for linearity (D-filter) are used:- Longest interval;

- Shortest interval for odd multiplicity;- Shortest interval for even multiplicity;

- Complex criteria [4, 5];

- Criteria for even multiplicity [6]; - Criteria for even multiplicity [7];

NiiiNii XXXXD 11 ninin-inii XXXXD 11 ni1ni1n-inii XXXXD

NiiiNiiNi

NiiiNiiNii

XX,,XX,XXmin

XX,,XX,XXmaxD

11

11

iNiiNii XXXXD 22

iNiiNii XXXXD 11

Transfer function of D-filter - longest interval

Requirements to the D-filter:Requirements to the D-filter:- Gain zero for - Gain zero for ff = 0 Hz; = 0 Hz;- Gain zero for - Gain zero for f = Ff = F..

fn

sinfD2

4 2

16



Different filters for interference removing (K-filter) are used:

- ‘three-point’ filter [11] -

Transfer function of K-filterRequirements to the Requirements to the KK-filter:-filter:- Unity gain for - Unity gain for ff = 0 Hz; = 0 Hz;- Gain zero for - Gain zero for f = F;f = F;- Linear phase response.- Linear phase response.

2.2. Interference extracting

n

njjii X

nY

12

1

fsin

fnsin

nfK

12

12

1

1

1 22

1 -n

-nj

ninijii

XXX

nY

fn

cosf

sin

fnsin

nfK

212

2

1

4

2 niinii

XXXY

fn

cosfK 2

- filter for odd multiplicity [8, 9] -

- filter for even multiplicity with a linear phase response [7, 8] -

The interference samples Bi are calculated (extracted) by:

iii Y-XB

17



Case of multiple sampling

Requirements to the B-filter:Requirements to the B-filter:- Unity gain for - Unity gain for ff = = FF;;- Linear phase response.- Linear phase response.

2.3. Interference restoring

The interference is cancelled by subtracting the corresponding interference value with appropriate phase taken from the FIFO buffer: N-ii BB

Case of non-multiple sampling

A digital filter is applied for the values in the buffer (B-filter) :

0

2 *NjjijBa

The restored value of the interference is calculated by:

N

NjjijNiF

i a

BaBB

B*

*

1

2

(N* is the rounded value of N)

Content of the temporary buffer

18

3. Experiments and Results



The experimental verification is carried out on large real ECG The experimental verification is carried out on large real ECG databases. American Heart Association Holter and 8-standard databases. American Heart Association Holter and 8-standard leads data bases are used.leads data bases are used.

Experiment with non-stationary interference. M=160 μV, F=50 Hz and Ф=400 Hz [6]

19



Experimental data, presented at the Academic Lecture of Prof. Ivan K. Daskalov, February, 27, 2003

20



How to test the method objectively?

AA Take a “clean” Take a “clean” ECG record without ECG record without interference.interference.BB Add to the record Add to the record an interference signal.an interference signal.CC Perform the Perform the filtration.filtration.DD Subtract the Subtract the corresponding corresponding samples of the clean samples of the clean and filtered record to and filtered record to obtain the error signal.obtain the error signal.EE Visually and Visually and statistically evaluate statistically evaluate the error signal.the error signal.

Test for odd multiple sampling, F=50 Hz and Ф=250 Hz.

21



Test for non-multiple sampling, F=60 Hz and Ф=250 Hz.

22

4. Acknowledgements



We would like to acknowledge the contributions of: Cveta We would like to acknowledge the contributions of: Cveta Georgieva Ph.D. student for her theoretical research of the Georgieva Ph.D. student for her theoretical research of the total transfer function of the subtraction method and Todor total transfer function of the subtraction method and Todor Stoyanov Ph.D student for research in software Stoyanov Ph.D student for research in software synchronisation of the sampling rate to the powerline synchronisation of the sampling rate to the powerline frequency.frequency.

23

5. Discussion and Conclusions



Ph.D and Dr.Sc thesises in BulgariaPh.D and Dr.Sc thesises in Bulgaria

[4] [4] Mihov GMihov G. . Software automated electronic devices for processing and visualization Software automated electronic devices for processing and visualization of electrocardiosignalsof electrocardiosignals. Ph.D. thesis, TU-Sofia, 1983 . Ph.D. thesis, TU-Sofia, 1983 ((in Bulgarian).in Bulgarian).[12] Levkov Ch. [12] Levkov Ch. Engineering support of automated electrocardiographyEngineering support of automated electrocardiography. D. Sc. . D. Sc. thesis, Sofia, 1987 thesis, Sofia, 1987 ((in Bulgarian)in Bulgarian)..[1[133] Dotsinsky I. ] Dotsinsky I. Methods and Devices for ECG Signal AnalysisMethods and Devices for ECG Signal Analysis. D.Sc. thesis, Sofia, . D.Sc. thesis, Sofia, 19861986 ( (in Bulgarian)in Bulgarian)..[1[144] Christov, I. ] Christov, I. ECG Signals Pick-up, Processing and DisplayECG Signals Pick-up, Processing and Display.. Ph.D. thesis, Sofia, Ph.D. thesis, Sofia, 19881988 ( (in Bulgarian)in Bulgarian)..

Contribution of the PaperContribution of the Paper

Theoretical generalization of the Subtraction Method for Removing Theoretical generalization of the Subtraction Method for Removing Powerline Interference from ECG and its Application.Powerline Interference from ECG and its Application.

24



Papers and PublicationsPapers and Publications

- Citations in 46 papers from Bulgaria, Germany, France, China, - Citations in 46 papers from Bulgaria, Germany, France, China, USA, Canada etc.USA, Canada etc.- The method has been applied from other scientists:- The method has been applied from other scientists:[[1515] Yan, X. G. ] Yan, X. G. Dynamic Levkov – Christov subtraction of mains interferenceDynamic Levkov – Christov subtraction of mains interference. . Medical & Biological Engineering & Computing, 1993,Medical & Biological Engineering & Computing, 1993, 31, 635-638.31, 635-638.[16] Gang Li, Lin Ling, Yu Qilian, Yu Xuemin. A new adaptive coherent model [16] Gang Li, Lin Ling, Yu Qilian, Yu Xuemin. A new adaptive coherent model algorithm for removal of power-line interference. J. of Clinical. Engng, 1995, 20, algorithm for removal of power-line interference. J. of Clinical. Engng, 1995, 20, No 2, 147-150.No 2, 147-150.[17] Monaco, A. [17] Monaco, A. Sviluppo di un modulo software per la gestione di unSviluppo di un modulo software per la gestione di un sistema sistema per il controllo remoto dei portatori di pacemakerper il controllo remoto dei portatori di pacemaker.. Tesi di laurea Tesi di laurea in tecnologie in tecnologie biomediche, Universitа degli studi di Napoli 'Federico II',biomediche, Universitа degli studi di Napoli 'Federico II', Facoltа di ingegneria, Facoltа di ingegneria, Dipartimento di Ingegneria Elettronica e delleDipartimento di Ingegneria Elettronica e delle Telecomunicazioni. 2000Telecomunicazioni. 2000 (in (in Italian).Italian).- The method has been applied for different type of signals:- The method has been applied for different type of signals:[1[188] K. E. Butler, R. D. Russell “Cancellation on Multiple Harmonic Noise Series ] K. E. Butler, R. D. Russell “Cancellation on Multiple Harmonic Noise Series in Geophysical Records”, Geophysics, Jan. 8, 2003, Canada.in Geophysical Records”, Geophysics, Jan. 8, 2003, Canada.

25



Future research and improvementsFuture research and improvements

- Improvement of criteria for linearity in order to avoid false recognition - Improvement of criteria for linearity in order to avoid false recognition of linear segments;of linear segments;- Theoretical evaluation of the total transfer function of the subtraction - Theoretical evaluation of the total transfer function of the subtraction method including non-linear segments;method including non-linear segments;- Real time processing for the case where the interference is non-- Real time processing for the case where the interference is non-stationary.stationary.- Real-time processing when ECG signal is sampled in very high rate - Real-time processing when ECG signal is sampled in very high rate (oversampling).(oversampling).

26

6. References



[1] Levkov C., Mihov G., Ivanov R., and Daskalov. I. Method and device for elimination of power-line frequency disturbance during analysis of biological signals. Bulgarian Patent No 30792 30.06.1980, (in Bulgarian).[2] Daskalov I., Ivanov R., Levkov C. and Mihov G. Elimination of power-line frequency disturbance by microprocessor. Proceedings of the XVI Scientific Session “Radio day ‘81”, May 1981, Sofia, 97-100 (in Bulgarian). [3] Daskalov I., Ivanov R. and Mihov G. Microprocessor controlled ECG slector for screening analysis. Proceedings of the XVI Scientific Session “Radio day ‘81”, May 1981, Sofia, 93-96 (in Bulgarian).[4] Mihov G. Software automated electronic devices for processing and visualization of electrocardiosignals. Ph.D. thesis, TU-Sofia, 1983 (in Bulgarian).[5] Levkov, C., Mihov, G., Ivanov, R., and Daskalov. I., 1984, Subtraction of 50 Hz interference from the electrocardiogram. Medical & Biological Engineering & Computing, 22, 371-373.[6] Christov, I. and Dotsinsky, I., 1988, New approach to the digital elimination of 50 Hz interference from the electrocardiogram. Medical & Biological Engineering & Computing, 26, 431-434.[7] Dotsinsky, I. A., I. K. Daskalov, Comments on ‘Dynamic Levkov-Christov subtraction of mains interference’. Med. & Biol. Eng. & Comput., 33, 1995, 360.

27



[8] Mihov G. Elimination of non-synchronized power-line frequency disturbance from ECG signal. Proceedings of the Anniversary “50 years TU-Sofia” National Scientific Conference “Electronics ’95, Sozopol, book 3, 197-202 (in Bulgarian).[9] Georgieva, T. and Mihov. G., 2002, On-line subtraction method for mains interference removing from ECG signals with signal processor. Proceedings of the 6th International Symposium ‘Microelectronics Technologies and Microsystems’, November 20-22, Lviv, Ukraine, pp. 87-93.[10] Mihov, G., 1996, Elimination of Mains Interference from the ECG in Non-Syncronized Sampling: A Theoretical Approach. Proceedings of the 13-th biennial international conference BIOSIGNAL’96, Brno, Czechoslovakia. pp. 189-191.[11] Levkov, C. and Mihov, G., 1996, Rejection – subtraction filter of mains interference from the ECG. Proceedings of the 13-th biennial international conference BIOSIGNAL’96, Brno, Czechoslovakia. pp. 183-185.[12] Levkov Ch. Engineering support of automated electrocardiography. D. Sc. thesis, Sofia, 1987 (in Bulgarian).[13] Dotsinsky I. Methods and Devices for ECG Signal Analysis. D.Sc. thesis, Sofia, 1986 (in Bulgarian).[14] Christov, I. ECG Signals Pick-up, Processing and Display. Ph.D. thesis, Sofia, 1988 (in Bulgarian).[16] Gang Li, Lin Ling, Yu Qilian, Yu Xuemin. A new adaptive coherent model algorithm for removal of power-line interference. J. of Clinical. Engng, 1995, 20, No 2, 147-150.

28



[17] Monaco, A. Sviluppo di un modulo software per la gestione di un sistema per il controllo remoto dei portatori di pacemaker. Tesi di laurea in tecnologie biomediche, Universitа degli studi di Napoli 'Federico II', Facoltа di ingegneria, Dipartimento di Ingegneria Elettronica e delle Telecomunicazioni. 2000 (in Italian).[18] K. E. Butler, R. D. Russell (2003) “Cancellation on Multiple Harmonic Noise Series in Geophysical Records”, Geophysics, Jan. 8, Canada.[1[199] Dotsinsky I.,] Dotsinsky I., I. Daskalov.I. Daskalov. Suppressing of Poverline Interference in ElectrocardiogramSuppressing of Poverline Interference in Electrocardiogram. . "Е+Е", 5-6/1996, 32 –38 (in Bulgarian)."Е+Е", 5-6/1996, 32 –38 (in Bulgarian).[20] Christov, I., 2000, Dynamic powerline interference subtraction from biosignals. Medical Engineering & Technology, 24, 169-172.[21] Daskalov, I., Dotsinsky I. Reader respond to “A new adaptive coherent model algorithm for removal of power-line interference” . J. of Clinical Engng., September/October, 1995, 357-358.[22] Bazhyna A., I. Christov, A. Gotchev, I. Daskalov, K. Egiazarian. Powerline Interference Suppression in High-Resolution ECG. Computers in Cardiology, 2003, 30, pp. 561-564.

The paper has been reviewed by Professor Ph.D. Stefan TabakovThe paper has been reviewed by Professor Ph.D. Stefan Tabakovfrom Technical University – Sofiafrom Technical University – Sofia

29



Assoc. Prof. D.Sc.Assoc. Prof. D.Sc.Chavdar LevkovChavdar Levkov

Assoc. Prof. Ph.D.Assoc. Prof. Ph.D.Georgy MihovGeorgy Mihov

Prof. Ph.D.Prof. Ph.D.Ratcho IvanovRatcho Ivanov

Prof. D.Sc.Prof. D.Sc.Ivan K. DaskalovIvan K. Daskalov

Prof. D.ScProf. D.Sc Ivan DotsinskyIvan Dotsinsky

Assoc. Prof. Ph.D.Assoc. Prof. Ph.D.Ivaylo ChristovIvaylo Christov

30

Subtraction Method for Removing Powerline Interference from ECG

Documents

Transcript of Subtraction Method for Removing Powerline Interference from ECG