ECDL 2006 An Exploration of Space-Time Constraints on Contextual Information in Image-based Testing...
-
date post
15-Jan-2016 -
Category
Documents
-
view
216 -
download
0
Transcript of ECDL 2006 An Exploration of Space-Time Constraints on Contextual Information in Image-based Testing...
ECDL 2006
An Exploration of Space-Time Constraints on Contextual
Information in Image-based Testing Interfaces
Unmil Karadkar, Marlo Nordt
Richard Furuta
Cody Lee
Christopher Quick
Texas A&M University
Center for the Study of Digital Libraries
The Department of Computer Science
Michael E. DeBakey Institute
Veterinary Physiology & Pharmacology
ECDL 2006
Digital Image Collections• Collections
– Museums• The Louvre
– Historical image archives• New York Times
• Supported tasks– Searching and browsing
• Typical interfaces– Table/grid
ECDL 2006
Specialized Tasks• Analytical tasks
– Examine image content– Interpret visible detail
• Assessment tasks– Estimate dimensions of
objects within images
• Associative tasks – Compare and contrast
multiple images
ECDL 2006
Images in Cardiovascular Research
• Artery / Vein recognition– Analytical task
• Estimate order of blood vessels– Assessment task
• Identification of Lymphatic vessels– Associative task
ECDL 2006
Montage Layout
• Spatial Dimensions (0)
• Context (user dependent)
• Temporal Dimension (used)
• Slowed down video
ECDL 2006
Scrolling Layout
• Spatial Dimensions (1)
• Context (4 images)
• Temporal Dimension (used)
• Smooth scrolling
ECDL 2006
Thumbnail Layout• Spatial Dimensions (2)
• Context (all images in set)
• Temporal Dimension (unused)
ECDL 2006
Properties of Image Layouts Montage ThumbnailScrolling
Spatial Dim.
Temporal Dim.
Context
0
Used
User-dependent
1
Used
4 images
2
Unused
16 images
ECDL 2006
Experimental Design• 3 image layouts
– Thumbnail, Scrolling, Montage
• 16 images per set
• 3 tasks– Artery/Vein recognition– Size Estimation– Lymphatic vessel wall identification
• Balanced across subjects – Ordering of tasks (AEL, ELA, LAE)– Ordering of image layouts (TMS, MST, STM)
ECDL 2006
Subjects
• Trained researchers in the bat lab– Available subject pool of 30
• 9 experts– Research experience of more than 2 semesters
• 6 novices– Research experience of 1 semester only
• Age group 18 to 35
• 4 female, 11 male
ECDL 2006
Data Collection• Responses to image-based
questions– Verbal and drawn– Validated by experts– Instrumental measurements
• Demographic questionnaire– Self evaluation of expertise
• Task reflection questionnaire– For each task
• Preferred layout• Perceived best performance
ECDL 2006
• Did not improve between sets
• Performance constrained by knowledge
• Improved between sets (p=0.0004)
• Performance constrained by layout
Artery/Vein RecognitionExperts
0%
20%
40%
60%
80%
100%
T M S
Layout
Per
form
ance
Round 1
Round 2
Novices
0%
20%
40%
60%
80%
100%
T M S
Layout
Pe
rfo
rma
nc
e
Round 1
Round 289.6%
91.4
%
• Performance differed significantly between groups• Both groups preferred the Thumbnail layout
(p=0.005)
ECDL 2006
Size EstimationExperts
0%
20%
40%
60%
80%
M1 M2 S1 S2 T1 T2
Round
Pe
rfo
rma
nc
e +/- 40%
+/- 30%
+/- 20%
+/- 10%
Novices
0%
10%
20%
30%
40%
M1 M2 S1 S2 T1 T2
Round
Pe
rfo
rma
nce +/- 40%
+/- 30%
+/- 20%
+/- 10%
General understanding of the order of blood vessels does not translate into accurate estimation of size
(p<0.002)
(p<0.002)
• Judged at various levels of accuracy (10% - 40%)• Performance differed between groups at all levels• Both groups preferred the Thumbnail layout
ECDL 2006
Lymphatic Vessel Wall Identification
Experts
0%
20%
40%
60%
80%
100%
M S T
Layout
Per
form
ance
Novices
0%
20%
40%
60%
M S T
Layout
Per
form
ance
• Preferred the Montage layout
• Also worked best with the Montage layout (p=0.035)
(p=0.03)87%
55%
Experts and Novices use different strategies for applying contextual information
• Preferred the Montage layout
• But worked best with the Thumbnail layout
ECDL 2006
Continuing Work
• User studies with Summer research students– 21 novices (9 male, 12 female)
• Comparison of training strategies– 10-week intensive vs. 14-week sporadic training
• Results– Consistent with those of the 6 novices presented
here (not statistically different)– Differences in training style did not affect
performance
ECDL 2006
Lessons Learned
• Digital Image Collections enable discipline-specific scholarly tasks– Typical collections provide only a thumbnail
layout view
• Performance on these tasks is affected by– Image layout properties– Task characteristics– User experience and expertise
Collections should provide different layouts
ECDL 2006
Future Work
• Investigate mechanisms for conveying dimensions of objects within images
• Explore other image layouts– Static and dynamic collages
• Train subjects on new interfaces– Improve learning time– Increase productivity
ECDL 2006
http://ebat.tamu.edu
Center for the Study of Digital Libraries
Michael E. DeBakey Institute
Further Information
Unmil Karadkar
Marlo Nordt
Richard Furuta
Christopher Quick [email protected]