Using Virtual Telepathology to Train Pathology Residents
Elizabeth A. Krupinski, PhD Allison Tillack, MS Lynne Richter, MT Jeffrey Henderson, MD Achyut Bhattacharyya, MD Katherine Scott, MD Anna Graham, MD Michael Descour, PhD John Davis, MD Ronald S. Weinstein, MD
Arizona Telemedicine Program University of Arizona
Rationale
• Little research in pathology on the nature & development of expertise • Much known in radiology from eyeposition studies with significant differences between novices & experts • Differences similar in pathology? • Can we utilize differences to develop training programs to foster expertise?
Virtual Slides
• Dmetrix virtual slide processor •Three 80-element 10 x 8 lenslet arrays stacked vertically with 80 each side • Each has 250 micron FOV • Scans @ 3 mm/sec yields 3000 images/sec/array • 58 sec for 2.25 cm2 section • 40X objective detail •1 image = 200 MB – 1 GB storage space
Eye-Position Study
• • • • 20 virtual slides (breast biopsy) IBM 9-MegaPixel color display Initial resolution only 9 observers
– 3 pathologists – 3 pathology residents – 3 medical students
• ASL 4000SU Eye Tracker • Pick 3 areas most want to zoom on
Marked Zoom Areas
Marked Zoom Areas
Eye-Tracking
Eye-Tracking Data
Cluster
Saccade
Fixation (size = dwell)
Med Student Resident Pathologist
Viewing Time
Mean Total View Time (sec)
14 12 10 8 6 4 2 0 Path Res Med
F = 29.880, p < 0.0001 All differ significantly
Number Fixations
50 45 40 35 30 25 20 15 10 5 0 Path Res Med
F = 47.891, p < 0.0001 All differ significantly
Mean Number Fixations
# Saccades
50
Average Saccades Generated per Image
40 30 20 10 0 Path Res Experience Level Med
Benign
Malignant
F = 36.063, p < 0.0001 All differ significantly
Saccade Length
0.6
Saccade Length (sec)
0.5 0.4 0.3 0.2 0.1 0 Path Res Med
F = 54.761, p < 0.0001 All differ significantly
Saccade Distance
Saccade Length Deg Visual Angle
6 5 4 3 2 1 0 Path Res Med
F = 11.975, p < 0.0001 All differ significantly
Discussion
• Trained pathologists are more efficient at extracting information from virtual slides than less experienced residents or med students • Field of attention expands with expertise taking in more relevant information in the global percept or gist followed by directed focal search
Discussion
• Possibly use expert scan paths as examples for learners to accelerate the adoption of better search strategies • Possibly utilize expert scans to automatically change magnifications & field selections • Possibly use in self-assessment or proficiency testing scenarios
THANK YOU!