Abstract
Purpose: :
Assessing the optic nerve head for diagnosis and management of glaucoma is subjective and requires pattern recognition and experience. Learning may be more effective if the residents’ or glaucoma fellows’ assessment can be compared objectively to a reference standard (rs). Longitudinal evaluation of the trainee’s progress may also be informative. We have developed an online tool (Truthseeker, http://webscreen.ophth.uiowa.edu/disc) to objectively compare residents’ and fellows’ planimetry with an expert’s evaluation of cup and rim on stereoscopic optic disc images. We evaluate the performance of residents and fellows using this tool.
Methods: :
Six ophthalmology residents and 3 glaucoma fellows performed online planimetry by delineating rim and cup on 57 sets of stereo color photographs of 57 patients with suspected or open-angle glaucoma at various stages of disease. The rs was created for this dataset by having 3 glaucoma fellowship trained experts evaluate these images (1). The fellows’ and residents’ planimetry were compared to the rs in 2 ways: linear cup-to-disc ratio and accuracy (number of pixels assigned correctly/total number of pixels) (1).
Results: :
The correlation of 3 1st year residents with the reference standard was 0.58, 0.72, and 0.77; of 2nd year residents 0.45, 0.66, and 0.75; and of the 3 fellows 0.73, 0.81, and 0.86, respectively. The accuracy of correctly assigning each pixel to the right structure (rim, cup, or background) for the 1st year residents was 0.90, 0.91, 0.93; of the 2nd years 0.92, 0.92, 0.94; and of the fellows 0.95, 0.95, and 0.96, respectively.
Keywords: optic disc • imaging/image analysis: clinical • learning