Abstract
Purpose :
To present the evolution of RetinAsk: The Queryable Atlas of the Retina
Methods :
Originally presented in 2000 (Sayegh et al, AAO 2000) RetinAsk is unique as it is a queryable retina atlas. Over the span of 15 years we were able to take advantage of new imaging technologies such as OCT as well as faster computer systems allowing implementation of sophisticated algorithms in real time. In 2015 we integrated OCT imaging into the database and implemented better search algorithms (Taibl and Sayegh, ARVO 2015). In 2016 we have started an new phase in the development of RetinAsk by implementing simplified forms of computer-assisted diagnosis (CAD) rooted in algorithms that map image features to language tags and integrate features from multiple imaging modalities. We also allowed for the front end to allow "superusers" to submit labeled images and for the program to group them with other, similarly tagged, images.
Results :
Initial programming focused on specific features such as disc detection, cup to disc ratio, probability of glaucoma, vessel detection, and retinal thickness. These are then integrated at a “higher level” of decision making to assist in the diagnosis. Several conditions have been modeled and underwent preliminary testing.
Conclusions :
The Queryable Atlas of the Retina is developing into a universal retina database for ophthalmologists, trainees, and photographers, allowing sharing, learning and collaboration and as a prototype of diagnostic support system.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.