Automated Detection of DR on Ultra-Widefield Pseudocolor Images

In this interview from ASRS 2016, Dr. Sirnivas Sadda discusses his pilot study using an existing automated algorithm to detect diabetic retinopathy (DR) lesions on ultra-widefield pseudo-color images. The algorithm showed high sensitivity (91.7%), but only moderate specificity (50.0%), which Dr. Sadda attributes to image artifact and eyelashes. With optimization, Dr. Sadda believes that widefield imaging may be used to stage retinopathy severity and could be used to develop a quantitative score in diabetic retinopathy.