Dr.TEJASWINI ANTARVEDI

DR. ARVIND KUMAR MORYA, DR BHARAT GURNANI, Dr. SIDDHARAM JANTI

Abstract

Deep Learning (DL) & Artificial Intelligence (AI) have become widespread due to the advanced technologies & availability of digital data.This study evaluates viability of an annotation tool which works on a smartphone & can be used in a healthcare setting. Method: We developed a smartphone-based grading system for grading multiple retinal fundi.The process consisted of designing the flow of user interface (UI) keeping in view feedback from experts. Quantitative & qualitative analysis done .The dataset size was approximately 2 lac images with adjudicated labels by a minimum of 3 doctors. Results: A binary referrable retinal DL model was used. 32 doctors graded all images. Data analytics suggested significant portability ,flexibility & sensitivity of the tool. Conclusion: The user feedback & feature usage confirm with high specificity in offline mode & remotest of areas.Same concept can be used for fundus images of other ocular diseases as well as other streams of medical science .

Full Text