DR. ARVIND KUMAR MORYA

Dr. VAISHALI LALIT UNE, Dr. SIDDHARAM JANTI, Dr. RAJENDRA PRASAD

Abstract

Introduction: World’s first smartphone based annotation- tool developed on the principle of Deep Learning & Artificial Intelligence ,Supervised learning algorithms used for better extractor-quantifier images of DR.

Methods: A smartphone-tool to grade multiple retinal fundi(55000) & designing the flow of user interface (UI) keeping in view feedback from experts. Quantitative and qualitative analysis of change in speed of a grader over time and feature usage statistics done.

Results: We created a DL model for a binary referrable DR Classification task. A total of 32 doctors used it for 55000 images. Analytics suggested significant portability and flexibility. Grader variability for images was in Mean agreement of 75.9% on annotation. Conclusion: Annotations of DR Images was faster & easier without quality degradation. Statistics confirm incorporation of brightness and contrast variations, green channels.

Keywords: artificial intelligence, deep learning, referrable diabetic retinopathy

Full Text