Diabetic Retinopathy Detection Using AI/ML

EasyChair Preprint 9524

Abstract

Diabetes, it’s a disease, that is caused by the hormone(insulin) imbalance int the body, leading to not synthesizing sugar properly in the human body. Diabetic Retinopathy is a type of debility, that is imitative by excess of habitual diabetes and its major damage is, blindness, if not timely treated. In order to stop proper blindness, it is necessary that recent therapeutic diagnosis of diabetic retinopathy and its red-carpet remedy is done in proper order & time, so as to resist the excessive side of retinopathy.
The physical detection of diabetic retinopathy, takes much time, the doctors physically take much time, and in this particular time duration, the patients need to suffer the most. If such a system that can detect diabetic retinopathy quickly, then the treatment process can start asap. This study conducted, proposes a machine learning method, that classifies the image mainly into 4 categories.
1) NO DR
2) MILD
3) MODERATE
4) PROLIFERATE DR
From statistical combination of machine learning algorithms, example K-nearest neighbor, random forest, logistic regression, multilayer perception artery, the highest accuracy that was achieved was around 70-75%. The approach combined took a total score of 0.8001 and the F-score took around 0.7938.

Keyphrases: Diabetes, Retinopathy, machine learning

@booklet{EasyChair:9524,
  author    = {Kshtij Sharma and Akshit Punj and Sarthak Srivastav},
  title     = {Diabetic Retinopathy Detection Using AI/ML},
  howpublished = {EasyChair Preprint 9524},
  year      = {EasyChair, 2023}}