Technical Description
The present invention provides a diabetic retinopathy detection system. It is a self-learned scalable system which is trained to classify labeled fundus images to predict diabetic retinopathy severity grade.
Problems Addressed
- High Cost
- Insufficient Labeled Data
- Need of Scalable System
- Supervised Pre-Training Requirement
- Lacks Pre-Trained Module
- Dependency on Human Labels
Tech Features
- Inexpensive
- Scalable System
- Improved Detection Accuracy
- Bypasses Cumbersome Image Labeling
- Minimum Human Intervention
- Incorporates Contrastive, Reconstruction & Clustering Method
- Self-Supervised Unlabeled Training
- Enhanced Medical Imaging
- Pre-Trained Severity Classifier
- Harness Raw Data
- Achieve Faster Convergence
Target Audience
- Pharmaceutical Industry
- Healthcare and Medicine
- Biomedical Device Sector
- Research and Development
Tech ID: 14-1149 TRL 4 Patent Status: Granted Available For Exclusive and Non-exclusive License
