This project is an implementation of the paper "Hybrid machine learning architecture for automated detection and grading of retinal images for diabetic retinopathy" using MATLAB for the detection and grading of Diabetic Retinopathy (DR) in retinal fundus images. Diabetic Retinopathy is a leading cause of vision loss and blindness in people with diabetes, and early detection can significantly reduce the risk of vision loss.
- Automated detection and grading of Diabetic Retinopathy using MATLAB.
- Integration of multiple deep learning models (VGG16 and InceptionV3) for feature extraction.
- Ensemble deep learning approach for improved classification accuracy.
- Utilizes the APTOS 2019 Diabetic Retinopathy dataset.
- Image Preprocessing: Input retinal fundus images are preprocessed to improve image quality using MATLAB's image processing toolbox.
- Feature Extraction: Pre-trained VGG16 and InceptionV3 models extract deep features from preprocessed images. Features from both models are combined to form a composite feature vector.
- Classification: The composite feature vector is fed into a Random Forest classifier for predicting the DR severity stage. (haven't done yet)
To run this project, you'll need:
- MATLAB R2020a or higher
- MATLAB's Deep Learning Toolbox
- MATLAB's Image Processing Toolbox
- Download these CNN's within matlab: Resnet18,Vgg16,Inception-v3,AlexNet
- Download and unzip the APTOS 2019 Diabetic Retinopathy dataset.
- Run the DR_Detection.mlx script to start the learning workflow on the APTOS dataset.
- Read the Report.pdf for the final results
- Class Activation Mapping Results are depicted in the following figure
Special thanks to Barath Narayanan Narayanan et al. for their research and the maintainers of the APTOS 2019 Diabetic Retinopathy dataset.
A python version of this project is in the future work list