Project title
Early Detection of Retinopathy with AI
Project aim
The goal of this project is to improve the accuracy and dependability of diabetic retinopathy classification by creating an innovative interactive tool that uses a hybrid machine learning model that combines the strengths of the ResNet-50 and InceptionV3 architectures.
Project outline
Early identification of diabetic retinopathy is essential to prevent serious vision impairment, as it is a significant cause of blindness among the diabetic population.
This hybrid approach is different from past work that mainly focused on algorithm efficiency without integrating it into user-friendly applications, as it blends improved image processing techniques and a user-centered interface.
Utilising the APTOS 2019 dataset, the model undergoes rigorous training and validation, focusing on optimizing image processing techniques and classification algorithms.
To make it easier for healthcare professionals to utilise the tool in clinical settings, it has an intuitive user-centric interface.
Evaluation metrics such as accuracy, precision, and F1 score indicate significant improvements over existing models, supporting the tool’s potential in transforming diabetic retinopathy screening practices.
The hybrid model achieved the highest average accuracy score which is 73.83% while, ResNet-50 achieved 58.32% and InceptionV3 achieved 69.74%.
Future studies will examine real-world deployment scenarios and broaden the model’s applicability to other retinal disorders. This approach solves a significant healthcare problem in a scalable manner while advancing the area of medical image analysis.
Images
