Published Paper Details:

Diabetic retinopathy (DR) is a serious eye disease that affects individuals with diabetes and can lead to vision loss and blindness. Preventing vision loss requires timely recognition and therapy of DR. Currently, DR detection is performed by trained ophthalmologists using fundus images of the eye. However, due to the increasing number of individuals with diabetes, the demand for DR screening exceeds the availability of trained professionals, making it necessary to develop automated DR detection systems. We provide a unique technique to DR detection based on multitask learning in this study. The proposed model is trained on a large dataset of fundus images to perform multiple tasks simultaneously, including DR classification, gaussian blur, and grey scaling. By jointly optimizing all tasks, the model is able to make more informed predictions and improve its performance compared to traditional single-task models. Experimental results on a freely available DR dataset show that our multitask learning approach outperforms state-of-the-art DR detection models in terms of F1-score and accuracy. Additionally, the gaussian blur capabilities of the model enable it to provide visual explanations for its predictions, which can be valuable for clinicians to understand the basis for DR diagnosis. In conclusion, our proposed multi-task learning approach for DR detection demonstrates promising results and has the potential to be used as an efficient and effective screening tool for DR in a clinical setting.