MobileNetV2-SVM Hybrid Model Classification of Diabetic Retinopathy effected Human Eye

In the realm of medical diagnostics, particularly in the ophthalmology sector, Diabetic Retinopathy (DR) stands as a formidable challenge. DR, a condition that can lead to blindness if not diagnosed and treated early. Traditional approaches, heavily reliant on the meticulous examination of fundus photographs by seasoned clinicians, are not only time-intensive but are also prone to inconsistencies due to subjective interpretations. Researchers have pioneered a novel method merging MobileNetV2 and Support Vector Machine (SVM) to enhance DR severity classification. Traditional DR diagnosis methods fall short due to their heavy reliance on large datasets and substantial computational power, often unfeasible in mobile settings. The innovative approach utilizes MobileNetV2 for its compact, efficient design, optimized for mobile applications through depthwise separable convolutions, paired with the accuracy and robustness of SVM. This hybrid model, trained on the enriched APTOS 2019 dataset through data augmentation and resampling, addresses the challenges of data scarcity and computational demands. Remarkably, it achieved over 90% accuracy in classifying DR severity, marking a significant leap in ophthalmic diagnostics. This model not only streamlines DR screening but also makes it accessible and scalable, potentially reducing vision loss risk for millions, demonstrating a groundbreaking advancement in leveraging machine learning for healthcare.