Application of Deep Learning and IoT for Detection of Diabetic Retinopathy and Diabetic Macular Edema

الملخص

In recent years, diabetes mellitus has been increasing rapidly, and due to that, around 380 million people around the globe have been affected. This disease may cause many people to become blind and other health issues. Diabetic Macular Edema (DME) and Diabetic Retinopathy (DR) are medical conditions in humans caused by prolonged high blood sugar levels and have a direct impact on human eyesight, which can subsequently lead to blindness. In the early stages, DR usually progresses without any remarkable symptoms, making early detection difficult. If left untreated for a prolonged period, it can result in permanent vision loss. To facilitate proper diagnosis and timely treatment, computer-based systems today often rely on clinical images. In fact, a vital indicator of DR is the presence of microaneurysms (MA), which are critical for identifying the onset of the disease. In line with the emergence of the Internet of Things (IoT), a wide range of electronic devices can be usefully interconnected and are very capable of collecting, transmitting, and responding to data in real time. In the field of human healthcare, such IoT-powered systems possess sufficient capabilities to support remote diagnosis, particularly through the use of medical sensors in telemedicine scenarios. Nonetheless, such a shift can lead to critical privacy issues for a patient. The protection of critical health-related information becomes particularly critical. Hence, the major challenge here is implementing remote systems to support remote diagnosis while ensuring strict confidentiality to protect the patient's privacy. In the present research work, an IoT-based deep learning approach achieving 98.86% accuracy for Diabetic Macular Edema (DME) and 86.04% for Diabetic Retinopathy is proposed.