An Explainable AI-Based Skin Disease Detection

Lack of understanding and ignorance has led to, skin illness being one of the most frequent and difficult diseases to diagnose. Dermatologists are frequently sought for in many underdeveloped countries for skin problems and prevention. People are unsure about the dermatologist’s medical prescriptions, and there is no reason for this method. Because skin plays a large role in defending the form against fungal and dangerous bacterial infections, it is critical to recognise disease early on and not ignore it. As a result, detecting disease and making a diagnosis early is critical. As a result, in order to provide a feasible and efficient system, particularly in the light of the growth of image processing-based disease analysis, more reminders may be provided. People can submit input through camera techniques, and image processing, and machine learning techniques are integrated; the respective skin problem is recognised, and diagnosis is sometimes indicated. Machine-driven identification of skin illness using deep learning/machine learning techniques is the main goal of this project. In this study, ResNet18 is trained, the model is valued, and the Grad-CAM model explainability technique is used to help dermatologists better understand the model’s predictions. The ResNet18 is the end product of deep learning. The accuracy of the classification model is 96%. This paper presents a survey of the deep learning clarifying literature on skin disease diagnosis using structural photos. The difference between what dermatologists consider explainable and what existing tactics provide is discussed, as well as future proposals for closing the gap.