Differential Diagnosis of Ringworm and Eczema Using Image Processing and Deep Learning

Abstract

Misdiagnosis of dermatological disorders is an ordinary incident among both doctors and dermatologists around the world. Among them, misdiagnosis of common dermatological disorders is exceptionally high since these types of common diseases appear visually similar on some occasions. Then these easily curable diseases can get devastatingly complicated due to the initial misdiagnosis and wrong treatments. Ringworm and eczema are two commonly misdiagnosed dermatological diseases that often display similar visual attributes as well as non-visual patient history questions. These two diseases were chosen after an investigation to differentially diagnose using an image-based system for this research. In the proposed system, the end-users are allowed to upload an image to the system, and using different image processing techniques; the lesion area will be detected. Then the image will be classified using a convolutional neural network that was trained using a dataset and will display the diagnosis result. To give a better experience, an android application was designed and developed as the frontend of the system. The proposed solution is entirely based on the visual aspects displayed on the lesion, which once trained, is capable of diagnosing with an accuracy of 82%.