Automated Skin Disease Diagnosis Using Deep Learning & Image Processing Techniques

Abstract

Skin conditions must be treated promptly and appropriately because if left untreated they can develop into more serious conditions like melanoma, or skin cancer. They must be treated promptly and appropriately. Manual examinations by doctors take time, and the outcomes can vary from one to the next. In order to solve this problem, this study created an automated system that recognizes skin conditions using machine learning and image processing methods. First, skin images are cleaned—hair is removed using digital hair removal technology, and image quality and clarity are improved using Gaussian filtering. Next, the GrabCut algorithm is used to precisely isolate only the affected area (lesion). The structure, color, texture, and statistical features of the area are then extracted, which are used to identify the disease. These features are analyzed by machine learning models such as SVM, KNN, and Decision Tree to determine which skin disease is present. This system helps increase diagnostic accuracy, reduce human error, and speed up the identification process. In study conducted on datasets such as ISIC 2019 and HAM10000, after the test the SVM model demonstrated the highest accuracy. The results shows that this technique can be very helpful for doctors diagnose skin diseases earlier, and automatically.