Skin Lesion Segmentation Using U-Net

An accurate investigation of skin lesions is essential for dermatological diagnosis and treatment to be successful. In this research, a thorough investigation that included segmenting skin lesions and subsequently classifying them is presented. On the ISIC 2018 dataset, we precisely distinguish lesion types using the U-Net architecture, obtaining a noteworthy accuracy of 92%. In addition, we classify the segmented lesions using a Convolutional Neural Network (CNN) model. As a result of the combined strategy, the CNN was able to classify the segmented images with an astounding accuracy of 95%. This achievement highlights the effectiveness of deep learning approaches. These results highlight the value of using cutting-edge machine learning techniques to dermatology, improving clinical judgment and patient care.