UNet Segmentation based Effective Skin Lesion Detection using Deep Learning

Abstract

Skin cancer is a common and possibly fatal condition. Effective therapy depends on early discovery and precise diagnosis. This study proposes a two-step, segmentation-and classification-based method for skin lesion analysis. The U-Net architecture, a semantic segmentation model based on deep learning, is used in the initial stage to segment skin lesions. On the test set, the suggested method achieves a promising segmentation accuracy of 94.88%. Precise segmentation helps separate the skin lesions from the surrounding environment and facilitates further classification. Using a Support Vector Machine (SVM) classifier, the segmented lesions are classified into benign and melanoma categories in the second stage. The classification results show a 78% accuracy rate, indicating that the suggested method has the capacity to differentiate between benign and malignant skin lesions.