SKIN CANCER DETECTION USING DEEP LEARNING

Skin cancer is one of the most prevalent and life-threatening malignancies worldwide, where early and accurate diagnosis significantly improves patient outcomes. This paper presents an automated skin lesion analysis system leveraging deep learning for simultaneous classification and segmentation of dermoscopic images. The proposed system utilizes the HAM10000 dataset comprising images across seven diagnostic categories: Melanoma (MEL), Melanocytic Nevi (NV), Basal Cell Carcinoma (BCC), Actinic Keratosis (AKIEC), Benign Keratosis (BKL), Dermatofibroma (DF), and Vascular Lesion (VASC). For classification, DenseNet121 and Vision Transformer (ViT) architectures are explored and compared, while U-Net and DeepLabV3+ with ResNet-50 backbone and Atrous Spatial Pyramid Pooling (ASPP) are evaluated for precise lesion segmentation. The best-performing models are deployed within a Flask-based web application that provides on-demand review of uploaded dermoscopic images, delivering instant lesion boundary visualization, disease classification, confidence scoring, severity assessment, and personalized clinical recommendations. This interactive diagnostic feedback system empowers both clinicians and patients with immediate, actionable insights, demonstrating that integrating convolutional and transformer-based deep learning models into an end-to-end web platform offers a reliable and scalable solution for early skin cancer detection.

Keywords: Skin cancer detection, dermoscopy, DenseNet121, Vision Transformer, U-Net, DeepLabV3+, HAM10000, web-based diagnosis, deep learning, Flask web application.