Published Paper Details:

Skin cancer is one of the most prevalent and life-threatening diseases worldwide, emphasizing the need for accurate and early diagnosis to improve patient outcomes. This research presents an advanced skin cancer detection frame- work that integrates the Swin Transformer (Shifted Window Transformer) with semantic segmentation techniques for the classification of dermatoscopic images. The Swin Transformer effectively captures hierarchical global features, enabling precise differentiation between melanoma and benign lesions. To enhance the model's generalizability, preprocessing tech- niques such as image resizing, normalization, and augmentation are applied to datasets including HAM10000, ISIC-2008, ISIC- 2019, and PH-2. Furthermore, Explainable AI (XAI) methodolo- gies, including Grad-CAM and Grad-CAM++, are incorporated to provide visual explanations, highlighting crucial regions influ- encing classification decisions. The model's performance is evalu- ated using accuracy, sensitivity, specificity, and the Dice similarity coefficient, ensuring reliable and interpretable diagnostics. Additionally, Bayesian Optimization is employed for hyperpa- rameter tuning, optimizing model performance while maintaining computational efficiency. This research contributes to the field of AI-driven dermatology by offering a transparent and robust diagnostic tool that assists dermatologists in early melanoma detection. Future enhancements include optimizing the model for real-time edge deployment, expanding datasets to improve generalizability across diverse skin tones, and integrating multi- modal data for enhanced diagnostic accuracy.