Impact of Fuzziness for Skin Lesion Classification using Vision Transformer

In dermatology, classifying skin lesions is essential for assisting in the early identification of disorders of the skin, including melanoma. Traditional deep learning models often struggle with uncertainty in classification, leading to potential misdiagnoses. This project integrates a Vision Transformer (ViT)-based deep learning model with fuzzy logic to enhance decision-making in skin lesion classification. The ViT model extracts high-level features from dermoscopic images, while the fuzzy logic system refines the final decision by incorporating confidence levels. The proposed approach aims to reduce false positives and negatives by handling uncertain cases more effectively. The findings from the experiments reveal that this combined approach enhances both the accuracy and reliability of classification, positioning it as a viable solution for practical clinical use.