A HYBRID APPROACH FOR DETECTION OF SKIN CANCER

Abstract

Cutaneous cancer is a common and sometimes fatal condition for which a timely and precise diagnosis is essential to the course of treatment. Automating the identification of skin cancer using dermatoscopic images has showed promise in recent breakthroughs in machine learning, especially in deep learning. However, due to their limitations in handling varied datasets and integrating clinical data, existing models frequently perform poorly in terms of accuracy and generalizability. Using the HAM10000 dataset, this research suggests a hybrid model that combines Transformers and Convolutional Neural Networks (CNNs) to enhance cutaneous cancer detection. The suggested model makes use of multi-modal data and sophisticated regularization techniques to fill in the gaps found in the previous five years of research. With an accuracy of over 95% , the proposed model showed promise for implication in clinical settings.