Hierarchical Deep Learning Model for Skin Cancer Detection and Skin Disease Diagnosis

This paper introduces a unified Deep-learning model incorporating Inception ResNet V2 architecture, selected for its ability to capture intricacies in medical imaging. The model is designed to deal with two challenges: skin cancer detection and the classification of 9 types of skin diseases addressing the critical necessity for timely and accurate diagnosis. The Initial stage centralizes on identifying Melanoma Skin Cancer, reaching an accuracy of 96%. In the second stage, images are classified into one of nine skin diseases including Acne, Eczema, Hair loss, Vitiligo etc., with the model attaining an accuracy of 84% and precision, recall, and F1 scores of 85%. The overall performance of integrating both models demonstrate an accuracy of 91.03%, with precision, recall, and F1 scores of 87.77%, 82.93%, and 83.80%, respectively. This comprehensive model design elevates accuracy and efficiency, constituting a major evolution in dermatological imaging and outstanding other state-of-the-art models.