Web-based PoCT for Preliminary Detection of Skin Cancer Using Machine Learning and Deep Learning Models and Their Analysis

Skin cancer is a prevalent and fatal illness, for which the treatment outcome is significantly enhanced with early diagnosis. Traditional diagnosis entails dermatological examinations, which are inaccessible owing to geographic and economic limitations. This study explores a new approach utilizing deep learning (DL) models to develop web-based self-diagnosis tools for detecting skin cancer. Various state-of-the-art models, including YOLOv8, ResNet18, InceptionV3, EfficientNet B1, and Vision Transformers, are used on individual websites so that users can upload a photo of their skin for self-diagnosis. The models identify conditions as cancer or non-cancer and forecast individual types of cancer like melanoma, basal cell carcinoma, and squamous cell carcinoma. The research contrasts model performance based on accuracy, precision, and computational expense to determine their applicability in real-world applications. The study provides insights in the form of AI-driven dermatological assessments that enhance accessibility and early detection. The research significantly advances AI-driven healthcare, creates awareness, and improves skin cancer diagnoses.