ARCHIVES

Skin diseases are among the most prevalent health concerns globally, affecting nearly 900 million people at any given time. I n India, the burden is further increased due to climatic conditions, high population density, and limited access to dermatological specialists, especially in rural areas of the country. This study presents Derma Nex, a multi-model deep learning framework designed for skin cancer detection and common dermatological disease classification.The system integrates four specialized models based on Efficient N et architectures to perform binary cancer screening, multi-class lesion classification, melanoma-versus-nevus differentiation, and common skin disease recognition. A three-model ensemble consensus mechanism combines multiple probabilistic predictions to improve the robustness and reliability. The proposed approach achieves strong performance, with high cancer detection sensitivity, 96% AUC-ROC in screening tasks, and up to 85.8% accuracy in the classification of common diseases. To enhance interpretability, the system incorporates a multi-layer Grad-CAM for a visual explanation of the predictions. The models were trained on a diverse dataset of over 50,000 images collected from multiple sources, ensuring improved generalization. Furthermore, the system was deployed as a full-stack web application with risk-level analysis and user-friendly interaction, making it suitable for real-world applications. Overall, DermaNex provides an efficient, scalable, and interpretable solution for AI-assisted dermatological diagnoses.