DermAssist AI is an advanced artificial intelligence-powered dermatological diagnostic assistance system that leverages deep learning and computer vision to analyse skin lesion images and provide preliminary diagnostic insights for a wide spectrum of common skin conditions. The system is built upon a Convolutional Neural Network (CNN) architecture enhanced with transfer learning from a pre-trained EfficientNetB3 model, trained on the HAM10000 dataset containing over ten thousand labelled dermatoscopic images spanning seven diagnostic categories: Melanocytic nevi, Melanoma, Benign keratosis-like lesions, Basal cell carcinoma, Actinic keratoses, Vascular lesions, and Dermatofibroma. The complete data science and software engineering lifecycle is implemented, encompassing systematic data preprocessing, augmentation, model training, performance evaluation using medical-grade metrics (AUC, sensitivity, specificity), and deployment as an interactive Flask web application. An explainability layer using Gradient-weighted Class Activation Mapping (Grad-CAM) highlights the specific image regions most influential in the diagnostic prediction. The system achieves a macro-averaged AUC of 0.889 across all seven classes, demonstrating strong generalisation capability. DermAssist AI represents a meaningful contribution to the democratisation of dermatological care through artificial intelligence.
Priya et al. (Thu,) studied this question.