An AI-based Skin Cancer Detection Using Raspberry Pi

Skin cancer is a major global health concern, necessitating early detection for improved patient outcomes. This study presents an AI-powered skin cancer detection system using a fine-tuned MobileNetV2 Convolutional Neural Network (CNN) integrated into a Raspberry Pi 5 microcontroller. The model was trained on an augmented dataset comprising images from the International Skin Imaging Collaboration (ISIC) and DermNet, classifying melanoma, basal cell carcinoma, squamous cell carcinoma, and non-cancerous lesions. The system achieved an impressive average accuracy of 96.71%, with precision, recall, and F1-score values of 88.62%, 92.48%, and 90.15%, respectively. Comparative analysis with prior studies demonstrates superior classification performance due to the optimized deep learning approach and enhanced computational capabilities of the Raspberry Pi 5. While the model exhibited high accuracy, challenges were noted in distinguishing squamous cell carcinoma, emphasizing further dataset expansion and model refinement. The developed system offers an efficient, low-cost, and portable preliminary skin cancer screening solution, particularly in resource-limited settings.