A Deep Learning Model for Skin Cancer Detection with PSO-Based Optimization

Skin cancer, particularly melanoma, is one of the most malignant and lethal cancers. Detection early and with precision is essential to enhance survival rates. Conventionally, methods of diagnosis rely on visual examination, which may be subjective and lengthy. This project introduces an automated skin cancer diagnostic system with Convolutional Neural Networks (CNNs) with the addition of Particle Swarm Optimization (PSO) for image preprocessing. Employing the ISIC dataset of 2,637 dermoscopic images tagged as benign or malignant, PSO enhances feature selection by paying attention to pertinent regions, optimizing CNN performance. The system delivers over 95% accuracy with high precision, recall, and F1-score. It facilitates early melanoma detection, facilitating timely treatment. This method presents a scalable and low-cost solution for clinical and remote healthcare, decreasing manual workload and enhancing diagnostic consistency in resource-limited settings.