Skin Cancer Lesion Detection Using Deep Learning

This final year research project explored the application of artificial intelligence in medical image analysis, specifically the classification of skin cancer lesions. The goal was to investigate the feasibility of using machine learning models, particularly convolutional neural networks (CNNs), to distinguish between different types of skin lesions from dermoscopic images.

The project began with a thorough literature review of existing work in AI-based dermatology, followed by the development of a research methodology involving dataset selection (ISIC archive), image preprocessing, and model evaluation. Several CNN architectures were experimented with and fine-tuned in Google Colab, with performance evaluated using accuracy, precision, AUC, confusion matrix, and F1-score metrics.

This project strengthened my understanding of deep learning theory and practical model training workflows. While it wasn’t deployed as a production system, it served as a robust foundation in applying AI to healthcare problems, with a strong emphasis on model interpretability, evaluation, and ethical considerations in medical AI.