The ISIC2024 Skin Cancer Detection project aims to develop machine learning algorithms that identify skin cancer from 3D total body photographs and patient metadata. The goal is to support early diagnosis of melanoma, basal cell carcinoma, and squamous cell carcinoma, improving patient outcomes.

Inspired by the ISIC2024 Kaggle competition, we used a dataset of skin lesion images and related clinical information from thousands of patients. We trained deep learning models, including ResNet50 and EfficientNetV2, to predict malignancy, with a focus on high sensitivity. The primary evaluation metric was the Partial Area Under the ROC Curve (pAUC) above an 80% True Positive Rate (TPR).

Our best model achieved a pAUC of 0.140, demonstrating the potential of AI in skin cancer detection. Future work will focus on refining the models, improving image preprocessing, and exploring new ensemble techniques.