Check out Medium for our full project report and findings
Skin cancer is one of the most prevalent cancers globally, and early detection significantly improves prognosis. This project aims to develop a machine learning pipeline for binary classification of skin lesions (cancerous vs. noncancerous) and explain the predictions using state-of-the-art interpretability techniques like Integrated Gradients and Guided Grad-CAM (via Captum). The goal is to provide a model that is not only accurate but also trustworthy for real-world healthcare applications.
- Deep Learning Model: A convolutional neural network (CNN) trained on the ISIC dataset.
- Hybrid Approach: Outputs from the CNN's dense layer are used as inputs for additional classifiers in a Mixture of Experts model.
- Interpretability: Attribution overlays generated using Captum to explain predictions.
- Class Imbalance Handling: Techniques to address the significant imbalance between cancerous and noncancerous samples in the dataset.
The dataset for this project is from the 2024 ISIC Competition on Kaggle, which can be found here: https://www.kaggle.com/competitions/isic-2024-challenge/data