a-adegite/Lightweight-Brain-Tumor-Segmentation-on-Low-Resource-Systems

A CPU-friendly 3D U-Net for brain tumor segmentation using the BraTS-Africa 2024 dataset optimized for low-resource systems and based on the MAI Lab & SPARK tutorial. # Lightweight-Brain-Tumor-Segmentation-on-Low-Resource-Systems This project implements a lightweight 3D U-Net for brain tumor segmentation on CPU-only and low-resource systems, based on the tutorial: **“Lightweight Brain Tumor Segmentation on Low-Resource Systems: A Step-by-Step Guide with 3D U-Net”** developed by **MAI Lab** and the **SPARK program**. ## 🚀 Project Highlights - Preprocessing and loading of 3D brain MRI data - Simplified and efficient 3D U-Net model - Dice loss optimization and patch-wise training - CPU-optimized training and inference - Based on the **BraTS-Africa 2024** dataset ## 🛠️ Technologies Used - Python 3.12.9 - PyTorch - MONAI - Nibabel - Matplotlib, Seaborn ## Reference If you use this repository, please cite the original tutorial: > Oladele, A., Akintoye, A., Folarin, T., & Adesina, O. (2024). > *Lightweight Brain Tumor Segmentation on Low-Resource Systems: A Step-by-Step Guide with 3D U-Net*. > MAI Lab & SPARK Program, Nigeria. > dx.doi.org