bjayadikary/Brain-Tumor-Segmentation-in-Sub-Saharan-Africa-Adult-Glioma-Dataset

This repository contains the implementation of the MedNeXt architecture with parameter-efficient fine-tuning (PEFT) using convolutional adapters for brain tumor segmentation. For more details: arxiv.org # 📌 Overview This project addresses the segmentation of multi-modal 3D MRI volumes, where the input consists of four MRI sequences: T1-weighted (T1w), T1-weighted contrast-enhanced (T1-c), T2-weighted (T2w), and T2-weighted FLAIR. The model outputs segmentation masks identifying four classes: Background, Enhancing Tumor (ET), Non-Enhancing Tumor Core (NETC), and Surrounding Non-Enhancing FLAIR Hyperintensity (SNFH). # 📌 Environment Setup To set up the environment, clone this repository and install the dependencies from requirements.txt # 📌 Dataset Preparation We use two publicly available datasets: * BraTS 2021 training data: 1251 adult glioma cases * BraTS Africa training data: 60 adult glioma cases with lower spatial resolution and unique characteristics such as late presentation ## Data Preparation for Sub-Saharan African 2023 Dataset The BraTS2023_SSA directory looks like this: ``` BraTS2023_SSA |-- BraTS2023_SSA_Training | |--ASNR-MICCAI-BraTS2023-SSA-Challenge-TrainingData_V2 | | |-- BraTS-SSA-00002-000 | | | |-- BraTS-SSA-00002-000-seg.nii.gz | | | |-- BraTS-SSA-00002-000-t1c.nii.gz | | | |-- BraTS-SSA-00002-000-t1n.nii.gz | | | |-- BraTS-SSA-00002-000-t2f.nii.gz | | | |-- BraTS-SSA-00002-000-t2w.nii.gz | | |-- BraTS-SSA-00007-000 | | | |-- BraTS-SSA-00007-000-seg.nii.gz | | | |-- BraTS-SSA-00007-000-t1c.nii.gz | | | |-- BraTS-SSA-00007-000-t1n.nii.gz | | | |-- BraTS-SSA-00007-000-t2f.nii.gz | | | |-- BraTS-SSA-00007-000-t2w.nii.gz | | |--... ``` ## Stacking the Files We have 60 folders in the `BraTS2023_SSA_Training/ASNR-MICCAI-BraTS2023-SSA-Challenge-TrainingData_V2` and 15 folders under `BraTS2023_SSA/BraTS2023_SSA_Validation` directory. Each folder contai …