hongweilibran / motum

the official repository for MOTUM dataset

This repository serves as the official source for the MOTUM dataset, a sustained effort to make a diverse collection of multi-origin brain tumor MRI scans from multiple centers publicly available, along with corresponding clinical non-imaging data, for research purposes.

Download

The latest dataset can be downloaded from the Harvard Dataverse repository.

Data format

The patients are grouped into several tumor categories based on their origins. Each patient for any tumor category has the same naming format, such as

001-Glioma-AYEY

    |  flair.nii.gz

    |  t1.nii.gz

    |  t2.nii.gz

    |  t1ce.nii.gz

    |  flair_seg_label1.nii.gz

    |  t1ce_seg_label2.nii.gz

In this case, 001-Glioma-AYEY is the general patient ID, 001 is the patient number, Glioma is the tumor category, AYEY is the center ID, flair_seg_label1 is the segmentation mask of Label 1 derived from the FLAIR sequence, and t1ce_seg_label2 is the segmentation mask of Label 2 derived from the T1-ce sequence.

Automated segmentation tool

We provide a Docker container to segment MR images on your side. segment_one_subject.sh is the shell script to run the container on one subject. Please make sure:

(a) The FLAIR, T1, T1-ce, and T2 images have the same naming format as the 001-Glioma-AYEY example before.

(b) The nvidia-container-toolkit package is installed corrected.

Preprocessing scripts

We provide the shell pre-processing scripts based on existing packages. Please check the 'preprocessing' folder.