BIOSE-Non-labelled MRI Dataset for Unsupervised Learning

Published: 12 June 2025| Version 1 | DOI: 10.17632/9mcp5pbtbr.1
Contributors:
Tahmid Noor Rahman, Tonmoy Paul, Tahiat Zarin Tasnim, Mehedi Hasan Shawon, Rifat Bin Reza, Fardin Hasan, Akter Sumaiya, Md Jobayer

Description

# BIOSE:Non-labelled MRI Dataset for Unsupervised Learning ## Overview This dataset contains anonymized multimodal brain MRI scans from 40 subjects, formatted according to the [BIDS (Brain Imaging Data Structure) 1.8.0](https://bids.neuroimaging.io/) standard. It is designed to support **unsupervised and self-supervised learning research** in medical imaging. --- ## Dataset Summary * **Subjects**: 40 (`sub-00` to `sub-39`) * **Sessions per Subject**: 1 (`ses-01`) * **Total Files**: 497 * **Total Size**: 817.65 MB * **Modalities**: * Anatomical MRI: T1w, T2w, FLAIR, T2star * Diffusion MRI (DWI): Includes `.nii.gz`, `.bval`, and `.bvec` files --- ## Directory Structure ``` dataset_root/ ├── dataset_description.json ├── participants.tsv ├── README.md ├── sub-00/ │ └── ses-01/ │ ├── anat/ │ │ ├── sub-00_ses-01_T1w.nii.gz │ │ ├── sub-00_ses-01_T2w.nii.gz │ │ ├── sub-00_ses-01_FLAIR.nii.gz │ │ └── sub-00_ses-01_T2star.nii.gz │ └── dwi/ │ ├── sub-00_ses-01_dwi.nii.gz │ ├── sub-00_ses-01_dwi.bval │ └── sub-00_ses-01_dwi.bvec ├── ... └── sub-39/ ``` --- ## Files in the Root Directory * `dataset_description.json`: Metadata describing the dataset name, authors, license, and funding. * `participants.tsv`: Table listing participants (subjects) and any relevant metadata (anonymized). * `README.md`: This file. --- ## Intended Use This dataset is best suited for: * **Unsupervised MRI feature learning** * **Contrastive or generative modeling** * **Modality synthesis and harmonization** * **Diffusion MRI pre-processing or denoising** No diagnostic labels or clinical metadata are provided. --- ## License **CC0 (Creative Commons Zero)** – No rights reserved. Free to use for research, education, and commercial purposes. --- ## Acknowledgements * MRI scans sourced from **Islami Bank Central Hospital** * Dataset organized and anonymized by researchers from **BRAC University** and the **Biomedical Science and Engineering (BIOSE) Research Group** --- ## Citation When using this dataset, please cite: Noor Rahman, Tahmid ; Paul, Tonmoy; Zarin Tasnim, Tahiat; Hasan Shawon, Mehedi ; Bin Reza, Rifat; Hasan, Fardin; Sumaiya, Akter; Jobayer, Md (2025), “BIOSE-Non-labelled MRI Dataset for Unsupervised Learning”, Mendeley Data, V1, doi: 10.17632/9mcp5pbtbr.1 ---

Files

Institutions

  • BRAC University

Categories

Neuroscience, Artificial Intelligence, Computer Vision, Biomedical Engineering, Medical Imaging, Magnetic Resonance Imaging, Unsupervised Learning, Brain, Brain Imaging, Neuroimaging, Diffusion Weighted Magnetic Resonance Imaging, Deep Learning, Self-Supervised Learning

Licence