# M4Raw: multi-contrast, multi-repetition four-channel k-space of brain MRI on a 0.3 T low-field scanner

## Facts

- Provider: Mengye Lyu et al.
- License: CC-BY-4.0 (https://zenodo.org/records/8056074)
- Size: 2,265 files
- Format: PNG, TSV (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/mylyu/M4Raw_brain
- Access: public, readable without an account
- DOI: https://doi.org/10.5281/zenodo.8056074
- Category: Neuroscience
- Source: https://doi.org/10.1038/s41597-023-02181-4
- Page: https://scigantic.com/archive/722c55ed-5f5e-5d19-9e94-5219b4e4b893
- Keywords: Neuroscience, MRI, PNG, TSV, Hugging Face

## About

M4Raw is a raw k-space dataset for low-field MRI reconstruction research, acquired on a 0.3 T whole-body MRI system with a four-channel head coil from healthy volunteers. Each examination provides T1-weighted, T2-weighted and FLAIR acquisitions with repeated averages (three repetitions for T1w and T2w, two for FLAIR; six and four in the expanded test cohort), plus GRE acquisitions for the original 183-participant cohort, and dedicated inter-scan and intra-scan motion subsets. Version 1.6 holds 2,230 HDF5 files from 208 unique participants: 1,024 training, 240 validation, 400 test, 200 motion and 366 GRE files, in multicoil_train/, multicoil_val/, multicoil_test/, multicoil_motion/ and multicoil_gre/, with manifests in metadata/ and portable readers in code/. The H5 payloads are byte-identical to the authors' Zenodo record 8056074 and store multicoil k-space with root-sum-of-squares (RSS) image slices; read them with h5py. Uses: denoising and super-resolution of low-SNR low-field images, accelerated reconstruction, multi-repetition self-supervised training and motion artifact studies. About 28 GB (measured 2026-11-17). Released by the original authors under CC BY 4.0; cite Lyu et al., Scientific Data 10, 264 (2023).

## Load the data

```python
import pandas as pd
df = pd.read_csv("hf://datasets/mylyu/M4Raw_brain/metadata/FLAIR_manifest.tsv", sep='\t')
df.head()
```

## Data structure

File types in sampled files: PNG, TSV

Example file: `metadata/FLAIR_manifest.tsv`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/722c55ed-5f5e-5d19-9e94-5219b4e4b893
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/722c55ed-5f5e-5d19-9e94-5219b4e4b893
- Over MCP: `get_schema_card` with id `722c55ed-5f5e-5d19-9e94-5219b4e4b893` at https://api.scigantic.com/mcp

## Related topics

- [Neuroscience datasets](https://scigantic.com/topics/neuroscience)
- [MRI datasets](https://scigantic.com/topics/mri)
- [PNG datasets](https://scigantic.com/topics/png)
- [TSV datasets](https://scigantic.com/topics/tsv)
