---
license: mit
language:
- en
tags:
- medical
- biology
- cartilage
- t2mapping
- OAI
pretty_name: OAI T2 maps with EPG fitting
---
# Osteoarthritis Initiative (OAI) T2 Maps – EPG Fit Dataset

This dataset repository contains T2 maps derived from the Multi-Echo Spin-Echo (MESE) MRI data in the Osteoarthritis Initiative (OAI). 
The maps were generated specifically for cartilage regions using the Extended Phase Graph (EPG) formalism, which improves the accuracy and reproducibility of cartilage T2 mapping, as detailed in the work of Marco Barbieri, Anthony A. Gatti, and Feliks Kogan (2024) [https://doi.org/10.1002/jmri.29646](https://doi.org/10.1002/jmri.29646).

The graphical abstract of the work is reported below, showing that EPG modeling improved reproducibility in cartilage T2 in a cohort of healthy subjects from the OAI dataset.

![Dataset Structure](images/Graphical_abstract_rescaled.png)

## Dataset Structure

### Files and Folders

The dataset is organized by acquisition timepoints. Each main folder represents a timepoint in the OAI dataset and contains subfolders for individual subjects.

- **Timepoints**: `00m`, `12m`, `24m`, `36m`, `48m`, `72m`, `96m`.
- **Subject folders**: Each folder name is the unique OAI subject ID (e.g., `9000099`).

Within each subject folder:
- **`t2.nii.gz`**: The T2 map computed using the EPG dictionary fitting method, specific to cartilage tissue.
- **`r2.nii.gz`**: The r-squared value of the fit (goodness of fit).

### MESE Data Location Files

For each acquisition timepoint (e.g., `00_month_mese_locations.csv`, `12_month_mese_locations.csv`, etc), a CSV file provides a mapping to the original MESE data within the OAI dataset. Each CSV file includes the following columns:

- **subject_id**: The unique identifier for each OAI subject.
- **visit**: The month corresponding to the acquisition timepoint (e.g., 36 for `36m`).
- **laterality**: Indicates whether the MESE data is from the **RIGHT** or **LEFT** knee.
- **dicom_mese_path**: The relative path to the original DICOM MESE data within the OAI dataset.
- **t2map_nifti_path**: The relative path to the computed T2 map for that subject, located in this dataset.

These CSV files will help locate the original MESE DICOM data within the OAI dataset.

### Features

- **Subject ID** (str): Unique identifier for each subject in the OAI study.
- **T2 Map (`t2.nii.gz`)**: Computed T2 map for cartilage using the EPG fitting method.
- **R-Squared Map (`r2.nii.gz`)**: Fit accuracy metric for the T2 computation.

## Cartilage-Specific T2 Mapping

The T2 map in this dataset is provided **only for cartilage regions**, as the EPG model used in the computation is specifically designed for cartilage MR properties. To speed up computation, we have exploited segmented cartilage regions from the femoral, tibial, and patellar regions. Here’s the complete mapping process:

1. **Cartilage Segmentation**: For each subject, the femoral, tibial, and patellar cartilage were segmented from the corresponding Double Echo Steady State (DESS) image using the [ShapeMedKneeModel](https://huggingface.co/aagatti/ShapeMedKnee).
  
2. **Registration to MESE Images**: The segmented cartilage masks were then registered to the MESE images using [Elastix](https://github.com/SuperElastix/elastix), ensuring anatomical alignment across sequences.

3. **Dilated Mask for T2 Mapping**: A dilated version of the cartilage mask was used during the T2 mapping process to allow researchers the flexibility to apply their segmentations if desired. This ensures that cartilage boundaries are fully captured while also accounting for anatomical variations.

The cartilage segmentations used for the OAI dataset are available in the public repository [ShapeMedKnee](https://huggingface.co/aagatti/ShapeMedKnee) and will be regularly maintained and updated there.

## Dataset Creation

The T2 maps in this dataset were generated from the MESE data in the OAI dataset using the Extended Phase Graph (EPG) fitting method as described in the work by [Barbieri, Gatti, and Kogan, published in *Journal of Magnetic Resonance Imaging* (2024)](https://doi.org/10.1002/jmri.29646). 
The code used to perform this fitting is open-source and accessible on GitHub at [EPGfit_for_cartilage_T2_mapping](https://github.com/barma7/EPGfit_for_cartilage_T2_mapping).

## Getting Started

### Installation

You can install and access the dataset using the `datasets` library:

```bash
pip install datasets
```

### Usage

Load and interact with the dataset in Python:

```python
from datasets import load_dataset

dataset = load_dataset("barma7/oai-t2maps-epgfit")

# Accessing a specific timepoint and subject data
print(dataset["00m"]["9000099"]["t2"])
print(dataset["00m"]["9000099"]["r2"])
```

## Dataset Details

- **File Size**: Each T2 map file (`t2.nii.gz`) and r-squared file (`r2.nii.gz`) are stored in compressed `.nii.gz` format, with sizes varying per subject and time point.
- **Number of Samples**: Covers subjects across seven OAI acquisition timepoints for which MESE was available.
- **File Format**: `.nii.gz` files.

## License

This dataset is licensed under the MIT License, which allows for free use, modification, and distribution with attribution. For full license details, please see the LICENSE file in this repository.

---

## Acknowledgments

This dataset was created based on the Osteoarthritis Initiative (OAI) dataset. The authors of this repository acknowledge the original OAI study and the contributions of all OAI collaborators.

## Citation

If you use this dataset in your research, please cite:

Barbieri, M., Gatti, A.A. and Kogan, F. (2024), Improving Accuracy and Reproducibility of Cartilage T2 Mapping in the OAI Dataset Through Extended Phase Graph Modeling. J Magn Reson Imaging. https://doi.org/10.1002/jmri.29646