# Enhanced MedMNIST Dataset

MedMNIST is a comprehensive collection of standardized biomedical images designed for various analytical tasks in the medical field. This dataset has been expanded to include three new subsets, broadening the range of imaging modalities and classification challenges available to researchers. These additions complement the existing MedMNIST collections, offering a more diverse set of resources for developing and evaluating machine learning models across various medical imaging applications.

## Overview of 



These datasets are integrated into the MedMNIST collection, which includes both 2D and 3D biomedical images across various modalities and tasks.

## MedMNIST2D Datasets and Statistics

MedMNIST2D comprises twelve pre-processed 2D datasets from selected sources, covering primary data modalities such as X-Ray, OCT, Ultrasound, and more. Each dataset is tailored for specific classification tasks, including binary and multi-class classifications, with varying numbers of samples.

| **Dataset Name**   | **Data Modality**       | **Task (Number of Classes)** | **Number of Samples** |
|--------------------|-------------------------|------------------------------|-----------------------|
|_2D Datasets_ | | |
| BreastMNIST        | Ultrasound       | Binary-Class (2)             | 780                   |
| OrganAMNIST        | Abdominal CT            | Multi-Class (11)             | 58,830                |
| OrganCMNIST        | Abdominal CT            | Multi-Class (11)             | 23,583                |
| OrganSMNIST        | Abdominal CT            | Multi-Class (11)             | 25,211                |
|_Additional 2D Datasets_ | | |
| Brain Tumor Dataset  | Magnetic Resonance      | Multi-Class (3)              | 3,064                  |
| Brain Dataset        | Magnetic Resonance      | Multi-Class (23)             | 1,600                  |
| Breast Cancer        | Ultrasound              | Binary-Class (2)             | 1,875                  |
|_Out-of-Distribution_ | | |
| BloodMNIST         | Blood Cell Microscope   | Multi-Class (8)              | 17,092                |
| PneumoniaMNIST     | Chest X-Ray             | Binary-Class (2)             | 5,856                 |
| DermaMNIST         | Dermatoscope            | Multi-Class (7)              | 10,015                |

These datasets are pre-processed into a standardized format, facilitating ease of use for machine learning applications.

## MedMNIST3D Datasets and Statistics

MedMNIST3D comprises six pre-processed 3D datasets from selected sources, covering primary data modalities such as CT and MRI scans. Each dataset is tailored for specific classification tasks, including binary and multi-class classifications, with varying numbers of samples.

| **Dataset Name**   | **Data Modality**       | **Task (Number of Classes)** | **Number of Samples** |
|--------------------|-------------------------|------------------------------|-----------------------|
| OrganMNIST3D       | Abdominal CT            | Multi-Class (11)             | 1,742                 |
| NoduleMNIST3D      | Chest CT                | Binary-Class (2)             | 1,633                 |
| AdrenalMNIST3D     | Abdominal CT            | Binary-Class (2)             | 1,584                 |
| FractureMNIST3D    | Chest CT                | Multi-Class (3)              | 1,370                 |
|_Out-of-Distribution_ | | |
| VesselMNIST3D      | Brain MRA               | Binary-Class (2)             | 1,908                 |
| SynapseMNIST3D     | Electron Microscope     | Binary-Class (2)             | 1,759                 |

These datasets are pre-processed into a standardized format, facilitating ease of use for machine learning applications.

## Accessing the Dataset

The enhanced MedMNIST collection, including both existing and new datasets, is accessible on Hugging Face. Researchers and practitioners can utilize this resource to develop and evaluate machine learning models across various medical imaging applications.

## Citation

If you find this dataset useful, please cite the following papers:

```
@inproceedings{yavuz2025policy,
  title={Policy Gradient-Driven Noise Mask},
  author={Yavuz, Mehmet Can and Yang, Yang},
  booktitle={International Conference on Pattern Recognition},
  pages={414--431},
  year={2025},
  organization={Springer}
}

@article{medmnistv2,
  title={MedMNIST v2-A large-scale lightweight benchmark for 2D and 3D biomedical image classification},
  author={Yang, Jiancheng and Shi, Rui and Wei, Donglai and Liu, Zequan and Zhao, Lin and Ke, Bilian and Pfister, Hanspeter and Ni, Bingbing},
  journal={Scientific Data},
  volume={10},
  number={1},
  pages={41},
  year={2023},
  publisher={Nature Publishing Group UK London}
}

@inproceedings{medmnistv1,
  title={MedMNIST Classification Decathlon: A Lightweight AutoML Benchmark for Medical Image Analysis},
  author={Yang, Jiancheng and Shi, Rui and Ni, Bingbing},
  booktitle={IEEE 18th International Symposium on Biomedical Imaging (ISBI)},
  pages={191--195},
  year={2021}
}
```

Please also cite the corresponding papers of the source data if you use any subset of MedMNIST, as per the description on the project website.

## License

This project is licensed under the MIT License. Each subset retains the same license as that of the source dataset. Please refer to the project website for more details.

*Note: This dataset is NOT intended for clinical use.*