MedMNIST Classification Decathlon: A Lightweight AutoML Benchmark for Medical Image Analysis


Shanghai Jiao Tong University, Shanghai, China

Abstract

We present MedMNIST, a collection of 10 pre-processed medical open datasets. MedMNIST is standardized to perform classification tasks on lightweight 28 * 28 images, which requires no background knowledge. Covering the primary data modalities in medical image analysis, it is diverse on data scale (from 100 to 100,000) and tasks (binary/multi-class, ordinal regression and multi-label). MedMNIST could be used for educational purpose, rapid prototyping, multi-modal machine learning or AutoML in medical image analysis. Moreover, MedMNIST Classification Decathlon is designed to benchmark AutoML algorithms on all 10 datasets; We have compared several baseline methods, including open-source or commercial AutoML tools.

Key Features

  • Educational: Our multi-modal data, from multiple open medical image datasets with Creative Commons (CC) Licenses, is easy to use for educational purpose.
  • Standardized: Data is pre-processed into same format, which requires no background knowledge for users.
  • Diverse: The multi-modal datasets covers diverse data scales (from 100 to 100,000) and tasks (binary/multiclass, ordinal regression and multi-label).
  • Lightweight: The small size of 28 × 28 is friendly for rapid prototyping and experimenting multi-modal machine learning and AutoML algorithms.

Please note that this dataset is NOT intended for clinical use.

Materials

An Overview of MedMNIST Dataset
Name Data Modality Tasks (# Classes/Labels) # Training # Validation # Test
PathMNIST Pathology Multi-Class (9) 89,996 10,004 7,180
ChestMNIST Chest X-ray Multi-Label (14) Binary-Class (2) 78,468 11,219 22,433
DermaMNIST Dermatoscope Multi-Class (7) 7,007 1,003 2,005
OCTMNIST OCT Multi-Class (4) 97,477 10,832 1,000
PneumoniaMNIST Chest X-ray Binary-Class (2) 4,708 524 624
RetinaMNIST Fundus Camera Ordinal Regression (5) 1,080 120 400
BreastMNIST Breast Ultrasound Binary-Class (2) 546 78 156
OrganMNIST_Axial Abdominal CT Multi-Class (11) 34,581 6,491 17,778
OragnMNIST_Coronal Abdominal CT Multi-Class (11) 13,000 2,392 8,268
OrganMNIST_Sagittal Abdominal CT Multi-Class (11) 13,940 2,452 8,829

Benchmarking

Methods PathMNIST ChestMNIST DermaMNIST OCTMNIST PneumoniaMNIST
AUC ACC AUC ACC AUC ACC AUC ACC AUC ACC
ResNet-18 (28) 0.972 0.844 0.706 0.947 0.899 0.721 0.951 0.758 0.957 0.843
ResNet-18 (224) 0.978 0.860 0.713 0.948 0.896 0.727 0.960 0.752 0.970 0.861
ResNet-50 (28) 0.979 0.864 0.692 0.947 0.886 0.710 0.939 0.745 0.949 0.857
ResNet-50 (224) 0.978 0.848 0.706 0.947 0.895 0.719 0.951 0.750 0.968 0.896
auto-sklearn 0.500 0.186 0.647 0.642 0.906 0.734 0.883 0.595 0.947 0.865
AutoKeras 0.979 0.864 0.715 0.939 0.921 0.756 0.956 0.736 0.970 0.918
Google AutoML Vision 0.982 0.811 0.718 0.947 0.925 0.766 0.965 0.732 0.993 0.941
Benchmarking Performance on MedMNIST Dataset
Methods RetinaMNIST BreastMNIST OrganMNIST (Axial) OrganMNIST (Coronal) OrganMNIST (Sagittal)
AUC ACC AUC ACC AUC ACC AUC ACC AUC ACC
ResNet-18 (28) 0.727 0.515 0.897 0.859 0.995 0.921 0.990 0.889 0.967 0.762
ResNet-18 (224) 0.721 0.543 0.915 0.878 0.997 0.931 0.991 0.907 0.974 0.777
ResNet-50 (28) 0.719 0.490 0.879 0.853 0.995 0.916 0.990 0.893 0.968 0.746
ResNet-50 (224) 0.717 0.555 0.863 0.833 0.997 0.931 0.992 0.898 0.970 0.770
auto-sklearn 0.694 0.525 0.848 0.808 0.797 0.563 0.898 0.676 0.855 0.601
AutoKeras 0.655 0.420 0.833 0.801 0.996 0.929 0.992 0.915 0.972 0.803
Google AutoML Vision 0.762 0.530 0.932 0.865 0.988 0.818 0.986 0.861 0.964 0.706

Citation and Licenses

If you find this project useful, please cite our paper as:

Jiancheng Yang, Rui Shi, Bingbing Ni. "MedMNIST Classification Decathlon: A Lightweight AutoML Benchmark for Medical Image Analysis," IEEE 18th International Symposium on Biomedical Imaging (ISBI), 2021.

or using bibtex:

@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}
}

Besides, please cite the corresponding paper if you use any subset of MedMNIST. Each subset uses the same license as that of the source dataset.

PathMNIST
License: CC BY 4.0
Jakob Nikolas Kather, Johannes Krisam, et al., "Predicting survival from colorectal cancer histology slides using deep learning: A retrospective multicenter study," PLOS Medicine, vol. 16, no. 1, pp. 1–22, 01 2019.
ChestMNIST
License: CC0 1.0
Xiaosong Wang, Yifan Peng, et al., "Chestx-ray8: Hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases," in CVPR, 2017, pp. 3462–3471.
DermaMNIST
License: CC BY-NC 4.0
Philipp Tschandl, Cliff Rosendahl, and Harald Kittler, "The ham10000 dataset, a large collection of multisource dermatoscopic images of common pigmented skin lesions," Scientific data, vol. 5, pp. 180161, 2018.
Noel Codella, Veronica Rotemberg, et al.: “Skin Lesion Analysis Toward Melanoma Detection 2018: A Challenge Hosted by the International Skin Imaging Collaboration (ISIC)”, 2018; arXiv:1902.03368.
OCTMNIST/PneumoniaMNIST
License: CC BY 4.0
Daniel S. Kermany, Michael Goldbaum, et al., "Identifying medical diagnoses and treatable diseases by image-based deep learning," Cell, vol. 172, no. 5, pp. 1122 – 1131.e9, 2018.
RetinaMNIST
License: CC BY 4.0
DeepDR Diabetic Retinopathy Image Dataset (DeepDRiD), "The 2nd diabetic retinopathy – grading and image quality estimation challenge," https://isbi.deepdr.org/data.html, 2020.
BreastMNIST
License: CC BY 4.0
Walid Al-Dhabyani, Mohammed Gomaa, Hussien Khaled, and Aly Fahmy, "Dataset of breast ultrasound images," Data in Brief, vol. 28, pp. 104863, 2020.
OrganMNIST_{Axial,Coronal,Sagittal}
License: CC BY 4.0
Patrick Bilic, Patrick Ferdinand Christ, et al., "The liver tumor segmentation benchmark (lits)," arXiv preprint arXiv:1901.04056, 2019.
Xuanang Xu, Fugen Zhou, et al., "Efficient multiple organ localization in ct image using 3d region proposal network," IEEE Transactions on Medical Imaging, vol. 38, no. 8, pp. 1885–1898, 2019.

Copyright © 2020-2021 MedMNIST Team

This website is hosted on GitHub. Updated on July, 2021.