MedMNIST v2 -- A large-scale lightweight benchmark for 2D and 3D biomedical image classification

TL;DR

MedMNIST v2 provides a comprehensive, standardized collection of lightweight biomedical image datasets for 2D and 3D classification tasks, facilitating research and education in biomedical imaging and machine learning.

Contribution

It introduces a large-scale, pre-processed biomedical image dataset collection with diverse modalities and tasks, along with benchmark results for various models.

Findings

Benchmarking shows competitive performance of baseline models.

Datasets support various classification tasks and scales.

Open-source code and data promote further research.

Abstract

We introduce MedMNIST v2, a large-scale MNIST-like dataset collection of standardized biomedical images, including 12 datasets for 2D and 6 datasets for 3D. All images are pre-processed into a small size of 28x28 (2D) or 28x28x28 (3D) with the corresponding classification labels so that no background knowledge is required for users. Covering primary data modalities in biomedical images, MedMNIST v2 is designed to perform classification on lightweight 2D and 3D images with various dataset scales (from 100 to 100,000) and diverse tasks (binary/multi-class, ordinal regression, and multi-label). The resulting dataset, consisting of 708,069 2D images and 10,214 3D images in total, could support numerous research / educational purposes in biomedical image analysis, computer vision, and machine learning. We benchmark several baseline methods on MedMNIST v2, including 2D / 3D neural networks and open-source / commercial AutoML tools. The data and code are publicly available at https://medmnist.com/.