MedMNIST

MedMNIST是由上海交通大学开发的包含10个预处理医学图像数据集的集合，旨在教育、标准化、多样化和轻量化，适用于医学图像分析中的快速原型设计和多模态机器学习/AutoML基准。数据集涵盖多种医学图像模态和任务，数据规模从100到100,000不等，支持二分类/多类、序数回归和多标签任务。创建过程涉及从多个开放医学图像数据集中选择数据，进行预处理和标准化，以适应28×28图像的分类任务。该数据集广泛应用于医学图像分析的教育、研究和AutoML算法的基准测试。

MedMNIST is a collection of 10 preprocessed medical image datasets developed by Shanghai Jiao Tong University. It is designed for education, standardization, diversification and lightweight utility, and is suitable for rapid prototyping in medical image analysis as well as serving as benchmarks for multimodal machine learning and AutoML. The dataset covers diverse medical image modalities and tasks, with data volumes ranging from 100 to 100,000, supporting binary/multi-class classification, ordinal regression and multi-label tasks. Its creation involves selecting data from multiple open-access medical image datasets, followed by preprocessing and standardization to adapt to 28×28 image classification tasks. This dataset is widely applied in education, research of medical image analysis and benchmark testing of AutoML algorithms.

搜集汇总

背景与挑战

背景概述

MedMNIST是一个包含18个标准化生物医学图像分类数据集的集合，涵盖2D和3D图像，提供多种尺寸选项（28x28到224x224）。数据集设计轻量、多样化且用户友好，适合机器学习和教育用途，但不适用于临床。