MedImg: An Integrated Database for Public Medical Images

Area of Science:

• Medical imaging
• Artificial intelligence
• Computer vision

Background:

• Deep learning shows promise in medical image analysis, often matching or exceeding human performance.
• Clinical translation of these AI models is hindered by the lack of large, well-characterized datasets for validation.

Purpose of the Study:

• To address the challenge of limited data availability for deep learning in medical imaging.
• To create a centralized, accessible resource for medical image datasets.

Main Methods:

• Curated 105 diverse medical image datasets from public sources.
• Compiled a total of 1,995,671 images spanning 14 modalities and 13 organs.
• Developed the MedImg online database to organize and provide open access to these curated datasets.

Main Results:

• Established MedImg, an open-access online platform (https://www.cuilab.cn/medimg/).
• The database contains 1,995,671 images from 105 datasets, covering 14 medical imaging modalities.
• The dataset includes images from 13 different organs, facilitating broad research applications.

Conclusions:

• MedImg provides a valuable, systematically organized resource to overcome data limitations in deep learning for medical image analysis.
• The platform aims to accelerate research and development of AI-driven diagnostic tools by enhancing data accessibility.