Chest X-Ray Images and Associated Text Reports

MIMIC-CXR contains over 100,000 chest X-ray images linked to radiology text reports. The dataset was created by the MIT Laboratory for Computational Physiology and is hosted on HuggingFace by itsanmolgupta, with a last update in December 2024. It supports multimodal analysis of medical images and their corresponding clinical interpretations.

Use Cases

• Train image classification models to detect pathologies like pneumothorax from chest X-ray images.
• Develop natural language processing models to generate or evaluate radiology report text from image features.
• Perform multimodal fusion to align visual findings in X-ray images with descriptive concepts in the associated reports.
• Benchmark vision-language models on the task of generating accurate medical reports from medical images.

Strengths

• Over 100,000 medical images provide a substantial base for model training.
• Multimodal pairing of images with text reports enables cross-domain analysis.

Limitations

• Specific row count, column details, and license information are not provided in the available metadata.
• The dataset's geographic and temporal coverage are unspecified, which may limit generalizability.

Provenance

Source

MIT Laboratory for Computational Physiology (inferred from MIMIC project).

Collection Method

Clinical data collection from hospital systems.

Freshness

Last updated on HuggingFace in December 2024.

Geography

United States (inferred from 'Regionus' platform tag).