NIH Chest X-rays Bbox version

Home » Dataset Download » NIH Chest X-rays Bbox version

NIH Chest X-rays Bbox version

Datasets

File

NIH Chest X-rays Bbox version

Use Case

NIH Chest X-rays Bbox version

Description

Explore the NIH Chest X-ray Dataset, featuring 112,120 annotated X-ray images from 30,805 patients. Leveraging Natural Language Processing for label accuracy, this dataset supports advanced research in medical imaging and weakly-supervised learning, overcoming challenges in chest X-ray diagnosis.

Description:

The NIH Chest X-ray Dataset is a significant resource for medical imaging research, addressing the challenges of diagnosing chest X-rays, which can be more complex than chest CT scans. Previously, Openi held the largest publicly available collection of chest X-ray images with 4,143 images. However, this NIH dataset expands the field dramatically, offering 112,120 X-ray images from 30,805 unique patients.

Download Dataset

To generate the disease labels for this dataset, researchers utilized Natural Language Processing to extract disease classifications from the corresponding radiological reports. This process ensures that the labels are over 90% accurate, making them highly suitable for weakly-supervised learning. While the original radiology reports are not publicly available, you can find detailed information about the labeling process in the Open Access paper: “ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases” by Wang et al.