Non-Contrast CT Images for Automated Detection of Non-Calcified Coronary Plaques

An automated pipeline combining deep learning and radiomics for detecting non-calcified plaques in the left anterior descending and right coronary arteries. The dataset, shared by Wen Chen under a CC-BY-4.0 license, includes non-contrast calcium score CT images from patients at two medical sites. Radiomics models built from the data demonstrated vessel-level diagnostic performance with AUCs ranging from 0.700 to 0.855.

Use Cases

• Training deep learning models for coronary artery segmentation based on manually annotated CACS images.
• Building radiomics models for plaque detection using features from coronary artery, pericoronary adipose tissue, and combined regions of interest.
• Validating automated pipelines for large-scale CAD screening using non-contrast chest CT scans.

Strengths

• Models achieved AUCs ranging from 0.700 to 0.855 across datasets, indicating moderate to good diagnostic performance.
• The pipeline was validated on data from two separate medical sites, suggesting multi-site applicability.
• The dataset is shared under a permissive CC-BY-4.0 license, facilitating reuse and research.

Limitations

• The dataset description does not specify the number of images, patients, or rows, limiting scale assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The primary file format is DOCX, which may not be the standard format for image data and could require extraction.

Provenance

Source

figshare, author Wen Chen.

Collection Method

Retrospectively enrolled patients undergoing coronary CT angiography for suspected CAD from two medical sites.

Freshness

Last updated 2026-06-01 04:20:37; freshness should be verified.