PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
June 1, 2026Frontiers in Cardiovascular Medicine0 citationsOpen Access

Automated deep learning–radiomics pipeline for non-calcified coronary plaque detection using non-contrast calcium score CT

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Does an automated deep learning-radiomics pipeline using non-contrast calcium score CT accurately detect non-calcified coronary plaques in patients with suspected CAD?

Population

Patients undergoing coronary CT angiography for suspected CAD from two medical sites

Comparison

Lesion vs control groups and radiomics models based on coronary artery, PCAT, or combined ROIs

Design

Retrospective two-center diagnostic study

Key result

An automated deep learning-radiomics pipeline using non-contrast calcium score CT detected non-calcified coronary plaques with moderate to good diagnostic performance, achieving AUCs ranging from 0.700 to 0.855 across datasets.

Authors

WCWen ChenQTQing TaoCCC Chen

Discussion

Loading...

Member takes

Overview

May support AI development for non-calcified plaque detection on calcium scoring CT; leaves open prospective validation before clinical use.

Study Design

Type

Observational (n=1,745)

Multicenter

Yes

Structured PICO

Does an automated deep learning-radiomics pipeline using non-contrast calcium score CT accurately detect non-calcified coronary plaques in patients with suspected CAD?

P
Population
1,745 patients with suspected coronary artery disease undergoing coronary CT angiography across two centers in China, retrospectively analyzed to develop and validate a deep learning-radiomics pipeline.
E
Exposure
Automated pipeline combining deep learning (SegResNet) for coronary segmentation and radiomics for detection of non-calcified plaques using non-contrast coronary artery calcium score (CACS) CT.
C
Comparator
Patients with no significant abnormalities on CCTA (control group), and comparison of different radiomics models (coronary artery, PCAT, and combined regions).
O
Outcome
Diagnostic performance for predicting non-calcified plaques, measured by area under the curve (AUC).surrogate

An automated deep learning and radiomics pipeline using non-contrast calcium score CT can effectively detect non-calcified coronary plaques, offering a potential tool for large-scale CAD screening.

Main Result

Effect estimate: AUC 0.700-0.855

Limitations

  • Lack of detailed clinical factors such as BMI, hypertension, diabetes, and smoking.
  • Exclusion of the left circumflex artery (LCX) due to variable anatomy and small caliber.
  • Use of ECG-gated CACS instead of routine chest CT scans.
  • Focus on vessel-level analysis rather than patient-level prediction.
  • Use of identical equipment and protocols across centers, limiting generalizability to other scanners.

Cite This Study

Chen et al. (2026) conducted an observational in Suspected coronary artery disease (n=1,745). Automated deep learning-radiomics pipeline using non-contrast CACS vs. Coronary CT angiography (reference standard) was evaluated on Detection of non-calcified coronary plaques (AUC 0.700-0.855). An automated deep learning-radiomics pipeline using non-contrast calcium score CT detected non-calcified coronary plaques with moderate to good diagnostic performance, achieving AUCs ranging from 0.700 to 0.855 across datasets.

synapsesocial.com/papers/6a346259096a8bf9ee71b790https://doi.org/10.3389/fcvm.2026.1794024
View Full Paper
Ask AI
Bookmark
Share