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

1

The study introduces an automated pipeline combining deep learning and radiomics to detect non-calcified coronary plaques using non-contrast CACS.
2

SegResNet models demonstrated optimal performance in coronary segmentation, enhancing the detection of non-calcified plaques.
3

Radiomics models achieved moderate to good diagnostic performance for non-calcified plaques, with AUCs ranging from 0.700 to 0.855.
4

Combined-region models generally outperformed or matched PCAT models in predicting non-calcified plaques across various datasets.
5

This automated approach may facilitate large-scale CAD screening by improving the detection of clinically significant non-calcified plaques.

Frontiers In Cardiovascular Medicine

by Wen Chen, Qing Tao, Can Chen, Su Hu, Feirong Yao, Jie Chen, Jinggang Zhang, Kejun Gu, Li Su, Wei Xing, Chunhong Hu
June 1, 2026

1