Back to all papers

Predicting rapid progression of non-target lesions after PCI using machine learning combining coronary CTA plaque features and clinical inflammatory markers.

November 20, 2026pubmed logopapers

Authors

Chen Z

Affiliations (1)

  • Department of Cardiology, College of the Third Clinical Medical College, Jinzhou Medical University, Jinzhou, Liaoning Province, People's Republic of China.

Abstract

This study investigates the value of a machine learning method combining coronary computed tomography angiography (CTA plaque features and clinical inflammatory markers for predicting rapid progression of non-target lesions after Percutaneous Coronary Intervention (PCI). A total of 331 post-PCI patients were retrospectively enrolled and divided into rapid progression (n = 78) and non-rapid progression (n = 253) groups based on follow-up CTA. Baseline clinical data, inflammatory markers (hs-CRP, IL-6, neutrophil-to-lymphocyte ratio), and CTA plaque features (low-attenuation plaque, positive remodeling, noncalcified plaque volume) were collected. Patients were randomly split 7:3 into training and validation sets. Least Absolute Shrinkage and Selection Operator regression identified predictors. Logistic regression, inflammation-only, CTA-only, and multimodal fusion models were developed and compared. Least Absolute Shrinkage and Selection Operator regression selected 8 independent predictors: hs-CRP, IL-6, neutrophil-to-lymphocyte ratio, low-attenuation plaque, positive remodeling, noncalcified plaque volume, napkin-ring sign, and diabetes history. In the validation set, the multimodal fusion model achieved an area under the curve of 0.89 (95%CI: 0.83-0.95), outperforming the inflammation-only (0.78), CTA-only (0.83), and Logistic regression (0.72) models (all P < .05). Its accuracy, sensitivity, and specificity were 92.9%, 95.2%, and 92.3%, respectively. The fusion model showed excellent calibration (HL P = .58), the highest net benefit across threshold probabilities of 8% to 35%, and stable performance across subgroups (area under the curve 0.86-0.92). The multimodal fusion machine learning model effectively predicts rapid progression of non-target lesions after PCI, outperforming single-modality models and supporting risk stratification and individualized secondary prevention.

Topics

Machine LearningPercutaneous Coronary InterventionComputed Tomography AngiographyPlaque, AtheroscleroticCoronary Artery DiseaseJournal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAISlice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.