Pericoronary Radiomics Signature for Non-Culprit Lesion Progression and Revascularization Decision in NSTE-ACS

Figures

• 

  Figure 1. Study Flow Diagram.

• 

  Figure 2. Discriminative performance and calibration of the predictive models for 4-year NCL-related MACE. (A,B) ROC curves comparing the clinical model, the isolated Rad model, and the combined clinic-radiomics model in the training (A) and internal validation (B) cohorts. The combined model consistently yields the highest AUC. (C,D) Calibration curves assessing the agreement between the combined model-predicted probabilities and the actual observed NCL-MACE frequencies in the training (C) and validation (D) cohorts. The diagonal dashed gray line denotes the ideal reference of perfect calibration. The solid red line represents the bias-corrected performance derived from 1000 bootstrap resamples. The Hosmer–Lemeshow goodness-of-fit test yielded p > 0.05 for both cohorts, indicating robust calibration. Abbreviations: AUC, area under the receiver operating characteristic curve.

• 

  Figure 3. Risk Stratification and Clinical Utility of the PCAT Radiomics Signature (A,B) Kaplan–Meier curves in training (A) and validation (B) cohort for cumulative incidence of NCL-MACE stratified by the optimal Rad-score cut-off value, determined using maximally selected rank statistics. Patients with a high Rad model exhibited significantly worse outcomes (log-rank p < 0.01). (C,D) Decision curve analysis (DCA) evaluates the net clinical benefit of the predictive models in the training (C) and validation (D) cohorts. The horizontal solid black line assumes no patients experience NCL-MACE (“Treat-none”), while the descending solid gray line assumes all patients experience NCL-MACE (“Treat-all”). The combined model (red line) provides a superior net clinical benefit across a broad spectrum of threshold probabilities relative to the clinical baseline model (blue line).

• 

  Figure 4. Association Between Baseline PCAT Radiomics Signature and Subsequent Plaque Progression. Analysis performed in the plaque progression subgroup. (A) Box plots comparing ΔNCPV/year between patients with high and low Rad model (dichotomized by the median). ** indicates p < 0.01. (B) Forest plot showing standardized regression coefficients (β) from linear regression models. Model 1: unadjusted; Model 2, adjusted for age and sex; Model 3, Model 2 plus adjusted for baseline NCPV% and clinical risk factors.