Epicardial adipose tissue radiomics predicts VR and MACE after AMI: a prospective cohort study

Figures

• 

  Figure 1. Study flow chart.

• 

  Figure 2. Predictive nomograms and variable importance for ventricular remodeling (VR) and major adverse cardiovascular events (MACE). (A) nomogram for predicting 6- and 12-month VR based on clinical and imaging predictors; (B) relative importance of each predictor in the VR model; (C) nomogram for predicting 6- and 12-month MACE based on clinical and imaging predictors; (D) relative importance of each predictor in the MACE model.

• 

  Figure 3. Selection of the epicardial adipose tissue radiomic features using LASSO regression. (A, C): LASSO coefficient profiles for VR (A) and MACE (C). As the penalty parameter (λ) increases, more coefficients shrink to zero, leaving only the most predictive features; (B, D): Optimal λ selection using 10-fold cross-validation for VR (B) and MACE (D). The vertical dotted line represents the λ value with the minimum binomial deviance, corresponding to the optimal number of non-zero coefficients selected for the predictive model.

• 

  Figure 4. Shapley additive explanations (SHAP) analysis of the top 5 radiomics feature variables for ventricular remodeling (VR) and major adverse cardiovascular events (MACE). (A, C): SHAP bar plots displaying the mean absolute SHAP values, indicating the average impact of each radiomics feature on the predictive model for VR (A) and MACE (C). (B, D): SHAP beeswarm plots demonstrating the distribution of SHAP values for individual feature contributions in VR (B) and MACE (D).

• 

  Figure 5. Nomograms and variable importance for predicting ventricular remodeling (VR) and major adverse cardiovascular events (MACE) using radiomics-based models. (A) nomogram for predicting VR based on EAT radiomics features combined with clinical parameters; (B) relative contribution of each predictor in the VR model; (C) nomogram for predicting MACE based on EAT radiomics features combined with clinical parameters; (D) relative contribution of each predictor in the MACE model.

• 

  Figure 6. ROC curves comparing the predictive performance of model 1 and model 2 for ventricular remodeling (VR) and major adverse cardiovascular events (MACE) at different follow-up periods. (A–D) ROC curves for VR prediction at 3 months (A), 6 months (B), 9 months (C), and 12 months (D); (E–H) ROC curves for MACE prediction at 3 months (E), 6 months (F), 9 months (G), and 12 months (H).

• 

  Figure 7. Performance evaluation of model 1 and model 2 for predicting ventricular remodeling (VR) and major adverse cardiovascular events (MACE). (A, D) time-dependent concordance index (C-index) comparisons among model 1, model 2, and the GRACE score for VR (A) and MACE (D) prediction, demonstrating the superior discriminatory ability of model 2 over time. (B, E) calibration curve for model 2 predicting VR (B) and MACE (E). (C, F) decision curve analysis to compare model 1, model 2, and the GRACE score for predicting VR (C) and MACE (F).

• 

  Figure 8. Internal validation of model 2 for predicting VR and MACE. (A) ROC curve of model 2 for VR prediction; (B) time-dependent C-index of model 2 for VR; (C) Calibration curve for VR; (D) DCA for VR; (E) ROC curve of model 2 for MACE prediction; (F) time-dependent C-index of model 2 for MACE; (G) calibration curve for MACE; (H) DCA for MACE.

• 

  Figure 9. EAT exacerbates cardiac injury and dysfunction in HFD + ISO-treated rats. (A–E). Echocardiographic analysis of heart rate, LVEDD, LVESD, EF, and FS. (F) EAT weight. (G) Heart weight. (H) Heart weight-to-body weight ratio (HW/BW). (I–K). Serum levels of lactate dehydrogenase (LDH), cardiac troponin I (cTnI), and creatine kinase-MB (CK-MB). Data are presented as mean ± SD (n=9). Statistical analysis was performed using one-way ANOVA with Bonferroni post hoc test. *P < 0.05, **P < 0.01 vs. ND group; #P < 0.05, ##P < 0.01 vs. HFD group.

• 

  Figure 10. EAT aggravates myocardial injury and remodeling in HFD+ISO-treated rats. (A) HE staining showing myocardial morphology across ND, ND+ISO, HFD, and HFD+ISO groups. (B) Masson staining for interstitial collagen deposition with quantitative analysis of fibrosis area. (C) WGA staining showing cardiomyocyte membranes (red) and nuclei (DAPI, blue); scale bars = 50 μm. Data are presented as mean ± SD (n=9). Statistical analysis was performed using one-way ANOVA with Bonferroni post hoc test. **P < 0.01 vs. ND group; ##P < 0.001 vs. HFD group.

• 

  Figure 11. EAT enhances H9c2 cardiomyocyte injury and apoptosis under H/R conditions. (A). TUNEL staining (green) showing apoptotic nuclei, with DAPI counterstaining (blue); Scale bars = 50 μm. (B–D) Levels of LDH, cTnI, and CK-MB levels in the culture supernatants. (E–G) Relative mRNA expression of ANP, BNP, and β-MHC detected by qRT-PCR. Data are presented as mean ± SD (n=3). Statistical analysis was performed using one-way ANOVA with Bonferroni post hoc test. **P < 0.01 vs. CON group; ##P < 0.001 vs. H/R group.