Phase entropy of gated SPECT-MPI for predicting major adverse cardiovascular events: incremental prognostic value beyond perfusion and function

Figures

• 

  Figure 1. Annualized incidence of major adverse cardiovascular events for entropy. The bar chart illustrates the annualized major adverse cardiovascular events incidence rates (%) across ten deciles of phase entropy (n = 1,674 total patients; n = 119–211 per decile). The orange line represents the mean entropy value for each decile. Annualized major adverse cardiovascular events rates were calculated as events per 100 person-years. Statistical comparisons across deciles were performed using the χ2 test for categorical variables. MACE, major adverse cardiovascular events; LVEF, left ventricular ejection fraction; TPD, total perfusion deficit.

• 

  Figure 2. Distribution of phase variables. Histograms showing the frequency distribution of (A) Bandwidth, (B) Phase SD, and (C) entropy in the study cohort (n = 1,674). Patients are stratified by the occurrence of MACE (No MACE: n = 1,012; MACE: n = 662). Descriptive statistics, including mean, standard deviation skewness, and excess kurtosis, are provided for each variable. Normality of the distributions was assessed using the Shapiro–Wilk test. SD, standard deviation; MACE, major adverse cardiovascular events.

• 

  Figure 3. Relationship between phase analysis parameters. Scatter plots showing the relationships between (A) Bandwidth and Phase SD, (B) Bandwidth and entropy, and (C) Phase SD and entropy. The red lines represent the linear regression fit. (n = 1,674). Statistical significance was assessed using Pearson correlation analysis. All correlations were statistically significant (P < 0.001). SD = standard deviation.

• 

  Figure 4. Correlation between entropy and TPD or LVEF. Scatter plots illustrating the associations between Entropy and (A) TPD and (B) LVEF. The black lines represent the linear regression fit. The color scale indicates the density of data points (log10 count). (n = 1,674). Statistical significance was assessed using Pearson correlation analysis. Both correlations were statistically significant (P < 0.001). LVEF, left ventricular ejection fraction; TPD, total perfusion deficit.

• 

  Figure 5. Adjusted hazard ratios for major adverse cardiovascular events of phase variables modeled with RCS. The HR (solid line) and 95%CI (shaded blue area) for major adverse cardiovascular events are plotted against increasing values of entropy (A), Bandwidth (B), and Phase SD (C) (n = 1,674). RCS analyses were performed to explore potential nonlinear relationships between each variable and the risk of major adverse cardiovascular events in Cox proportional hazards models. The overall and nonlinear associations were tested using likelihood ratio tests. HR and their 95% CI were estimated, with HR = 1.0 as the reference. MACE, major adverse cardiovascular events; RCS, restricted cubic splines; SD, standard deviation.

• 

  Figure 6. Kaplan–Meier curves for major adverse cardiovascular events by entropy, TPD and LVEF. Kaplan–Meier curves stratified by the groups based on Entropy (A), LVEF and Entropy (B), TPD and Entropy (C) are shown. (n = 1,674). Kaplan–Meier survival analysis was performed to estimate major adverse cardiovascular events free survival rates across groups. The log-rank test was used to compare survival distributions between strata. Patients were stratified by entropy tertiles (A) and combined risk categories (B, C) to evaluate incremental prognostic value. MACE, major adverse cardiovascular events; LVEF, left ventricular ejection fraction; TPD, total perfusion deficit.

• 

  Figure 7. Incremental value of phase variables for prediction of major adverse cardiovascular events. Model 0: TPD; Model 1: TPD + LVEF; Model 2: TPD + LVEF + Phase SD; Model 3: TPD + LVEF + Entropy; Model 4: TPD + LVEF + Bandwidth; Model 5: TPD + LVEF + Phase SD + Entropy + Bandwidth. The incremental prognostic value of phase entropy was assessed using nested Cox proportional hazards models. The global likelihood ratio chi-square test was used to compare model fit across specifications, with P < 0.05 considered statistically significant. (n = 1,674). MACE, major adverse cardiovascular events; SD, standard deviation; LVEF, left ventricular ejection fraction; TPD, total perfusion deficit.

• 

  Figure 8. NRI and IDI statistics for the addition of SPECT variables to major adverse cardiovascular events prediction. Time-dependent ROC curve analysis was performed to assess the incremental prognostic value of each variable. The C-index (concordance index), IDI, continuous NRI and median risk score improvement were calculated to quantify the improvement in predictive performance of nested models over baseline models. (n = 1,674). IDI, integrated discrimination improvement; NRI, net reclassification improvement; SD, standard deviation; LVEF, left ventricular ejection fraction; TPD, total perfusion deficit.