Time-series clustering of RV free-wall strain identified phenogroups with differing CV risk; cluster 5 had a 22% higher adjusted risk of cardiac events (HR 1.22, P=0.029).
Does time-series clustering of right ventricular free-wall strain features predict cardiac events in the general population?
Time-series clustering of right ventricular free-wall strain features identifies distinct phenogroups that predict long-term cardiac events in the general population.
Absolute Event Rate: 0% vs 0%
Abstract Aims – Current clinical practice evaluates right ventricular (RV) function primarily by assessing peak RV strain, mainly in symptomatic individuals. However, the potential of time-series RV strain recordings for cardiovascular (CV) risk assessment in the general population remains unexplored. Therefore, this study investigates whether time-series clustering of RV free-wall strain (RVFWS) can enhance CV risk stratification. Materials and Methods – We included 404 FLEMENGHO participants randomly recruited from the general population (mean age 56.0±15.2, 49.5% women). We measured echocardiographic and clinical parameters including the RVFWS and collected cardiac outcome 8.4 years later. We employed Gaussian Mixture Model on 16 features derived from basal, mid and apical RVFWS curves. These features included the slopes during systole, early and late diastole, the height of the diastasis, and the peak RVFWS, while the duration of diastasis was represented as a single feature. Results – Guided by a hybrid cluster validity index, we separated the dataset into 5 phenogroups. Clusters 1 and 2 showed the most favourable clinical profiles, characterized by significantly lower CV risk scores. On the other hand, clusters 3, 4, and 5 showed worse clinical profiles with increased CV risk. Moreover, Cluster 5 was associated with a significantly higher prevalence of left ventricular (LV) abnormalities, including diastolic dysfunction. The Kaplan-Meier curve and Cox regression showed that cluster 5 had a significantly increased risk of cardiac events compared to the average population risk (adjusted HR: 1.22; 95% CI: 1.02-1.47; P = 0.029). Conclusion – Using unsupervised learning on features extracted from RVFWS curves, we identified clinically meaningful phenogroups with additional prognostic information. The significant differences in the prevalence of LV abnormalities among clusters suggest that RV function may begin to deteriorate even during the subclinical stages of LV maladaptation.
Ntalianis et al. (Sat,) reported a other. Time-series clustering of RV free-wall strain identified phenogroups with differing CV risk; cluster 5 had a 22% higher adjusted risk of cardiac events (HR 1.22, P=0.029).