Clustering analysis of left atrial strain for risk assessment in hypertrophic cardiomyopathy

Abstract Background Although left atrial (LA) strain, diameter and volume are shown to be predictors of adverse events in patients with hypertrophic cardiomyopathy (HCM), the LA strain curves morphology have not been fully exploited to characterise the risk of major adverse cardiovascular events (MACE). Aim : To investigate the relation between LA strain features and MACE incidence in HCM patients. Methods Three hundred and nineteen HCM Patients were included and followed longitudinally (mean duration 6 years). Clinical, electrocardiographic, and echocardiographic data were collected. LA longitudinal strain signals were also analyzed with custom-made algorithms to extract LA strain morphological features associated with reservoir, conduit and pump phases. Slope, intercepts, and curves integrals were calculated for each phase. After filtering correlated variables using a cut-off Pearsons’ coefficient>0.8, eighteen LA strains features were selected and analyzed using a clustering approach (k-means clustering). The primary endpoint was MACE, including supraventricular and ventricular arrhythmias, heart failure hospitalization, embolic stroke and death. Results The population had a mean age of 56±14 years. 55% (177 patients) had MACE within the 6-year follow-up. Three clusters were identified with 102 patients (32%) in cluster 1, 131 patients (41%) in cluster 2, and 86 patients (27%) in cluster 3. MACE incidence was 64% (65/102) in cluster 1, 56% (73/131) in cluster 2, and 45% (39/86) in cluster 3. The lower risk cluster (C3) is associated with modifications of LA strain slopes: significantly higher in the reservoir phase, lower in conduit and pumps phases. Patients from cluster 2 have a higher risk of atrial arrythmia (28%, 37/131) and are associated with reduced integral values and increased slope of the pump phase. Conclusion Three clusters are described based only on LA-strain patterns, defining groups associated with different risk levels. Our results underscore the importance of LA function for risk stratification in HCM patients. Automatic quantitative analysis of LA strain curves appears as a promising tool to improve understanding of LA mechanics and patient characterization in HCM.Features extraction pipeline K-means clusters and incidence of events