"Unsupervised Phenotypic Clustering of Cardiac MRI Data Reveals Distinct Subgroups and Predicts Outcomes in Ischemic Cardiomyopathy"

<b>Background: </b>Ischemic cardiomyopathy (ICM) exhibits significant heterogeneity in clinical outcomes, challenging conventional risk stratification methods. Cardiac magnetic resonance (CMR) imaging provides detailed insights into myocardial structure and function, yet integrating these multidimensional data remains complex. Unsupervised machine learning offers a promising approach to identify distinct phenotypic subgroups and improve prognostic accuracy.<br/><br/><b>Methods:</b> This study included 319 clinically stable ICM patients from the UHSM CMR Study. CMR-derived variables, including left ventricular ejection fraction (LVEF), ventricular volumes, and myocardial scar burden, were analysed using the KAMILA clustering algorithm, which handles mixed continuous and categorical data. The optimal number of clusters was determined through silhouette analysis, within-cluster sum of squares, and gap statistics. Principal Component Analysis (PCA) visualized the clustering results, and prognostic value was assessed using Cox regression and Kaplan-Meier survival analysis. SHAP (SHapley Additive exPlanations) values were used to evaluate feature importance.<br/><br/><b>Results:</b> Two distinct phenotypic clusters were identified. Cluster 1 (n=219) exhibited better cardiac function, with higher LVEF, smaller ventricular volumes, and lower scar burden. Cluster 2 (n=100) showed advanced disease, with lower LVEF, larger volumes, higher scar burden, and greater midwall fibrosis. PCA confirmed clear separation between clusters, explaining 62.6% of the variance. Patients in Cluster 2 had a significantly higher risk of adverse outcomes (HR = 3.96, p &lt; 0.0001). SHAP analysis highlighted ischemic scar burden, sphericity index, and midwall fibrosis as key predictors of outcomes.<br/><br/><b>Conclusion: </b>Unsupervised clustering of CMR-derived variables identified distinct ICM phenotypes with significant prognostic implications. This approach enhances risk stratification and may guide personalized treatment strategies, underscoring the potential of machine learning in understanding ICM heterogeneity. Further validation in multicentre cohorts is warranted.<br/>