Mechanical signatures of left ventricular ejection fraction and mass index in a community cohort (ACE 1950)

This study investigated how cardiovascular magnetic resonance feature tracking (CMR-FT)-derived myocardial mechanics, beyond global longitudinal strain (GLS), relate to left ventricular ejection fraction (LVEF) and left ventricular mass index (LVMI). In 186 community-dwelling individuals, global circumferential and longitudinal CMR-FT metrics were extracted, including strain, strain rates, displacement, velocities, time-to-peak measures, mechanical dispersion. Univariate associations with LVEF and LVMI were assessed using Spearman correlation. Multivariable associations between a broad set of mechanical metrics and LVEF and LVMI were evaluated using supervised machine-learning models interpreted with Shapley Additive Explanations (SHAP). Among 17 mechanical metrics, the strongest correlated with LVEF were global circumferential strain (GCS, ρ = -0.71), longitudinal systolic strain rate (sSRlong, ρ = -0.53), circumferential diastolic strain rate (dSRcirc, ρ = 0.50), and GLS (ρ = -0.48) (all p < 0.001). In multivariable models, GCS and sSRlong remained most influential for LVEF, with GLS dropping to least influential (SHAP rank #17/17). For LVMI, longitudinal diastolic strain rate (dSRlong) showed the strongest univariate association (ρ = -0.50), followed by GLS (ρ = 0.41) and dSRcirc (ρ = -0.40) (all p < 0.001). SHAP ranked dSRlong and dSRcirc highest, followed by GLS. These findings indicate that feature importance in multivariable SHAP analyses differed from univariate correlation rankings, suggesting that metrics substantially correlated with LVEF or LVMI may contribute less information once myocardial mechanics are considered jointly. LVEF was most strongly associated with GCS and peak systolic strain rate, while LVMI with peak diastolic strain rates and GLS. Future work should test whether integrated mechanical profiles offer added prognostic and diagnostic value beyond single-metric biomarkers.