Equivalence of feature-tracking-derived myocardial strain across spatial resolution and compressed sensing acceleration.

Reference values for myocardial strain as measured by feature-tracking (FT) MRI cine images are known. As these values may be impacted by source image resolution, we aimed at assessing strain parameters across low-resolution single-shot and high-resolution segmented cine images using different FT software. Thirty-seven patients (median age 40 years [IQR: 26-52]; 13 female) with same-session segmented high-resolution (SEG) and compressed sensing-accelerated single-shot (CS) cines from a 1.5T system (Magnetom Sola, Siemens Healthineers) were retrospectively included. Commercial (CVI42, Circle) and prototype software (TrufiStrain, Siemens Healthineers), FT and DRA, respectively, were used for analysis. Peak circumferential and radial strain were calculated globally and segmentally using individual segment peaks (peak-segment) or segment strain at the global peak frame (peak-frame). Equivalence was tested with the two one-sided test (TOST) for equivalence (90% CI), and Spearman correlation (ρ) with left ventricular ejection fraction (LVEF) was calculated. A significance level of α = 0.05 was used. Strain values differed significantly between SEG and CS images for most combinations, except peak-segment global circumferential strain (GCS) using DRA (SEG: -16.0% [- 17.5, - 14.4] vs. CS: -16.8% [- 17.8, - 14.9]; p = 0.072), which met equivalence criteria (mean difference - 0.38%, CI90: -0.72% to - 0.05%). Using CS images, DRA-derived GCS correlated more strongly with LVEF (ρ = -0.780) than did FT-derived GCS (ρ = -0.514). In most cases, myocardial strain was not equivalent across accelerated low-resolution and high-resolution segmented images. Only one of eight tested combinations (strain parameter, software, peak definition) met statistical equivalence, showcasing the need for novel processing methods to mitigate variability.