A comparison of the reproducibility of two cine-derived strain software programmes in disease states

Background Systolic strain and peak-early diastolic strain rate (PEDSR) measure subclinical cardiac dysfunction. These parameters can be derived from cardiovascular magnetic resonance (CMR) cine images using new software packages, but the comparative test-retest reproducibility of these software in disease states is unknown. This study compared the test-retest reproducibility of strain measures derived from two software packages (feature-tracking software (FT) and tissue-tracking (TT)) in disease populations with preserved ejection fractions. Methods This was a prospective study of 10 patients with aortic stenosis (AS), 10 haemodialysis patients and 10 diabetic patients at 1.5 and 3-Tesla. 30 subjects underwent test-retest reproducibility scans of global circumferential strain (GCS), global longitudinal strain (GLS), circumferential-PEDSR and longitudinal-PEDSR calculated using TT and FT software. Results Test-retest reproducibility of GCS and GLS were similar for FT and TT across patient groups. Coefficient of variability (CoV) for FT-derived GCS 8.1%, 5% and 7.9% for AS, diabetic and haemodialysis patients, compared to 3.3%, 9.2% and 5.4% for TT-derived GCS, with CoV for FT-derived GLS 8%, 6.4% and 8.2% for AS, diabetic and haemodialysis patients, compared to 5.3%, 4.8% and 7% for TT-derived GLS). Reproducibility of FT-derived circumferential and longitudinal-PEDSR was worse than TT-derived circumferential and longitudinal-PEDSR (CoV for FT-derived circumferential-PEDSR 18.2%, 18% and 17.4% for AS, diabetic and haemodialysis patients, compared to 6.1%, 11.7% and 11% for TT-derived circumferential-PEDSR with CoV for FT-derived longitudinal PEDSR 18.2%, 18.9%, 18.3% for AS, diabetic and haemodialysis patients, compared to 8.9%, 9.1% and 11.4% for TT-derived longitudinal-PEDSR). Bland-Altman analysis revealed no systematic bias with tighter limits of agreement for TT-derived strain measures. Conclusions Reproducibility of GCS and GLS are excellent with FT and TT software across diseases. TT had superior test-retest reproducibility for quantification of longitudinal and circumferential-PEDSR than FT-derived PEDSR across diseases.