Feasibility of strain-encoded magnetic resonance at 0.55T

Background: Low-field (<1.0T) wide-bore cardiovascular magnetic resonance (CMR) has the potential to improve patient accessibility; however intrinsically reduced signal-to-noise ratio may affect techniques such as strain-encoded magnetic resonance (SENC), a method to quantify regional strain. We sought to characterize the performance of SENC on a low-field system in a phantom, healthy subjects, and a porcine model of myocardial infarction (MI). Methods: A prototype SENC sequence was implemented on 0.55T and 1.5T systems and used to scan a phantom and 16 healthy volunteers. 10 subjects underwent repeat scans at each field strength for scan-rescan repeatability testing. T-tests were used to compare global strain values; reproducibility between field strengths and scan-rescan repeatability were assessed via Bland-Altman and intra-class correlation (ICC). Adjunctive SENC followed by late gadolinium enhancement (LGE) was acquired at 0.55T in a porcine MI model (n = 6). Left ventricular (LV) segments were categorized by LGE, and segmental strain was compared via one-way analysis of variance. Results: Phantom strain showed no significant differences between field strengths (p > 0.10). In volunteers, mean LV global longitudinal (GLS) and circumferential strain (GCS) were -19.4% ±1.1 and -20.4% ± 0.9 at 0.55T compared to -18.7 ±1.4% and -19.2% ±1.6 at 1.5T (p > 0.10). LS proved to have better agreement than CS, and mean biases were low for both global and segmental comparisons throughout. Limits of agreement were good for global strain but wider for segmental measurements. Pooled LV segmental strain ICC showed good reproducibility for LS between field strengths (0.78) and good repeatability at 0.55T (0.89); however, reproducibility for CS was fair (0.60), as was repeatability at 0.55T (0.64). In the porcine infarct model, segmental LS in LGE+ segments (-10.8% ±4.0) was less negative than remote segments (-16.8% ± 5.1), p < 0.001. Similarly, segmental CS in LGE+ vs remote segments was -11.9% ± 2.7 vs -14.6% ± 2.7; p = 0.0011. Conclusion: Our results support the feasibility of SENC at 0.55T, with accurate phantom measurements, good agreement of global values with 1.5T in human volunteers, and correlates of functional impairment with known MI. Reproducibility showed minimal systemic bias but at times substantial limits of agreement. Repeatability of global and segmental LS at 0.55T was similar to established 1.5T performance, although CS was notably worse than LS. LV CS may lack sufficient reliability in its current implementation for use at 0.55T.