Left ventricular volume measurement using cardiac axis nuclear magnetic resonance imaging. Validation by calibrated ventricular angiography.

Proton nuclear magnetic resonance (NMR) imaging has the potential to serially assess left ventricular (LV) volumes with optimal accuracy because it is a high-resolution, three-dimensional, noninvasive modality. Previous NMR studies to assess LV volumes have been suboptimal, as they have used either planes aligned with the axes of the body, which are compromised by partial volume effects, or spin-echo techniques that have been time-consuming to acquire and analyze. Accordingly, for LV volume measurement, we developed a gradient-echo (cine) NMR strategy that uses two orthogonal planes intersecting along the intrinsic long axis of the heart (two-chamber and four-chamber). This approach was validated against calibrated contrast biplane LV cineangiography (CATH) and also compared with a previously reported short-axis spin-echo NMR method. Twenty-one patients underwent CATH and NMR (long-axis, n = 21; short-axis, n = 14) within a 3-day interval. Although both long- and short-axis NMR LV volumes and ejection fractions correlated well with CATH (r greater than 0.90, p less than 0.001 in all), end-diastolic volumes by both long-axis (161 +/- 85 ml) and short-axis (151 +/- 81 ml) NMR were systematically less than those by CATH (182 +/- 85 ml) (p less than 0.05). Consequently, ejection fractions by long-axis (48 +/- 17%) and short-axis (49 +/- 17%) NMR consistently underestimated those by CATH (54 +/- 16%, p less than 0.05). End-systolic volumes by long-axis (94 +/- 71 ml) and short-axis (87 +/- 72 ml) NMR were not significantly different from those by CATH (92 +/- 69 ml). Both NMR techniques had low intraobserver and interobserver variation (less than 11%); however, short-axis spin-echo NMR involved longer acquisition/reconstruction (35 versus 18 minutes) and analysis (25 versus 10 minutes) times. We conclude that both short-axis spin-echo and long-axis gradient-echo NMR approaches reliably estimate LV volumes. Currently, the long-axis strategy appears more practical for clinical use because the scan and analysis times are relatively short.