Single breath‐hold three‐dimensional whole‐heart T2 mapping with low‐rank plus sparse reconstruction

Abstract

AbstractThe purpose of the current study was to develop and evaluate a three‐dimensional single Breath‐hOLd cardiac T2 mapping sequence (3D BOLT) with low‐rank plus sparse (L + S) reconstruction for rapid whole‐heart T2 measurement. 3D BOLT collects three highly accelerated electrocardiogram‐triggered volumes with whole‐heart coverage, all within a single 12‐heartbeat breath‐hold. Saturation pulses are performed every heartbeat to prepare longitudinal magnetization before T2 preparation (T2‐prep) or readout, and the echo time of T2‐prep is varied per volume for variable T2 weighting. Accelerated volumes are reconstructed jointly by an L + S algorithm. 3D BOLT was optimized and validated against gradient spin echo (GraSE) and a previously published approach (three‐dimensional free‐breathing cardiac T2 mapping [3DFBT2]) in both phantoms and human subjects (11 healthy subjects and 10 patients). The repeatability of 3D BOLT was validated on healthy subjects. Retrospective experiments indicated that 3D BOLT with 4.2‐fold acceleration achieved T2 measurements comparable with those obtained with fully sampled data. T2 measured in phantoms using 3D BOLT demonstrated good accuracy and precision compared with the reference (R2 > 0.99). All in vivo imaging was successful and the average left ventricle T2s measured by GraSE, 3DFBT2, and 3D BOLT were comparable and consistent for all healthy subjects (47.0 ± 2.3 vs. 47.7 ± 2.7 vs. 48.4 ± 1.8 ms) and patients (50.8 ± 3.0 vs. 48.6 ± 3.9 vs. 49.1 ± 3.7 ms), respectively. Myocardial T2 measured by 3D BOLT had excellent agreement with 3DFBT2 and there was no significant difference in mean, standard deviation, and coefficient of variation. 3D BOLT showed excellent repeatability (intraclass correlation coefficient: 0.938). The proposed 3D BOLT achieved whole‐heart T2 mapping in a single breath‐hold with good accuracy, precision, and repeatability on T2 measurements.