High-resolution three-dimensional ANGIE T1 mapping of the whole heart

Abstract

Background: Assessment of fibrosis in thinner myocardial structures such as the right ventricular (RV) and left atrial walls would be valuable in disorders such as pulmonary hypertension, congenital heart disease, and atrial fibrillation. We recently developed a novel technique, termed ANGIE, which provides high-resolution 2D T1 mapping for assessment of thin structures such as the wall of the RV. However, high resolution 2D imaging is limited by lower SNR and time efficiency. A 3D image acquisition provides higher SNR and complete coverage of the heart. However standard 3D navigated techniques have prohibitively long scan times. Thus, the aim of the present study was to extend ANGIE to perform highresolution three-dimensional (3D) T1 mapping of the whole heart within a clinically acceptable scan time. Methods: A novel image reconstruction algorithm, which uses compressed sensing (CS) with local low-rank sparsity in conjunction with partial Fourier and parallel imaging, was developed to achieve high acceleration (R=10) for 3D ANGIE. High-resolution (1.4×1.4×4mm3) T1 mapping of the whole heart was performed using 3D ANGIE in three healthy volunteers on a 1.5T system (Avanto Siemens). Imaging parameters included: matrix size=224×224×26, number of inversion times=12, and partial Fourier reduction factor of 3/4 along phase and partition encode. Additionally, 2D MOLLI (2.5×2×4mm3) and 2D ANGIE (1.3x1.3x4mm3) scans were performed at three different short-axis slice positions. Results: Figure 1 illustrates example reconstructed (A-D), fully sampled (E-H) and reconstruction error (I-L) images from a retrospectively accelerated (R=10) 3D ANGIE dataset. The image reconstruction algorithm effectively suppresses aliasing artifacts and retains spatial resolution, illustrating the capability of achieving a high acceleration rate (R=10) using the proposed reconstruction algorithm. Figure 2 illustrates example T1 maps from a healthy volunteer acquired using all three techniques with prospective acceleration at three different slice positions across the ventricle. 2D MOLLI (Fig. 2G-I) provides high SNR but incomplete definition of the RV wall, whereas 2D ANGIE (Fig. 2D-F) provides good definition of the RV wall but lower SNR. However, 3D ANGIE (Fig. 2A-C) provides both good definition of the RV wall as well as high SNR. The left ventricular myocardium and blood T1 estimates of 3D ANGIE (LV: 1013±67ms, blood: 1501±44ms) were in close agreement with 2D MOLLI (LV: 994±81ms, blood: 1507±68ms). The RV T1 estimates of 3D ANGIE (1022±85ms) were in close agreement with 2D ANGIE (1028±89ms). The acquisition time and navigator acceptance for 3D ANGIE was 6.6±2.7mins and 67±19%, while for 2D ANGIE was 2.2± 1.2mins (per slice) and 60±16%. Conclusions: 3D ANGIE provides accurate high-resolution (1.4×1.4×4mm3) native T1 maps of the whole heart within a clinically acceptable scan time. (Figure Presented).