Contrast-free high-resolution 3D magnetization transfer imaging for simultaneous myocardial scar and cardiac vein visualization

Figures

• 

  Fig. 1. a Acquisition scheme for the two 3D segmented (electrocardiogram-triggered) datasets (with bSSFP), the reference (REF) dataset (without preparation) and the MT-prepared dataset (with MT preparation) are acquired sequentially. Both acquisitions include 2D image navigators (iNAVs) obtained by encoding the first 14 start-up pulses of the imaging module. b Scheme of the MT preparation module which consisted of 20 off-resonance pulses (20 ms) separated by pause periods (1.5 ms) where 3D spoiler gradients were played out

• 

  Fig. 2. a Data flow during image acquisition and reconstruction. From the top left, raw data and image navigators (iNAVs) were obtained from the scanner; iNAVS were used for translational motion correction in two ways: (1) to the end expiration phase for translational motion correction (MC); (2) to the centre of each respiratory phase (or bin), followed by non-rigid registration of the bins and generation of 3D motion fields. Non-rigid MC was performed using a generalized matrix description that included the calculated motion fields, sampling mask and coil maps. b Data flow during image analysis. MT and REF datasets were reformatted to either a 2D coronary projection image for vessel sharpness calculation, or to left-ventricle short axis view for manual segmentation and creation of a mask. MTR maps were calculated from non-reformatted MT and REF datasets. A blurring Gaussian filter was applied in the MTR maps before segmentation measurements

• 

  Fig. 3. a Simulation of signal (Mxy) of the proposed sequence in myocardium-like tissue with balanced steady-state free precession, as a function of B0 inhomogeneity. Signal from the MT and REF acquisitions (left axis) showed ‘stop’ bands (1/2TR) approximately at ± 150 Hz, while the MTR (right axis) showed a maximum reduction of up to 8% at B0 < ± 150 Hz compared to B0 = 0. b Simulations as described in a, except with an SPGR imaging module instead. The robustness of SPGR is shown in the flat profile obtained for the REF dataset. MTR signal is only marginally affected (< 3%) throughout the B0 inhomogeneity range that was in the range of ± 200 Hz

• 

  Fig. 4. a MTR sensitivity to the free pool relaxation rates, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${T}_{1}^{A}$$\end{document}T1A and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${T}_{2}^{A}$$\end{document}T2A, comparing bSSFP and SPGR imaging modules. SPGR showed a flatter profile with MTR variations less than 2% across a common range of relaxation rates. b MTR sensitivity to the exchange rate (R) and the pool size ratio (PSR) for bSSFP and SPGR. Both imaging modules showed similar increase in MTR as a function of PSR, as well as a more moderate increase in MTR as a function of R

• 

  Fig. 5. a MTR dependency as a function of the off-resonance frequency (ΔF) showed strong dependency at low ΔF, i.e. ΔF < 1.5 kHz, for a phantom made of 2.5% concentration of agar. b MTR as function of MT flip angle showed steeper slope for low ΔF (400 Hz) than that of higher ΔF (3000 Hz) with nonlinear dependency at high flip angles (> 800˚), for a phantom made of 2.5% concentration of agar. Measured and simulated data were in good agreement in a and b. c MTR images of the four agar-based samples (1, 2, 2.5, and 5% concentration) with bSSFP and SPGR imaging modules using different MT flip angles and fixed ΔF = 400 Hz. Red dashed arrows show an intensity gradient from centre to periphery in the bSSFP (not observed in SPGR). Likewise, red dashed squares show other artefacts seen only in the bSSFP acquisition

• 

  Fig. 6. Co-registered coronal slices comparing MTR maps with LGE images in three pigs, 6 weeks after left anterior descending artery occlusion. In all pigs, the MTR maps show significant reduction in areas that overlap with signal enhancement in the LGE images, allowing clear localization of myocardial scar from the contrast-free MTR maps alone

• 

  Fig. 7. Left and middle columns show a coronal slice in a representative healthy subject, comparing the MT-prepared dataset with the MTR map. bSSFP and SPGR imaging modules are shown in each row. The MTR map looks more homogenous with SPGR than with bSSFP, where there is an artefact in the lateral region (red arrow). The right column shows MTR mean values per segment of the American Heart Association (AHA) 16-segment model, averaged over the entire cohort of healthy subjects (n = 10). The SPGR imaging module showed reduced intersegment variation compared to bSSFP (red arrows)

• 

  Fig. 8. Main coronary veins and structures such as coronary sinus (CS), the great cardiac vein (GCV) and the posterior branch of the cardiac vein (PCV) are shown in three different subjects, comparing the two motion correction (MC) approaches. The non-rigid MC improves the visualization with respect to the translational MC in all cases, recovering blurred features and increasing vessel sharpness