Interrogation of living myocardium in multiple static deformation states with diffusion tensor and diffusion spectrum imaging

Abstract

Diffusion tensor magnetic resonance imaging (MRI) reveals valuable insights into tissue histo-anatomy and microstructure, and has steadily gained traction in the cardiac community. Its wider use in small animal cardiac imaging in vivo has been constrained by its extreme sensitivity to motion, exaggerated by the high heart rates usually seen in rodents. Imaging of the isolated heart eliminates respiratory motion and, if conducted on arrested hearts, cardiac pulsation. This serves as an important intermediate step for basic and translational studies. However, investigating the micro-structural basis of cardiac deformation in the same heart requires observations in different deformation states. Here, we illustrate the imaging of isolated rat hearts in three mechanical states mimicking diastole (cardioplegic arrest), left-ventricular (LV) volume overload (cardioplegic arrest plus LV balloon inflation), and peak systole (lithium-induced contracture). An optimised MRI-compatible Langendorff perfusion setup with the radio-frequency (RF) coil integrated into the wet chamber was developed for use in a 9.4T horizontal bore scanner. Signal-to-noise ratio improved significantly, by 75% compared to a previous design with external RF coil, and stability tests showed no significant changes in mean T1, T2 or LV wall thickness over a 170 min period. In contracture, we observed a significant reduction in mean fractional anisotropy from 0.32 ± 0.02 to 0.28 ± 0.02, as well as a significant rightward shift in helix angles with a decrease in the proportion of left-handed fibres, as referring to the locally prevailing cell orientation in the heart, from 24.9% to 23.3%, and an increase in the proportion of right-handed fibres from 25.5% to 28.4%. LV overload, in contrast, gave rise to a decrease in the proportion of left-handed fibres from 24.9% to 21.4% and an increase in the proportion of right-handed fibres from 25.5% to 26.0%. The modified perfusion and coil setup offers better performance and control over cardiac contraction states. We subsequently performed high-resolution diffusion spectrum imaging (DSI) and 3D whole heart fibre tracking in fixed ex vivo rat hearts in slack state and contracture. As a model-free method, DSI augmented the measurements of water diffusion by also informing on multiple intra-voxel diffusion orientations and non-Gaussian diffusion. This enabled us to identify the transition from right- to left-handed fibres from the subendocardium to the subepicardium, as well as voxels in apical regions that were traversed by multiple fibres. We observed that both the mean generalised fractional anisotropy and mean kurtosis were lower in hearts in contracture compared to the slack state, by 23% and 9.3%, respectively. While its heavy acquisition burden currently limits the application of DSI in vivo, ongoing work in acceleration techniques may enable its use in live animals and patients. This would provide access to the as yet unexplored dimension of non-Gaussian diffusion that could serve as a highly sensitive marker of cardiac micro-structural integrity.

Keywords: Cardiac MRI; Diffusion spectrum imaging; Diffusion tensor imaging; Isolated heart; Langendorff perfusion; Magnetic resonance imaging.

Fig. 1

Optimised setup for magnetic resonance imaging of the perfused rat heart. A: Design of the perfusion setup. The transmit/receive resonator was integrated inside the perfusion chamber to improve the filling factor and thus the SNR. A bubble trap was built into the perfusion head to prevent air bubbles from reaching the heart, and a balloon was added to the system for improved control over the state of contraction. B: Photograph of the perfusion setup prior to assembly. The heart is first connected to the cannula and the balloon then inserted in the left ventricle. The perfusion head (a), internal resonator (b) and perfusion chamber (c) are then assembled in a concentric fashion.

Fig. 2

SNR improved with the integration of the RF coil inside the perfusion chamber. Typical images obtained using a previous perfusion setup where the resonator was placed outside the perfusion chamber (left) and the proposed design (right). The images are displayed at the same scale.

Fig. 3

Anatomical images and DTI maps in three mechanical states obtained in the same perfused rat heart. Apparent diffusion coefficient (ADC), fractional anisotropy (FA) and helix angle maps from the live isolated rat heart in slack (left), in volume overload conditions (middle) and after induction of contracture (right). The images and maps are displayed at the same scale. Note: slices are not exactly matching due to different orientations of states.

Fig. 4

Summary of changes in helix angles in multiple mechanical states. Mean proportions of left-handed fibres (LHF: α ≤ −30°), circumferential fibres (CF: −30° < α < 30°) and right-handed fibres (RHF: α ≥ 30°) observed across the 4 perfused hearts. The changes in the partition of fibres between the three mechanical states were statistically significant (Chi-square: P < 10⁻⁴).

Fig. 5

3D fibre tracking in hearts in slack (top) and contractured (bottom) states shows progression from right to left helical fibres, moving from the subendocardium to the subepicardium. Tracts were progressively seeded with a spherical ROI of increasing radius, normalised to wall thickness and centered in the left ventricular cavity. Tracts were colour coded by helix angle where negative and positive angles correspond to left handed and right handed fibres, respectively.

Fig. 6

Region-of-interest (ROI) based fibre tracking illustrates transition from subepicardial (left) to subendocardial (right) fibres in the lateral wall of the left ventricle (Figs. 6A & A′) and the septal wall of the left ventricle (Figs. 6B & B′). Tracts were seeded with a single voxel radius ROI in a mid-ventricular short-axis slice, and colour coded by helix angle. These are overlaid on single slice masks. Fibre tracking in an apical slice reflects the presence of regions of multiple predominant cell orientations (Figs. 6C & C′). It is important not to confuse this tracking of locally prevailing cell orientations with physically continuous heart muscle strands as cardiomyocytes are discrete and only about 200 μm long.

Fig. 7

DSI parameter maps including mean squared length (MSL) (top), generalised fractional anisotropy (GFA) (middle) and mean kurtosis (MK) (bottom). Mid-sagittal long-axis views of ex vivo rat hearts in slack (left) and contractured (middle) states are presented, alongside normalised histograms from 3D whole heart data (right) from 2 hearts each (solid and dashed lines) in slack state (black) and in contracture (green). Hearts in contracture show evidence of decreased GFA and MK as well as larger variation in MSL and MK.