Dark blood motion compensated spin echo cardiac diffusion tensor imaging

Background: Motion compensated spin echo (MCSE) sequences efficiently interrogate in-vivo cardiac microstructure[1]. However, typical 2^nd-order motion compensated diffusion encoding results in residual blood signal. Blood signal makes delineation of the endocardial borders a difficult trade-off between including edge pixels with artefactually high diffusivities due to blood and losing the endocardial detail of the cardiomyocyte and sheetlet orientation.

Here we attenuate blood signal by modifying the diffusion encoding waveforms.

Methods: Ventricular blood flow is predominantly longitudinal and the first and second order motion compensation result in residual blood signal in MCSE. Therefore, by introducing a small non-zero first and second order moment to the diffusion encoding in the through-plane direction (in short axis), blood signal is attenuated.

An offset in gradient amplitude was added to the longer of the two pairs of MCSE encoding gradients(figure 1) in the slice direction. This design maintains zeroth order moment nulling.

Dark blood encoding was incorporated into an interleaved-slice MCSE research sequence with reduced phase field of view, using the flip back approach and water selective excitation[2].

A three-slice scheme was used with variable gradient offset (dark blood encoding) by slice. Stronger offsets are required apically where blood flow is less, and lower offsets were required basally where longitudinal motion is greater.

The additional diffusion encoding from the gradient offsets was incorporated into the b-matrix in the DICOM header and accounted for when processing (https://github.com/ImperialCollegeLondon/INDI).

Data were acquired using a clinical 3T 200mT/m gradient performance scanner (MAGNETOM Cima.X, Siemens Healthineers, Forchheim,Germany). Field of view 360x125mm², 2.4x2.4mm² acquired resolution, 8mm slices, TE=55ms, SENSEx2, TR=3RR-intervals, 12 directions, b=450smm^-2 and b=50smm^-2. The offset gradient was optimised slicewise to provide dark blood at b=50smm^-2 without motion-induced signal loss.

Results: 5 healthy volunteers were imaged. Figure 2 demonstrates the optimisation of the dark blood offset gradient. Figure 3 demonstrates the effective blood signal suppression in cDTI maps and reduction in MD consistent with the reduction in blood signal. In this example mean LV MD in the 3 slices was 1.46±0.18, 1.48±0.15 and 1.53±0.24 for bright blood vs. 1.38±0.18, 1.28±0.12 and 1.37±0.08 dark blood (x10^-3mm²s^-1 basal, mid, apical respectively).

Conclusion: A modification of second order motion compensated diffusion encoding gradients can supress blood signal in cDTI and removes high MD values in blood pool borders. This method avoids the timing difficulties associated with double inversion dark blood[3] and is compatible with interleaved multislice imaging.

Dark blood MCSE technique will enable more precise quantification of MD and other microstructural parameters in the endocardium, for example after myocardial infarction.

Figure 1: The modified second order motion compensated gradient waveforms with a gradient offset in the through-slice direction (z) that results in a small offset in the first (m1) and (m2) second order gradient moments at the end of the diffusion encoding (right hand zoomed region) while ensuring that the total area (zeroth moment, m0) is nulled and all three moments, m0-m2 are nulled in the in-plane directions (x and y).
Figure 2: Optimising the gradient offset used to minimise blood signal in the b=50smm-2 diffusion weighted images (higher b-value images typically have less residual blood signal). The original second order motion compensated images include substantial residual blood signal (top right), making segmentation of the endocardium and right ventricular septal border difficult without partial volume effects. Applying the gradient offset in-plane (phase direction, bottom left) results in motion induced signal loss (magenta arrows) while providing in complete blood suppression (mauve arrow). Applying the offset gradient through plane (slice direction, bottom right) with a large magnitude results in motion related signal loss.
Figure 3: A comparison of mean diffusivity (MD, units x10-3mm2s-1) and helix angle (HA, units degrees) acquired in an a different example subject in three short axis slices. Even in the basal slice where the blood signal is minimal even without the dark blood offset gradient, the MD maps show increased MD values in myocardium bordering on blood pool (arrows highlight increased MD due to blood signal), which is removed with the dark blood preparation. HA maps appear similar in both cases, although the blood suppression allows endocardial regions to be analysed without blood signal contamination.