Cardiac-induced brain tissue volumetric strain (an important metric reflecting microvascular blood volume pulsation) exhibits large inter-subject variability. We measured the intra-subject repeatability of brain tissue volumetric strain over the cardiac cycle in healthy subjects at 7T using a high resolution retrospectively-gated DENSE and found that the strain curve shape and peak value were very consistent between measurements. Furthermore, we validated the peak brain tissue volume change against measured spinal CSF stroke volume at C2-C3, and observed a strong correlation. The consistent findings strengthens the potential of volumetric strain as biomarker for microvascular function in the ageing brain.
Measurement
8 healthy subjects (3 females, mean age: 26.8±6) provided informed consent to participate in the study which was approved by the Ethical Review Board of our institution. 4D retrospectively-gated DENSE was implemented on a 7T scanner (Philips Healthcare) with a 32-channel head coil (Nova Medical), and used to measure the cardiac-induced brain tissue displacement in the RL, AP and FH directions. Cardiac gating was done with a pulse oximeter on the index finger. Subjects were shortly taken out of the scanner and then re-scanned with the same DENSE protocol. CSF flow into the spine was measured using a 2D cine-PCMRI acquisition located at the C2-C3 level. Additionally, a 3D T1-weighted FFE acquisition was made for registration and segmentation of brain tissue. See Table 1 for all scan parameters.
Analysis
Raw phase images were converted to displacement maps as previously described3. Displacement maps were registered to the T1-weighted image space using Elastix6,7. CSF, white and grey matter probability masks were created using the Computational Anatomy Toolbox (Jena University Hospital) for SPM12.
CSF, noise and artifact regions were removed from all displacement maps before computing the volumetric strain as follows. The difference between the displacement maps at the beginning and end of the CC was computed for all encoding directions. Any voxel whose absolute displacement exceeded one standard deviation of this difference map was removed. Furthermore, any voxel outside the tissue mask (having a probability less than 0.95 in the sum of the grey and white matter masks, or a non-zero value in the CSF mask) was also removed.
Volumetric strain was computed as the divergence of the masked displacement field for each cardiac phase (see Figure 1 for example images), and the mean value within the brain tissue for each cardiac phase was found. Repeatability was assessed using a Bland-Altman plot of the peak tissue volume change (pTVC) of the two measurements (calculated as the product of peak volumetric strain and tissue volume within the tissue mask). The sCSF was assessed by integrating the CSF flow over the CC8, and compared to the mean pTVC of the two measurements with a correlation plot.
Volumetric strain and CSF flow measurements were successfully attained for all subjects. The general shape of the volumetric strain curves was similar for all measurements and subjects (Figure 2). The mean absolute difference between repeated measurements was 0.14mL (see Bland-Altman plot, Figure 3). pTVC and sCSF are shown in Table 2. A strong correlation was found between sCSF and pTVC.
The results of the Bland-Altman analysis suggests good agreement between repeated measurements. The similarity between the tissue volume change (equivalently, volumetric strain) curve shapes observed amongst all subjects reflects a consistent inter-subject pattern of cardiac induced brain tissue volume change. The inter-subject peak volume change is noticeably variable. Nevertheless, the repeatability of this finding and the correlation to sCSF shows that this variation probably reflects physiological variation between subjects rather than measurement errors. This strengthens the case for brain tissue displacement as a possible biomarker for microvasculature blood volume changes, thereby providing an important window to the microvasculature which is notoriously challenging to image directly in-vivo. Thus, this metric holds potential as a means for studying blood volume pulsations in the elderly population (since arterial properties change with age), or in diseases such as cerebral small vessel disease. Calculation of volumetric strain was hampered by image artifacts3, particularly in the FH acquisitions, which currently limits regional analysis of brain tissue volumetric strain. Further improvements to this technique are necessary to remove these artifacts and boost SNR, which would allow regional strain analyses.
Table 1. Imaging parameters used in the study.