Introduction

Vasomotion, defined as an ultra-low frequency (≈0.08 Hz) spontaneous oscillation of blood vessels, has been hypothesized as a mechanism for resting-state BOLD functional connectivity¹ and for driving paravascular waste clearance in the brain². While vasomotion oscillations have been successfully detected in the rodent brain where they can be entrained by functional stimuli modulated at near the vasomotion frequency¹, direct localized measurement in humans has not been demonstrated.
We propose here a proof-of-concept for measuring vasomotion in the human brain by Fourier decomposition of high frame-rate Cine-FLASH data.

Methods

Concept: In a Cine-FLASH sequence, stationary spins get saturated because of shortly repetition time RF pulses especially if the flip-angle is above Ernst angle³. In a region of interest encompassing a pulsating structure, the total signal in that ROI will be modulated by the pulsation frequency. Therefore, if the temporal resolution is sufficient to respect Nyquist criteria (leading to a sampling rate of at least twice the frequency of the signal to be detected), vascular oscillations could potentially be detected by performing a Fourier decomposition⁴ of a temporal series acquired in a slice containing a vascular region of interest.
Experiments: Two healthy adults (54 and 64 yo males) were scanned at 3T (MAGNETOM Prisma, Siemens Healthineers, Erlangen, Germany) using a 64-ch head/neck coil. After localizers, we acquired a 3D Time-of-Flight angiogram (TOF-MRA) accelerated with Compressed Sensing⁵ to localize a slice of interest (0.4mm isotropic, R=10.3), containing branches of both middle cerebral arteries (MCA) and posterior cerebral arteries (PCA). Once the spatial location was selected, an axial single slice, single-shot Cine-FLASH sequence was acquired continuously for 2.7min (512 phases) as illustrated in Figure 1. In order to increase temporal resolution as much as possible, we used parallel imaging (GRAPPA6 with R=2), phase Partial Fourier (6/8) as well as an asymmetric echo (TE = 2.06ms). With a matrix of 160x160 and FOV = 180x180mm², bandwidth = 947Hz/pixel, we reached a minimum TR = 315ms for a 3-mm thick slice. We used a flip angle alpha = 35 degrees to saturate parenchymal signal (assuming T1 = 0.8/1.6s in white and gray matter⁷). In one of the of the volunteers, we presented a visual stimulation consisting of 16 Hz flicker with a modulation frequency of 0.1Hz as an attempt to amplify the vasomotion using neural stimulation. Starting in total dark, we acquired Cine-FLASH data for 512 phases (2.7min) prior to switching on the visual stimulation for the same amount of time.
Processing: Cine time series were saved as DICOM images and processed offline. After realignment of the time series using mcflirt (FSL)⁸, a 512-points discrete Fast Fourier Transform (FFT) was performed in MATLAB on a time-series spatially averaged in a manually defined ROI, leading to a spectral resolution of 0.0062Hz. A 3-points sliding window averaging was used for noise reduction.

Results and Discussion

An example of the raw time series is shown in two different ROIs placed in the parenchyma and PCA. When looking at their respective Fourier decomposition, while the parenchymal ROI does not highlight any specific feature, the one from the PCA shows specific peaks reflective of cardiac pulsation at 0.9Hz (corresponding to RR=1111ms, HR=54bpm), but also a peak in the ultra-low frequency range ≈ 0.05-0.1 Hz consistent with vasomotion.
When looking at the FD-FLASH with and without visual stimulation, an increase in the power spectrum in the 0.1Hz frequency range, with a shift from 0.08 to 0.1Hz with amplification by 154% of the 0.1Hz signal is observed (Figure 3). Interestingly, we could also observe a different signal time course with a signal drift and shift of the cardiac peak in the frequency spectrum suggesting systemic modification of hemodynamic parameters during visual stimulation.

Conclusions

In this preliminary pilot work, an ultra-low frequency signal suggestive of vasomotion could be detected in vivo in the human brain using a Fourier decomposition of Cine-FLASH T1-weighted MRI, as well as potential sign of vasomotion modulation with functional stimulation. Future work will be geared towards improving acquisition SNR and temporal resolution using for example Non-Cartesian radial trajectories. Additionally, modulation of the visual stimulation frequency will also be explored to assess whether we can detect associated vasomotion modulations.

Acknowledgements

No acknowledgement found.