Introduction

Recently, a direct link between cerebrospinal fluid (CSF) fluctuations in the 4^th-ventricle with low-frequency hemodynamic oscillations in the cortex was found using an accelerated resting-state fMRI sequence^1,2. This MRI-technique is based on fresh CSF inflow (feet-head) into the bottom slice of the imaging volume. It therefore cannot capture CSF outflow dynamics nor measure flow quantitatively. Quantifying CSF flow in real-time could elucidate CSF flow dynamics and potentially provide a better picture of the interaction between CSF and BOLD fluctuations coupling. Hence, we accelerated a real-time phase contrast (rtPC)³ sequence planned on the 4^th-ventricle and interleaved each dynamic with cortical BOLD echo-planar imaging (EPI). We show that rtPC and inflow-sensitized EPI scans in the 4^th-ventricle measure similar CSF dynamics and compare the BOLD-CSF coupling of both approaches.

Methods

Five healthy volunteers (age= 27±4y, 2 female) signed informed consent and laid supine in a 7 Tesla Philips (Best, The Netherlands) MR scanner (dual-transmit, 32-channel receive coil (Nova Medical Inc.)). Respiration signal, pulse plethysmography and a full-brain anatomical scan were acquired. We first interleaved rtPC with an inflow-sensitized EPI, both positioned on the 4th-ventricle for direct method comparison (experiment 1, Fig.1). Subsequently, the following sets of sequence combinations were used during a breathing task (experiment 2), a visual stimulation (experiment 3) and at rest (experiment 4):

- rtPC positioned on the 4^th-ventricle interleaved with a BOLD scan positioned on the cortex for capturing cortical hemodynamic fluctuations
- a standard inflow-sensitized EPI for BOLD-CSF coupling for comparison.

500 dynamics (+10 dummies) were acquired for each scan (2.58 min per scan). All inflow-sensitized EPI scans (TR/TE=348/22ms, FA=30°, resolution=2.5x2.5x3mm3, FOV=190x230x63mm3, 21 slices, EPI-factor=45, SENSE=2, multiband-factor=3) were aligned perpendicular to the midsection of the 4^th-ventricle. rtPC-BOLD interleaved scans were acquired with a 10mm 2D rtPC slab centered on the 4^th-ventricle coinciding with the lowest inflow-EPI slice (VencFH=5cm/s, TR/TE=7.8/5.7ms, Tdyn=202ms, FA=4°, FOV=110x110mm, resolution=2x2mm, SENSE=3, partial-Fourier=70%). The interleaved BOLD-EPI volumes were centered on the visual cortex (parameters: TR/TE=146/22ms, FA=30°, resolution=2.5x2.5x3mm voxels, FOV=190x230x27mm, 9 slices, EPI-factor=31, SENSE=3, multiband-factor=3). The combined dynamic scan time of the interleaved sequences matched the inflow-sensitized EPI TR.

Paced breathing drives CSF dynamics, as seen on rtPC³, so we visually guided paced breathing (5s in, 5s out) to drive respiratory-linked CSF fluctuations. Additionally, as BOLD-CSF coupling is shown to be manipulated by visual stimulation¹, an 8Hz-flickering checkerboard was shown to induce visual cortex functional hyperemia (10s-ON, 10s-OFF) using PsychoPy⁴.

Cortex ribbon and visual area were segmented (FreeSurfer⁵). BOLD-EPI (pre)-processing included motion-correction (SPM12), demeaning and detrending, filtering (if indicated), and negative temporal derivative estimation (-ΔBOLD/dt). Negative values in -ΔBOLD/dt were set to zero before cross-correlation with inflow-EPI, but not for rtPC since the latter allows measurement of flow in both directions. 4^th-ventricle masks were delineated on the time-averaged lowest slice of the inflow-EPI and PC magnitude image. Spectral analysis was performed using five-taper² Fourier transform (Chronux toolbox⁶).

Results and Discussion

By directly interleaving rtPC and inflow-EPI at the 4^th-ventricle level during a paced breathing paradigm, we show that both methods can capture similar signal patterns, including cardiac pulsations, respiratory and low-frequency oscillations (Fig.2A&B). rtPC CSF signal is mainly cardiac driven, as expected from previous research³. In separate paced breathing inflow-EPI and rtPC scans, signal correspondence remains high between CSF measurements (Fig.2D) but with a slight time lag of the inflow-EPI (Fig.2C).

Paced breathing modulated CSF dynamics and lead to changes in BOLD during inspiration (Fig.3), as dictated by the Monroe-Kelly doctrine. During this paradigm, rtPC provided stronger and more coherent BOLD-CSF coupling across participants (Fig.3B) than inflow-EPI (Fig.3D). The “bimodal” CSF peaks across blocks as observed in inflow-EPI scans could be explained by tissue shifting following inspiration.

During visual stimulation, cardiac signals dominate the unfiltered BOLD timeseries in the visual cortex (Fig.4A&B). Nonetheless, coupling to CSF-flow in the 4^th-ventricle and -ΔBOLD/dt could be observed, indicating an effect of neurovascular coupling on CSF flow.

Paced breathing did not directly improve BOLD-CSF coupling compared to rest, possibly due to increased tissue motion that made extracted signals from the static cortical and 4^th-ventricle masks more variable.

Conclusion

We demonstrated that real-time phase contrast (rtPC) MRI can be interleaved with inflow-EPI, and showed that both methods acquire similar CSF pulsation patterns in the 4^th-ventrivbe. The quantitative nature and the insensitivity to B0 fluctuations of rtPC is an important advantage when studying BOLD-CSF coupling. In multiple experiments, rtPC captured BOLD-CSF coupling better than inflow-sensitized EPI, overcoming the inflow-pulse limitations of EPI to reliably show signal correspondence between BOLD and CSF during in- and out-flow dynamics.

Acknowledgements

This project was made possible by grants from the Dutch research council (NWO), the joint program neurodegenerative diseases (JPND-program human brain clearance imaging) and Alzheimer Netherlands.