0344

Investigation of the dependence of free water and pseudo-diffusion MRI estimates on the cardiac cycle
Alberto De Luca1, Suzanne Franklin1,2, Carlo Lucci3, Jeroen Hendrikse3, Martijn Froeling3, and Alexander Leemans1

1Image Sciences Institute, UMC Utrecht, Utrecht, Netherlands, 2C. J. Gorter Center for High Field MRI, Leiden University Medical Center, Leiden, Netherlands, 3Department of Radiology, UMC Utrecht, Utrecht, Netherlands

Synopsis

With diffusion MRI (dMRI) data at multiple diffusion weightings it is possible to quantify the relative fractions of multiple water pools. In this work, we investigated changes in free water diffusion and micro- and macro-vascular pseudo-diffusion during the cardiac cycle. Further, we propose a data driven method to bin the dMRI signals according to the cardiac phase. dMRI at 4 diffusion weightings was acquired 80 times with short repetition time. A multi-exponential fit of the binned data showed increases of free water in white matter and periventricular areas, and opposite increases/decreases for micro- and macro-vascular pseudo-diffusion in grey matter, respectively.

Introduction

The diffusion MRI (dMRI) signal measured with Pulsed Gradient Spin-Echo (PGSE) sequences is sensitive to multiple diffusion domains, including intra/extra-cellular diffusion, free water, and blood pseudo-diffusion1. A previous study2 showed that pseudo-diffusion estimates in the brain depend on whether the dMRI signal was measured in the systole or the diastole. In this work, we investigated whether changes in free water and micro- and macro-vascular diffusion are affected by the cardiac cycle. Further, we propose a data-driven method to synchronize the dMRI signal to the cardiac phase.

Methods

One subject (F, 27 years) underwent a 3T MRI. The protocol included a 1mm isotropic T1W image (TE=3.7ms, TR=8.1ms, SENSE=2+1.4) and a dMRI acquisition with 13 volumes (1 b=0s/mm2, 3xb=50,100,300,800s/mm2 with orthogonal gradients, TE=96ms, TR=1s, SENSE=2, 14 slices) at 2.5mm isotropic resolution.The acquisition was repeated for 80 dynamics to enable the acquisition of volumes randomly sampled over the different phases of the cardiac cycle, which was recorded with a peripheral pulse unit (PPU).

dMRI data were processed for subject motion and eddy currents3, geometrically averaged, and fit with a four-exponentials model. The exponentials were centered at diffusion values DT=0.7x10-3mm2/s, DFW=3x10-3mm2/s, DMIC=20x10-3mm2/s, DMAC=200x10-3mm2/s, to determine signal fractions associated to white/grey matter tissue (fT), free water4 (fFW), micro-vascular pseudo-diffusion (fMIC) and macro-vascular pseudo-diffusion5 (fMAC), respectively.

dMRI data were binned with two methods. In the first method, the inter-peak time of the PPU signal was divided in 10 bins, and the slices binned accordingly. In the second method, named “self-synchronized” and shown in Figure 1, the average signal of the first slice of each volume was detrended6 and processed to obtain normalized signal variations, which were divided in 10 bins. This method assumes a direct link between signal variations and blood flow changes during the cardiac cycle.

Pearson correlations were computed between signal changes and the equally sampled PPU signal. Each slice of the dataset was binned accordingly by interpolating the bin assignment over the slice acquisition order. Consequently, the four exponentials fit was repeated on data corresponding to each bin. Three regions of interest (ROIs) were manually drawn over multiple slices in i) the ventricles ii) a pure white matter region iii) bi-lateral grey matter areas (with some partial volume of fluids). Statistics of the signal fractions were computed within the ROIs. Z-tests were performed between signal fractions obtained with the binned data and the results from a fit of the whole data. Further, the fractions and their difference from the average were visually compared between the two methods.

Results

Figure 2 shows the normalized signal variations corresponding to each diffusion weighting (Figure 2A), their concatenation and binning (Figure 2B), and a comparison with the PPU signal in a short time frame (Figure 2C). Large fluctuations above 1 standard deviation were observed throughout the dynamics. Significant correlations between the signals from PPU and and dMRI were observed, as reported in Table 1.

Figure 3 shows the changes in signal fractions over the cardiac phase. Estimates of fMIC and fMAC appeared to be coupled with both synchronization methods, implying that increases in fMIC reflected in decreases of fMAC, and viceversa. The largest changes were observed for fFW, with a decrease/increase in WM. Figure 4 shows an example slice of the fractional maps obtained with the two methods and their evolution over the cardiac cycle. For both methods, remarkable fFW increments were observed in WM and peri-ventricular areas. fMIC changed mostly in GM and deep GM areas, whereas the biggest changes in fMAC were located in vessels and their surroundings.

Discussion

The dMRI signal depends on the cardiac cycle over multiple diffusion weightings, as proven by the correlation with the PPU signal. At ROI level, the self-synchronized method provided similar estimates to the PPU synchronization, allowing to detect characteristic changes of free water diffusion in WM. At voxel level, both methods showed similar cyclic patterns over time, but a possible time-shift was observed. Additionally, stronger changes in fFW were observed with self-synchronization, suggesting sensitivity of the method beyond pure blood flow. In this experiment, we did not control for the delay between the PPU unit and blood arrival in the brain, which should be addressed in future work. Further, we decoupled changes in micro and macro-vascular pseudo-diffusion5, which had opposite trends that might cancel out when considered together2.

Conclusions

We have shown that measures of free water diffusion, micro and macro-vascular diffusion are sensitive to physiological changes during the cardiac cycle, and introduced an acquisition and processing strategy to investigate such changes without external synchronization.

Acknowledgements

The authors acknowledge dr. Frank Zijlstra for advices on the PPU signal optimization.

References

1. De Luca, A., Leemans, A., Bertoldo, A., Arrigoni, F. & Froeling, M. A robust deconvolution method to disentangle multiple water pools in diffusion MRI. NMR Biomed. e3965 (2018). doi:10.1002/nbm.3965

2. Federau, C. et al. Dependence of brain intravoxel incoherent motion perfusion parameters on the cardiac cycle. PLoS One 8, e72856 (2013).

3. Leemans, A., Jeurissen, B., Sijbers, J. & Jones, D. K. ExploreDTI: a graphical toolbox for processing, analyzing, and visualizing diffusion MR data. in 17th annual meeting of the International Society for Magnetic Resonance in Medicine, Honolulu, Hawaii, USA 3537 (2009).

4. Pasternak, O., Sochen, N., Gur, Y., Intrator, N. & Assaf, Y. Free water elimination and mapping from diffusion MRI. Magn. Reson. Med. 62, 717–30 (2009).

5. Fournet, G. et al. A two-pool model to describe the IVIM cerebral perfusion. J. Cereb. Blood Flow Metab. 37, 2987–3000 (2017).

6. Vos, S. B. et al. The importance of correcting for signal drift in diffusion MRI. Magn. Reson. Med. 22, 4460 (2016).

Figures

Figure 1: A schematic overview of the processing steps to perform data binning according to the “self-synchronization” of the dMRI signal. The raw data, consisting of 13 volumes repeatedly acquired over 80 dynamics, underwent common pre-processing and geometric averaging. All data were fit with a four-exponentials model, to determine the signal fractions associated to four tissue classes. A mask of the macro-vasculature was derived by thresholding fMAC above 20%. The signal of the first slice of each volume was averaged within the mask, detrended, normalized and divided in 10 bins.

Figure 2: A) Detrended and normalized signal variations of data at b=0s/mm2 (solid black line) and of each diffusion weighting (colored lines) over the acquired dynamics. All data showed signal variations up to 3 standard deviations across dynamics. B) The signal variations determined for the first slice of each diffusion weighting were concatenated according to acquisition order, then subdivided in 10 bins. C) A comparison between the PPU and the self-synchronization signals within 100 seconds (100 volumes).

Table 1: Correlation coefficient between the PPU signal and the signal from the first slice of dMRI data at multiple b-values.

Figure 3: Average ± standard deviation of the signal fractions within 3 ROIs according to the PPU synchronized (first row) and the self-synchronized (second row) data binning methods. Dashed lines represent the average fraction values when considering all data without binning. Asterisks mark points significantly different from the dashed line. The values of the vascular components were very similar for both methods across ROIs. In WM, fFW decreased up to half cycle, then approximately recovered its initial value, with more marked changes with self-synchronization. Increase in fMAC generally corresponded to decrease in fMIC, as in GM, and viceversa.

Figure 4: Average fractional maps (AVG) of free water (fFW), micro (fMIC) and macro-vascular (fMAC) diffusion, and the difference of the subsequent time-points (ΔTi) to the average with the two synchronization methods. With both methods, clear changes in fFW were observed in WM, ventricular and peri-ventricular areas. Changes in fMIC were mostly located in GM and deep GM, whereas fMAC mostly changed around major vessels. A time-shift between PPU and self-synchronization was appreciable.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)
0344