Measuring changes in neonatal cerebral blood flow (CBF) and microvessel density are critical tools for studying neurodevelopment. Post-mortem studies have shown changes in microvessel density over the first two years of life¹, but are difficult to measure. Existing noninvasive measurement techniques for neonatal CBF and volume (CBV) include arterial spin labeling (ASL)². ASL suffers from difficulty in obtaining white matter (WM), gray matter (GM) and basal ganglia (BG) relative CBF measurements in the same acquisition without measuring transit delays and re-tuning acquisition parameters³.

The intravoxel incoherent motion (IVIM) model provides a way to noninvasively characterize microvascular blood volume and flow in the brain with diffusion weighted MRI (dMRI)⁴. Compared to other methods for measuring blood flow in the brain, such as ASL, IVIM can characterize both blood volume, through the perfusion fraction, f, and flow, by the pseudodiffusion coefficient, D*, while modeling tissue through the diffusion coefficient, D.

$$ S=S_0 [f\ exp(-bD^*)+(1-f)\ exp(-bD)] $$

As IVIM is based on the motion of blood in vessels rather than bulk transfer of blood like ASL, it is suggested that IVIM reflects microvessel geometry and volume as well as flow^4,5.

In addition, IVIM is applied by fitting images from dMRI techniques, and thus can leverage advanced acquisition techniques that are available on clinical scanners, such as simultaneous multislice (SMS) acquisitions. Furthermore, acquisition can be integrated into protocols for diffusion tensor imaging sequences with very minor protocol modifications.

A cohort of 12 healthy volunteer infants were scanned at 2 and 14 weeks post-natal in compliance with IRB guidelines with written and signed informed consent. An SMS sequence was used to acquire dMRI images with TE = 104ms, TR = 3.4s, FA = 90, and 2mm isotropic resolution with multiband factor 2⁶. 5 subjects were included after discarding 7 subjects for excessive motion. Time point 1 (TP1) scans occurred at (25.2 +/- 7.8) days and time point 2 (TP2) scans occurred at (99 +/- 12.7) days. Seven b-values were acquired between .5- 700 s/mm2 with gradient direction <1,1,1> in 33s. The IVIM model was fitted using a biexponential model with a maximum penalized likelihood estimation cost function⁷ minimized in MATLAB R2015a. Brain extracted images and subject specific atlases with WM, GM and BG labels were made longitudinally over both time points from fractional anisotropy data, T1-weighted, and T2-weighted scans using iBEAT^8-10 and BET¹¹. IVIM parameters were averaged inside of each tissue class to ensure normality for statistical analysis.

Perfusion fraction, pseudodiffusion coefficient and diffusion coefficient maps were estimated for each subject and each time point. In addition, fD* maps were calculated from the product of f and D* maps to evaluate the blood flow. Figure 1 shows representative maps for the two time points from a single subject.

Pairwise t-tests were performed to assess longitudinal changes in the IVIM parameters for the different tissue classes across both time points (Minitab Ver. 17) as listed in Figure 2. A significant decrease (p<.05) (1.12%) was observed in f between TP1 and TP2 in WM but not GM nor BG. Significant changes (p<.01) were found in the D in all three tissue classes between TP1 and TP2. Significant changes were not seen in the D* coefficient for WM, GM, or BG. Separately, one-way ANOVA tests were run to detect differences between tissue classes inside of each time point and shown in table form in Figure 2 and graphically in Figure 3.

The longitudinal decrease observed in perfusion fraction in WM may be associated with the microvessel density in the WM dominating over the GM at 40 weeks gestation as compared to developmental trend of GM microvessel density dominating at two years¹. Higher blood flow, as reflected by fD*, was seen in GM vs WM although significant longitudinal changes were not seen in this study. The BG showed an intermediate level of blood flow that agrees with another study examining ASL perfusion measures on neonates¹².

Near significant longitudinal changes in the flow related parameters fD* and D* were observed in the subjects suggesting a follow up study with more subjects may be warranted. IVIM is a feasible measure of perfusion in neonates in a very short scan time, with the caveat that sensitivity to motion corruption is high, as with all blood flow imaging. Increasing the number of b-values acquired or incorporating repeated measures may lead to a higher success rate through increased redundancy by rejecting motion corrupted volumes while obtaining acceptable data fits.