Intravoxel incoherent motion imaging (IVIM) has been implemented in various organs to measure perfusion and diffusion simultaneously with promising results, but its application in the brain is still lagging behind, probably due to the small cerebral perfusion fraction resulting in low SNR. The basic assumption of IVIM is that it contains two compartments: a slow compartment D, where the signal decays slowly as a function of b-value due to Gaussian diffusion; and a fast compartment D*, where the signal drops much faster as a function of b-value due to pseudo-random capillary blood flow ¹. A major concern of the validity of this model is that a distribution of capillary velocities as well as non-random orientation could result in a more complex relation than the single D* assumption.

By employing an ASL-preparation module before an IVIM-readout (ASL-IVIM) the arterial perfusion contribution can be isolated. In this way, the arterial IVIM signal was studied as a function of post-labeling delay (PLD), and compared to the results from conventional IVIM.

Six volunteers were scanned on a 3T MRI scanner (Philips Healthcare). For ASL-IVIM, ASL-preparation was achieved by time-encoded pseudo-continuous ASL (te-pCASL) which enables temporal ASL-data acquisition in a time-efficient manner ^2,3. te-pCASL was performed with a T₁-adjusted 12-block-hadamard matrix (894,579,429,340,283,241,211,187,168,153,115 ms, interval between labeling- and readout-module 49ms), IVIM readout was achieved by a bipolar-gradient corresponding to 8 b-values (0,4,8,12,15,20,80,168 s/mm² in the phase-encoding direction; 6 repeats per b-value; ~40min; single-shot EPI, 3.75*3.75*7mm³). For conventional IVIM 30 b-values were acquired (0,5,7,10,20,30,40,50,60,70,80,90,100,120,140,160,180,200,250,300,350,400,450,500,550,600,650,700,750,800 s/mm² in the phase-encoding direction; 10 repeats per b-value; ~10min; single-shot spin-echo DWI, 3.75*3.75*7mm³).

Raw images were first eddy current corrected and subsequently motion corrected in FSL (FMRIB, Oxford, UK). For ASL-IVIM, the images were subsequently Hadamard-decoded in Matlab. The ASL signal as a function of b-value was fitted to a mono-exponential model: S(b)/S₀=f•exp(-b•D*). For conventional IVIM, the signals with b>=200s/mm² were fitted to a mono-exponential model to estimate the diffusion coefficient D. Subsequently, the signal for all b values were fitted to a bi-exponential model ⁴: S(b)/S₀=f•exp(-b•D*)+(1-f)•exp(-b•D). D: diffusion coefficient; D*: pseudo-diffusion coefficient, f: perfusion fraction, S₀: signal intensity at b=0. IVIM-perfusion signal was isolated by subtracting the mono-exponential signal of the slow-diffusion compartment (D) from the total signal.

Figure 1 shows the Hadamard decoded perfusion weighted images corresponding to the different sub-boli (each having different labeling durations and PLD) for different b-values. Figure 2 shows the normalized gray matter perfusion-signal (signal of a sub-boli divided by the corresponding labeling duration) as a function of PLD, as well as how this signal varies for different b-values. Signals with very short PLD were neglected considering the low SNR. Figure 3 shows for the different PLDs, the pseudo-diffusion coefficient D* (a) and natural logarithm of D* (b) as calculated from the ASL-IVIM signal. The quantified values from conventional IVIM were found to be: f = 0.10±0.02; D = 0.85±0.06 (10^-3 mm²/s); D* = 10.1±1.2 (10^-3 mm²/s). Figure 4 shows fitted signal of the fast compartment as a function of b-value for different PLDs acquired by ASL-IVIM and IVIM, respectively.

D* as calculated from ASL-IVIM data was found to be highly dependent on the PLD. Very high D*-values were observed for PLDs<883ms, probably due to vascular signal, which is in good agreement with the ATT in gray matter. As the labeled blood travels further down the vascular tree, D* decreases gradually (Figure 3) suggesting that more-and-more of the signal originates from the microvasculature. Interestingly, the signal from the conventional IVIM scan shows an intermediate D*-value corresponding to the ASL-IVIM-signal for a PLD of approximately 1750ms. This might indicate that the IVIM signal does not only originate from the microvasculature, but also contains a certain amount of vascular signal. An alternative explanation, would be that for larger PLDs some of the ASL-signal has already extravasated into the extravascular compartment and therefore reflects tissue diffusion. However, the fact that the observed D* at these PLDs plateaus at a factor 3~4 higher than the diffusion coefficient of the slow compartment seems to be incompatible with this explanation.

One limitation of this study was that all gradients for IVIM readout module were all along a single direction (the phase-encoding direction), which may underestimate the complexity of the blood flow in the capillaries. Furthermore, with ASL only the arterial component of the IVIM-signal can be studied.

ASL-IVIM was successfully implemented to investigate the perfusion and diffusion characteristics of the arterial blood flow showing a much more complicated process than conventionally assumed in IVIM.