Synopsis

Keywords: Diffusion/other diffusion imaging techniques, Diffusion/other diffusion imaging techniques, FEXI, trans-membrane exchange

We explored how compartmental anisotropy affects filter-exchange imaging (FEXI) data in white matter by applying different types of tissue water magnetization filters, including multi-orientation and spherical tensor encoded diffusion filters, as well as a T2 filter. The results show that, depending on the mutual orientation of the filter and detection gradient pairs relative to fiber orientation, filtering efficiencies can be positive or negative. We conclude that the fast compartment reduced during diffusional filtering need not be extracellular, but can also reflect intra-axonal signal losses with effect size determined by the diffusion tensor components perpendicular and parallel to the fibers.

Introduction

$$${\;\;\;\;}$$$Trans-membrane exchange can be an important indicator of tissue functioning. To quantify cellular trans-membrane exchange using MRI, Lasic et al.¹ proposed the filter-exchange imaging (FEXI) technique for measuring exchange between slow and fast diffusion compartments. Based on in-vitro cell experiments^2,3, the FEXI method assumes that diffusion is slower inside the cell than outside and that the diffusion filter suppresses the extracellular compartment. Apparent exchange rates (AXRs) measured by FEXI have been reported to distinguish abnormalities in tumor tissue⁴ and subtypes^5,6.
$$${\;\;\;\;}$$$Recently, it was demonstrated that AXR is affected not only by filter strength⁷ but also diffusion anisotropy, especially in white matter^8,9. This indicates that diffusion tensor components may be relevant in addition to diffusion compartments. Here, we explore how compartmental anisotropy affects FEXI results in white matter by applying different filtering and detection gradient orientations. We also investigate the effect of T2 filtering alone, i.e., with zero diffusion filter strength.

Methods

$$${\;\;\;\;}$$$The FEXI signal from a double-pulsed-gradient-spin-echo (PGSE) sequence (Fig. 1a) is^1,4:$${S(b_d,t_m)\;=\;S'(t_m)exp(-b_d{\cdot}ADC'(t_m))\;\;\;\;\;\;[Eq.1]\\ADC'(t_m)\;=\;ADC\cdot(1-{\sigma}exp(-t_m{\cdot}AXR)\;\;\;\;\;\;[Eq.2]\\{\sigma}\;=\;(D_{fast}-D_{slow})(f_{fast}^{eq}-f_{fast}^0)/ADC\;\;\;\;\;\;[Eq.3]\\}$$ $$${ADC',\;t_m,\;{\sigma},\;and\;f_{fast}^{eq/0}\;}$$$indicate ADC measured during detection (Fig. 1b), mixing time, filtering efficiency, fast-/slow-diffusion coefficients, and equilibrium/filtered fractions of the fast-diffusing compartment, respectively. To estimate ADC’(t_m), diffusion-weighted signals were voxel-wise fitted to Eq. 1; and AXR were estimated using Eq.2 (Fig.1).
$$${\;\;\;\;}$$$Human studies were approved by the local IRB; all volunteers signed informed consent. Data were acquired at 3T (Philips Ingenia-Elition-RX, 95mT/m) with TE=45ms, 5mm single-slice, resolution=3×3mm², relaxation-delay=4000ms, 4 averages, σ_f/σ_d/Δ_f/Δ_d=13.2/9.8/16.2/21.9ms, b-value for diffusion filter (b_f)=900s/mm², b-value for detection (b_d)={10,300,600,900}s/mm²; ={15,100,200,400}ms. Detection gradient (G_d) was applied along X,Y, and Z for each filter gradient direction (G_f =X,Y,Z), giving 9 combinations (Fig.1e). Other experiments included spherical-b-tensor encoding for trace-weighted diffusion filtering (b_f=900s/mm²)^10,11 (Fig.4c), and omitting G_f for T2 filtering only (Fig.4e), both acquired with separate orthogonal diffusion readouts (G_d=X,Y,Z) to generate trace-weighted images.
$$${\;\;}$$$Images from different acquisitions were co-registered linearly. Trace images from the same G_f and G_d directions⁴ (FEXI_same;Fig.1f) and individual X, Y, Z G_f directions (FEXI_X/Y/Z; Fig.1g-i) were generated by averaging DWIs acquired from three G_d directions and used for FEXI fitting.
$$${\;\;\;\;}$$$ROI analysis was performed in cortical spinal tract, CST, for coherently oriented white matter fibers along Z-direction (defined as having an angle <15° between the principal eigenvector and Z)¹².

Results

$$${\;\;\;\;}$$$When reconstructing FEXI using FEXI_same, frontal white matter results (AXR = 1.3 s^-1 and σ = 0.24; Fig.1) are comparable to the literature.
$$${\;\;\;\;}$$$Fig.2 shows FEXI parameters for CST where fiber is parallel to Z direction. Importantly, individual combinations of G_f and G_d directions show strong effects of tissue diffusional anisotropy. When applying G_f and G_d with the same direction relative to the fiber orientation (i.e. both perpendicular or both parallel), σ values are positive as expected [Eq. 3], while five of six AXR values are small (0-0.7s^-1, Fig.2a), agreeing with a previous report⁹. Interestingly, when the direction relative to the fiber differs between G_f and G_d, negative σ values are measured consistently (Fig. 2a). Since, according to the FEXI model, σ reflects the relative saturation of the fast and slow diffusion components, this important observation would indicate that the slow diffusion component is filtered out more strongly than the fast component. This seemingly “impossible” effect is direct experimental evidence that, at least for FEXI in white matter, the description needs to be adjusted to account for the signal loss due to selective filtering of diffusion tensor components of at least two restricted compartments with a different diffusion tensor. This interpretation is supported by small σ-values when using trace-weighted detection (σ < 0.07) (Fig.2b). Also, when studying DWIs, filtering along the fiber direction shows strong signal reduction, indicating more filtering of intra-axonal signal than extra-cellular (Fig.3).
$$${\;\;\;\;}$$$Fig.4 compares three filter types, all with trace-weighted (X,Y,Z-averaged) detection: (a,b) PGSE-trace-weighted diffusion-filter (FEXI_same); (c,d) spherically-encoded diffusion-filter; (e,f) T2-filter. Regardless of filter type, the results show a comparable ADC' dependence on t_m (ΔADC'=0.02-0.04μm²/ms) despite the strongly varying filter efficiencies σ (Fig.4). For the single-shot spherical-tensor-encoded trace-filter, σ is small compared to the PGSE-trace-filter based on averaging over separately applied gradient directions (X,Y,Z) (Figs.4b vs. d), in line with predictions by Khateri et al.¹³ These data suggest that to obtain measurable FEXI exchange in white matter, it is necessary to apply anisotropic diffusion filters to label the two compartments separately.
$$${\;\;\;\;}$$$The T2-filtering efficiency was negligible, i.e. ADC’ values with and without filter were equal, but ADC' still increased as a function of t_m (Fig.4f). Such an increase may indicate removal of slower diffusing components within a compartment, possibly due to faster T1 decay during t_m. This ADC’ evolution observed after applying a T2 filter of length equal to the echo time of the diffusion filter is likely to strongly affect AXR estimations.

Conclusion and discussion

$$${\;\;\;\;}$$$We explored the effects of diffusion anisotropy and T2-filtering on FEXI parameters in white matter. The results indicate that FEXI experiments in white matter need to account for and exploit differences in anisotropy between the exchanging diffusion compartments. FEXI measurements in the diseased brain need to be interpreted carefully because the physical and cellular changes in a pathological condition can alter the contribution diffusion anisotropy in tissue.

Acknowledgements

No acknowledgement found.