Introduction

Blood-oxygenation-level-dependent (BOLD) fMRI is widely used to study brain function. However, the BOLD-fMRI signal represents an integration of several hemodynamic parameters and therefore provides only indirect information on neuronal functioning[1]. To disentangle hemodynamic parameters -cerebral blood volume(CBV) and oxygen saturation(SO₂)-, studies have explored the use of gas-challenges to manipulate vessel caliber(CBV, modulated by CO₂) and hemoglobin saturation state(O₂)[2,9,10]. Nonetheless, arterial gas manipulations alone are insufficient to understand the relationship between vascular properties and the resulting hemodynamic signals. For this, computational modeling is necessary. This work aims to simulate the BOLD-fMRI signal change for different CBV and SO₂ changes based on virtual vascular networks. We present look-up tables that determine the possible arterial/venous contributions to CO₂/O₂-inspired BOLD-fMRI signal changes for gradient-echo(GE) and spin-echo(SE) at 7T.

Methods

We computed the relative BOLD-fMRI signal change in two vascular models. (1) We simulated susceptibility and diffusion effects in an artificial vascular network(AVN) to validate our computational method. (2) We created a representative realistic vascular network(RVN) to simulate susceptibility effects for arteries and veins. Randomly oriented cylinders represent the AVN for vessel sizes ranging from 1µm-100µm fulfilling a specific volume fraction in an isotropic voxel of 1mm³. Levels of SO₂ ranging from 60-90% were separately imposed with a hematocrit level of 45%. Diffusion motion was simulated with a Monte Carlo method(1µm²/ms) and induced-dephasing was recorded for an ensemble of spins(1E9 spins) for GE-BOLD(TE=27ms) and SE-BOLD(TE=45ms) at 7T[3,4]. We adopted an SO₂ of 60% as basal-state. To study vessel dilation effects, we use a basal CBV of 3%, and simulated CBV changes ranging from -33%(constriction) to 100%(dilation). Only extravascular effects were considered since intravascular contributions are assumed to be negligible at 7T[5]. An RVN was generated based on a modified mouse vascular network[6]; an artery-to-vein ratio of 3:1 was applied simulating the human cerebrovasculature[7]. Vectorial description of the RVN provides detailed information regarding the spatial distribution and vessel properties in an isotropic 1mm³ simulation voxel. The susceptibility-induced inhomogeneous magnetic field was computed assuming a separate contribution of arteries and veins. Variable SO₂ levels for arteries and veins were simulated to examine the impact of each compartment to the BOLD-fMRI signal change. CBV changes in arteries and veins were investigated considering only an increase (dilation) in vessel radius. Based on the Fourier transform of the inhomogeneous magnetic field, we calculated the signal decay for GE-BOLD(TE=27ms) at 7T[8]. Since the Fourier method assumes a static dephasing regime, the contribution of capillaries can be superimposed assuming a weighted-microvascular contribution to the computed BOLD signal change. However, we only considered the contribution of the arterial and venous compartments.

Results

Figure1 shows the simulation obtained with an AVN model for GE (top row) and SE (bottom row). Our computational method reproduces the well-known sensitivities of GE and SE to all and small vessels respectively, as reported in[3,4](Figure1.a;1.d). Simulation of changes in SO₂ for different vessel sizes shows that about 10% or larger SO₂ changes are necessary to saturate the BOLD-fMRI signal change for GE(Figure1.b). For SE, small vessel contribution to the BOLD-fMRI signal is relatively larger due to the refocusing pulse. Hence, larger SO₂ changes are required to obtain a comparable GE-like BOLD-fMRI signal change(Figure1.e). GE has a high sensitivity to CBV changes(Figure1.c) in contrast to SE(Figure1.f). Figure2 displays the look-up tables of the BOLD-fMRI signal change in which several SO₂ levels and CBV changes are simulated for arteries and veins (RVN model). A characteristic non-linear BOLD-fMRI signal behavior is displayed in Figure2.a which can be attributed to the arterial-to-venous ratio. Hence, reduction in arterial SO₂ sets the negative BOLD-fMRI signal change; only with this condition set, the venous SO₂ change also starts to play a role in showing negative BOLD responses. For null or increases in arterial SO₂, the BOLD signal is positive for most of the venous SO₂ changes. Figure2.b displays the impact of the SO₂ and CBV change in the venous compartment while arterial SO₂ and CBV is maintained. Figure2.c shows the absolute difference between the arterial basal-state(Figure2.b) and the dilated arterial compartment(dilation of 12.5%, 25%, 37.5% and 50%), while keeping constant the SO₂(98%), for different CBV and SO₂ changes in the venous compartment. Arterial dilation affects the BOLD signal only when the degree of dilation is considerable, i.e. dilation larger than 25% the intrinsic vessel size.

Discussion / Conclusion

We present look-up tables for the relative BOLD-fMRI signal change for different hemodynamic parameters (i.e. CBV and SO2) to investigate the arterial and venous contribution in CO₂/O₂ gas-challenge fMRI experiment. Thus, we have performed simulations in artificial and realistic vascular models for several SO₂ levels and CBV changes. Notably a non-linear relation between the contribution of arteries and veins is displayed by the RVN model. Larger contributions are obtained from the venous side for small SO₂ and CBV changes. Arterial dilation affects the BOLD signal only when the degree of dilation is considerable, i.e. dilation larger than 25% the intrinsic vessel size. Relative BOLD signal changes (±5-10%) are reported [9,10] under controlled CO2/O2-gas stimulus in fMRI measurements. Given the presented look-up tables it may be possible to infer the vascular contributions to the measured BOLD signals and/or particular hemodynamic basal states

Acknowledgements

This work was supported by the National Institute of Mental Health of the National Institutes of Health under the Award Number R01MH111417

References

[1] Ogawa, S. et al. Functional brain mapping by blood oxygenation level-dependent contrast magnetic resonance imaging. A comparison of signal characteristics with a biophysical model. Biophys. J. 1993; 64:803–812. [2] Faraco CC. et al. The cumulative influence of hyperoxia and hypercapnia on blood oxygenation and R*₂. J Cereb Blood Flow Metab. 2015; 35:2032-42. [3] Kiselev V.G., Posse S. Analytical model of susceptibility-induced MR signal dephasing: effect of diffusion in a microvascular network. Magn Reson Med 1999; 41:499-509. [4] Weisskoff, R., et al. Microscopic susceptibility variation and transverse relaxation: Theory and experiment. Magn. Reson. Med 1994; 31:601–610. [5] Uludag K. et al. An integrative model for neuronal activity-induced signal changes for gradient and spin echo functional imaging. Neuroimage 2009; 48:150-165. [6] Báez-Yáñez M.G. et al. A statistical 3D model of the human cortical vasculature to compute the hemodynamic fingerprint of the BOLD fMRI signal. bioRxiv 2020.10.05.326512 [7] Duvernoy, H.M. et al. Cortical blood vessels of the human brain. Brain Res. Bull. 1981; 7:519–579 [8] Ziener C. et al. Frequency distribution and signal formation around a vessel. Magn Res Mat in Phys, Bio and Med 2005; 225-230 [9] Bhogal A. et al. Investigating the non-linearity of the BOLD cerebrovascular reactivity response to targeted hypo/hypercapnia at 7T. Neuroimage 2014; 98:296-305 [10] Schellekens W. et al. Vascular and neurovascular effects on micro- and macro-vascular compartments in human visual cortex. Proceedings of the ISMRM 2020. Abstract 3831.