0323

A novel biophysical simulation framework for intravascular MRI signals using 3D Vascular Anatomical Networks applied to VASO-fMRI

Grant Hartung^1,2, Daniel Gomez^1,2,3, Avery Berman^4,5, and Jonathan R. Polimeni^1,2,6
¹A.A. Martinos Center For Biomedical Imaging, MGH, Charlestown, MA, United States, ²Department of Radiology, Harvard Medical School, Boston, MA, United States, ³Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States, ⁴Physics, Carleton University, Ottowa, ON, Canada, ⁵Royal Ottowa Mental Health Centre, Ottowa, ON, Canada, ⁶Massachusetts Institute of Technology, Cambridge, MA, United States

Synopsis

Keywords: fMRI Acquisition, Blood vessels, brain, contrast mechanisms, flow, fMRI (task based), gray matter, high-field MRI, in silico, modelling, signal modeling

Motivation: The emerging fMRI method VASO provides improved neuronal specificity compared to BOLD, however the precise interpretation of its origins and principled means to optimize this sequence is not straightforward.

Goal(s): To use biophysical models to investigate the origins of the VASO signal and compare it with direct estimates of CBV.

Approach: We extend our 3D biophysical Vascular Anatomical Network framework to incorporate intravascular signals undergoing inversion recovery to model the VASO sequence.

Results: The VASO signal appears sensitive to slab thickness, and activation biases occur if the slab is too thin. Simulated profiles of VASO differ from measurements, possibly due to model simplifications.

Impact: Our new methodology enables biophysical simulations of fMRI based on inverting blood. Our findings may provide a deeper understanding of the hemodynamic origins of VASO and provide guidance for optimizing SS-SI VASO protocols to yield veridical representation of neural activity.

Introduction

Functional MRI tracks hemodynamic changes associated with neuronal activity. These hemodynamics are shaped by microvascular anatomy, thus fully interpreting fMRI signals requires linking observable fMRI measurements to underlying microvasculature. The most popular contrast is blood-oxygenation-level-dependent (BOLD), which reflects an interplay between blood flow, volume, and oxygenation changes. Other functional contrasts specific to flow or volume may relate more to microvascular dynamics, and appear to provide enhanced neuronal specificity¹. The VAscular Space Occupancy (VASO) approach² measures cerebral blood volume (CBV), and empirically provides sufficient specificity to detect functional differences between cortical layers³. However, the leading VASO approach, slice-selective slab-inversion (SS-SI) VASO^4,5 utilizes inversion pulses to exploit the non steady-state magnetization of fresh inflowing blood. It also requires several assumptions⁴ for interpretation and optimization such as all blood in the voxel being inverted exactly once.

Here we extend a biophysical fMRI model that uses realistic microvascular anatomy and dynamics based on Vascular Anatomical Networks (VANs)^6–8. This framework represents every blood vessel inside a voxel, relating microvascular changes directly to measured fMRI data. While this framework has been applied to interpreting BOLD^6,7,9,10 and VASO¹¹, here we incorporate longitudinal magnetization (the basis of the VASO signal) and blood arrival time (AT, the time for blood to reach the voxel) and transit times (TT, the amount of time between entering the voxel and reaching the individual microvessel segment). This allows us to test the assumptions behind SS-SI VASO and compare its cortical depth response profiles against physiological CBV responses to neuronal activity.

Methods

We extended the VAN framework^6,7 to model longitudinal magnetization, Mz, inversion recovery, and blood AT and TT. The VAN models are reconstructed from a mouse somatosensory cortical voxel^6,12. We impose arterial dilations from in-vivo single-vessel microscopy after 2-s forepaw stimulation representing the vascular response to neuronal activity^13,14. We use the Balloon Model¹⁵ to passively dilate capillaries and veins^6,7.

The SS-SI VASO sequence⁴ uses an inversion pulse (assuming 95% efficiency⁴) followed by two excitation pulses at TI=1348 ms for “blood null” and 1848 ms for “BOLD correction” images and 3000 ms between inversions (TR) (Figure 1).

Simulated blood perfusion is 100 ml/100g/min during baseline (t=0 s post-stimulus) and 176 ml/100g/min during activation (t=3.5 s). We calculated the percentage of incoming flow and corresponding TT of each vessel to account for the mixing of blood with different transit times.

Delivery time (DT) is the sum of the AT and TT. If DT is short (DT<TR), the segment has once-inverted blood. If the DT is long (DT≥TR), the blood is at least twice inverted. The “perfect” slab thickness corresponds to an equal AT and TR (all incoming blood is once-inverted). We also model a “too thin” slab with the final 30% of inlet blood being un-inverted and a “too thick” slab with the first 30% of blood being twice-inverted. We discard the first inversion-recovery period.

Results

Baseline analysis. The perfect slab thickness (Figure 3) with matching AT and TR results in some vessels having a long TT causing a long DT and thus blood with multiple inversions. Segments with multiple inversions are more common with a too-thick slab. A too-thin slab results in a large portion of un-inverted blood.

Activation analysis. During activation, blood velocities increase. This translates to faster AT (a decrease of ~27% in AT, Figure 3). A slab with perfect thickness at baseline results in significant un-inverted blood volume during activation. Increasing baseline slab-thickness by 30% reduces this effect. Conversely, decreasing slab thickness exacerbates this effect. We also show results from an idealized scenario with a dynamic slab (consistently perfect thickness), representing the “best-case” scenario. This case shows little bias between baseline and active states. The cortical depth profiles of each case were compared to CBV (Figure 4). The VASO ΔS profile best matches the ΔCBV profile in Cases 2 and 4, suggesting a thicker VASO slab best reflects CBV changes. None of the VASO profiles match the ΔCBV profile of capillaries only.

Discussion

Our results suggest that the perfect slab thickness results in some blood with nonzero Mz during the blood nulling time. This bias can be reduced by adopting larger thicknesses.

Conclusion

We proposed a model incorporating blood Mz dynamics with a VAN for interpreting the SS-SI-VASO sequence. This model can guide optimizing VASO protocols, and may aid interpretation of other, related techniques such as ASL. Our simulated VASO profiles do not perfectly match those from experimental data, which may suggest model improvements with more accurate physics (e.g., water exchange with tissue^4,16) and/or physiology (e.g., active capillary dilation^4,16).

Acknowledgements

We thank Profs. David Boas and David Kleinfeld for providing their Vascular Anatomical Network models. This work was supported in part by the NIH NIBIB (grants P41-EB030006, R01-EB019437, and R01-EB03274, R01-EB033206), NINDS (grant R01-NS128843) the BRAIN Initiative (NIH NIMH grants R01-MH111419 and F32-MH125599 and NINDS grant U19-NS123717), and the MGH/HST Athinoula A. Martinos Center for Biomedical Imaging.

References

1. Huber, L. et al. Ultra-high resolution blood volume fMRI and BOLD fMRI in humans at 9.4 T: Capabilities and challenges. NeuroImage 178, 769–779 (2018).

2. Lu, H., Golay, X., Pekar, J. J. & Zijl, P. C. M. van. Functional magnetic resonance imaging based on changes in vascular space occupancy. Magn. Reson. Med. 50, 263–274 (2003).

3. Prsa, M., Galiñanes, G. L. & Huber, D. Rapid integration of artificial sensory feedback during operant conditioning of motor cortex neurons. Neuron 93, 929–939 (2017).

4. Huber, L. et al. Slab-selective, BOLD-corrected VASO at 7 Tesla provides measures of cerebral blood volume reactivity with high signal-to-noise ratio. Magn. Reson. Med. 72, 137–148 (2014).

5. Kim, T. & Kim, S.-G. Temporal Dynamics and Spatial Specificity of Arterial and Venous Blood Volume Changes during Visual Stimulation: Implication for Bold Quantification. J. Cereb. Blood Flow Metab. 31, 1211–1222 (2011).

6. Gagnon, L. et al. Quantifying the Microvascular Origin of BOLD-fMRI from First Principles with Two-Photon Microscopy and an Oxygen-Sensitive Nanoprobe. J. Neurosci. 35, 3663–3675 (2015).

7. Hartung, G., Pfannmoeller, J. P., Berman, A. J. & Polimeni, J. R. Simulated fMRI responses using human Vascular Anatomical Network models with varying architecture and dynamics. in (2022).

8. Hartung, G. et al. Voxelized simulation of cerebral oxygen perfusion elucidates hypoxia in aged mouse cortex. PLoS Comput. Biol. 17, e1008584 (2021).

9. Pfannmoeller, J. et al. Simulations of the BOLD non-linearity based on a viscoelastic model for capillary and vein compliance. ISMRM Annu. Conf. (2021).

10. Báez-Yáñez, M. G., Siero, J. C. & Petridou, N. A statistical 3D model of the human cortical vasculature to compute the hemodynamic fingerprint of the BOLD fMRI signal. bioRxiv (2020).

11. Genois, É., Gagnon, L. & Desjardins, M. Modeling of vascular space occupancy and BOLD functional MRI from first principles using real microvascular angiograms. Magn. Reson. Med. 85, 456–468 (2021).

12. Blinder, P. et al. The cortical angiome: an interconnected vascular network with noncolumnar patterns of blood flow. Nat. Neurosci. 16, 889–897 (2013).

13. Tian, P. et al. Cortical depth-specific microvascular dilation underlies laminar differences in blood oxygenation level-dependent functional MRI signal. Proc. Natl. Acad. Sci. 107, 15246–15251 (2010).

14. Uhlirova, H. et al. Cell type specificity of neurovascular coupling in cerebral cortex. Elife 5, e14315 (2016).

15. Boas, D. A., Jones, S. R., Devor, A., Huppert, T. J. & Dale, A. M. A vascular anatomical network model of the spatio-temporal response to brain activation. NeuroImage 40, 1116–1129 (2008).

16. Krieger, S. N., Streicher, M. N., Trampel, R. & Turner, R. Cerebral blood volume changes during brain activation. J. Cereb. Blood Flow Metab. 32, 1618–1631 (2012).

Figures

Figure 1. Overview of our experimental setup. A) Our VASO model applies the inversion pulses over a volume (slab) and then records the voxel from the imaging slice. B) Inversions (black squares) and excitation pulses (black stars) are applied every TR (3 s). We also show the magnetization (Mz) recovery for blood and gray matter with seven points marked. C) These timepoints are then visualized across a VAN showing how the inflow of once-inverted blood replaces twice-inverted blood in some vessels.

Figure 2. Overview of our simulation framework. We apply biophysical simulations to the VAN structure to generate changes in blood flow, volume and pressure which is then used to compute subsequent oxygenation changes and BOLD time course. We also use these flow and volume properties to compute the transit time for every segment. These transit times are then used to compute the magnetization in each vessel.

Figure 3. Mz under four slab thicknesses for (top row) baseline and (bottom row) active states with nulled vessels in grey. Ideally, all vessels are nulled. During activation, feeding arterial diameters expand, generating faster flow and shortening the arrival time, as pictured for Case 1. Case 1 is a perfect slab at baseline, whereas Case 2 is for a slab with an overestimated thickness of 30%, Case 3 is for a slap with underestimated slab thickness of 30%, and Case 4 is an idealized, dynamically-adapted thickness that is optimally chosen at baseline and active states.

Figure 4. Cortical-depth profiles of the VASO signal and ΔCBV responses. (A) The simulated change of the VASO signal (ΔS) is plotted against cortical depth. Case 1 (black) uses slab thickness tuned at baseline while Case 2 (blue) overestimates the slab thickness by 30%, Case 3 (cyan) underestimates the slab by 30% and Case 4 (green) has a perfect slab at baseline and activation states. (B) The change in blood volume, ΔCBV is calculated for reference based on their dilations for all vessels in the VAN (orange solid line) and for capillaries only (dashed orange line).

Proc. Intl. Soc. Mag. Reson. Med. 32 (2024)

0323

DOI: https://doi.org/10.58530/2024/0323