3772
Monte Carlo simulation of VASO fMRI from real microvascular angiograms of the mouse cortex1Université Laval, Quebec, QC, Canada, 2Massachussets General Hospital, Boston, MA, United States, 3University of California, San Diego, San Diego, CA, United States, 4Boston University, Boston, MA, United States
Synopsis
The vascular space occupancy (VASO) fMRI method probes changes in cerebral blood volume (CBV) but subsequent BOLD response contaminates the VASO signal reducing its SNR. Quantifying this BOLD contamination for different pulse-sequence parameters would improve our ability to measure CBV changes and optimize their quantification in a variety of physiological states and diseases. We have implemented a Monte Carlo method to model such effect from experimental cortical angiograms measured with two-photon microscopy on mice brains. Our results indicate that at 3T the BOLD contamination is relatively low (below 10%) but that this contamination is much higher at 7T and above.
Introduction
The vascular space occupancy (VASO) fMRI method1 probes changes in cerebral blood volume (CBV) under various physiological states, including neuronal activation in humans. To achieve this task, a careful choice of the pulse-sequence parameters must be taken because the blood oxygen-level dependent (BOLD) fMRI signal contaminates the VASO signal and reduces our ability to assess these CBV changes2. Quantifying this BOLD contamination for different pulse-sequence parameters would improve our ability to measure CBV changes and optimize their quantification in a variety of physiological states and diseases. However, this task is challenging and requires a detailed knowledge of the cerebral vascular geometry of the MRI voxel. Towards this end, two-photon microscopy can provide high resolution 3D images of the cortical microvasculature in mice (i.e. angiograms). Combined with Monte Carlo simulations of proton diffusion, these angiograms can further be used to model quantitatively the extravascular BOLD response3. In this work, we extend our previous model to include the VASO effect i.e. the decrease in the number of extravascular protons arising from the dilation of blood vessels during neural activation. Our new model allows to (1) accurately quantify the BOLD contamination in the VASO response and to (2) accurately extract the intra- and extravascular contributions to the BOLD response.Methods
A schematic overview of the modeling procedure is presented in Fig. 1 and the details of the processing of the experimental two-photon microscopy data can be found in our previous publication3. Compared to our previous approach, a substantial difference relates to the oxygen distribution of the microvascular compartments. In the current work, the oxygen content was populated homogenously for arterioles, capillaries and veinules from tabulated experimental data, both for baseline and for the activated states4. The volume change of each vascular compartment was also taken from the literature5. This method allowed us to extend the model to the full field-of-view of the two-photon angiograms, and especially to include all of the large pial vessels that strongly contribute to the BOLD-VASO signals. Oxygen distribution of these vessels with the previous modeling approach was less accurate due to shadowing.
Our previous Monte Carlo model3 was modified by normalizing the extravascular BOLD signal using the same number of protons in the baseline and in the activated state. The loss in signal due to the decrease in the number of extravascular protons in the activated state was interpreted as the VASO response (see Fig. 2 for details).
Results
Simulated VASO responses averaged across N=5 angiograms are shown in Fig. 3 as a function of TE for both Gradient Echo (GRE) and Spin Echo (SE). The averaged initial loss in signal due to the VASO effect was -0.5 %. As TE increases, the BOLD effect, which is of opposite sign compared to the VASO signal, reduces the amplitude of the VASO signal. At 3T, up to TE=6 ms, this BOLD contamination is relatively small i.e. under 10% for GRE and under 2% for SE. This BOLD contamination increases with B0-field strength. At 7T, 25% of the VASO signal is lost at TE=6 ms for GRE and 10% for SE. At very high field (14T), the BOLD contamination is much stronger. At TE=6 ms, 100% of the VASO response is killed with GRE and the response is reduced by 50% with SE.
Compartmental contribution of the extravascular BOLD response for different B0-field strengths is shown in Fig. 4. Good agreement was obtained between our simulated results and experimental values provided in the literature6-8. Our results indicate that at 3T, 45% of the signal originates from the extravascular compartment for GRE and this value increases to 55% for SE.
Discussion
The higher contamination with increasing B0-field strength is explained by the stronger BOLD response at higher B0 for the same VASO effect. This result has important implications for the development of VASO fMRI at ultra high field fMRI (>7T). It shows that TE must be kept very short in order to avoid a significant reduction in the amplitude (and therefore SNR) of the VASO response.Conclusion
Our results provide a framework to accurately model BOLD contamination in the VASO signal for different sequence parameters and different B0-field strengths. Such information is valuable to optimize the pulse sequence timing in human VASO and BOLD fMRI, leading the way to a wider application of these fMRI techniques in healthy and diseased brain. With the development of fMRI at higher B0 fields, our results demonstrate that care must be taken when choosing sequence parameters to keep BOLD contamination at minimum.Acknowledgements
This work was funded by the Natural Sciences and Engineering Research Council of Canada (NSERC/CRSNG) and the Québec Bio-Imaging Network (QBIN/RBIQ).References
1. Lu H, Golay X, Pekar JJ, Van Zijl PCM, Functional magnetic resonance imaging based on changes in vascular space occupancy, Mag. Res. Med. 2003;50(2) 263-274.
2. Lu H, Hua J, van Zijl PCM, Non-invasive functional imaging of Cerebral Blood Volume with Vascular-Space-Occupancy (VASO) MRI. NMR biomed. 2013;26(8) 932-948.
3. Gagnon L, Sakadžić S, Lesage F, Musacchia JJ, Lefebvre J, Fang Q, et al. Quantifying the microvascular origin of BOLD-fMRI from first principles with two-photon microscopy and an oxygen-sensitive nanoprobe. J Neurosci. 2015;35(8):3663‑75.
4. Sakadžić S, Mandeville ET, Gagnon L, Musacchia JJ, Yaseen MA, Yucel MA, et al. Large arteriolar component of oxygen delivery implies a safe margin of oxygen supply to cerebral tissue. Nat Commun. 2014;5:5734.
5. Drew PJ, Shih AY, Kleindeld D, Fluctuating and sensory-induced vasodynamics in rodent cortex extend arteriole capacity, PNAS 2011;108(20), 8473-8478.
6. Duong TQ, Yacoub E, Adriany G, Hu X, Ugurbil L, Kim SG, Microvascular BOLD contribution at 4 and 7 T in the Human Brain: Gradient-Echo and Spin-Echo fMRI with suppression of blood Effects, Mag. Res. Med. 2003;49:1019–1027.
7. Jochimsen TH, Norris DG, Mildner T, Moller HE, Quantifying the Intra- and Extravascular Contributions to Spin-Echo at 3T, Mag. Res. Med. 2004;52:724 –732.
8. Boxerman JL, Bandettini PA, Kwong KK, Baker JR, Davis TL, Rosen BR, et al. The intravascular contribution to fMRI signal change: Monte Carlo modeling and diffusion-weighted studies in vivo. Mag. Res. Med. 1995;34(1):4‑10.
Figures
