Synopsis

Keywords: Data Analysis, Blood vessels

To obtain cerebral vasculature related information, various kinds of cerebrovascular magnetic resonance imaging (MRI) technique are being applied. However, it turned out that cerebral blood volume (CBV) and vessel size index (VSI) values differ depending on their spatial resolutions. To find an adjustment solution for this phenomenon, the resolution dependency must be confirmed in advance. Therefore, this study aims to investigate the resolution dependency of CBV and VSI through whole brain and region of interest based analysis. The results show that micro CBV and VSI values are resolution-dependent, while total CBV did not show any distinct patterns.

Introduction

Non-invasive cerebral perfusion mapping is important for determining perfusion changes due to molecular and neuropathological modifications. Perfusion magnetic resonance imaging (MRI) provides quantitative information related to the cerebral vasculature non-invasively. In particular, cerebral blood volume (CBV) and vessel size index (VSI) are considered important physiological indicators in that they are used to evaluate abnormalities in tissue and vascular reactivity as well as the microvascular structure of tumors¹. These measurements are known to depend on spatial resolution, but to the best of our knowledge they have not been systematically studied. Here, we investigate the resolution dependence of perfusion MRI measured using exogenous contrast agents. We compared rat CBV and VSI for three different spatial resolutions by region-by-region basis. The results of this study will be helpful in quantitatively analyzing the cerebral perfusion values measured at various spatial resolutions.

Methods

Data acquisition
MRI datasets were acquired from 20 rats (Wistar, 210‒400 g, under 2.0‒1.5% isoflurane anesthesia) on a 7 T Bruker scanner. For ΔR₂^* and ΔR₂ mapping, multi-echo gradient-echo (MGE) and multi-slice multi-echo (MSME) pulse sequence were performed before and after the contrast agent administration (300 µmol Fe/Kg of monocrystalline iron oxide nanoparticles (MION) were administered as an intravenous bolus injection). Three different in-plane resolution data (125, 250, and 500 µm² in-plane resolutions with 1.5 mm slice thickness) were measured.
Determination of perfusion parameters
ΔR₂^* and ΔR₂ maps were calculated by using the mono-exponential and decay equation with one fixed echo time (ΔR₂^*,ΔR₂ = 1/TE · ln(DATA_pre/DATA_post))². ΔR₂^* and ΔR₂ were normalized to the total CBV and micro CBV, and VSI was calculated by the proportional equation (VSI ≈ (ΔR₂^*/ΔR₂)^3/2) ³. CBV and VSI maps of each animal were then registered to the Waxholm Space (WHS) rat brain atlas^4,5 via Advanced Normalization Tools (ANTs) registration software⁶. The co-registered CBV and VSI maps of 3 different resolutions were used for the whole brain and ROI analysis (cortex, corpus callosum, striatum, thalamus, and hippocampus). The declination of ΔR₂^* and ΔR₂ due to washout of the contrast agent was corrected based on the general assumption that ΔR₂^* is linearly proportional to the tissue contrast agent concentration.

Results

Figure 1 shows the voxel-wise mean total CBV, micro CBV, and VSI maps of twenty rats (n=20). Differences in perfusion values in cortical and white matter regions are clearly distinguished. The white matter corpus callosum showed low perfusion values, while some brain regions had high perfusion values possibly due to the high sensitivity of gradient-echo imaging to large vessels. In both total CBV and micro CBV maps, the lower the spatial resolution, the wider the area with the lower perfusion value at high resolution. On the other hand, in the VSI map, as the resolution decreased, regions with high values tended to spread.
The difference map for each resolution is shown in Figure 2. In the difference map, positive values indicate that the CBV and VSI increase with decreasing resolution, and negative values indicate that the values decrease with decreasing resolution. Total CBV tended to decrease as the brain in-plain resolution decreased from 0.125 µm² to 0.250 µm². On the other hand, when the in-plane resolution was decreased from 0.125 µm² to 0.500 µm² and from 0.250 µm² to 0.500 µm², an increasing portion and a decreasing portion were mixed. As the in-plain resolution decreased, micro CBV showed a tendency to decrease overall in the brain, and VSI showed a tendency to increase overall.
Figure 3 shows the whole brain and selected ROIs (left side) and boxplot analysis results for the ROIs (right side). The Kruskal-Wallis test was used to calculate the p-value for each difference in the three resolution data. All p-values were close to 0, except for one case of total CBV in the thalamus, indicating that the change in the value according to the resolution change was statistically significant. The median of the total CBV showed a tendency to gradually decrease as the resolution decreased and showed a tendency to increase exceptionally in the corpus callosum region. The median value of micro CBV showed a tendency to decrease as the resolution decreased and decreased faster than the total CBV. As a result, the median value of VSI showed a tendency to increase as the resolution decreased in all ROIs.

Conclusion and discussion

In this study, the resolution dependence of total CBV (ΔR₂^*), micro CBV (ΔR₂) and VSI (ΔR₂^*/ΔR₂) maps was investigated. Both total CBV and micro-CBV (except for the corpus callosum) tended to decrease with decreasing resolution throughout the brain, with micro-CBV decreasing more rapidly. As a result, VSI showed a tendency to increase as the resolution decreased. This may be a result of changes in the distribution, angle, or other structural parameters of blood vessels within a voxel as the size of the imaged voxel increases as the resolution decreases. For further investigation, additional ROI analysis or Monte-Carlo simulations taking into account the vascular characteristics of the region will be required. Increasing the resolution of CBV and VSI using the deep learning network SRGAN (Super-Resolution Generative Adversarial Network)⁷ is effective in solving the resolution-dependent problem by correcting underestimated and overestimated values.

Acknowledgements

This work was supported by the Korea Health Industry Development Institute by the 2019 Research Fund (HI18C0713).