The influence of draining veins on apparent grey matter volume changes caused by hypercapnia
Julia Huck1, Christopher J. Steele2,3, Anna-Thekla Jäger3, Audrey P. Fan4, Sophia Grahl3, Christine L. Tardif5,6, Uta Schneider3, Arno Villringer3, Pierre-Louis Bazin3,7, and Claudine J. Gauthier1,8

1Physics, Concordia University, Montreal, QC, Canada, 2Psychology, Concordia University, Montreal, QC, Canada, 3Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 4Stanford University, Stanford, CA, United States, 5Biomedical Engineering, McGill University, Montreal, QC, Canada, 6Montreal Neurological Institute, Montreal, QC, Canada, 7Social and Behavioural Sciences, University of Amsterdam, Amsterdam, Netherlands, 8Montreal Heart Institute, Montreal, QC, Canada


Past studies have shown that T1-weighted measures of grey matter volume (GMV) can be biased by differences in blood volume. Here, we investigate the vascular compartments associated with this bias by quantifying the spatial relationship between t-values for the apparent GMV increase observed during hypercapnia and the location of draining veins. Draining veins were identified using the VENAT atlas. Overall, the results of this analysis demonstrate that while proximity to veins is related to the presence of higher t-values (larger apparent GMV change during hypercapnia), large veins themselves are unlikely to be the main cause of this bias; suggesting that smaller veins or arteries may have a larger role in the observed bias.


T1 weighted images are used in morphometric studies to identify differences in grey matter volume (GMV) and thickness across groups or following an intervention. However, the T1 weighted contrast has been shown to be influenced by differences in blood volume and flow1. Previous work has demonstrated increased blood volume and flow due to CO2-induced vasodilation causes overestimation of GMV1. The precise vascular compartments responsible for this bias are at presently unclear, however, vascular atlases can be used to identify the contribution of specific vascular compartments. Here, we use the recently released VENAT atlas2 to investigate the impact of draining veins on this blood-volume bias in GMV estimation.


GMV differences between air breathing and hypercapnia (5% CO2), were identified in 12 young healthy participants. Images were acquired on a 7T Siemens MR system (1mm³ MP2RAGE: TI1/TI2/TR/TE=900/2750/5000/2.35ms, α1/α2=5°/3°, Grappa=3). Participants breathed each gas (air or 5% CO2) starting 30s before and during each acquisition. The UNI images of Air and CO2 were rigidly co-aligned, segmented, and non-linearly registered to the MNI152 template using the Computational Anatomy Toolbox (CAT12) within SPM12. GMV was calculated3 and smoothed (5mm kernel). Differences between Air and CO2 were assessed with randomise (5000 permutations) using threshold-free cluster enhancement (TFCE)4 in SPM12. The VENAT atlas was created from 20 healthy volunteers who were scanned 5 times at 7T. Flow-compensated 3D multi-echo gradient echo images were acquired with TR/TE1/TE2=29/8.16/18.35ms; Grappa=3; 0.6mm³. The atlas was created from 3D segmentations of the venous vasculature. The vessels were segmented5 on reconstructed6 quantitative susceptibility maps (QSM) and co-registered with ANTs 7 across the 5 scans. Participant averages were co-registered and used to create a population partial volume map of the veins, which was then binarized (threshold>=0.15) and converted to a distance map. We explored the link between draining vein location and GMV estimation bias during hypercapnia by correlating the distance to the closest vein (VENAT) with t-statistics for the CO2>Air comparison (GMV).


The overlay of the venous atlas with t-values shows that regions with high t-values are more often close to veins, and regions further away tend to show lower t-values (Fig. 1). Fig. 2 shows that while there is a wide range of vein diameter and t-value combinations, there is a tendency for larger t-values closer to veins (r=-0.11). Calculating the median t-value for different distance ranges revealed that the t-values were lower for distances close to veins (0-3mm), peaked at distances between 3-5mm, and continued to decline with greater distance (r=-0.97, p<0.05 for distances from 5-27mm) (Fig. 3). To understand the direct contribution of veins at distance 0, a correlation of the t-values to vein diameter is shown in Fig. 4, where vessels with smaller diameter have a larger t-value (r=-0.45, p<0.05).


Fig. 1 shows that most veins do not directly overlap with higher t-values. However, results from Fig. 3 show that higher t-stats tend to occur within 3-5mm and Fig. 4 shows that smaller diameters are associated with larger t-values. These results are consistent with a more prominent role of smaller veins in changes to blood volume that lead to the GMV bias observed. One limitation of the current analysis is that smaller veins are not typically consistent enough across participants to be included in the multi-participant atlas. Thus with the current data we cannot assess this effect of smaller veins. Further studies should create average diameter maps across subjects that include the smaller veins from the high-quality single participant averages that are currently missing in the VENAT atlas to address this question. Furthermore, as arteries are the main locus of vasodilation8, and given the local excitation used for the T1-weighted acquisition at 7T, it is possible that arteries play a larger role in this vascular bias. Future work could also use a recently-published arterial atlas9, in combination with the venous atlas used here to further probe these effects.


Previous work has shown that T1-weighted measures of GMV may be biased by the vasculature, creating problems for comparisons between age groups and with diseases that affect the vasculature. Here we sought to identify the role of the venous compartment in this bias. Our results are consistent with the hypothesis that while changes in venous blood volume may affect GMV measurements, this bias is most likely to occur in smaller veins or arteries. Future work should seek to measure the contributions of these compartments using complementary vascular atlases.


The authors thank Domenica Wilfling and Elisabeth Wladimirov for their help with data acquisition and logistics of the multi-modal plasticity initiative (mMPI) dataset, and Dr. Andreas Schäfer for help with the sequence parameter choice. This work was supported by the Alexander von Humboldt Foundation (C.J.G), the Fonds de Recherche Québécois en Santé (C.J.G), the Canadian National Sciences and Engineering Research Council (RGPIN-2015-04665, C.J.G.), the Heart and Stroke Foundation of Canada (N.I.A. C.J.G.), the National Institute of Health (1K99NS102884, A.P.F.) and the Quebec Bio-Imaging Network (QBIN) for the scholarship for Training course abroad (J.H.).


  1. Tardif CL, Steele CJ, Lampe L, et al. Investigation of the confounding effects of vasculature and metabolism on computational anatomy studies. Neuroimage. 2017; 149:233–243. doi: 10.1016/j.neuroimage.2017.01.025.
  2. Huck J, Wanner Y, Fan AP, et al. High resolution atlasing of the venous brain vasculature from 7T quantitative susceptibility. 2008; bioRxiv 444349. doi: 10.1101/444349.
  3. Ashburner J, Friston KJ. Unified segmentation. Neuroimage. 2005; 26:839–851. doi: 10.1016/J.NEUROIMAGE.2005.02.018.
  4. Smith SM, Nichols TE. Threshold-free cluster enhancement: Addressing problems of smoothing, threshold dependence and localisation in cluster inference. Neuroimage. 2009; 44:83–98. doi: 10.1016/J.NEUROIMAGE.2008.03.061.
  5. Bazin P-L, Plessis V, Fan AP, et al. Vessel segmentation from quantitative susceptibility maps for local oxygenation venography. In: 2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI). IEEE. 2016; pp 1135–1138
  6. Bilgic B, Fan AP, Polimeni JR, et al. Fast quantitative susceptibility mapping with L1‐regularization and automatic parameter selection. Magnetic resonance in medicine. 2014 Nov 1;72(5):1444-59.
  7. Avants BB, Epstein CL, Grossman M, et al. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Medical image analysis. 2008 Feb 29;12(1):26-41.
  8. Girouard H, Iadecola C. Neurovascular coupling in the normal brain and in hypertension, stroke, and Alzheimer disease. J Appl Physiol. 2006; 100:328–335. doi: 10.1152/japplphysiol.00966.2005
  9. Bernier M, Cunnane SC, Whittingstall K. The morphology of the human cerebrovascular system. Hum Brain Mapp. 2008; doi: 10.1002/hbm.24337.


Overlay of the CO2>Air t-value maps and the binary mask of the VENAT atlas.T-statistics are in semi-transparent red overlaid on VENAT atlas in white. While larger veins do not overlap directly with higher t-values, veins are typically located close to high t-statistics regions.

2D histogram of the CO2>Air t-values against distance in mm from the closest vein. The red line denotes the correlation between the CO2>Air t-values and distance. The distance histogram at the top of the figure is displayed using a logarithmic scale.

Median of CO2>Air t-values for different 2mm distance bins. This figure shows lower t-values within 3mm of veins and a peak in t-values at distances of 3-5mm. This may indicate that arteries or smaller veins are more likely to be involved in the GMV vascular bias.

Correlation between CO2>Air t-values and vein diameter (r=-0.45). This figure emphasizes the role of smaller veins with large t-statistics in the significant range being mainly associated with veins of diameter less than 1mm.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)