Age-related differences in human cortical microstructure depend on the distance to the nearest vein

Abstract Age-related differences in cortical microstructure are used to understand the neuronal mechanisms that underlie human brain ageing. The cerebral vasculature contributes to cortical ageing, but its precise interaction with cortical microstructure is poorly understood. In a cross-sectional study, we combine venous imaging with vessel distance mapping to investigate the interaction between venous distances and age-related differences in the microstructural architecture of the primary somatosensory cortex, the primary motor cortex and additional areas in the frontal cortex as non-sensorimotor control regions. We scanned 18 younger adults and 17 older adults using 7 Tesla MRI to measure age-related changes in longitudinal relaxation time (T1) and quantitative susceptibility mapping (QSM) values at 0.5 mm isotropic resolution. We modelled different cortical depths using an equi-volume approach and assessed the distance of each voxel to its nearest vein using vessel distance mapping. Our data reveal a dependence of cortical quantitative T1 values and positive QSM values on venous distance. In addition, there is an interaction between venous distance and age on quantitative T1 values, driven by lower quantitative T1 values in older compared to younger adults in voxels that are closer to a vein. Together, our data show that the local venous architecture explains a significant amount of variance in standard measures of cortical microstructure and should be considered in neurobiological models of human brain organisation and cortical ageing.


Introduction
Age-related differences in cortical microstructure are used to understand the neuronal mechanisms that underlie human brain ageing and neurodegeneration.Vascular supply patterns in the brain vary between individuals (Bernier et al., 2018;Huck et al., 2019), influence cortical microstructure and functional decline (Perosa et al., 2020;Vockert et al., 2021), and explain brain pathology (Schreiber et al. 2023).However, the contribution of the local vascular architecture on age-related changes in cortical microstructure is poorly understood.This knowledge gap prevents us from understanding the mechanisms that underlie human cortical ageing and potential protective factors.
The human cortex is permeated by a dense net of blood vessels that is highly variable in location and diameter (Huck et al., 2019).Vascular changes are a risk factor for all-cause dementia (Ungvari et al. 2021), and antihypertensive treatment can lower dementia incidence (Peters et al. 2022).A human MRI study reported that participants with cerebral small vessel disease who show a double-supply pattern of hippocampal vascularization perform better in cognitive tests compared to those with a single-supply pattern (Perosa et al., 2020).Vascular health is regarded as a key driver for resilience in ageing, for example by maintaining blood flow and neuronal metabolism (Arenaza-Urquijo et al. 2019).
However, it is still unclear how individual vascular supply patterns interact with age-related cortical degeneration.More specifically, cortical locations that are more distant to the next larger blood vessel might be more prone to age-related structural alterations, such as myelin loss (Cho et al. 1997;Callaghan et al. 2014;Grydeland et al. 2019) and iron accumulation (Callaghan et al. 2014;Acosta-Cabronero et al. 2016;Betts et al. 2016;Northall et al. 2023a), in case ageing blood vessels particularly reduce supply to locations that are farther away from major branches (Hypothesis 1, H1).Alternatively, myelin loss and iron accumulation might be more pronounced at cortical locations that are closer to blood vessels, because age-related substance accumulation occurs near blood vessels (Farkas et al. 2006) (Hypothesis 2, H2).Finally, one could hypothesise that cortical degeneration occurs independent of the distance to the next blood vessel (Hypothesis 3, H3).
To test these three hypotheses, we compared two extreme age groups (healthy younger adults < 30 years of age, healthy older adults > 65 years of age) using ultra-high resolution magnetic resonance imaging at 7 Tesla (7T MRI).Quantitative T1 values (qT1) were used as validated proxy for cortical myelin content (Dinse et al., 2015;Stüber et al., 2014;Haast et al. 2016), and quantitative susceptibility mapping (QSM) was used to quantify cortical iron (Langkammer et al., 2012;Deistung et al. 2013;Hametner et al. 2018;Acosta-Cabronero et al. 2016;Betts et al. 2016).We focus on primary somatosensory cortex (S1) and primary motor cortex (M1) as both areas play an important role in ageing and neurodegeneration, and venous density in S1 and M1 has been associated with myelin loss and decline of sensorimotor function in older adults (Schreiber et al. 2023).Ultra-high resolution QSM images were used for vein identification, and Vessel Distance Mapping (VDM) for linking age-related changes in layer-specific cortical microstructure to the distance to the nearest vein (Mattern et al. 2021a;Garcia-Garcia et al. 2023).
Together, we combine MRI proxies of cortical myelin and iron with VDM to test whether age-related changes in S1 and M1 microstructure depend on the distance to the nearest vein.This work contributes to our understanding of how the individual vascular architecture relates to microstructural variance in the ageing cortex, and provides critical insights into the mechanisms of cortical ageing.
Ultra-high field MRI contraindications (e.g.presence of active implants, non-removable metallic objects, tinnitus, tattoos/permanent makeup, claustrophobia, pregnancy, consumption of alcohol or drugs), chronic illness, psychiatric or neurological disorders, central acting medications and impaired hand function were exclusion criteria.1/17 older adults underwent carpal tunnel surgery on both hands.However, no complaints were reported and no outliers were detected in sensorimotor tasks (see Supplementary Table 1).
With respect to vascular health, 6/17 older adults reported a history of hypertension that was well-controlled with specific medication in all cases (see Table 1).Two experienced neurologists evaluated vascular health of older adults according to the STandards for ReportIng Vascular changes on nEuroimaging (STRIVE-2, Wardlaw et al. 2013, Duering et al. 2023) for the detection of cerebral small vessel diseases, and the guideline by Chen et al. (2022) for the detection of thromboses.We detected microbleeds, cortical superficial siderosis, lacunae and/or enlarged perivascular spaces in 6/17 older adults (see Table 1).All neurological anomalies were detected outside S1 and M1.In total, 9/17 older adults showed vascular health risk factors (i.e., hypertension and/or signs of cerebral small vessel diseases).Besides, participants showed no medical illnesses.As indicated by the 'Montreal Cognitive Assessment' (MoCA; Nasreddine et al., 2005), there were no signs of cognitive impairments (mean score 28.33 ± 1.45), except for one older adult who had a MoCA score of 21 (MoCA ≥ 26 is considered normal; Nasreddine et al., 2005; see Table 1).Because this participant performed equally well in sensorimotor tasks compared with all other older adults (i.e., no extreme outliers, see Supplementary Table 1), we kept the data in the analysis.There were no professional musicians among the participants (Elbert et al., 1995;Ragert et al., 2004;Schwenkreis et al., 2007).
All participants were paid for their attendance and provided written informed consent.The study was approved by the local Ethics committee of the Otto-von-Guericke University Magdeburg, Germany.Structural S1 data of younger adults (Doehler et al., 2023) and structural M1 data of younger and older adults (Northall et al., 2023a) were published previously.However, in both studies, the factor of venous architecture was not investigated.

MRI Data Acquisition
All participants took part in one structural 7T-MRI scanning session.The data were acquired at a Siemens MAGNETOM scanner located in Magdeburg, Germany, using a 32-channel head coil.

Image Preprocessing
We used the same preprocessing pipeline as described previously (Doehler et al. 2023;Northall et al. 2023).In short, data quality was ensured by two independent raters.QSM images were reconstructed using the Bayesian multi-scale dipole inversion (MSDI) algorithm as implemented in qsmbox (version v2.0, freely available for download: https://gitlab.com/acostaj/QSMbox,Acosta-Cabronero et al., 2018).MP2RAGE and non-normalized QSM data (Betts et al., 2016) were processed using the Java Image Science Toolkit (JIST; Lucas et al., 2010) and CBS Tools (Bazin et al., 2014)  Dura mater masks were manually checked and eventually refined to ensure that all non-brain tissue was removed.QSM images were registered to the merged whole brain qT1 images using the software ITK-SNAP (version 3.6.0,freely available for download at www.itksnap.org).
Cortical segmentation was calculated on the UNI image, using the Topology-preserving Anatomy-Driven Segmentation (TOADS) algorithm (Bazin & Pham, 2007).To estimate the boundaries between the white matter (WM) and the grey matter (GM), and between the GM and cerebrospinal fluid (CSF), the CRUISE (Cortical Reconstruction Using Implicit Surface Evolution) algorithm (Han et al, 2004) was used.Resulting level set images (surfaces in Cartesian space using a level set framework; Sethian, 1999) were optimised to precisely match M1 and S1 by thresholding the inner (WM-GM) and outer (CSF-GM) border to a maximum value of -5.0 and -2.0 for M1 and -2.8 and -0.2 for S1, respectively (based on empirical optimization; Northall et al., 2023;Doehler et al., 2023).
The cortex was divided into 21 cortical depths using the validated equi-volume model (Waehnert et al. 2014).After removing the three most superficial and the two deepest cortical depths to reduce partial volume effects (Kuehn et al., 2017;Tardif et al., 2015)

Vessel segmentation and Vessel Distance Mapping (VDM)
To extract vessels from the QSMs, we used the Filtered Vessels module (v3.0.4;CBS Tools; Bazin et al., 2014).The vasculature was segmented using a MRF (Markov Random Field) diffusion technique.Previous validation studies revealed that this technique captures venous architectures with high precision (Bazin et al., 2016).The resulting vessel probability map represents the probability of each voxel belonging to a vessel (1=highest probability, 0=lowest probability).
For computing VDMs (Garcia-Garcia et al. 2023), individual vessel probability maps were binarized.For thresholding, we used the openly available sMall vEsseL sEgmenTa-Tion pipelinE (OMELETTE, Mattern, 2021b, https://gitlab.com/hmattern/omelette). Thresholds were set automatically using a 3-class Otsu histogram analysis (returning a lower and upper threshold).Voxels above the upper threshold were considered as vessels, voxels below the lower threshold were considered as background.To classify vessels between the lower and the upper threshold, hysteresis thresholding was used (Fraz et al., 2011), which considers a voxel belonging to a vessel only if it is connected to a voxel with a value above the upper threshold.A Euclidean distance transformation (Maurer et al., 2003) was applied to the thresholded data to compute the distance to the closest segmented vein per voxel, generating vessel distance maps (VDMs).VDMs were multiplied with binarized cortical depths masks and S1/M1 masks, resulting in one VDM per cortical depth, participant, and brain area (see Figure 1 for overview).
qT1 images were multiplied with inverted binarized vessel probability masks (extracting values not overlapping with identified veins).Reconstructed QSM images were multiplied with inverted binarized vessel probability masks after thresholding them for positive values (pQSM values).Resulting qT1 and pQSM values were sampled at different cortical depths.Masks were applied to extract data from the primary motor cortex (M1, red mask) and the primary somatosensory cortex (S1, blue mask).Quantitative Susceptibility Maps (QSM; displayed as Maximum Intensity Projections, MIP) were used to extract the vessel probability maps (including intensity values between 0 and 1) to identify venous vasculature.A hysteresis filter was used to generate a binarized vein mask before a Euclidean distance transformation was applied to compute the Vessel Distance Map (VDM).Individual VDMs were multiplied with binarized cortical depths masks and then applied to the qT1 image before resulting parameter maps were multiplied with binarized M1/S1 masks (analysis pathway shown as black lines).Please note that the same analysis pathway was applied to the pQSM data (black dotted line).B: Shown are masks covering left M1 (red) and left S1 (blue) together with the vessel probability map of one example participant.Magnified images show extracted cortical depth compartments in relation to the distance to the nearest vein.

Anatomical definition of S1 and M1
M1 and S1 masks were manually generated using an operational definition based on anatomical landmarks extracted from cytoarchitectonic (Geyer et al., 1999), fMRI (Yousry et al., 1997) and multimodal parcellation studies (Glasser et al. 2016, Glasser andVan Essen 2011), following a standardised procedure (for details see Northall et al., 2023a;Northall et al. 2023b;Doehler et al., 2023) using the manual segmentation tool in ITK-SNAP (v3.8.0, Yushkevich et al., 2006).The hand knob served as a localiser to identify the hand area in M1 (97-100% accuracy, Yousry et al., 1997).We masked all axial slices in which the hand knob was visible, and extended the masks inferior and superior to the hand knob until the precentral gyrus was completely covered (e.g.Northall et al., 2023a;Northall et al. 2023b).S1 masks were drawn from the crown of the postcentral gyrus to the fundus of the central sulcus (Doehler et al., 2023), covering area 3b, and parts of area 1 and area 3a (Geyer et al., 1999).Only those slices were masked in which the postcentral gyrus was visible.For M1 masks also include medial M1, whereas S1 masks mainly cover lateral S1.Resulting masks were plotted in reference to co-registered Freesurfer labels to ensure that the delineated regions overlap with those outlined by automated approaches.As normalised anatomical data often falls short in capturing nuanced inter-individual distinctions in high-resolution MRI data due to smoothing effects and/or anatomical landmarks deviating from functional circuit structures (Guntupalli et al., 2016), we chose an individual segmentation approach, where no normalisation is needed and the individual anatomy is considered.

Distance Binning
Each cortical VDM was discretized into 5 bins, which represent levels of different distances to the nearest vein (i.e., 0.01-2 mm, 2.01-4 mm, 4.01-6 mm, 6.01-8 mm and 8.01-10 mm), for each given voxel in M1 and S1.Binning was used to obtain different distance conditions for statistical group analysis.We chose five bins to include a minimum of 4 voxels per distance condition.
Resulting binned data (5 VDMs per cortical depth and brain region) were multiplied with qT1 and pQSM images.

Vascular Health Evaluation
Two experienced neurologists evaluated vascular health of older adults based on visual inspection of MR scans.Following the guideline by Chen et al. (2022), thromboses were investigated using part-brain QSM data in comparison with part-brain SWI data.According to the STRIVE-2 recommendations, microbleeds and cortical superficial siderosis were investigated based on the part-brain SWI data, whereas lacunae and microinfarcts were examined using whole-brain T1-weighted images (MP2RAGE).

Behavioural Tasks
Tactile detection thresholds of the right index fingertip were assessed using fine hair stimuli

Statistical Analyses
To investigate the interaction between venous distance and cortical microstructure, we calculated Eta-squared (η² G ) was calculated as an effect size estimator for ANOVAs (Lakens, 2013).Cohen's benchmarks of 0.06 and 0.14 for medium and large effects, respectively, were applied (Cohen, 1988).
To control for a possible effect of vascular health risk on qT1 values and a possible interaction effect between vascular health risk and distance to the nearest vein in older adults, we calculated linear regression models (random intercept models), using the lmer function from the lme 4 package (version 1.1.33)in R (version 4.2.2,R Core Team, 2022).We compared two models

Age-related differences in cortical myelination are driven by the distance to the nearest vein
To test for the effect of venous distance on age-related changes in cortical myelination, we computed an ANOVA with the factors, age (younger, older), brain area (S1, M1), cortical depth (SF, OM, IM, DP) and venous distance (0.01-2 mm, 2.01-4 mm, 4.01-6 mm, 6.01-8 mm, 8.01-10 mm) on qT1 values.As expected, there is a significant main effect of brain area (see Table 2) driven by lower qT1 values (higher myelination) in M1 compared to S1 (see Figure 2B), and a significant main effect of cortical depth driven by more myelinated deep layers in M1 and S1 (see

Table 2, Figure 2C).
There is also a significant main effect of venous distance (see  2 for means and standard errors of compared conditions).This shows that, in both younger and older adults, there is greater cortical myelination with greater distance to the nearest vein (see Figure 2D).
The significant interaction between brain area and venous distance on qT1 values (see  2F).
We then explored potential age effects and detected no significant main effect of age (see Table 2, Figure 2A) and no significant interaction between age and brain area (see Table 2), but a significant interaction between age and venous distance on qT1 values (see  2E).
This shows that older adults have higher cortical myelination than younger adults close to a vein, but lower cortical myelination than younger adults more distant to a vein.
Critically, the interaction between age and venous distance remains significant when excluding the closest distance of 0.01-2 mm that is most prone to partial volume effects (see

Supplementary Table 2).
Finally, a significant three-way interaction between age, brain area and venous distance (see

Cortical iron content depends on the distance to the nearest vein
To test for the effect of venous distance on age-related changes in cortical iron content, we computed the same ANOVA on pQSM values.As expected, there is a significant main effect of brain area (see Table 3) with higher pQSM values (more iron) in M1 compared to S1 (see Figure 4B), and a significant main effect of cortical depth driven by less iron in deeper layers in M1 and S1 (see Table 3, Figure 4C).
There is also a significant main effect of venous distance (see  4D).There is also a trend towards significantly higher pQSM values at distances 2.01-4 mm compared to distances 4.01-6 mm (t(34)=1.715,p=0.095).This shows that in both younger and older adults, there is a U-shaped relationship between cortical iron content and venous distance.Other than for cortical myelin, this effect is not mediated by brain area (see Figure 4F, Table 3).
We then explored potential age effects and detected a significant main effect of age (see Table 3) with higher pQSM values (more iron) in older compared to younger adults in both M1 and S1 (see Figure 4A).However, other than for cortical myelin, the age effect is not mediated by the distance to the nearest vein (see Figure 4E, Table 3); instead, age effects differ significantly between brain areas (see

Discussion
We here provide insights that the distance to the nearest vein is related to age-related differences in cortical microstructure.Specifically, we tested three hypotheses: Cortical locations that are more distant to the next larger blood vessel are more prone to age-related structural alterations (H1), cortical locations that are closer to the next larger blood vessel are more prone to age-related structural alterations (H2), age-related cortical degeneration is independent of the distance to the next blood vessel (H3).Our data confirm H1 for cortical myelin and H3 for cortical iron.In addition, our data shows that in both younger and older adults, there is a linear negative relation between cortical myelination and venous distance, and a U-shaped relation between cortical iron and venous distance.Together, this data shows that the local venous architecture explains a significant amount of variance in standard measures of cortical microstructure in younger adults and older adults, and needs to be considered in neurobiological models of human brain organisation and cortical ageing.
We report a significant interaction between age and venous distance on qT1 values (reflecting cortical myelination).This effect is driven by older adults showing lower qT1 values (more myelin) than younger adults in voxels closer to a vein, but showing a trend towards higher qT1 values (less myelin) in voxels farther away from a vein.Prior studies show either reduced or higher cortical myelination in older compared to younger adults (Seiler et al. 2020;Cho et al. 1997;Grydeland et al. 2013;Grydeland et al. 2019;Callaghan et al. 2014) or no significant differences at all (Northall et al. 2023a).Those contrary findings may be related to hidden variables that explain a significant amount of variance but have been disregarded so far.Our data indicate that the venous architecture, specifically the distance to the nearest vein, acts as such a hidden variable, and that age-related myelin loss specifically occurs more distant from a vein (confirming H1).Given the pQSM effect is reversed (more iron accumulation in older adults in M1 at voxels more distant from a vein), the myelin effect is unlikely to be driven by microvessels that increase pQSM values.Possible neuronal mechanisms driving the distance-dependent myelin loss in ageing are that increased metabolism close to blood vessels may prevent age-related myelin loss, and/or ageing blood vessels particularly reduce supply to locations that are farther away from major branches.Myelin itself can be a driver for increased metabolism, since it is a highly demanding structure (Hirrlinger & Nave, 2014).The venous architecture would in this view be a protective mechanism for older adults' cortical microstructure, specifically with respect to myelin loss.Perosa et al. (2020) hypothesised that in the hippocampus, a steady arterial vascular supply constitutes a 'vascular reserve' mechanism in cases of pathology, relating to larger hippocampal volume (Vockert et al. 2021) and preserved cognitive abilities in patients with cerebral small vessel disease (Pohlack et al., 2014).Our study indicates that a similar vascular reserve mechanism can be assigned to veins.
Our data shows that the dependence of age-related myelin changes on the distance to the nearest vein is mainly driven by M1 and to a lesser extent by S1.The reason for this difference is at present unclear.Given S1 has significantly higher qT1 values (less myelin) and significantly lower pQSM values (less iron) than M1, one could assume that possible effects of the venous distance on S1 microstructure are harder to detect.Alternatively, the effect of the venous architecture on cortical microstructure might be particularly strong in a motor output area with a pronounced layer V and high cell and myelin density.M1 is more metabolically active than S1 because of its cell structure and function, and may need a much faster turnover of metabolites, thus depending more on intact vessel function.
Such a relationship between age-dependent alterations in cortical microstructure and venous distance, as described above for cortical myelin, cannot be detected for iron, in principle confirming H3.Both younger and older adults show a U-shaped relationship between pQSM values (iron) and venous distances, driven by highest pQSM values (most iron) at cortical voxels closest and farthest away from a vein.This was not modulated by age.However, methodological limitations need to be considered: We here focus on veins at the mesoscopic scale (resolution: 0.5 mm isotropic), and we cannot comment on microvessels below that resolution.Therefore, it is possible that microvessels located close to larger veins increase pQSM values close to larger veins (i.e., at distances 0.01-2 mm).Increased pQSM values at distances close to a vein need to be treated with caution and could potentially be driven by partial volume artefacts.Therefore, we recalculated all analyses taking out the closest distance.When doing so, the cortical myelin effects as described above do not change, but there is an interaction between age, brain area, and venous distance on pQSM values.This is driven by opposing relationships between age-related iron accumulation and venous distance in M1 versus S1: Whereas age-related iron accumulation is larger at greater distances in M1, it is larger at smaller distances in S1.This would confirm H1 and H2 for M1 and S1, respectively, rather than confirming H3.Given the resolution was 0.5 mm isotropic, which limited the amount of small veins that could be classified, the present data can not completely clarify the relationship between age-related iron accumulation and venous distance, but warrant further research in this area using even higher resolutions (i.e., 0.35 mm resolution or higher; van Gelderen et al. 2023).However, if future studies confirm H1 in M1 also for iron accumulation, this would strengthen the argument that the vascular architecture acts as protective mechanisms in ageing.
With respect to study limitations, the participant number was relatively low but motivated by previous layer-dependent 7T MRI studies using quantitative in-vivo proxies to describe the microstructural cortex architecture (Dinse et al., 2015, Kuehn et al., 2017), and is well above previously reported sample sizes.To reduce type I errors, we restricted the number of statistical tests to the minimum number needed (i.e., two ANOVAs and two regression analyses).Moreover, the image resolution of 0.5 mm isotropic did not allow to identify the entire venous network, as already pointed out above.Furthermore, for a comprehensive investigation of the neuronal mechanisms that underlie the interaction between blood vessels and cortical microstructure, also the arterial blood supply patterns need to be investigated, and should be complemented by blood flow assessments.Additional analyses of time-of-flight MRI measurements would benefit future studies.Moreover, previous studies pointed out that hypertension and alterations in brain vasculature relate to myelin loss and M1 pathology (Jacków-Nowicka et al. 2021;Tsushima, et al. 2003;Dobrynina et al. 2018, Li et al. 2023;Schreiber et al. 2023).Here, 9/17 older adults showed global signs of vascular health risk (i.e., presence of hypertension and/or signs of cerebral small vessel diseases outside S1 and M1).However, these indicators did not significantly predict S1 and M1 microstructure in older adults, and may therefore not account for the main findings.We instead argue in favour of a protective mechanism of veins preserving the myeloarchitecture in older age, which may only hold in cases of intact local vascular architecture.Assessing the potential impact of hypertension and other vascular disorders on cortical microstructure is an important topic for future studies.
Taken together, we show that the venous architecture interacts with cortical microstructure and age-related changes thereof.This is an important insight as most studies do not analyse cortical alterations as a function of venous or arterial distance, which may lead to misinterpretations of present (or absent) effects.Our study provides evidence for the importance of the human venous architecture in understanding cortical function in health and disease.duration of 20 seconds.One run contained 15 trials with intertrial intervals of 10 seconds, lasting about 8 minutes in total.All participants performed two runs that were separated by a 5-minute resting period.After each trial, participants received feedback about their individual performance level on screen.We monitored the time (in seconds) the controllable bar was within a given percentage above (2.5%) and below (2.5%) the target line (upper edge of the reference bar) (Vieluf et al., 2013;Voelcker-Rehage & Alberts, 2005) (y = yes, n = no), visually detected thrombosis following the guideline by Chen et al. (2022) and the STRIVE-2 criteria (Duering et al. 2023; cSS = cortical superficial siderosis; EPVS = enlarged perivascular spaces).In parentheses, the number of lesions are detailed.The composite score reflects the sum of positive STRIVE-2 criteria (1 = no positive criterion, 2 = one positive criteria, 3 = two positive criteria, 4 = three positive criteria).

Figure 1 :
Figure1: Overview Methodology and Analysis.A: qT1 and QSM images were sampled at different cortical depths (cortical depths masks) after the cortex segmentation (cortex masks).Masks were applied to extract data from the primary motor cortex (M1, red mask) and the primary somatosensory cortex (S1, blue mask).Quantitative Susceptibility Maps (QSM; displayed as Maximum Intensity Projections, MIP) were used to extract the vessel probability maps (including intensity values between 0 and 1) to identify venous vasculature.A hysteresis filter was used to generate a binarized vein mask before a Euclidean distance transformation was applied to compute the Vessel Distance Map (VDM).Individual VDMs were multiplied with binarized cortical depths masks and then applied to the qT1 image before resulting parameter maps were multiplied with binarized M1/S1 masks (analysis pathway shown as black lines).Please note that the same analysis pathway was applied to the pQSM data (black dotted line).B: Shown are masks covering left M1 (red) and left S1 (blue) together with the vessel probability map of one example participant.Magnified images show extracted cortical depth compartments in relation to the distance to the nearest vein.

Figure 2 .
Figure 2. Effects of age, brain area, cortical depth and venous distance on cortical myelination.qT1 values are given in milliseconds (lower values indicate higher myelin content).A: Box plots show no significant main effect of age on qT1 values.Shown are medians, interquartile ranges, and lower and upper quartiles for younger (n=18, light grey) and older (n=17, dark grey) adults.Dots above a box mark outliers.B: Box plots show a significant main effect of brain area (i.e., M1 (red) and S1 (blue)) on qT1 values.C: Line graph shows the main effect of cortical depth on qT1 values; dots represent mean qT1 values for the different cortical depths averaged across age groups, distances, and brain areas.Error bars show standard errors of the mean (SEM).D: Line graph shows main effect of venous distance on qT1 values (0 -2 = 0.01 -2 mm, 2 -4 = 2.01 -4 mm, 4 -6 = 4.01 -6 mm, 6 -8 = 6.01 -8 mm, 8 -10 = 8.01 -10 mm).E: Line graph shows interaction effect between venous distance and age on qT1 values.F: Line graphs show interaction effect between venous distance and brain area on qT1 values.G / H: Line graphs show interaction effects between venous distance, age and brain area (G: data for M1, H: data for S1).Significant results are marked by asterisks: * p≤0.05, ** p≤0.005, *** p≤0.0005 (uncorrected).Trends above p=0.05are marked by a T.

Figure 3 .
Figure 3. Distribution of vascular health risk in older adults.Individual vascular health risk (i.e., binary variable indicating the presence of hypertension and/or signs of cerebral small vessel diseases according to the STRIVE-2 criteria; green: no risk, magenta: increased risk) shown in relation to the distance to the nearest vein (distance, given in millimetres).(A) Distribution of vascular health risk for older adults (n=17) in primary motor cortex (M1) in relation to quantitative T1 values (qT1, given in milliseconds).(B) Distribution of vascular health risk for older adults in primary somatosensory cortex (S1) in relation to quantitative T1 values (qT1, given in milliseconds).(C) Distribution of vascular health risk for older adults in M1 in relation to positive QSM values (pQSM, given in parts per million, i.e. ppm).(D) Distribution of vascular health risk for older adults in S1 in relation to positive QSM values (pQSM, given in parts per million, i.e. ppm).

Figure 4 .
Figure 4. Effects of age, brain region, cortical depth and venous distance on pQSM values.A: Box plots show the main effect of age on pQSM values.Shown are medians, interquartile ranges, and lower and upper quartiles for younger (n=18, light grey) and older (n=17, dark grey) adults.Dots mark outliers.pQSM values are given in ppm (higher values indicate higher iron content).B: Box plots show the main effect of brain area (M1 (red) versus S1 (blue)) on pQSM values.C: The line graph shows the main effect of cortical depth on pQSM values.Dots indicate mean pQSM values for different cortical depths averaged across age groups, distances, and brain areas.Error bars indicate standard errors of the mean (SEM).D: The line graph shows the main effect of venous distance on pQSM values (0-2 = 0.01-2 mm, 2-4 = 2.01-4 mm, 4-6 = 4.01-6 mm, 6-8 = 6.01-8 mm, 8-10 = 8.01-10 mm).E: The line graph shows the interaction effect between venous distance and age on pQSM values.F: The line graph shows the interaction effect between venous distance and brain area on pQSM values.G / H: Line graphs show the interaction effect between venous distance, age and brain area (G: data for M1, H: data for S1).Significant results are marked by asterisks: * p≤0.05, ** p≤0.005, *** p≤0.0005 (uncorrected).Trends above p=0.05are marked by a T.

Table 1 . MoCA scores and vascular health rating for older adults
. The 'Montreal Cognitive Assessment' (MoCA) score is an indicator of cognitive impairment.A MoCA score ≥ 26 is considered normal.Vascular health is given as self-reported history of hypertension for 4 distance conditions (i.e., excluding the closest distance condition).The significance level was set to p=0.05.In addition to uncorrected p-values, we report Holm-Bonferroni corrected p-values for post-hoc tests (i.e., multiple comparisons).Generalised mixed-effects ANOVAs on qT1 and pQSM values with age group(younger, older)as between-subjects factor, and brain area (M1, S1), cortical depth (SF, OM, IM, DP) and distance to the nearest vein (0.01-2 mm, 2.01-4 mm, 4.01-6 mm, 6.01-8 mm, 8.01-10 mm; as well as: 2.01-4 mm, 4.01-6 mm, 6.01-8 mm, 8.01-10 mm) as within-subjects factors.Associated statistical analyses were performed using SPSS (IBM Corp. Released 2012.IBM SPSS Statistics for Windows, Version 21.0.Armonk, NY: IBM Corp.).Sample distributions were tested for normality using Shapiro-Wilk's test in combination with visual inspections (see Supplementary Figure1, Supplementary Figure2).Homogeneity of variances was tested with Levene's test.In case of sphericity violations, Greenhouse-Geisser-corrected results were used.Post-hoc tests were performed as two-tailed paired-samples t-tests to follow up within-subjects effects and as distance (i.e., 0.01-2mm), using the method of multiple monotone imputation based on linear regression, and

Table 2 . ANOVA Results for qT1 values.
Results for the mixed-effects ANOVA on qT1 values with factors age

Table 3
This leads to age-related differences in cortical iron content that are dependent on the distance to the nearest vein, but this effect is different in M1 versus S1 (M1: more iron in

Table 2 . ANOVA Results for qT1 values.
. Higher values reflect better sensorimotor integration.Results for the ANOVA on qT1 values with factors age (younger, older), cortical depth (SF, OM, IM, DP), brain area (S1, M1) and venous distance (4 distances, excluding the distance 0.01 -2 mm).Given are mean qT1 values in milliseconds (Mean), degrees of freedom of the numerator (DFn), degrees of freedom of the denominator (DFd), p-values (p), and effect sizes (nG²).P-values below 0.05 are considered as significant.