Search

Applying deep learning in population genomics is challenging because of computational issues and lack of interpretable models. Here, we propose GenNet, a novel open-source deep learning framework for predicting phenotypes from genetic variants. In this framework, interpretable and memory-efficient neural network architectures are constructed by embedding biologically knowledge from public databases, resulting in neural networks that contain only biologically plausible connections. We applied the framework to seventeen phenotypes and found well-replicated genes such as HERC2 and OCA2 for hair and eye color, and novel genes such as ZNF773 and PCNT for schizophrenia. Additionally, the framework identified ubiquitin mediated proteolysis, endocrine system and viral infectious diseases as most predictive biological pathways for schizophrenia. GenNet is a freely available, end-to-end deep learning framework that allows researchers to develop and use interpretable neural networks to obtain novel insights into the genetic architecture of complex traits and diseases., van Hilten and colleagues present GenNet, a deep-learning framework for predicting phenotype from genetic data. This framework generates interpretable neural networks that provide insight into the genetic basis of complex traits and diseases.

Background and Purpose: The prevalence of unruptured intracranial aneurysms (UIAs) in the adult population is ≈3%. Rupture of an intracranial aneurysm can have devastating consequences, which emphasizes the importance of identification of potentially modifiable determinants for the presence and size of UIAs. Our aim was to study the association of a broad spectrum of potential determinants with the presence and size of UIAs in a general adult population. Methods: Between 2005 and 2015, 5841 participants from the population-based Rotterdam Study (mean age, 64.4 years, 45.0% male) underwent brain magnetic resonance imaging (1.5T). These scans were evaluated for the presence of incidental UIAs. We determined number and volume of the UIAs. Using logistic and linear regression models, we assessed the association of cardiovascular, lifestyle and emerging inflammatory and hormonal determinants with the presence and volume of UIAs. Results: In 134 (2.3%) participants, ≥1 UIAs were detected (149 UIAs in total), with a median volume of 61.1 mm 3 (interquartile range, 33.2–134.0). In multivariable models, female sex (odds ratio, 1.92 [95% CI, 1.33–2.84]), hypertension (odds ratio, 1.73 [95% CI, 1.13–2.68]), and current smoking (odds ratio, 3.75 [95% CI, 2.27–6.33]) were associated with the presence of UIAs. We found no association of alcohol use, physical activity, or diet quality with UIA presence. Finally, we found white blood cell count to relate to larger aneurysm volume (difference in volume of 33.6 mm 3 per 10 9 /L increase in white blood cell [95% CI, 3.92–63.5]). Conclusions: In this population-based study, female sex, hypertension, and smoking, but no other lifestyle determinants, were associated with the presence of UIAs. White blood cell count is associated with size of UIAs. Preventive strategies should focus on treating hypertension and promoting cessation of smoking.

textabstractObjective: To investigate the association of migraine genetic variants with cerebral blood flow (CBF). Background: Migraine is a common disorder with many genetic and non-genetic factors affecting its occurrence. The exact pathophysiological mechanisms underlying the disease remain unclear, but are known to involve hemodynamic and vascular disruptions. Recent genome-wide association studies have identified 44 genetic variants in 38 genetic loci that affect the risk of migraine, which provide the opportunity to further disentangle these mechanisms. Methods: We included 4665 participants of the population-based Rotterdam Study (mean age 65.0 ± 10.9 years, 55.6% women). Cross-sectional area (mm2), flow velocity (mm/s), and blood flow (mL/min) were measured in both carotids and the basilar artery using 2-dimensional phase-contrast magnetic resonance imaging. We analyzed 43 previously identified migraine variants separately and calculated a genetic risk score (GRS). To assess the association with CBF, we used linear regression models adjusted for age, sex, and total brain volume. Hierarchical clustering was performed based on the associations with CBF measures and tissue enrichment. Results: The rs67338227 risk allele was associated with higher flow velocity and smaller cross-sectional area in the carotids (Pminimum = 3.7 × 10−8). Other variants were related to CBF with opposite directions of effect, but not significantly after multiple testing adjustments (P

The gap between predicted brain age using magnetic resonance imaging (MRI) and chronological age may serve as a biomarker for early-stage neurodegeneration. However, owing to the lack of large longitudinal studies, it has been challenging to validate this link. We aimed to investigate the utility of such a gap as a risk biomarker for incident dementia using a deep learning approach for predicting brain age based on MRI-derived gray matter (GM). We built a convolutional neural network (CNN) model to predict brain age trained on 3,688 dementia-free participants of the Rotterdam Study (mean age 66 ± 11 y, 55% women). Logistic regressions and Cox proportional hazards were used to assess the association of the age gap with incident dementia, adjusted for age, sex, intracranial volume, GM volume, hippocampal volume, white matter hyperintensities, years of education, and APOE ε4 allele carriership. Additionally, we computed the attention maps, which shows which regions are important for age prediction. Logistic regression and Cox proportional hazard models showed that the age gap was significantly related to incident dementia (odds ratio [OR] = 1.11 and 95% confidence intervals [CI] = 1.05–1.16; hazard ratio [HR] = 1.11, and 95% CI = 1.06–1.15, respectively). Attention maps indicated that GM density around the amygdala and hippocampi primarily drove the age estimation. We showed that the gap between predicted and chronological brain age is a biomarker, complimentary to those that are known, associated with risk of dementia, and could possibly be used for early-stage dementia risk screening.

Introduction: Our aim was to study whether systemic metabolites are associated with magnetic resonance imaging (MRI) measures of brain and hippocampal atrophy and white matter hyperintensities (WMH). Methods: We studied associations of 143 plasma-based metabolites with MRI measures of brain and hippocampal atrophy and WMH in three independent cohorts (n = 3962). We meta-analyzed the results of linear regression analyses to determine the association of metabolites with MRI measures. Results: Higher glucose levels and lower levels of three small high density lipoprotein (HDL) particles were associated with brain atrophy. Higher glucose levels were associated with WMH. Discussion: Glucose levels were associated with brain atrophy and WMH, and small HDL particle levels were associated with brain atrophy. Circulating metabolites may aid in developing future intervention trials.

Background and Purpose: Periventricular white matter hyperintensities (WMH; PVWMH) and deep WMH (DWMH) are regional classifications of WMH and reflect proposed differences in cause. In the first study, to date, we undertook genome-wide association analyses of DWMH and PVWMH to show that these phenotypes have different genetic underpinnings. Methods: Participants were aged 45 years and older, free of stroke and dementia. We conducted genome-wide association analyses of PVWMH and DWMH in 26,654 participants from CHARGE (Cohorts for Heart and Aging Research in Genomic Epidemiology), ENIGMA (Enhancing Neuro-Imaging Genetics Through Meta-Analysis), and the UKB (UK Biobank). Regional correlations were investigated using the genome-wide association analyses -pairwise method. Cross-trait genetic correlations between PVWMH, DWMH, stroke, and dementia were estimated using LDSC. Results: In the discovery and replication analysis, for PVWMH only, we found associations on chromosomes 2 ( NBEAL ), 10q23.1 ( TSPAN14/FAM231A ), and 10q24.33 ( SH3PXD2A). In the much larger combined meta-analysis of all cohorts, we identified ten significant regions for PVWMH: chromosomes 2 (3 regions), 6, 7, 10 (2 regions), 13, 16, and 17q23.1. New loci of interest include 7q36.1 ( NOS3 ) and 16q24.2. In both the discovery/replication and combined analysis, we found genome-wide significant associations for the 17q25.1 locus for both DWMH and PVWMH. Using gene-based association analysis, 19 genes across all regions were identified for PVWMH only, including the new genes: CALCRL (2q32.1), KLHL24 (3q27.1), VCAN (5q27.1), and POLR2F (22q13.1). Thirteen genes in the 17q25.1 locus were significant for both phenotypes. More extensive genetic correlations were observed for PVWMH with small vessel ischemic stroke. There were no associations with dementia for either phenotype. Conclusions: Our study confirms these phenotypes have distinct and also shared genetic architectures. Genetic analyses indicated PVWMH was more associated with ischemic stroke whilst DWMH loci were implicated in vascular, astrocyte, and neuronal function. Our study confirms these phenotypes are distinct neuroimaging classifications and identifies new candidate genes associated with PVWMH only.

Deep learning is rarely used in population genomics because of the computational burden and challenges in interpreting neural networks. Here, we propose GenNet, a novel open-source deep learning framework for predicting phenotypes from genetic variants. In this framework, interpretable and memory-efficient neural network architectures are constructed by embedding biological knowledge from public databases, resulting in neural networks that contain only biological plausible connections.We applied the framework to seventeen phenotypes from a case-control study, a population-based study and the UK Biobank. Interpreting the networks revealed well-replicated genes such as HERC2 and OCA2 for hair and eye color and novel genes such as ZNF773 and PCNT for schizophrenia. Additionally, the framework obtained an AUC of 0.74 in the held-out test set and identified ubiquitin mediated proteolysis, endocrine system and viral infectious diseases as most predictive biological pathways for schizophrenia.GenNet is a freely available, end-to-end deep learning framework that allows researchers to develop and use interpretable neural networks to obtain novel insights into the genetic architecture of complex traits and diseases.

Objective: Multiple sclerosis (MS) is a neurological disease with a substantial genetic component and immune-mediated neurodegeneration. Patients with MS show structural brain differences relative to individuals without MS, including smaller regional volumes and alterations in white matter (WM) microstructure. Whether genetic risk for MS is associated with brain structure during early neurodevelopment remains unclear. In this study, we explore the association between MS polygenic risk scores (PRS) and brain imaging outcomes from a large, population-based pediatric sample to gain insight into the underlying neurobiology of MS. Methods: We included 8- to 12-year-old genotyped participants from the Generation R Study in whom T1-weighted volumetric (n = 1,136) and/or diffusion tensor imaging (n = 1,088) had been collected. PRS for MS were calculated based on a large genome-wide association study of MS (n = 41,505) and were regressed on regional volumes, global and tract-specific fractional anisotropy (FA), and global mean diffusivity using linear regression. Results: No associations were observed for the regional volumes. We observed a positive association between the MS PRS and global FA (β = 0.098, standard error [SE] = 0.030, p = 1.08 × 10 −3). Tract-specific analyses showed higher FA and lower radial diffusivity in several tracts. We replicated our findings in an independent sample of children (n = 186) who were scanned in an earlier phase (global FA; β = 0.189, SE = 0.072, p = 9.40 × 10 −3). Interpretation: This is the first study to show that greater genetic predisposition for MS is associated with higher global brain WM FA at an early age in the general population. Our results suggest a preadolescent time window within neurodevelopment in which MS risk variants act upon the brain. ANN NEUROL 2020;87:774–787.

Enlarged perivascular spaces (EPVS) in the brain are an emerging imaging marker for cerebral small vessel disease, and have been shown to be related to increased risk of various neurological diseases, including stroke and dementia. Automatic quantification of EPVS would greatly help to advance research into its etiology and its potential as a risk indicator of disease. We propose a convolutional network regression method to quantify the extent of EPVS in the basal ganglia from 3D brain MRI. We first segment the basal ganglia and subsequently apply a 3D convolutional regression network designed for small object detection within this region of interest. The network takes an image as input, and outputs a quantification score of EPVS. The network has significantly more convolution operations than pooling ones and no final activation, allowing it to span the space of real numbers. We validated our approach using a dataset of 2000 brain MRI scans scored visually. Experiments with varying sizes of training and test sets showed that a good performance can be achieved with a training set of only 200 scans. With a training set of 1000 scans, the intraclass correlation coefficient (ICC) between our scoring method and the expert's visual score was 0.74. Our method outperforms by a large margin - more than 0.10 - four more conventional automated approaches based on intensities, scale-invariant feature transform, and random forest. We show that the network learns the structures of interest and investigate the influence of hyper-parameters on the performance. We also evaluate the reproducibility of our network using a set of 60 subjects scanned twice (scan-rescan reproducibility). On this set our network achieves an ICC of 0.93, while the intrarater agreement reaches 0.80. Furthermore, the automatic EPVS scoring correlates similarly to age as visual scoring.

Background Total hippocampal volume has been consistently linked to cognitive function and dementia. Yet, given its complex and parcellated internal structure, the role of subregions of the hippocampus in cognition and risk of dementia remains relatively underexplored. We studied subregions of the hippocampus in a large population-based cohort to further understand their role in cognitive impairment and dementia risk. Methods We studied 5035 dementia- and stroke-free persons from the Rotterdam Study, aged over 45 years. All participants underwent magnetic resonance imaging (1.5 T) between 2005 and 2015. Automatic segmentation of the hippocampus and 12 of its subregions was performed using the FreeSurfer software (version 6.0). A cognitive test battery was performed, and participants were followed up for the development of dementia until 2015. Associations of hippocampal subregion volumes with cognition and incident dementia were examined using linear and Cox regression models, respectively. All analyses were adjusted for age, sex, education, and total hippocampal volume. Results Mean age was 64.3 years (SD 10.6) with 56% women. Smaller volumes of the hippocampal fimbria, presubiculum and subiculum showed the strongest associations with poor performance on several cognitive domains, including executive function but not memory. During a mean follow-up of 5.5 years, 76 persons developed dementia. Smaller subiculum volume was associated with risk of dementia adjusted for total volume (hazard ratio per SD decrease in volume: 1.75, 95% confidence interval 1.35; 2.26). Conclusions In a community-dwelling non-demented population, we describe patterns of association between hippocampal subregions with cognition and risk of dementia. Specifically, the subiculum was associated with both poorer cognition and higher risk of dementia.

Objective: To investigate the association of brain volumes, white matter lesion (WML) volumes, and lacunes, with cognitive decline in a population-based cohort of nondemented persons. Methods: Within the Rotterdam Study, 3624 participants underwent brain magnetic resonance imaging. Cognition was evaluated at baseline (2005 to 2009) and at the follow-up visit (2011 to 2013). We used a test battery that tapped into domains of executive function, information processing speed, motor speed, and memory. The volumetric measures assessed were total brain volume, lobar (gray matter and white matter) volumes, and hippocampal volumes. We also studied the association of WML volumes and lacunes with cognitive decline using linear regression models. Results: Total brain volume was associated with decline in global cognition, information processing, and motor speed (P

ObjectiveTo investigate the association of enlarged perivascular spaces (ePVS) with cognition in elderly without dementia.MethodsWe included 5 studies from the Uniform Neuro-Imaging of Virchow-Robin Space Enlargement (UNIVRSE) consortium, namely the Austrian Stroke Prevention Family Study, Study of Health in Pomerania, Rotterdam Study, Epidemiology of Dementia in Singapore study, and Risk Index for Subclinical Brain Lesions in Hong Kong study. ePVS were counted in 4 regions (mesencephalon, hippocampus, basal ganglia, and centrum semiovale) with harmonized rating across studies. Mini-Mental State Examination (MMSE) and general fluid cognitive ability factor (G-factor) were used to assess cognitive function. For each study, a linear regression model was performed to estimate the effect of ePVS on MMSE and G-factor. Estimates were pooled across studies with the use of inverse variance meta-analysis with fixed- or random-effect models when appropriate.ResultsThe final sample size consisted of 3,575 persons (age range 63.4–73.2 years, 50.6% women). Total ePVS counts were not significantly associated with MMSE score (mean difference per ePVS score increase 0.001, 95% confidence interval [CI] −0.007 to 0.008, p = 0.885) or G-factor (mean difference per ePVS score increase 0.002, 95% CI −0.001 to 0.006, p = 0.148) in age-, sex-, and education-adjusted models. Adjustments for cardiovascular risk factors and MRI markers did not change the results. Repeating the analyses with region-specific ePVS rendered similar results.ConclusionsIn this study, we found that ePVS counts were not associated with cognitive dysfunction in the general population. Future studies with longitudinal designs are warranted to examine whether ePVS contribute to cognitive decline.

Introduction: Exploring the role of Alzheimer's disease (AD) implicated pathways in the predementia phase may provide new insight for preventive and clinical trials targeting disease specific pathways. Methods: We constructed weighted Genetic risk scores, first based on 20 genome-wide significant AD risk variants and second clustering these variants within pathways. Risk scores were investigated for their association with AD, mild cognitive impairment, and brain magnetic resonance imaging phenotypes including white matter lesions, hippocampal volume, and brain volume. Results: The risk score capturing endocytosis pathway was significantly associated with mild cognitive impairment (P = 1.44 × 10−4). Immune response (P =.016) and clathrin/AP2 adaptor complex pathway (P = 3.55 × 10−3) excluding apolipoprotein E also showed modest association with white matter lesions but did not sustain Bonferroni correction (P = 9.09 × 10−4). Discussion: Our study suggests that the clinical spectrum of early AD pathology is explained by different biological pathways, in particular, the endocytosis, clathrin/AP2 adaptor complex, and immune response pathways, that are independent of apolipoprotein E (APOE).

ObjectiveTo test the hypothesis that the inflammatory marker plasma soluble CD14 (sCD14) associates with incident dementia and related endophenotypes in 2 community-based cohorts.MethodsOur samples included the prospective community-based Framingham Heart Study (FHS) and Cardiovascular Health Study (CHS) cohorts. Plasma sCD14 was measured at baseline and related to the incidence of dementia, domains of cognitive function, and MRI-defined brain volumes. Follow-up for dementia occurred over a mean of 10 years (SD 4) in the FHS and a mean of 6 years (SD 3) in the CHS.ResultsWe studied 1,588 participants from the FHS (mean age 69 ± 6 years, 47% male, 131 incident events) and 3,129 participants from the CHS (mean age 72 ± 5 years, 41% male, 724 incident events) for the risk of incident dementia. Meta-analysis across the 2 cohorts showed that each SD unit increase in sCD14 was associated with a 12% increase in the risk of incident dementia (95% confidence interval 1.03–1.23; p = 0.01) following adjustments for age, sex, APOE ε4 status, and vascular risk factors. Higher levels of sCD14 were associated with various cognitive and MRI markers of accelerated brain aging in both cohorts and with a greater progression of brain atrophy and a decline in executive function in the FHS.ConclusionsCD14 is an inflammatory marker related to brain atrophy, cognitive decline, and incident dementia.

Finding automatically multiple lesions in large images is a common problem in medical image analysis. Solving this problem can be challenging if, during optimization, the automated method cannot access information about the location of the lesions nor is given single examples of the lesions. We propose a new weakly supervised detection method using neural networks, that computes attention maps revealing the locations of brain lesions. These attention maps are computed using the last feature maps of a segmentation network optimized only with global image-level labels. The proposed method can generate attention maps at full input resolution without need for interpolation during preprocessing, which allows small lesions to appear in attention maps. For comparison, we modify state-of-the-art methods to compute attention maps for weakly supervised object detection, by using a global regression objective instead of the more conventional classification objective. This regression objective optimizes the number of occurrences of the target object in an image, e.g. the number of brain lesions in a scan, or the number of digits in an image. We study the behavior of the proposed method in MNIST-based detection datasets, and evaluate it for the challenging detection of enlarged perivascular spaces - a type of brain lesion - in a dataset of 2202 3D scans with point-wise annotations in the center of all lesions in four brain regions. In the brain dataset, the weakly supervised detection methods come close to the human intrarater agreement in each region. The proposed method reaches the best area under the curve in two out of four regions, and has the lowest number of false positive detections in all regions, while its average sensitivity over all regions is similar to that of the other best methods. The proposed method can facilitate epidemiological and clinical studies of enlarged perivascular spaces., New formatting. A few changes in introduction, discussion and conclusion

Despite a substantial genetic component, efforts to identify common genetic variation underlying depression have largely been unsuccessful. In the current study we aimed to identify rare genetic variants that might have large effects on depression in the general population. Using high-coverage exome-sequencing, we studied the exonic variants in 1265 individuals from the Rotterdam study (RS), who were assessed for depressive symptoms. We identified a missense Asn396Ser mutation (rs77960347) in the endothelial lipase (LIPG) gene, occurring with an allele frequency of 1% in the general population, which was significantly associated with depressive symptoms (P-value=5.2 × 10 -08, β=7.2). Replication in three independent data sets (N=3612) confirmed the association of Asn396Ser (P-value=7.1 × 10 -03, β=2.55) with depressive symptoms. LIPG is predicted to have enzymatic function in steroid biosynthesis, cholesterol biosynthesis and thyroid hormone metabolic processes. The Asn396Ser variant is predicted to have a damaging effect on the function of LIPG. Within the discovery population, carriers also showed an increased burden of white matter lesions (P-value=3.3 × 1 -02) and a higher risk of Alzheimer's disease (odds ration=2.01; P-value=2.8 × 10 -02) compared with the non-carriers. Together, these findings implicate the Asn396Ser variant of LIPG in the pathogenesis of depressive symptoms in the general population.

Contains fulltext : 202907.pdf (Publisher’s version ) (Open Access) OBJECTIVE:: Attention deficit hyperactivity disorder (ADHD) is a common and highly heritable neurodevelopmental disorder with a complex pathophysiology. Intracranial volume (ICV) and volumes of the nucleus accumbens, amygdala, caudate nucleus, hippocampus, and putamen are smaller in people with ADHD compared with healthy individuals. The authors investigated the overlap between common genetic variation associated with ADHD risk and these brain volume measures to identify underlying biological processes contributing to the disorder. METHODS:: The authors combined genome-wide association results from the largest available studies of ADHD (N=55,374) and brain volumes (N=11,221-24,704), using a set of complementary methods to investigate overlap at the level of global common variant genetic architecture and at the single variant level. RESULTS:: Analyses revealed a significant negative genetic correlation between ADHD and ICV (rg=-0.22). Meta-analysis of single variants revealed two significant loci of interest associated with both ADHD risk and ICV; four additional loci were identified for ADHD and volumes of the amygdala, caudate nucleus, and putamen. Exploratory gene-based and gene-set analyses in the ADHD-ICV meta-analytic data showed association with variation in neurite outgrowth-related genes. CONCLUSIONS:: This is the first genome-wide study to show significant genetic overlap between brain volume measures and ADHD, both on the global and the single variant level. Variants linked to smaller ICV were associated with increased ADHD risk. These findings can help us develop new hypotheses about biological mechanisms by which brain structure alterations may be involved in ADHD disease etiology.

Key PointsQuestionIs the gap between brain age predicted from MRI and chronological age associated with incident dementia in a general population of Dutch adults?FindingsBrain age was predicted using a deep learning model, using MRI-derived grey matter density maps. In a population based study including 5496 participants, the observed gap was significantly associated with the risk of dementia.MeaningThe gap between MRI-brain predicted and chronological age is potentially a biomarker for dementia risk screening.AbstractImportanceThe gap between predicted brain age using magnetic resonance imaging (MRI) and chronological age may serve as biomarker for early-stage neurodegeneration and potentially as a risk indicator for dementia. However, owing to the lack of large longitudinal studies, it has been challenging to validate this link.ObjectiveWe aimed to investigate the utility of such a gap as a risk biomarker for incident dementia in a general Dutch population, using a deep learning approach for predicting brain age based on MRI-derived grey matter maps.DesignData was collected from participants of the cohort-based Rotterdam Study who underwent brain magnetic resonance imaging between 2006 and 2015. This study was performed in a longitudinal setting and all participant were followed up for incident dementia until 2016.SettingThe Rotterdam Study is a prospective population-based study, initiated in 1990 in the suburb Ommoord of in Rotterdam, the Netherlands.ParticipantsAt baseline, 5496 dementia- and stroke-free participants (mean age 64.67±9.82, 54.73% women) were scanned and screened for incident dementia. During 6.66±2.46 years of follow-up, 159 people developed dementia.Main outcomes and measuresWe built a convolutional neural network (CNN) model to predict brain age based on its MRI. Model prediction performance was measured in mean absolute error (MAE). Reproducibility of prediction was tested using the intraclass correlation coefficient (ICC) computed on a subset of 80 subjects. Logistic regressions and Cox proportional hazards were used to assess the association of the age gap with incident dementia, adjusted for years of education, ApoEε4 allele carriership, grey matter volume and intracranial volume. Additionally, we computed the attention maps of CNN, which shows which brain regions are important for age prediction.ResultsMAE of brain age prediction was 4.45±3.59 years and ICC was 0.97 (95% confidence interval CI=0.96-0.98). Logistic regression and Cox proportional hazards models showed that the age gap was significantly related to incident dementia (odds ratio OR=1.11 and 95% confidence intervals CI=1.05-1.16; hazard ratio HR=1.11 and 95% CI=1.06-1.15, respectively). Attention maps indicated that grey matter density around the amygdalae and hippocampi primarily drive the age estimation.Conclusion and relevanceWe show that the gap between predicted and chronological brain age is a biomarker associated with risk of dementia development. This suggests that it can be used as a biomarker, complimentary to those that are known, for dementia risk screening.

Christina M. Lill, who contributed to analysis of data, was inadvertently omitted from the author list in the originally published version of this article. This has now been corrected in both the PDF and HTML versions of the article.

Cognition in adults shows variation due to developmental and degenerative components. A recent genomewide association study identified genetic variants for general cognitive function in 148 independent loci. Here, we aimed to elucidate possible developmental and neurodegenerative pathways underlying these genetic variants by relating them to functional, clinical and neuroimaging outcomes. This study was conducted within the population-based Rotterdam Study (N=11,496, mean age 65.3±9.9 years, 58.0% female). We used lead variants for general cognitive function to construct a polygenic score (PGS), and additionally excluded developmental variants at multiple significance thresholds. A higher PGS was related to more years of education (β=0.29, p=4.3×10 -7 ) and a larger intracranial volume (β=0.05, p=7.5×10 -4 ). To a smaller extent, the PGS was associated with less cognitive decline (βΔG-factor=0.03, p=1.3×10 -3 ), which became non-significant after adjusting for education (p=1.6×10 -2 ). No associations were found with daily functioning, dementia, parkinsonism, stroke or microstructural white matter integrity. Excluding developmental variants attenuated nearly all associations. In conclusion, this study suggests that the genetic variants identified for general cognitive function are acting mainly through the developmental pathway of cognition. Therefore, cognition, assessed cross-sectionally, seems to have limited value as a biomarker for neurodegeneration.

Enlarged perivascular spaces (PVS) are structural brain changes visible in MRI, are common in aging, and are considered a reflection of cerebral small vessel disease. As such, assessing the burden of PVS has promise as a brain imaging marker. Visual and manual scoring of PVS is a tedious and observer-dependent task. Automated methods would advance research into the etiology of PVS, could aid to assess what a “normal” burden is in aging, and could evaluate the potential of PVS as a biomarker of cerebral small vessel disease. In this work, we propose and evaluate an automated method to quantify PVS in the midbrain, hippocampi, basal ganglia and centrum semiovale. We also compare associations between (earlier established) determinants of PVS and visual PVS scores versus the automated PVS scores, to verify whether automated PVS scores could replace visual scoring of PVS in epidemiological and clinical studies. Our approach is a deep learning algorithm based on convolutional neural network regression, and is contingent on successful brain structure segmentation. In our work we used FreeSurfer segmentations. We trained and validated our method on T2-contrast MR images acquired from 2115 subjects participating in a population-based study. These scans were visually scored by an expert rater, who counted the number of PVS in each brain region. Agreement between visual and automated scores was found to be excellent for all four regions, with intraclass correlation coefficients (ICCs) between 0.75 and 0.88. These values were higher than the inter-observer agreement of visual scoring (ICCs between 0.62 and 0.80). Scan-rescan reproducibility was high (ICCs between 0.82 and 0.93). The association between 20 determinants of PVS, including aging, and the automated scores were similar to those between the same 20 determinants of PVS and visual scores. We conclude that this method may replace visual scoring and facilitate large epidemiological and clinical studies of PVS.

Localization of focal vascular lesions on brain MRI is an important component of research on the etiology of neurological disorders. However, manual annotation of lesions can be challenging, time-consuming and subject to observer bias. Automated detection methods often need voxel-wise annotations for training. We propose a novel approach for automated lesion detection that can be trained on scans only annotated with a dot per lesion instead of a full segmentation. From the dot annotations and their corresponding intensity images we compute various distance maps (DMs), indicating the distance to a lesion based on spatial distance, intensity distance, or both. We train a fully convolutional neural network (FCN) to predict these DMs for unseen intensity images. The local optima in the predicted DMs are expected to correspond to lesion locations. We show the potential of this approach to detect enlarged perivascular spaces in white matter on a large brain MRI dataset with an independent test set of 1000 scans. Our method matches the intra-rater performance of the expert rater that was computed on an independent set. We compare the different types of distance maps, showing that incorporating intensity information in the distance maps used to train an FCN greatly improves performance.

Brain imaging data are increasingly made publicly accessible, and volumetric imaging measures derived from population-based cohorts may serve as normative data for individual patient diagnostic assessment. Yet, these normative cohorts are usually not a perfect reflection of a patient's base population, nor are imaging parameters such as field strength or scanner type similar. In this proof of principle study, we assessed differences between reference curves of subcortical structure volumes of normal controls derived from two population-based studies and a case-control study. We assessed the impact of any differences on individual assessment of brain structure volumes. Percentile curves were fitted on the three healthy cohorts. Next, percentile values for these subcortical structures for individual patients from these three cohorts, 91 mild cognitive impairment and 95 Alzheimer's disease cases and patients from the Alzheimer Center, were calculated, based on the distributions of each of the three cohorts. Overall, we found that the subcortical volume normative data from these cohorts are highly interchangeable, suggesting more flexibility in clinical implementation.

Deep learning techniques are often criticized to heavily depend on a large quantity of labeled data. This problem is even more challenging in medical image analysis where the annotator expertise is often scarce. We propose a novel data-augmentation method to regularize neural network regressors that learn from a single global label per image. The principle of the method is to create new samples by recombining existing ones. We demonstrate the performance of our algorithm on two tasks: estimation of the number of enlarged perivascular spaces in the basal ganglia, and estimation of white matter hyperintensities volume. We show that the proposed method improves the performance over more basic data augmentation. The proposed method reached an intraclass correlation coefficient between ground truth and network predictions of 0.73 on the first task and 0.84 on the second task, only using between 25 and 30 scans with a single global label per scan for training. With the same number of training scans, more conventional data augmentation methods could only reach intraclass correlation coefficients of 0.68 on the first task, and 0.79 on the second task., accepted in MICCAI 2019

Although several genome-wide association studies (GWAS) have investigated the genetics of pulmonary ventilatory function, little is known about the genetic factors that influence gas exchange. The aim of the study was to investigate the heritability of, and genetic variants associated with the diffusing capacity of the lung.GWAS was performed on diffusing capacity of the lung measured by carbon monoxide uptake (DLCO) and per alveolar volume (VA) using the single-breath technique, in 8372 individuals from two population-based cohort studies, the Rotterdam Study and the Framingham Heart Study. Heritability was estimated in related (n=6246) and unrelated (n=3286) individuals.Heritability of DLCO and DLCO/VA ranged between 23% and 28% in unrelated individuals and between 45% and 49% in related individuals. Meta-analysis identified a genetic variant in ADGRG6 that is significantly associated with DLCO/VA. Gene expression analysis of ADGRG6 in human lung tissue revealed a decreased expression in patients with chronic obstructive pulmonary disease (COPD) and subjects with decreased DLCO/VA.DLCO and DLCO/VA are heritable traits, with a considerable proportion of variance explained by genetics. A functional variant in ADGRG6 gene region was significantly associated with DLCO/VA. Pulmonary ADGRG6 expression was decreased in patients with COPD.

Introduction: We investigated whether a risk score based on genetic risk variants for Parkinson's disease (PD) is associated with the risk and improves prediction of incident PD, and whether the risk score is associated with basic activities of daily living (BADL) in healthy individuals. Methods: Within the population-based Rotterdam Study, we genotyped 26 independent risk variants for PD and constructed a genetic risk score in 7167 participants who were free of parkinsonism and dementia at baseline (1990 or 2000). Participants were followed for a maximum of twenty years for the onset of parkinsonism, dementia or death until January 1, 2011 (median follow-up 12.1 years). We studied the relationship between the genetic risk score and incident PD with adjustment for age, sex, smoking and parental history. In an independent sample of 2997 persons free of parkinsonism and dementia, we studied whether the PD risk score was associated with impaired BADL. Results: During follow-up (median 12.1 years), 99 persons were diagnosed with incident PD. The genetic risk score was associated with incident PD (hazard ratio per standard deviation risk 1.25 [95% confidence interval = 1.02; 1.551), but did not substantially improve prediction (change in C-statistic 0.687 [0.628; 0.745] to 0.698 [0.635; 0.760], Delta C = 0.011 [ 0.011; 0.0331). The genetic risk score was associated with a higher probability of any impairment in BADL (odds ratio = 1.11 [1.00; 1.23]). Conclusion: Genetic variants for PD are associated with the risk of incident PD in the general population and with impairment in daily functioning in individuals without clinical parkinsonism, but do not improve the clinical prediction of PD. However, we were probably underpowered to detect a small improvement in PD prediction. (C) 2016 Elsevier Ltd. All rights reserved.

The benefit of radical resections for glioblastoma patients remains a source of contention in the literature. Few studies have been conducted in pediatric patients, and it is becoming increasingly evident that data regarding adult glioblastoma (GB) patients cannot be generalized to pediatric patients affected by this neoplasm. A comparative effectiveness study is performed for different extent of resection (EOR) groups in the largest cohort of pediatric GB (pGB) patients. The Surveillance, Epidemiology, and End Results (SEER) cancer registry was used to identify pGB patients from 1988 through 2009. Multivariate- and multiple propensity score (mPS)-adjusted analyses were used to determine the effect of gross total resection (GTR), partial resection (PR), and biopsy (Bx) on overall survival. Survival prospects were summarized using direct adjusted survival curves. A total of 342 pGB patients were identified, and 35.4 % of patients received a GTR, 28.8 % PR, 17.3 % Bx, and 17.0 % did not undergo surgery. In our cohort, a median overall survival of 12 months was observed with 1-, 2-, and 5-year survival rates of 51.7, 28.3, and 15.7 %, respectively. EOR was a predictor of survival in both the multivariate- (P

Background and Purpose— Intracranial carotid artery calcification (ICAC) is one of the most important risk factors for stroke. Although several environmental risk factors for ICAC have been identified, its genetic background remains unclear. Methods— Between 2003 and 2006, 2034 participants from the prospective population-based Rotterdam study (mean age: 69.6±6.8 years; 51.7% female) underwent computed tomography to quantify vascular calcification in the intracranial internal carotid artery. Blood samples were drawn for genotyping. Genotypes of the participants were imputed to the 1000 Genomes reference panel to generate genetic relationship matrices for the estimation of the heritability of ICAC volume. Adjustments were made for age and sex. Subsequently, genome-wide association analyses were performed to identify specific variants. Results— The age- and sex-adjusted heritability ( h 2 ) of ICAC was 47% [standard error (SE): 19%; P =0.009]. Genome-wide association analyses identified a variant on chromosome 9p21.3 (rs1537372; N=2034; P =4.75×10 −9 ) and 1 variant on chromosome 11p11.2 (rs11038042, N=2034; P =3.27×10 −8 ) that were significantly associated with ICAC volume. Rs1537372 replicated in an independent sample of 716 stroke patients ( P combined =1.38×10 −10 ). Conclusions— ICAC volume is a heritable trait, which is partly explained by common genetic variation. We identified specific genetic variants associated with ICAC, which given the importance of ICAC in stroke risk, needs replication in larger-scale studies to further elucidate its genetic basis.

Background and Purpose— Genome-wide association studies have identified single-nucleotide polymorphisms (SNPs) for intracranial aneurysms in clinical samples. In addition, SNPs have been discovered for blood pressure, one of the strongest risk factors for intracranial aneurysms. We studied the role of these genetic variants on occurrence and size of unruptured intracranial aneurysms, discovered incidentally in a general community-dwelling population. Methods— In 4890 asymptomatic participants from the Rotterdam Study, 120 intracranial aneurysms were identified on brain imaging and segmented for maximum diameter and volume. Genetic risk scores (GRS) were calculated for intracranial aneurysms (10 SNPs), systolic blood pressure (33 SNPs), and diastolic blood pressure (41 SNPs). Results— The GRS for intracranial aneurysms was not statistically significantly associated with presence of aneurysms in this population (OR, 1.16; 95% CI, 0.96–1.40; P =0.119), but showed a significant association with both maximum diameter (difference in log-transformed mm per SD increase of GRS, 0.10; 95% CI, 0.02–0.19; P =0.018) and volume (difference in log-transformed µL per SD increase of GRS, 0.21; 95% CI, 0.01–0.41; P =0.040) of aneurysms. GRSs for blood pressures were associated with neither presence nor size of aneurysms. Conclusions— Genetic variants previously identified for intracranial aneurysms in clinical studies relate to the size rather than the presence of incidentally discovered, unruptured intracranial aneurysms in the general population.

Human gait is a complex neurological and musculoskeletal function, of which the genetic basis remains largely unknown. To determine the influence of common genetic variants on gait parameters, we studied 2,946 participants of the Rotterdam Study, a population-based cohort of unrelated elderly individuals. We assessed 30 gait parameters using an electronic walkway, which yielded seven independent gait domains after principal component analysis. Genotypes of participants were imputed to the 1,000 Genomes reference panel for generating genetic relationship matrices to estimate heritability of gait parameters, and for subsequent genome-wide association scans (GWASs) to identify specific variants. Gait domains with the highest age- and sex-adjusted heritability were Variability (h (2) = 61%), Rhythm (37%), and Tandem (32%). For other gait domains, heritability estimates attenuated after adjustment for height and weight. Genome-wide association scans identified a variant on 1p22.3 that was significantly associated with single support time, a variable from the Rhythm domain (rs72953990; N = 2,946; beta [SE] = 0.0069 (0.0012), p = 2.30x10(-8)). This variant did not replicate in an independent sample (N = 362; p = .78). In conclusion, human gait has highly heritable components that are explained by common genetic variation, which are partly attributed to height and weight. Collaborative efforts are needed to identify robust single variant associations for the heritable parameters.

Background and Purpose— Dilated Virchow–Robin spaces are an emerging neuroimaging biomarker, but their assessment on MRI needs standardization. Methods— We developed a rating method for dilated Virchow–Robin spaces in 4 brain regions (centrum semiovale, basal ganglia, hippocampus, and mesencephalon) and tested its reliability in a total of 125 MRI scans from 2 population-based studies. Six investigators with varying levels of experience performed the ratings. Intraclass correlation coefficients were calculated to determine intra- and interrater reliability. Results— Intrarater reliability was excellent for all 4 regions (intraclass correlation coefficient, >0.8). Interrater reliability was excellent for the centrum semiovale and hippocampus (intraclass correlation coefficient, >0.8) and good for the basal ganglia and mesencephalon (intraclass correlation coefficient, 0.6–0.8). This did not differ between the cohorts or experience levels. Conclusions— We describe a reliable rating method that can facilitate pathogenic and prognostic research on dilated Virchow–Robin spaces using MRI.

Background: Vascular dementia is a common disorder resulting in considerable morbidity and mortality. Determining the extent to which genes play a role in disease susceptibility and their pathophysiological mechanisms could improve our understanding of vascular dementia, leading to a potential translation of this knowledge to clinical practice. Discussion: In this review, we discuss what is currently known about the genetics of vascular dementia. The identification of causal genes remains limited to monogenic forms of the disease, with findings for sporadic vascular dementia being less robust. However, progress in genetic research on associated phenotypes, such as cerebral small vessel disease, Alzheimer’s disease, and stroke, have the potential to inform on the genetics of vascular dementia. We conclude by providing an overview of future developments in the field and how such work could impact patients and clinicians. Conclusion: The genetic background of vascular dementia is well established for monogenic disorders, but remains relatively obscure for the sporadic form. More work is needed for providing robust findings that might eventually lead to clinical translation.

Purpose To present an updated prevalence estimate for incidental findings on brain magnetic resonance (MR) images and provide information on clinical relevance, including natural course, over a period of up to 9 years. Materials and Methods This study was approved by the institutional review board and all participants gave informed consent. In a prospective population-based setting, structural brain MR imaging was performed in 5800 participants (mean age, 64.9 years; 3194 women [55.1%]). Trained reviewers recorded abnormalities, which were subsequently evaluated by neuroradiologists. The prevalence with 95% confidence interval (CI) of incidental findings was determined, and clinical management of findings that required the attention of a medical specialist was followed. Follow-up imaging in the study context provided information on the natural course of findings that were not referred. Results In 549 of 5800 participants (9.5% [95% CI: 8.7%, 10.3%]), incidental findings were found, of which meningiomas (143 of 5800; 2.5% [95% CI: 2.1%, 2.9%]) and cerebral aneurysms (134 of 5800; 2.3% [95% CI: 2.0%, 2.7%]) were most common. A total of 188 participants were referred to medical specialists for incidental findings (3.2% [95% CI: 2.8%, 3.7%]). Of these, 144 (76.6% [95% CI: 70.1%, 82.1%]) either underwent a wait-and-see policy or were discharged after the initial clinical visit. The majority of meningiomas and virtually all aneurysms not referred or referred but untreated remained stable in size during follow-up. Conclusion Incidental findings at brain MR imaging that necessitate further diagnostic evaluation occur in over 3% of the general middle-aged and elderly population, but are mostly without direct clinical consequences. © RSNA, 2016.

Effective prevention of Alzheimer's disease (AD) requires the development of risk prediction tools permitting preclinical intervention. We constructed a genetic risk score (GRS) comprising common genetic variants associated with AD, evaluated its association with incident AD and assessed its capacity to improve risk prediction over traditional models based on age, sex, education, and APOE epsilon 4. In eight prospective cohorts included in the International Genomics of Alzheimer's Project (IGAP), we derived weighted sum of risk alleles from the 19 top SNPs reported by the IGAPGWAS in participants aged 65 and older without prevalent dementia. Hazard ratios (HR) of incident AD were estimated in Cox models. Improvement in risk prediction was measured by the difference in C-index (Delta-C), the integrated discrimination improvement (IDI) and continuous net reclassification improvement (NRI>0). Overall, 19,687 participants at risk were included, of whom 2,782 developed AD. The GRS was associated with a 17% increase in AD risk (pooled HR=1.17; 95% CI = [1.13-1.21] per standard deviation increase in GRS; p-value = 2.86x10(-16)). This association was stronger among persons with at least one APOE epsilon 4 allele (HRGRS = 1.24; 95% CI = [1.15-1.34]) than in others (HRGRS = 1.13; 95% CI = [1.08-1.18]; p(interaction) = 3.45x10(-2)). Risk prediction after seven years of follow-up showed a small improvement when adding the GRS to age, sex, APOE epsilon 4, and education (Delta-Cindex = 0.0043 [0.0019-0.0067]). Similar patterns were observed for IDI and NRI>0. In conclusion, a risk score incorporating common genetic variation outside the APOE epsilon 4 locus improved AD risk prediction and may facilitate risk stratification for prevention trials.

Introduction\ud \ud Many consequences of cerebrovascular disease are identifiable by magnetic resonance imaging (MRI), but variation in methods limits multicenter studies and pooling of data. The European Union Joint Program on Neurodegenerative Diseases (EU JPND) funded the HARmoNizing Brain Imaging MEthodS for VaScular Contributions to Neurodegeneration (HARNESS) initiative, with a focus on cerebral small vessel disease.\ud \ud \ud \ud Methods\ud \ud Surveys, teleconferences, and an in-person workshop were used to identify gaps in knowledge and to develop tools for harmonizing imaging and analysis.\ud \ud \ud \ud Results\ud \ud A framework for neuroimaging biomarker development was developed based on validating repeatability and reproducibility, biological principles, and feasibility of implementation. The status of current MRI biomarkers was reviewed. A website was created at www.harness-neuroimaging.org with acquisition protocols, a software database, rating scales and case report forms, and a deidentified MRI repository.\ud \ud \ud \ud Conclusions\ud \ud The HARNESS initiative provides resources to reduce variability in measurement in MRI studies of cerebral small vessel disease.

Background and Purpose— Perivascular enlargement in the brain is a putative imaging marker for microvascular brain damage, but this link has not yet been confirmed using direct in vivo visualization of small vessels. We investigated the relation between microvascular calibers on retinal imaging and enlarged perivascular spaces (ePVSs) on brain magnetic resonance imaging. Methods— We included 704 participants from the Rotterdam study. Retinal arteriolar and venular calibers were measured semiautomatically on fundus photographs. ePVSs were counted in the centrum semiovale, basal ganglia, hippocampus, and mesencephalon, using a standardized rating method. We determined the association between retinal microvascular calibers and ePVSs with negative binomial regression models, adjusting for age, sex, the other vascular caliber, structural brain magnetic resonance imaging markers, and cardiovascular risk factors. Results— Both narrower arteriolar and wider venular calibers were associated with more ePVSs in the centrum semiovale and hippocampal region. Rate ratios (95% confidence interval) for arterioles in the centrum semiovale and hippocampus were 1.07 (1.01–1.14) and 1.13 (1.04–1.22), respectively, and for venules 1.08 (1.01–1.16) and 1.09 (1.00–1.18), respectively. These associations were independent from other brain magnetic resonance imaging markers and cardiovascular risk factors. Conclusions— Retinal microvascular calibers are related to ePVSs, confirming the putative link between microvascular damage and ePVSs.

Background: Serum total cholesterol and its fractions are inversely associated with intracerebral hemorrhages (ICH) and their potential subclinical precursor, cerebral microbleeds. To ascertain whether there is a genetic basis for this inverse association, we studied established genetic loci for serum total, LDL, and HDL cholesterol, and triglycerides in their association with ICH and microbleeds. Methods: Data on 161 genetic variants for serum lipids was collected in 9011 stroke-free participants (mean age 65.8, SD 10.2; 57.9% women) of the population-based Rotterdam Study. Participants were followed from baseline (1997-2005) up to 2013 for the occurrence of ICH. A subset of 4179 participants underwent brain MRI for microbleed assessment between 2005 and 2011. We computed genetic risk scores (GRS) for the joint effect of lipid variants. Cox proportional hazards and logistic regression models were used to investigate the association of GRS of lipid fractions with ICH and microbleeds. Results: After a mean follow-up of 8.7 (SD 4.1) years, 67 (0.7%) participants suffered an ICH. Microbleed prevalence was 19.6%. Higher genetic load for high serum total and LDL cholesterol was associated with an increased risk of ICH. Higher genetic load for high serum LDL cholesterol was borderline associated with a higher prevalence of multiple lobar microbleeds. Conclusions: Genetic susceptibility for high serum total and LDL cholesterol is positively associated with incident ICH and borderline associated with multiple lobar microbleeds. We did not find a genetic basis for the previously reported inverse association between serum lipid levels and ICH. (C) 2016 Elsevier Ireland Ltd. All rights reserved.

Background and Purpose— White matter lesion (WML) progression on magnetic resonance imaging is related to cognitive decline and stroke, but its determinants besides baseline WML burden are largely unknown. Here, we estimated heritability of WML progression, and sought common genetic variants associated with WML progression in elderly participants from the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium. Methods— Heritability of WML progression was calculated in the Framingham Heart Study. The genome-wide association study included 7773 elderly participants from 10 cohorts. To assess the relative contribution of genetic factors to progression of WML, we compared in 7 cohorts risk models including demographics, vascular risk factors plus single-nucleotide polymorphisms that have been shown to be associated cross-sectionally with WML in the current and previous association studies. Results— A total of 1085 subjects showed WML progression. The heritability estimate for WML progression was low at 6.5%, and no single-nucleotide polymorphisms achieved genome-wide significance ( P −8 ). Four loci were suggestive ( P −5 ) of an association with WML progression: 10q24.32 (rs10883817, P =1.46×10 −6 ); 12q13.13 (rs4761974, P =8.71×10 −7 ); 20p12.1 (rs6135309, P =3.69×10 −6 ); and 4p15.31 (rs7664442, P =2.26×10 −6 ). Variants that have been previously related to WML explained only 0.8% to 11.7% more of the variance in WML progression than age, vascular risk factors, and baseline WML burden. Conclusions— Common genetic factors contribute little to the progression of age-related WML in middle-aged and older adults. Future research on determinants of WML progression should focus more on environmental, lifestyle, or host-related biological factors.

Background— The burden of cerebral white matter hyperintensities (WMH) is associated with an increased risk of stroke, dementia, and death. WMH are highly heritable, but their genetic underpinnings are incompletely characterized. To identify novel genetic variants influencing WMH burden, we conducted a meta-analysis of multiethnic genome-wide association studies. Methods and Results— We included 21 079 middle-aged to elderly individuals from 29 population-based cohorts, who were free of dementia and stroke and were of European (n=17 936), African (n=1943), Hispanic (n=795), and Asian (n=405) descent. WMH burden was quantified on MRI either by a validated automated segmentation method or a validated visual grading scale. Genotype data in each study were imputed to the 1000 Genomes reference. Within each ethnic group, we investigated the relationship between each single-nucleotide polymorphism and WMH burden using a linear regression model adjusted for age, sex, intracranial volume, and principal components of ancestry. A meta-analysis was conducted for each ethnicity separately and for the combined sample. In the European descent samples, we confirmed a previously known locus on chr17q25 ( P =2.7×10 −19 ) and identified novel loci on chr10q24 ( P =1.6×10 −9 ) and chr2p21 ( P =4.4×10 −8 ). In the multiethnic meta-analysis, we identified 2 additional loci, on chr1q22 ( P =2.0×10 −8 ) and chr2p16 ( P =1.5×10 −8 ). The novel loci contained genes that have been implicated in Alzheimer disease (chr2p21 and chr10q24), intracerebral hemorrhage (chr1q22), neuroinflammatory diseases (chr2p21), and glioma (chr10q24 and chr2p16). Conclusions— We identified 4 novel genetic loci that implicate inflammatory and glial proliferative pathways in the development of WMH in addition to previously proposed ischemic mechanisms.

INTRODUCTION: Virchow-Robin spaces (VRS), or perivascular spaces, are compartments of interstitial fluid enclosing cerebral blood vessels and are potential imaging markers of various underlying brain pathologies. Despite a growing interest in the study of enlarged VRS, the heterogeneity in rating and quantification methods combined with small sample sizes have so far hampered advancement in the field.METHODS: The Uniform Neuro-Imaging of Virchow-Robin Spaces Enlargement (UNIVRSE) consortium was established with primary aims to harmonize rating and analysis (www.uconsortium.org). The UNIVRSE consortium brings together 13 (sub)cohorts from five countries, totaling 16,000 subjects and over 25,000 scans. Eight different magnetic resonance imaging protocols were used in the consortium.RESULTS: VRS rating was harmonized using a validated protocol that was developed by the two founding members, with high reliability independent of scanner type, rater experience, or concomitant brain pathology. Initial analyses revealed risk factors for enlarged VRS including increased age, sex, high blood pressure, brain infarcts, and white matter lesions, but this varied by brain region.DISCUSSION: Early collaborative efforts between cohort studies with respect to data harmonization and joint analyses can advance the field of population (neuro)imaging. The UNIVRSE consortium will focus efforts on other potential correlates of enlarged VRS, including genetics, cognition, stroke, and dementia.