The Neurobiology of Life Course Socioeconomic Conditions and Associated Cognitive Performance in Middle to Late Adulthood.

Despite major advances, our understanding of the neurobiology of life course socioeconomic conditions is still scarce. This study aimed to provide insight into the pathways linking socioeconomic exposures-household income, last known occupational position, and life course socioeconomic trajectories-with brain microstructure and cognitive performance in middle to late adulthood. We assessed socioeconomic conditions alongside quantitative relaxometry and diffusion-weighted magnetic resonance imaging indicators of brain tissue microstructure and cognitive performance in a sample of community-dwelling men and women (N = 751, aged 50-91 years). We adjusted the applied regression analyses and structural equation models for the linear and nonlinear effects of age, sex, education, cardiovascular risk factors, and the presence of depression, anxiety, and substance use disorders. Individuals from lower-income households showed signs of advanced brain white matter (WM) aging with greater mean diffusivity (MD), lower neurite density, lower myelination, and lower iron content. The association between household income and MD was mediated by neurite density (B = 0.084, p = 0.003) and myelination (B = 0.019, p = 0.009); MD partially mediated the association between household income and cognitive performance (B = 0.017, p < 0.05). Household income moderated the relation between WM microstructure and cognitive performance, such that greater MD, lower myelination, or lower neurite density was only associated with poorer cognitive performance among individuals from lower-income households. Individuals from higher-income households showed preserved cognitive performance even with greater MD, lower myelination, or lower neurite density. These findings provide novel mechanistic insights into the associations between socioeconomic conditions, brain anatomy, and cognitive performance in middle to late adulthood.

Degeneracy and disordered brain networks in psychiatric patients using multivariate structural covariance analyzes.

Introduction: In this study, we applied multivariate methods to identify brain regions that have a critical role in shaping the connectivity patterns of networks associated with major psychiatric diagnoses, including schizophrenia (SCH), major depressive disorder (MDD) and bipolar disorder (BD) and healthy controls (HC). We used T1w images from 164 subjects: Schizophrenia (n = 17), bipolar disorder (n = 25), major depressive disorder (n = 68) and a healthy control group (n = 54). Methods: We extracted regions of interest (ROIs) using a method based on the SHOOT algorithm of the SPM12 toolbox. We then performed multivariate structural covariance between the groups. For the regions identified as significant in t term of their covariance value, we calculated their eigencentrality as a measure of the influence of brain regions within the network. We applied a significance threshold of p = 0.001. Finally, we performed a cluster analysis to determine groups of regions that had similar eigencentrality profiles in different pairwise comparison networks in the observed groups. Results: As a result, we obtained 4 clusters with different brain regions that were diagnosis-specific. Cluster 1 showed the strongest discriminative values between SCH and HC and SCH and BD. Cluster 2 had the strongest discriminative value for the MDD patients, cluster 3 - for the BD patients. Cluster 4 seemed to contribute almost equally to the discrimination between the four groups. Discussion: Our results suggest that we can use the multivariate structural covariance method to identify specific regions that have higher predictive value for specific psychiatric diagnoses. In our research, we have identified brain signatures that suggest that degeneracy shapes brain networks in different ways both within and across major psychiatric disorders.

Statistical analyses of motion-corrupted MRI relaxometry data computed from multiple scans.

BACKGROUND: Consistent noise variance across data points (i.e. homoscedasticity) is required to ensure the validity of statistical analyses of MRI data conducted using linear regression methods. However, head motion leads to degradation of image quality, introducing noise heteroscedasticity into ordinary-least square analyses. NEW METHOD: The recently introduced QUIQI method restores noise homoscedasticity by means of weighted least square analyses in which the weights, specific for each dataset of an analysis, are computed from an index of motion-induced image quality degradation. QUIQI was first demonstrated in the context of brain maps of the MRI parameter R2 * , which were computed from a single set of images with variable echo time. Here, we extend this framework to quantitative maps of the MRI parameters R1, R2 * , and MTsat, computed from multiple sets of images. RESULTS: QUIQI restores homoscedasticity in analyses of quantitative MRI data computed from multiple scans. QUIQI allows for optimization of the noise model by using metrics quantifying heteroscedasticity and free energy. COMPARISON WITH EXISTING METHODS: QUIQI restores homoscedasticity more effectively than insertion of an image quality index in the analysis design and yields higher sensitivity than simply removing the datasets most corrupted by head motion from the analysis. CONCLUSION: QUIQI provides an optimal approach to group-wise analyses of a range of quantitative MRI parameter maps that is robust to inherent homoscedasticity.

Parkinson's disease may disrupt overlapping subthalamic nucleus and pallidal motor networks.

There is an ongoing debate about differential clinical outcome and associated adverse effects of deep brain stimulation (DBS) in Parkinson's disease (PD) targeting the subthalamic nucleus (STN) or the globus pallidus pars interna (GPi). Given that functional connectivity profiles suggest beneficial DBS effects within a common network, the empirical evidence about the underlying anatomical circuitry is still scarce. Therefore, we investigate the STN and GPi-associated structural covariance brain patterns in PD patients and healthy controls. We estimate GPi's and STN's whole-brain structural covariance from magnetic resonance imaging (MRI) in a normative mid- to old-age community-dwelling cohort (n = 1184) across maps of grey matter volume, magnetization transfer (MT) saturation, longitudinal relaxation rate (R1), effective transversal relaxation rate (R2*) and effective proton density (PD*). We compare these with the structural covariance estimates in patients with idiopathic PD (n = 32) followed by validation using a reduced size controls' cohort (n = 32). In the normative data set, we observed overlapping spatially distributed cortical and subcortical covariance patterns across maps confined to basal ganglia, thalamus, motor, and premotor cortical areas. Only the subcortical and midline motor cortical areas were confirmed in the reduced size cohort. These findings contrasted with the absence of structural covariance with cortical areas in the PD cohort. We interpret with caution the differential covariance maps of overlapping STN and GPi networks in patients with PD and healthy controls as correlates of motor network disruption. Our study provides face validity to the proposed extension of the currently existing structural covariance methods based on morphometry features to multiparameter MRI sensitive to brain tissue microstructure.

Topography of associations between cardiovascular risk factors and myelin loss in the ageing human brain.

Our knowledge of the mechanisms underlying the vulnerability of the brain's white matter microstructure to cardiovascular risk factors (CVRFs) is still limited. We used a quantitative magnetic resonance imaging (MRI) protocol in a single centre setting to investigate the cross-sectional association between CVRFs and brain tissue properties of white matter tracts in a large community-dwelling cohort (n = 1104, age range 46-87 years). Arterial hypertension was associated with lower myelin and axonal density MRI indices, paralleled by higher extracellular water content. Obesity showed similar associations, though with myelin difference only in male participants. Associations between CVRFs and white matter microstructure were observed predominantly in limbic and prefrontal tracts. Additional genetic, lifestyle and psychiatric factors did not modulate these results, but moderate-to-vigorous physical activity was linked to higher myelin content independently of CVRFs. Our findings complement previously described CVRF-related changes in brain water diffusion properties pointing towards myelin loss and neuroinflammation rather than neurodegeneration.

CYP2C19 expression modulates affective functioning and hippocampal subiculum volume-a large single-center community-dwelling cohort study.

Given controversial findings of reduced depressive symptom severity and increased hippocampus volume in CYP2C19 poor metabolizers, we sought to provide empirical evidence from a large-scale single-center longitudinal cohort in the community-dwelling adult population-Colaus|PsyCoLaus in Lausanne, Switzerland (n = 4152). We looked for CYP2C19 genotype-related behavioral and brain anatomy patterns using a comprehensive set of psychometry, water diffusion- and relaxometry-based magnetic resonance imaging (MRI) data (BrainLaus, n = 1187). Our statistical models tested for differential associations between poor metabolizer and other metabolizer status with imaging-derived indices of brain volume and tissue properties that explain individuals' current and lifetime mood characteristics. The observed association between CYP2C19 genotype and lifetime affective status showing higher functioning scores in poor metabolizers, was mainly driven by female participants (ß = 3.9, p = 0.010). There was no difference in total hippocampus volume between poor metabolizer and other metabolizer, though there was higher subiculum volume in the right hippocampus of poor metabolizers (ß = 0.03, pFDRcorrected = 0.036). Our study supports the notion of association between mood phenotype and CYP2C19 genotype, however, finds no evidence for concomitant hippocampus volume differences, with the exception of the right subiculum.

Abnormal brain iron accumulation in obstructive sleep apnea: A quantitative MRI study in the HypnoLaus cohort.

Obstructive sleep apnea syndrome (OSA) may be a risk factor for Alzheimer's disease. One of the hallmarks of Alzheimer's disease is disturbed iron homeostasis leading to abnormal iron deposition in brain tissue. To date, there is no empirical evidence to support the hypothesis of altered brain iron homeostasis in patients with obstructive sleep apnea as well. Data were analysed from 773 participants in the HypnoLaus study (mean age 55.9 ± 10.3 years) who underwent polysomnography and brain MRI. Cross-sectional associations were tested between OSA parameters and the MRI effective transverse relaxation rate (R2*) - indicative of iron content - in 68 grey matter regions, after adjustment for confounders. The group with severe OSA (apnea-hypopnea index ≥30/h) had higher iron levels in the left superior frontal gyrus (F3,760  = 4.79, p = 0.003), left orbital gyri (F3,760  = 5.13, p = 0.002), right and left middle temporal gyrus (F3,760  = 4.41, p = 0.004 and F3,760  = 13.08, p < 0.001, respectively), left angular gyrus (F3,760  = 6.29, p = 0.001), left supramarginal gyrus (F3,760  = 4.98, p = 0.003), and right cuneus (F3,760  = 7.09, p < 0.001). The parameters of nocturnal hypoxaemia were all consistently associated with higher iron levels. Measures of sleep fragmentation had less consistent associations with iron content. This study provides the first evidence of increased brain iron levels in obstructive sleep apnea. The observed iron changes could reflect underlying neuropathological processes that appear to be driven primarily by hypoxaemic mechanisms.

Restoring statistical validity in group analyses of motion-corrupted MRI data.

Motion during the acquisition of magnetic resonance imaging (MRI) data degrades image quality, hindering our capacity to characterise disease in patient populations. Quality control procedures allow the exclusion of the most affected images from analysis. However, the criterion for exclusion is difficult to determine objectively and exclusion can lead to a suboptimal compromise between image quality and sample size. We provide an alternative, data-driven solution that assigns weights to each image, computed from an index of image quality using restricted maximum likelihood. We illustrate this method through the analysis of quantitative MRI data. The proposed method restores the validity of statistical tests, and performs near optimally in all brain regions, despite local effects of head motion. This method is amenable to the analysis of a broad type of MRI data and can accommodate any measure of image quality.

Application of Mass Multivariate Analysis on Neuroimaging Data Sets for Precision Diagnostics of Depression.

We used the Mass Multivariate Method on structural, resting-state, and task-related fMRI data from two groups of patients with schizophrenia and depression in order to define several regions of significant relevance to the differential diagnosis of those conditions. The regions included the left planum polare (PP), the left opercular part of the inferior frontal gyrus (OpIFG), the medial orbital gyrus (MOrG), the posterior insula (PIns), and the parahippocampal gyrus (PHG). This study delivered evidence that a multimodal neuroimaging approach can potentially enhance the validity of psychiatric diagnoses. Structural, resting-state, or task-related functional MRI modalities cannot provide independent biomarkers. Further studies need to consider and implement a model of incremental validity combining clinical measures with different neuroimaging modalities to discriminate depressive disorders from schizophrenia. Biological signatures of disease on the level of neuroimaging are more likely to underpin broader nosological entities in psychiatry.

Clinical phenotype modulates brain's myelin and iron content in temporal lobe epilepsy.

Temporal lobe epilepsy (TLE) is associated with brain pathology extending beyond temporal lobe structures. We sought to look for informative patterns of brain tissue properties in TLE that go beyond the established morphometry differences. We hypothesised that volume differences, particularly in hippocampus, will be paralleled by changes in brain microstructure. The cross-sectional study included TLE patients (n = 25) from a primary care center and sex-/age-matched healthy controls (n = 55). We acquired quantitative relaxometry-based magnetic resonance imaging (MRI) data yielding whole-brain maps of grey matter volume, magnetization transfer (MT) saturation, and effective transverse relaxation rate R2* indicative for brain tissue myelin and iron content. For statistical analysis, we used the computational anatomy framework of voxel-based morphometry and voxel-based quantification. There was a positive correlation between seizure activity and MT saturation measures in the ipsilateral hippocampus, paralleled by volume differences bilaterally. Disease duration correlated positively with iron content in the mesial temporal lobe, while seizure freedom was associated with a decrease of iron in the very same region. Our findings demonstrate the link between TLE clinical phenotype and brain anatomy beyond morphometry differences to show the impact of disease burden on specific tissue properties. We provide direct evidence for the differential effect of clinical phenotype characteristics on processes involving tissue myelin and iron in mesial temporal lobe structures. This study offers a proof-of-concept for the investigation of novel imaging biomarkers in focal epilepsy.

Machine Learning for Health: Algorithm Auditing & Quality Control.

Developers proposing new machine learning for health (ML4H) tools often pledge to match or even surpass the performance of existing tools, yet the reality is usually more complicated. Reliable deployment of ML4H to the real world is challenging as examples from diabetic retinopathy or Covid-19 screening show. We envision an integrated framework of algorithm auditing and quality control that provides a path towards the effective and reliable application of ML systems in healthcare. In this editorial, we give a summary of ongoing work towards that vision and announce a call for participation to the special issue  Machine Learning for Health: Algorithm Auditing & Quality Control in this journal to advance the practice of ML4H auditing.

Apolipoprotein E allele 4 effects on Single-Subject Gray Matter Networks in Mild Cognitive Impairment.

There is evidence that gray matter networks are disrupted in Mild Cognitive Impairment (MCI) and associated with cognitive impairment and faster disease progression. However, it remains unknown how these alterations are related to the presence of Apolipoprotein E isoform E4 (ApoE4), the most prominent genetic risk factor for late-onset Alzheimer's disease (AD). To investigate this topic at the individual level, we explore the impact of ApoE4 and the disease progression on the Single-Subject Gray Matter Networks (SSGMNets) using the graph theory approach. Our data sample comprised 200 MCI patients selected from the ADNI database, classified as non-Converters and Converters (will progress into AD). Each group included 50 ApoE4-positive ('Carriers', ApoE4 + ) and 50 ApoE4-negative ('non-Carriers', ApoE4-). The SSGMNets were estimated from structural MRIs at two-time points: baseline and conversion. We investigated whether altered network topological measures at baseline and their rate of change (RoC) between baseline and conversion time points were associated with ApoE4 and disease progression. We also explored the correlation of SSGMNets attributes with general cognition score (MMSE), memory (ADNI-MEM), and CSF-derived biomarkers of AD (Aß42, T-tau, and P-tau). Our results showed that ApoE4 and the disease progression modulated the global topological network properties independently but not in their RoC. MCI converters showed a lower clustering index in several regions associated with neurodegeneration in AD. The SSGMNets' topological organization was revealed to be able to predict cognitive and memory measures. The findings presented here suggest that SSGMNets could indeed be used to identify MCI ApoE4 Carriers with a high risk for AD progression.

Brain tissue properties link cardio-vascular risk factors, mood and cognitive performance in the CoLaus|PsyCoLaus epidemiological cohort.

Given the controversy about the impact of modifiable risk factors on mood and cognition in ageing, we sought to investigate the associations between cardio-vascular risk, mental health, cognitive performance and brain anatomy in mid- to old age. We analyzed a set of risk factors together with multi-parameter magnetic resonance imaging (MRI) in the CoLaus|PsyCoLaus cohort (n > 1200). Cardio-vascular risk was associated with differences in brain tissue properties - myelin, free tissue water, iron content - and regional brain volumes that we interpret in the context of micro-vascular hypoxic lesions and neurodegeneration. The interaction between clinical subtypes of major depressive disorder and cardio-vascular risk factors showed differential associations with brain structure depending on individuals' lifetime trajectory. There was a negative correlation between melancholic depression, anxiety and MRI markers of myelin and iron content in the hippocampus and anterior cingulate. Verbal memory and verbal fluency performance were positively correlated with left amygdala volumes. The concomitant analysis of brain morphometry and tissue properties allowed for a neuro-biological interpretation of the link between modifiable risk factors and brain health.

Gradient of electro-convulsive therapy's antidepressant effects along the longitudinal hippocampal axis.

Despite decades of successful treatment of therapy-resistant depression and major scientific advances in the field, our knowledge about electro-convulsive therapy's (ECT) mechanisms of action is still scarce. Building on strong empirical evidence for ECT-induced hippocampus anatomy changes, we sought to test the hypothesis that ECT has a differential impact along the hippocampus longitudinal axis. We acquired behavioural and brain anatomy magnetic resonance imaging (MRI) data in patients with depressive episode undergoing ECT (n = 9) or pharmacotherapy (n = 24) and healthy controls (n = 30) at two time points 3 months apart. Using whole-brain voxel-based statistical parametric mapping and topographic analysis focused on the hippocampus, we observed ECT-induced gradient of grey matter volume increase along the hippocampal longitudinal axis with predominant impact on its anterior portion. Clinical outcome measures showed strong correlations with both baseline volume and rate of ECT-induced change exclusively for the anterior, but not posterior hippocampus. We interpret our findings confined to the anterior hippocampus and amygdala as additional evidence of the regional specific impact of ECT that unfolds its beneficial effect on depression via the "limbic" system. Main limitations of the study are patients' polypharmacy, heterogeneity of psychiatric diagnosis, and long-time interval between scans.

Apolipoprotein E4 effects on topological brain network organization in mild cognitive impairment.

The Apolipoprotein E isoform E4 (ApoE4) is consistently associated with an elevated risk of developing late-onset Alzheimer's Disease (AD); however, less is known about the potential genetic modulation of the brain networks organization during prodromal stages like Mild Cognitive Impairment (MCI). To investigate this issue during this critical stage, we used a dataset with a cross-sectional sample of 253 MCI patients divided into ApoE4-positive (?Carriers') and ApoE4-negative ('non-Carriers'). We estimated the cortical thickness (CT) from high-resolution T1-weighted structural magnetic images to calculate the correlation among anatomical regions across subjects and build the CT covariance networks (CT-Nets). The topological properties of CT-Nets were described through the graph theory approach. Specifically, our results showed a significant decrease in characteristic path length, clustering-index, local efficiency, global connectivity, modularity, and increased global efficiency for Carriers compared to non-Carriers. Overall, we found that ApoE4 in MCI shaped the topological organization of CT-Nets. Our results suggest that in the MCI stage, the ApoE4 disrupting the CT correlation between regions may be due to adaptive mechanisms to sustain the information transmission across distant brain regions to maintain the cognitive and behavioral abilities before the occurrence of the most severe symptoms.

Mapping grip force to motor networks.

AIM: There is ongoing debate about the role of cortical and subcortical brain areas in force modulation. In a whole-brain approach, we sought to investigate the anatomical basis of grip force whilst acknowledging interindividual differences in connectivity patterns. We tested if brain lesion mapping in patients with unilateral motor deficits can inform whole-brain structural connectivity analysis in healthy controls to uncover the networks underlying grip force. METHODS: Using magnetic resonance imaging (MRI) and whole-brain voxel-based morphometry in chronic stroke patients (n=55) and healthy controls (n=67), we identified the brain regions in both grey and white matter significantly associated with grip force strength. The resulting statistical parametric maps (SPMs) provided seed areas for whole-brain structural covariance analysis in a large-scale community dwelling cohort (n=977) that included beyond volume estimates, parameter maps sensitive to myelin, iron and tissue water content. RESULTS: The SPMs showed symmetrical bilateral clusters of correlation between upper limb motor performance, basal ganglia, posterior insula and cortico-spinal tract. The covariance analysis with the seed areas derived from the SPMs demonstrated a widespread anatomical pattern of brain volume and tissue properties, including both cortical, subcortical nodes of motor networks and sensorimotor areas projections. CONCLUSION: We interpret our covariance findings as a biological signature of brain networks implicated in grip force. The data-driven definition of seed areas obtained from chronic stroke patients showed overlapping structural covariance patterns within cortico-subcortical motor networks across different tissue property estimates. This cumulative evidence lends face validity of our findings and their biological plausibility.

Brain plasticity dynamics during tactile Braille learning in sighted subjects: Multi-contrast MRI approach.

A growing body of empirical evidence supports the notion of diverse neurobiological processes underlying learning-induced plasticity changes in the human brain. There are still open questions about how brain plasticity depends on cognitive task complexity, how it supports interactions between brain systems and with what temporal and spatial trajectory. We investigated brain and behavioural changes in sighted adults during 8-months training of tactile Braille reading whilst monitoring brain structure and function at 5 different time points. We adopted a novel multivariate approach that includes behavioural data and specific MRI protocols sensitive to tissue properties to assess local functional and structural and myelin changes over time. Our results show that while the reading network, located in the ventral occipitotemporal cortex, rapidly adapts to tactile input, sensory areas show changes in grey matter volume and intra-cortical myelin at different times. This approach has allowed us to examine and describe neuroplastic mechanisms underlying complex cognitive systems and their (sensory) inputs and (motor) outputs differentially, at a mesoscopic level.

Multivariate Analysis of Structural and Functional Neuroimaging Can Inform Psychiatric Differential Diagnosis.

Traditional psychiatric diagnosis has been overly reliant on either self-reported measures (introspection) or clinical rating scales (interviews). This produced the so-called explanatory gap with the bio-medical disciplines, such as neuroscience, which are supposed to deliver biological explanations of disease. In that context the neuro-biological and clinical assessment in psychiatry remained discrepant and incommensurable under conventional statistical frameworks. The emerging field of translational neuroimaging attempted to bridge the explanatory gap by means of simultaneous application of clinical assessment tools and functional magnetic resonance imaging, which also turned out to be problematic when analyzed with standard statistical methods. In order to overcome this problem our group designed a novel machine learning technique, multivariate linear method (MLM) which can capture convergent data from voxel-based morphometry, functional resting state and task-related neuroimaging and the relevant clinical measures. In this paper we report results from convergent cross-validation of biological signatures of disease in a sample of patients with schizophrenia as compared to depression. Our model provides evidence that the combination of the neuroimaging and clinical data in MLM analysis can inform the differential diagnosis in terms of incremental validity.

Medical Informatics Platform (MIP): A Pilot Study Across Clinical Italian Cohorts.

Introduction: With the shift of research focus to personalized medicine in Alzheimer's Dementia (AD), there is an urgent need for tools that are capable of quantifying a patient's risk using diagnostic biomarkers. The Medical Informatics Platform (MIP) is a distributed e-infrastructure federating large amounts of data coupled with machine-learning (ML) algorithms and statistical models to define the biological signature of the disease. The present study assessed (i) the accuracy of two ML algorithms, i.e., supervised Gradient Boosting (GB) and semi-unsupervised 3C strategy (Categorize, Cluster, Classify-CCC) implemented in the MIP and (ii) their contribution over the standard diagnostic workup. Methods: We examined individuals coming from the MIP installed across 3 Italian memory clinics, including subjects with Normal Cognition (CN, n = 432), Mild Cognitive Impairment (MCI, n = 456), and AD (n = 451). The GB classifier was applied to best discriminate the three diagnostic classes in 1,339 subjects, and the CCC strategy was used to refine the classical disease categories. Four dementia experts provided their diagnostic confidence (DC) of MCI conversion on an independent cohort of 38 patients. DC was based on clinical, neuropsychological, CSF, and structural MRI information and again with addition of the outcome from the MIP tools. Results: The GB algorithm provided a classification accuracy of 85% in a nested 10-fold cross-validation for CN vs. MCI vs. AD discrimination. Accuracy increased to 95% in the holdout validation, with the omission of each Italian clinical cohort out in turn. CCC identified five homogeneous clusters of subjects and 36 biomarkers that represented the disease fingerprint. In the DC assessment, CCC defined six clusters in the MCI population used to train the algorithm and 29 biomarkers to improve patients staging. GB and CCC showed a significant impact, evaluated as +5.99% of increment on physicians' DC. The influence of MIP on DC was rated from "slight" to "significant" in 80% of the cases. Discussion: GB provided fair results in classification of CN, MCI, and AD. CCC identified homogeneous and promising classes of subjects via its semi-unsupervised approach. We measured the effect of the MIP on the physician's DC. Our results pave the way for the establishment of a new paradigm for ML discrimination of patients who will or will not convert to AD, a clinical priority for neurology.

Remodeling of brain morphology in temporal lobe epilepsy.

BACKGROUND: Mesial temporal lobe epilepsy (TLE) is one of the most widespread neurological network disorders. Computational anatomy MRI studies demonstrate a robust pattern of cortical volume loss. Most statistical analyses provide information about localization of significant focal differences in a segregationist way. Multivariate Bayesian modeling provides a framework allowing inferences about inter-regional dependencies. We adopt this approach to answer following questions: Which structures within a pattern of dynamic epilepsy-associated brain anatomy reorganization best predict TLE pathology. Do these structures differ between TLE subtypes? METHODS: We acquire clinical and MRI data from TLE patients with and without hippocampus sclerosis (n = 128) additional to healthy volunteers (n = 120). MRI data were analyzed in the computational anatomy framework of SPM12 using classical mass-univariate analysis followed by multivariate Bayesian modeling. RESULTS: After obtaining TLE-associated brain anatomy pattern, we estimate predictive power for disease and TLE subtypes using Bayesian model selection and comparison. We show that ipsilateral para-/hippocampal regions contribute most to disease-related differences between TLE and healthy controls independent of TLE laterality and subtype. Prefrontal cortical changes are more discriminative for left-sided TLE, whereas thalamus and temporal pole for right-sided TLE. The presence of hippocampus sclerosis was linked to stronger involvement of thalamus and temporal lobe regions; frontoparietal involvement was predominant in absence of sclerosis. CONCLUSIONS: Our topology inferences on brain anatomy demonstrate a differential contribution of structures within limbic and extralimbic circuits linked to main effects of TLE and hippocampal sclerosis. We interpret our results as evidence for TLE-related spatial modulation of anatomical networks.