Resting-state Functional Connectomes Research Articles

EEG alpha band functional brain network correlates of cognitive performance in children after perinatal stroke

Mechanisms underlying cognitive impairment after perinatal stroke could be explained through brain network alterations. With aim to explore this connection, we conducted a matched test-control study to find a correlation between functional brain network properties and cognitive functions in children after perinatal stroke. First, we analyzed resting-state functional connectomes in the alpha frequency band from a 64-channel resting state EEG in 24 children with a history of perinatal stroke (12 with neonatal arterial ischemic stroke and 12 with neonatal hemorrhagic stroke) and compared them to the functional connectomes of 24 healthy controls. Next, all participants underwent cognitive evaluation. We analyzed the differences in functional brain network properties and cognitive abilities between groups and studied the correlation between network characteristics and specific cognitive functions. Functional brain networks after perinatal stroke had lower modularity, higher clustering coefficient, higher interhemispheric strength, higher characteristic path length and higher small world index. Modularity correlated positively with the IQ and processing speed, while clustering coefficient correlated negatively with IQ. Graph metrics, reflecting network segregation (clustering coefficient and small world index) correlated positively with a tendency to impulsive decision making, which also correlated positively with graph metrics, reflecting stronger functional connectivity (characteristic path length and interhemispheric strength). Our study suggests that specific cognitive functions correlate with different brain network properties and that functional network characteristics after perinatal stroke reflect poorer cognitive functioning.

Disruption in functional networks mediated tau spreading in Alzheimer's disease.

Alzheimer's disease may be conceptualized as a 'disconnection syndrome', characterized by the breakdown of neural connectivity within the brain as a result of amyloid-beta plaques, tau neurofibrillary tangles and other factors leading to progressive degeneration and shrinkage of neurons, along with synaptic dysfunction. It has been suggested that misfolded tau proteins spread through functional connections (known as 'prion-like' properties of tau). However, the local effect of tau spreading on the synaptic function and communication between regions is not well understood. I aimed to investigate how the spreading of tau aggregates through connections can locally influence functional connectivity. In total, the imaging data of 211 participants including 117 amyloid-beta-negative non-demented and 94 amyloid-beta-positive non-demented participants were recruited from the Alzheimer's Disease Neuroimaging Initiative. Furthermore, normative resting-state functional MRI connectomes were used to model tau spreading through functional connections, and functional MRI of the included participants was used to determine the effect of tau spreading on functional connectivity. I found that lower functional connectivity to tau epicentres is associated with tau spreading through functional connections in both amyloid-beta-negative and amyloid-beta-positive participants. Also, amyloid-beta-PET in tau epicentres mediated the association of tau spreading and functional connectivity to epicentres suggesting a partial mediating effect of amyloid-beta deposition in tau epicentres on the local effect of tau spreading on functional connectivity. My findings provide strong support for the notion that tau spreading through connection is locally associated with disrupted functional connectivity between tau epicentre and non-epicentre regions independent of amyloid-beta pathology. Also, I defined several groups based on the relationship between tau spreading and functional disconnection, which provides quantitative assessment to investigate susceptibility or resilience to functional disconnection related to tau spreading. I showed that amyloid-beta, other copathologies and the apolipoprotein E epsilon 4 allele can be a leading factor towards vulnerability to tau relative functional disconnection.

STREM2 is associated with attenuated tau aggregate accumulation in the presence of amyloid-β pathology.

Triggering Receptor Expressed on Myeloid Cell 2 (TREM2) plays a crucial role in the transition of microglia from a state of homeostasis to a state associated with the disease. Mutations in TREM2 are strongly linked with a higher risk of developing neurodegenerative diseases, including Alzheimer's disease. There have been contradictory findings regarding the potential detrimental or protective effects of microglial activation and TREM2-related microglial responses in Alzheimer's disease. Although previous studies reported increased CSF soluble TREM2 (sTREM2) in different clinical stages of Alzheimer's disease, the exact association between Alzheimer's disease hallmarks such as amyloid-beta and tau pathology remains unclear. In the present study, I aimed to investigate the association between TREM2-related microglial responses and tau accumulation in the presence and absence of amyloid-beta pathology in order to give a better view of the role of microglial activation in Alzheimer's disease development. Imaging data of 178 non-demented participants including 107 amyloid-beta-negative participants, 71 amyloid-beta-positive were recruited from Alzheimer's disease Neuroimaging Initiative. The CSF sTREM2 was used as an in vivo indicator of microglial responses associated with TREM2. Furthermore, I used longitudinal tau-PET and resting-state functional MRI connectomes in order to investigate the association of TREM2-related microglial activation and tau spreading through functional connections. A higher level of sTREM2 was associated with slower tau aggregate accumulation in non-demented amyloid-beta-positive. Furthermore, measuring the tau spreading through inter-connected regions using functional MRI connectomes confirms that the TREM2-related microglial activity might be a protective factor against tau pathology in brain tissue. These findings demonstrate that in individuals with initial amyloid-beta abnormalities, TREM2-related microglial activation is linked to reduced regional accumulation of tau aggregates and also, spreading across inter-connected brain regions, as evaluated through functional MRI connectomes during the early stages of Alzheimer's disease.

Predicting an individual's cerebellar activity from functional connectivity fingerprints

The cerebellum is gaining scientific attention as a key neural substrate of cognitive function; however, individual differences in the cerebellar organization have not yet been well studied. Individual differences in functional brain organization can be closely tied to individual differences in brain connectivity. ‘Connectome Fingerprinting’ is a modeling approach that predicts an individual's brain activity from their connectome. Here, we extend ‘Connectome Fingerprinting’ (CF) to the cerebellum. We examined functional MRI data from 160 subjects (98 females) of the Human Connectome Project young adult dataset. For each of seven cognitive task paradigms, we constructed CF models from task activation maps and resting-state cortico-cerebellar functional connectomes, using a set of training subjects. For each model, we then predicted task activation in novel individual subjects, using their resting-state functional connectomes. In each cognitive paradigm, the CF models predicted individual subject cerebellar activity patterns with significantly greater precision than did predictions from the group average task activation. Examination of the CF models revealed that the cortico-cerebellar connections that carried the most information were those made with the non-motor portions of the cerebral cortex. These results demonstrate that the fine-scale functional connectivity between the cerebral cortex and cerebellum carries important information about individual differences in cerebellar functional organization. Additionally, CF modeling may be useful in the examination of patients with cerebellar dysfunction, since model predictions require only resting-state fMRI data which is more easily obtained than task fMRI.

Multivariate resting-state functional connectomes predict and characterize obesity phenotypes.

The univariate obesity-brain associations have been extensively explored, while little is known about the multivariate associations between obesity and resting-state functional connectivity. We therefore utilized machine learning and resting-state functional connectivity to develop and validate predictive models of 4 obesity phenotypes (i.e. body fat percentage, body mass index, waist circumference, and waist-height ratio) in 3 large neuroimaging datasets (n = 2,992). Preliminary evidence suggested that the resting-state functional connectomes effectively predicted obesity/weight status defined by each obesity phenotype with good generalizability to longitudinal and independent datasets. However, the differences between resting-state functional connectivity patterns characterizing different obesity phenotypes indicated that the obesity-brain associations varied according to the type of measure of obesity. The shared structure among resting-state functional connectivity patterns revealed reproducible neuroimaging biomarkers of obesity, primarily comprising the connectomes within the visual cortex and between the visual cortex and inferior parietal lobule, visual cortex and orbital gyrus, and amygdala and orbital gyrus, which further suggested that the dysfunctions in the perception, attention and value encoding of visual information (e.g. visual food cues) and abnormalities in the reward circuit may act as crucial neurobiological bases of obesity. The recruitment of multiple obesity phenotypes is indispensable in future studies seeking reproducible obesity-brain associations.

Functional Network Alterations Associated with Cognition in Pre-Clinical Alzheimer's Disease.

Objective: Accumulation of cerebral amyloid-β (Aβ) is a risk factor for cognitive decline and defining feature of Alzheimer's disease (AD). Aβ is implicated in brain network disruption, but the extent to which these changes correspond with observable cognitive deficits in pre-clinical AD has not been tested. This study utilized individual-specific functional parcellations to sensitively evaluate the relationship between network connectivity and cognition in adults with and without Aβ deposition. Participants and Methods: Cognitively unimpaired adults ages 45-85 completed amyloid positron emission tomography, resting-state-functional magnetic resonance imaging (fMRI), and neuropsychological tests of episodic memory and executive function (EF). Participants in the upper tertile of mean standard uptake value ratio were considered Aβ+ (n = 50) while others were Aβ- (n = 99). Individualized functional network parcellations were generated from resting-state fMRI data. We examined the effects of group, network, and group-by-network interactions on memory and EF. Results: We observed several interactions such that within the Aβ+ group, preserved network integrity (i.e., greater connectivity within specific networks) was associated with better cognition, whereas network desegregation (i.e., greater connectivity between relative to within networks) was associated with worse cognition. This dissociation was most apparent for cognitive networks (frontoparietal, dorsal and ventral attention, limbic, and default mode), with connectivity relating to EF in the Aβ+ group specifically. Conclusions: Using an innovative approach to constructing individual-specified resting-state functional connectomes, we were able to detect differences in brain-cognition associations in pre-clinical AD. Our findings provide novel insight into specific functional network alterations occurring in the presence of Aβ that relate to cognitive function in asymptomatic individuals. Impact statement Elevated cerebral amyloid-β is a biomarker of pre-clinical Alzheimer's disease (AD). Associations between amyloidosis, functional network disruption, and cognitive impairment are evident in the later stages of AD, but these effects have not been substantiated in pre-clinical AD. Using individual-specific parcellations that maximally localize functional networks, we identify network alterations that relate to cognition in pre-clinical AD that have not been previously reported. We demonstrate that these effects localize to networks implicated in cognition. Our findings suggest that there may be subtle, amyloid-related alterations in the functional connectome that are detectable in pre-clinical AD, with potential implications for cognition in asymptomatic individuals.

Decoding Population Variance in Resting-State Functional Connectomes

Decoding Population Variance in Resting-State Functional Connectomes

Optimizing differential identifiability improves connectome predictive modeling of cognitive deficits from functional connectivity in Alzheimer's disease.

Functional connectivity, as estimated using resting state functional MRI, has shown potential in bridging the gap between pathophysiology and cognition. However, clinical use of functional connectivity biomarkers is impeded by unreliable estimates of individual functional connectomes and lack of generalizability of models predicting cognitive outcomes from connectivity. To address these issues, we combine the frameworks of connectome predictive modeling and differential identifiability. Using the combined framework, we show that enhancing the individual fingerprint of resting state functional connectomes leads to robust identification of functional networks associated to cognitive outcomes and also improves prediction of cognitive outcomes from functional connectomes. Using a comprehensive spectrum of cognitive outcomes associated to Alzheimer's disease (AD), we identify and characterize functional networks associated to specific cognitive deficits exhibited in AD. This combined framework is an important step in making individual level predictions of cognition from resting state functional connectomes and in understanding the relationship between cognition and connectivity.

GEFF: Graph embedding for functional fingerprinting

It has been well established that Functional Connectomes (FCs), as estimated from functional MRI (fMRI) data, have an individual fingerprint that can be used to identify an individual from a population (subject-identification). Although identification rate is high when using resting-state FCs, other tasks show moderate to low values. Furthermore, identification rate is task-dependent, and is low when distinct cognitive states, as captured by different fMRI tasks, are compared. Here we propose an embedding framework, GEFF (Graph Embedding for Functional Fingerprinting), based on group-level decomposition of FCs into eigenvectors. GEFF creates an eigenspace representation of a group of subjects using one or more task FCs (Learning Stage). In the Identification Stage, we compare new instances of FCs from the Learning subjects within this eigenspace (validation dataset). The validation dataset contains FCs either from the same tasks as the Learning dataset or from the remaining tasks that were not included in Learning. Assessment of validation FCs within the eigenspace results in significantly increased subject-identification rates for all fMRI tasks tested and potentially task-independent fingerprinting process. It is noteworthy that combining resting-state with one fMRI task for GEFF Learning Stage covers most of the cognitive space for subject identification. Thus, while designing an experiment, one could choose a task fMRI to ask a specific question and combine it with resting-state fMRI to extract maximum subject differentiability using GEFF. In addition to subject-identification, GEFF was also used for identification of cognitive states, i.e. to identify the task associated to a given FC, regardless of the subject being already in the Learning dataset or not (subject-independent task-identification). In addition, we also show that eigenvectors from the Learning Stage can be characterized as task- and subject-dominant, subject-dominant or neither, using two-way ANOVA of their corresponding loadings, providing a deeper insight into the extent of variance in functional connectivity across individuals and cognitive states.

Functional and structural connectivity substrates of cognitive performance in relapsing remitting multiple sclerosis with mild disability.

Multiple Sclerosis (MS) is the most common chronic inflammatory and neurodegenerative disease of the central nervous system (CNS), which can lead to severe cognitive impairment over time. Magnetic resonance imaging (MRI) is currently the best available biomarker to track MS pathophysiology in vivo and examine the link to clinical disability. However, conventional MRI metrics have limited sensitivity and specificity to detect direct associations between symptoms and their underlying CNS substrates. In this study, we aimed to investigate structural and resting state functional connectomes and subnetworks associated with neuropsychological (NP) performance using a graph theoretical approach. A comprehensive NP test battery was administered in a sample of patients with relapsing remitting MS (RRMS) and mild disability [n=33, F/M=20/13, age=40.9±9.7, median [Expanded Disability Status Scale] (EDSS)=2, range =0-4] and compared to healthy controls (HC) [n=29, F/M=19/10, age=41.0±8.5] closely matched for age, sex, and level of education. The NP battery comprised the most relevant domains of cognitive dysfunction in MS including attention, processing speed, verbal and spatial learning and memory, and executive function. While standard MRI metrics showed good correlations with TAP Alertness test, disease duration and neurological exams, structural networks showed closer associations with 9-hole peg test and cognitive performances. Decreased graph strength was associated with two out of the 5 NP tests in the spatial learning and memory domain specified by BVMT [Sum 1-3] and BVMT [Recall], and with also SDMT which is one out of the 9 NP tests in the attention/processing speed domain, while no correlation was found between these scores and functional connectivity. Nodal strength was decreased in all subnetworks based on Yeo atlas in patients compared to HC; however, no difference was observed in nodal level of functional connectivity between the groups. The difference in structural and functional nodal connectivity between the groups was also observed in the relationship between structural and functional connectivity within the groups; the relationship between nodal degree and nodal strength was reversed in patients but positive in controls. On a nodal level, structural and functional networks (mainly the default mode network) were correlated with more than one cognitive domain rather than one specific network for each domain within patients. Interestingly, poorer cognitive performance was mostly correlated with increased functional connectivity but decreased structural connectivity in patients. Increased functional connectivity in the default mode network had both positive as well as negative associations with verbal and spatial learning and memory, possibly indicating adaptive and maladaptive mechanisms. In conclusion, our results suggest that cognitive performance, even in patients with RRMS and very mild disability, may reflect a loss of structural connectivity. In contrast, widespread increases in functional connectivity may be the result of maladaptive processes.

Uncovering multi-site identifiability based on resting-state functional connectomes

Multi-site studies are becoming important to increase statistical power, enhance generalizability, and to improve the likelihood of pooling relevant subgroups together—activities which are otherwise limited by the availability of subjects or funds at a single site. Even with harmonized imaging sequences, site-dependent variability can mask the advantages of these multi-site studies. The aim of this study was to assess multi-site reproducibility in resting-state functional connectivity “fingerprints”, and to improve identifiability of functional connectomes. The individual fingerprinting of functional connectivity profiles is promising due to its potential as a robust neuroimaging biomarker with which to draw single-subject inferences. We evaluated, on two independent multi-site datasets, individual fingerprints in test-retest visit pairs within and across two sites and present a generalized framework based on principal component analysis to improve identifiability. Those principal components that maximized differential identifiability of a training dataset were used as an orthogonal connectivity basis to reconstruct the individual functional connectomes of training and validation sets. The optimally reconstructed functional connectomes showed a substantial improvement in individual fingerprinting of the subjects within and across the two sites and test-retest visit pairs relative to the original data. A notable increase in ICC values for functional edges and resting-state networks were also observed for reconstructed functional connectomes. Improvements in identifiability were not found to be affected by global signal regression. Post-hoc analyses assessed the effect of the number of fMRI volumes on identifiability and showed that multi-site differential identifiability was for all cases maximized after optimal reconstruction. Finally, the generalizability of the optimal set of orthogonal basis of each dataset was evaluated through a leave-one-out procedure. Overall, results demonstrate that the data-driven framework presented in this study systematically improves identifiability in resting-state functional connectomes in multi-site studies.

Topological disintegration of resting state functional connectomes in coma

Graph theory has been playing an increasingly important role in understanding the organizational properties of brain networks, subsequently providing new tools for the search of neural correlates of consciousness, particularly in the context of patients recovering from severe brain injury. However, this approach is not without challenges, as it usually relies on arbitrarily fixing a threshold in order to retain the strongest connections proportionally equal across subjects. This method increases the comparability between individuals or groups but it risks the inclusion of false positive and therefore spurious connections, especially in the context of brain disorders.Resting state data acquired in 25 coma patients and 22 healthy subjects was compared. We obtained a representative fixed density of significant connections by first applying a p-value-based threshold on healthy subjects' networks and then choosing a threshold at which all individuals exhibited meaningful connections. The obtained threshold (i.e. 10%) was used to construct graphs in the patient group. The findings showed that coma patients have lower number of significant connections with approximately 50% of them not fulfilling the criteria of the fixed density threshold. The remaining patients with relatively preserved global functional connectivity had sufficient significant connections between regions, but showed signs of major whole-brain network reorganization. These results warrant careful consideration in the construction of functional connectomes in patients with disorders of consciousness and set the scene for future studies investigating potential clinical implications of such an approach.

A novel joint HCPMMP method for automatically classifying Alzheimer’s and different stage MCI patients

A 360-area surface-based cortical parcellation was recently generated using multimodal data in a group average of 210 healthy young adults from the Human Connectome Project (HCP). In order to automatically and accurately identify mild cognitive impairment (MCI) at its two levels (early MCI and late MCI), Alzheimer's disease (AD) and healthy control (HC), a novel joint HCP MMP method was first proposed to delineate the cortical architecture and function connectivity in a group of non healthy adults. The proposed method was applied to register a dataset of 96 resting-state functional connectomes from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) to Connectivity Informatics Technology Initiative (CIFTI) space and parcellated brain into human connectome project multi-modal parcellation (HCPMMP) with 360 areas. Various network features in each node of the connectivity network were considered as the candidate features for classification.The fine-grained multi-modal based on HCP-MMP combined with machine learning in identification for EMCI, LMCI, AD and HC. Applying various network features, including strength, betweenness centrality, clustering coefficient, local efficiency, eigenvector centrality, etc, we trained and tested several machine learning models. Thousands of features were processed by filter and wrapper feature selection procedures, and finally there were thirty features to be selected to achieve classification accuracies of 93.8% for EMCI vs. HC, 95.8% for LMCI vs. HC, 95.8% for AD vs. HC, and 91.7% for LMCI vs. AD, respectively by using support vector machine (SVM) algorithm. Most of the selected features locate in the region of temporal or cingulate cortex. Compared with previous studies, our results demonstrate the superiority of the proposed method over existing techniques.

Basic Units of Inter-Individual Variation in Resting State Connectomes

Resting state functional connectomes are massive and complex. It is an open question, however, whether connectomes differ across individuals in a correspondingly massive number of ways, or whether most differences take a small number of characteristic forms. We systematically investigated this question and found clear evidence of low-rank structure in which a modest number of connectomic components, around 50–150, account for a sizable portion of inter-individual connectomic variation. This number was convergently arrived at with multiple methods including estimation of intrinsic dimensionality and assessment of reconstruction of out-of-sample data. In addition, we show that these connectomic components enable prediction of a broad array of neurocognitive and clinical symptom variables at levels comparable to a leading method that is trained on the whole connectome. Qualitative observation reveals that these connectomic components exhibit extensive community structure reflecting interrelationships between intrinsic connectivity networks. We provide quantitative validation of this observation using novel stochastic block model-based methods. We propose that these connectivity components form an effective basis set for quantifying and interpreting inter-individual connectomic differences, and for predicting behavioral/clinical phenotypes.

Predictive connectome subnetwork extraction with anatomical and connectivity priors

We present a new method to identify anatomical subnetworks of the human connectome that are optimally predictive of targeted clinical variables, developmental outcomes or disease states. Given a training set of structural or functional brain networks, derived from diffusion MRI (dMRI) or functional MRI (fMRI) scans respectively, our sparse linear regression model extracts a weighted subnetwork. By enforcing novel backbone network and connectivity based priors along with a non-negativity constraint, the discovered subnetworks are simultaneously anatomically plausible, well connected, positively weighted and reasonably sparse. We apply our method to (1) predicting the cognitive and neuromotor developmental outcomes of a dataset of 168 structural connectomes of preterm neonates, and (2) predicting the autism spectrum category of a dataset of 1013 resting-state functional connectomes from the Autism Brain Imaging Data Exchange (ABIDE) database. We find that the addition of each of our novel priors improves prediction accuracy and together outperform other state-of-the-art prediction techniques. We then examine the structure of the learned subnetworks in terms of topological features and with respect to established function and physiology of different regions of the brain.

Altered Functional Brain Connectomes between Sporadic and Familial Parkinson's Patients.

Familial Parkinson's disease (PD) is often caused by mutation of a certain gene, while sporadic PD is associated with variants of genes which can influence the susceptibility to PD. The goal of this study was to investigate the difference between the two forms of PD in terms of brain abnormalities using resting-state functional MRI and graph theory. Thirty-one familial PD patients and 36 sporadic PD patients underwent resting-state functional MRI scanning. Frequency-dependent functional connectivity was calculated for each subject using wavelet-based correlations of BOLD signal over 246 brain regions from Brainnetome Atlas. Graph theoretical analysis was then performed to analyze the topology of the functional network, and functional connectome differences were identified with a network-based statistical approach. Our results revealed a frequency-specific (0.016 and 0.031 Hz) connectome difference between familial and sporadic forms of PD, as indicated by an increase in assortativity and decrease in the nodal strength in the left medial amygdala of the familial PD group. In addition, the familial PD patients also showed a distinctive functional network between the left medial amygdala and regions related to retrieval of motion information. The present study indicates that the medial amygdala might be most vulnerable to both sporadic and familial PD. Our findings provide some new insights into disrupted resting-state functional connectomes between sporadic PD and familial PD.

Genetic and age-related variance of resting-state functional connectomes

Francois Chouinard-Decorte, Yassine Benhajali, Jack Kent, Melanie A. Carless, Joanne E. Curran, Tom Dyer, Harold Goring, Rene Olvera, Peter Fox, Laura Almasy, Ravi Duggirala, John Blangero, Alan C. Evans, David C. Glahn and Pierre Bellec 1 McConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University, Montreal, QC H3A 2B4, Canada 2 Centre de recherche de l'institut universitaire de geriatrie de Montreal, Montreal, QC H3W 1W5, Canada 3 Departement d'Informatique et de recherche operationnelle, Universite de Montreal, Montreal, QC H3C 3J7, Canada 4 Department of Genetics, Texas Biomedical Research Institute, University of Texas Health Science 5 Department of Psychiatry, University of Texas Health Science Center San Antonio, San Antonio, Texas Center San Antonio, San Antonio, Texas 6 Research Imaging Institute, University of Texas Health Science Center San Antonio, San Antonio, Texas 7 Department of Psychiatry, Yale University School of Medicine, New Haven, CT 06571, USA

Disease prediction based on functional connectomes using a scalable and spatially-informed support vector machine

Substantial evidence indicates that major psychiatric disorders are associated with distributed neural dysconnectivity, leading to a strong interest in using neuroimaging methods to accurately predict disorder status. In this work, we are specifically interested in a multivariate approach that uses features derived from whole-brain resting state functional connectomes. However, functional connectomes reside in a high dimensional space, which complicates model interpretation and introduces numerous statistical and computational challenges. Traditional feature selection techniques are used to reduce data dimensionality, but are blind to the spatial structure of the connectomes. We propose a regularization framework where the 6-D structure of the functional connectome (defined by pairs of points in 3-D space) is explicitly taken into account via the fused Lasso or the GraphNet regularizer. Our method only restricts the loss function to be convex and margin-based, allowing non-differentiable loss functions such as the hinge-loss to be used. Using the fused Lasso or GraphNet regularizer with the hinge-loss leads to a structured sparse support vector machine (SVM) with embedded feature selection. We introduce a novel efficient optimization algorithm based on the augmented Lagrangian and the classical alternating direction method, which can solve both fused Lasso and GraphNet regularized SVM with very little modification. We also demonstrate that the inner subproblems of the algorithm can be solved efficiently in analytic form by coupling the variable splitting strategy with a data augmentation scheme. Experiments on simulated data and resting state scans from a large schizophrenia dataset show that our proposed approach can identify predictive regions that are spatially contiguous in the 6-D “connectome space,” offering an additional layer of interpretability that could provide new insights about various disease processes.