Interacting Brain Regions Research Articles

Connectome-based disentangling of epilepsy networks from insular stereoelectroencephalographic leads.

Epilepsy is considered as a network disorder of interacting brain regions. The propagation of local epileptic activity from the seizure onset zone (SOZ) along neuronal networks determines the semiology of seizures. However, in highly interconnected brain regions such as the insula, the association between the SOZ and semiology is blurred necessitating invasive stereoelectroencephalography (SEEG). Normative connectomes on MRI data enable to link different symptoms and lesion locations to a common functional network. The present study applied connectomics to disentangle epilepsy networks from insular SEEG recordings and to describe their relationship to seizure semiology. We retrospectively extracted functional networks by normative connectome analysis from 118 insular contacts depicting epileptic discharges during SEEG in 20 epilepsy patients. The resulting epilepsy networks were correlated to the corresponding semiology by voxel-wise regression and multivariate analyses of variances. Epileptic foci were found in the posterior insula for somatosensory, other sensory and motor seizures, while cognitive and autonomic symptoms were related to the anterior insula. We identified insular connections to the superior temporal gyrus and heschl gyrus in sensory seizures and projections to the somatosensory cortex in somatosensory seizures. Insula-basal ganglia pathways were found in cognitive seizure manifestations, while insular connectivity to fronto-basal regions were strongest in patients with autonomic seizures. The semiology of seizures is mirrored in the functional connectivity of insular epileptic discharges. Combining SEEG and connectomics could provide additional information about seizure propagation within the epilepsy network and might enable new treatment options in the future like deep brain stimulation.

Pairwise hemorrhage-brain region interaction-driven hemorrhagic stroke assessment in CT.

Hemorrhagic stroke is a major global health problem requiring rapid and accurate diagnosis for effective treatment. Despite advances, current Computer-Aided Diagnosis (CAD) frameworks rarely account for the functional impacts of hemorrhages on specific critical brain regions and lack the detailed assessments essential for precise treatment. To provide detailed insights into hemorrhages, we aim to propose a CAD framework for in-depth hemorrhagic stroke assessment in CT. The framework includes segmenting hemorrhages and critical brain regions, intraparenchymal hemorrhage (IPH) classification for identifying hemorrhages in critical brain regions and detecting hemorrhage volume. To capture the complex interactions between hemorrhages and critical brain regions, we developed the Pairwise Hemorrhage-Brain Regions Interaction (PHRI) Network. Its emphasis a novel interaction head that integrates feature on hemorrhages, brain regions, and their interdependencies, enabling the model to learn these relationships during training. In addition, a Global-Local Fusion Unit was introduced to provide with image-wide contextual information, and an uncertainty-weighted loss method was utilized to simultaneously optimise the multitask framework. With Institutional Review Board (IRB) approval, an in-house hemorrhagic stroke dataset was collected, including 2,764 CT slices from 99 patients. A five-fold cross-validation was used to train and test the models. The proposed PHRI network was experimentally validated to effectively extract hemorrhage brain region interactions, thereby significantly outperforming several state-of-the-art models in both hemorrhage and critical brain region segmentation, with an average Dice of 0.9064 ± 0.1079 (P < 0.05), as well as in IPH classification, with an F1-Score of 0.8366. Additionally, the framework demonstrated good performance in hemorrhage volume detection, with an intraclass correlation coefficient of 0.981. &#xD;Significance. This study introduces a CAD framework for hemorrhagic stroke assessment, offering a novel approach that emphasizes relationships between hemorrhages and critical brain regions.

Motor learning is modulated by dopamine availability in the sensorimotor putamen.

Successful motor skill acquisition requires the dynamic interaction of multiple brain regions, with the striatum playing a critical role in this network. Animal studies suggest that dopaminergic mechanisms are involved in the regulation of motor learning-associated striatal plasticity. In humans, however, the contribution of nigrostriatal dopaminergic transmission to motor learning remains elusive beyond its well-characterized role in initiation and fluent execution of movements. In this prospective observational study, we investigated motor sequence learning in individuals who had undergone 123I-N-ω-fluoropropyl-2β-carbomethoxy-3β-(4-iodophenyl)nortropane single-photon emission computed tomography for the differential diagnosis of Parkinson's disease (n = 41) and age-matched healthy controls (n = 20). We found that striatal dopamine transporter depletion exhibited distinct spatial patterns that were associated with impairments in motor sequence learning and the manifestation of Parkinsonian motor symptoms, respectively. Specifically, significant associations between striatal dopamine transporter depletion and impairments in motor sequence learning were confined to posterior putaminal regions, whereas significant associations of striatal dopamine transporter depletion with Parkinsonian motor symptom severity showed a widespread spatial pattern across the entire striatal volume with an anterior maximum. Normative functional connectivity analysis revealed that both behavioural domains shared largely overlapping connectivity patterns with the basal ganglia and supplementary motor area. However, apart from connectivity with more posterior parts of the supplementary motor area, significant functional connectivity with primary motor cortical areas was only present for striatal dopamine transporter availability-related modulation of online motor learning. Our findings indicate that striatal dopaminergic signalling plays a specific role in motor sequence learning beyond its influence on mere motor execution, implicating learning-related sensorimotor striatum recruitment and cortico-striatal plasticity as dopamine-dependent mechanisms.

David Ferrier’s “complex whole”: Early traces of a “brain network” concept

ABSTRACT Currently, the idea that the brain is a complex network of interacting brain regions is hardly controversial. The rapid development of this field is often attributed to the emergence of powerful brain-imaging techniques and, around the millennium, the merging of the neuroscience of brain networks with modern mathematical graph theory. However, little is known about the historical roots of this concept. It is interesting to know when the first traces of a concept of brain networks can be found in the work of early neuroscientists, how this concept evolved over time, and what factors may have influenced this evolution. This study aims to set a first step in addressing these questions by a detailed analysis of David Ferrier’s classic study, The Functions of the Brain. From this analysis it will become clear that, in addition to a clear notion of localized functions in the brain, Ferrier speculated in several places about the need for several of these brain regions to communicate and interact in order to bring about higher brain functions. He referred to this perspective on the brain as a “complex whole,” which could be interpreted as an early precursor of the modern concept of brain networks.

Transformer-aided dynamic causal model for scalable estimation of effective connectivity

Abstract Dynamic Causal Models (DCMs) in functional Magnetic Resonance Imaging (fMRI) decipher causal interactions, known as Effective Connectivity, among neuronal populations. However, their utility is often constrained by computational limitations, restricting analysis to a small subset of interacting brain areas, typically fewer than 10, thus lacking scalability. While the regression DCM (rDCM) has emerged as a faster alternative to traditional DCMs, it is not without its limitations, including the linearization of DCM terms, reliance on a fixed Hemodynamic Response Function (HRF), and an inability to accommodate modulatory influences. In response to these challenges, we propose a novel hybrid approach named Transformer encoder DCM decoder (TREND), which combines a Transformer encoder with state-of-the-art physiological DCM (P-DCM) as decoder. This innovative method addresses the scalability issue while preserving the nonlinearities inherent in DCM equations. Through extensive simulations, we validate TREND’s efficacy by demonstrating its ability to accurately predict effective connectivity values with dramatically reduced computational time relative to original P-DCM even in networks comprising up to, for instance, 100 interacting brain regions. Furthermore, we showcase TREND on an empirical fMRI dataset demonstrating the superior accuracy and/or speed of TREND compared with other DCM variants. In summary, by amalgamating P-DCM with Transformer, we introduce and validate a pioneering approach for determining effective connectivity values among brain regions, extending its applicability seamlessly to large-scale brain networks.

Integrative intrinsic brain activity and molecular analyses of the interaction between first-episode depression and age

BackgroundNumerous studies have underscored the presence of abnormal intrinsic neural activity (INA) in individuals with depression. However, recognizing that the age stage may influence the pathophysiology of depression, our study sought to delve into the interplay of depression and age on INA and molecular architecture. MethodsOne hundred and thirty-eight first-episode depression patients and 120 healthy controls (HC) were recruited and underwent resting-state functional magnetic resonance imaging. The participants were stratified into four groups based on age. Utilizing amplitude of low-frequency fluctuation (ALFF) analyses, we employed an ANCOVA to compare INA patterns in four groups. Additionally, we conducted correlation analyses between ALFF and neurotransmitter maps to elucidate molecular underpinnings of INA abnormalities. ResultsIn comparison to adolescents with early-onset depression and adult HC, adult-onset depression exhibited increased ALFF in the right paracentral lobule. Conversely, early-onset depression, when contrasted with adolescent HC, displayed reduced ALFF in the right paracentral lobule. The interactive brain regions affected by ALFF alterations were associated with serotonergic, GABAergic, and opioid neurotransmitter systems. LimitationsThe present study was limited to its cross-sectional design. ConclusionsThis study illuminates an antagonistic effect of depression and age on brain activity in paracentral lobule and provides molecular underpinnings of the corresponding INA abnormalities related to key neurotransmitter systems. These insights may prove valuable in the development of neuromarkers for clinical intervention and treatment of depression.

"Witzelsucht". On the introduction and meaning of the term by Hermann Oppenheim in 1890 and its distinction from "moria"

Hermann Oppenheim (1858-1919) was a German neurologist without an academic career, who in his productive period around 1900 made aname for himself during his lifetime as amajor player in the history of German neurology with his many contributions to multiple sclerosis, syphilis and the controversial study of traumatic neurosis; however, it is almost unknown that in 1890 he introduced the term "witzelsucht", which is still used internationally today. Moritz Jastrowitz dealt with behavioral abnormalities due to frontal brain injuries 1year earlier and used the term "moria" for aform of mental disorder associated with akind of childish behavior and inappropriate jocularity. Oppenheim was critical of this and differentiated his "witzelsucht" from this. With this term he wanted to describe humoristic feeble-mindedness in a much narrower sense, which stands in striking contrast to the usual symptoms in cases of cerebral tumors. Oppenheim recognized the frontal brain, particularly the right brain, to be an important functional unit for humorous behavior. Modern research has confirmed that the processing of humor requires a complex interaction of multiple brain regions. Damage to the right frontal lobe or to connecting structures can lead to the disorder "witzelsucht". Whether asimultaneous damage to the left hemisphere must be present or if this is dependent on the individual dominant hemisphere, needs further research.

Small non-coding RNA transcriptomic profiling in adult and fetal human brain

Small non-coding RNAs (sncRNAs) make up ~1% of the transcriptome; nevertheless, they play significant roles in regulating cellular processes. Given the complexity of the central nervous system, sncRNAs likely hold particular importance in the human brain. In this study, we provide sncRNA transcriptomic profiles in a range of adult and prenatal brain regions, with a focus on piRNAs, due to their underexplored expression in somatic cells and tissue-specific nature. Using the WIND workflow, which combines two detection methods, we found 1333 (731 miRNAs, 249 piRNAs, 285 snoRNAs, and 68 other sncRNAs) and 1445 unique sncRNAs (770 miRNAs, 307 piRNAs, 289 snoRNAs, and 79 other sncRNAs) in developing and adult brains, respectively. Significant variations were found upon comparison of fetal and adult brain groups, with 82 miRNAs, 17 piRNAs, and 70 snoRNAs enriched in fetal brains and 22 miRNAs, 11 piRNAs in adult brains. This dataset represents a valuable resource for exploring the sncRNA roles in brain function, their involvement in neurological diseases, and the molecular mechanisms behind brain region interactions.

SHIELD: Skull-shaped hemispheric implants enabling large-scale electrophysiology datasets in the mouse brain

To understand the neural basis of behavior, it is essential to measure spiking dynamics across many interacting brain regions. Although new technologies, such as Neuropixels probes, facilitate multi-regional recordings, significant surgical and procedural hurdles remain for these experiments to achieve their full potential. Here, we describe skull-shaped hemispheric implants enabling large-scale electrophysiology datasets (SHIELD). These 3D-printed skull-replacement implants feature customizable insertion holes, allowing dozens of cortical and subcortical structures to be recorded in a single mouse using repeated multi-probe insertions over many days. We demonstrate the procedure's high success rate, biocompatibility, lack of adverse effects on behavior, and compatibility with imaging and optogenetics. To showcase SHIELD's scientific utility, we use multi-probe recordings to reveal novel insights into how alpha rhythms organize spiking activity across visual and sensorimotor networks. Overall, this method enables powerful, large-scale electrophysiological experiments for the study of distributed neural computation.

Rescuing impaired hippocampal-cortical interactions and spatial reorientation learning and memory during sleep in a mouse model of Alzheimer's disease using hippocampal 40 Hz stimulation.

One of the earliest symptoms of Alzheimer's disease (AD) is getting lost in space or experiencing deficits in spatial navigation, which involve navigation computations as well as learning and memory. We investigated cross brain region interactions supporting memory formation as a potential causative factor of impaired spatial learning and memory in AD. To assess this relationship between AD pathophysiology, brain changes, and behavioral alterations, we used a targeted approach for clearing amyloid beta and tau to rescue functional interactions in the brain. This research strongly connects brain activity patterns during sleep to tau and amyloid accumulation, and will aid in understanding the mechanisms underlying cognitive dysfunction in AD. Furthermore, the results offer insight for improving early identification and treatment strategies.

Identifying Diagnostic Biomarkers for Autism Spectrum Disorder From Higher-order Interactions Using the PED Algorithm.

In the field of neuroimaging, more studies of abnormalities in brain regions of the autism spectrum disorder (ASD) usually focused on two brain regions connected, and less on abnormalities of higher-order interactions of brain regions. To explore the complex relationships of brain regions, we used the partial entropy decomposition (PED) algorithm to capture higher-order interactions by computing the higher-order dependencies of all three brain regions (triads). We proposed a method for examining the effect of individual brain regions on triads based on the PED and surrogate tests. The key triads were discovered by analyzing the effects. Further, the hypergraph modularity maximization algorithm revealed the higher-order brain structures, of which the link between right thalamus and left thalamus in ASD was more loose compared with the typical control (TC). Redundant key triad (left cerebellum crus 1 and left precuneus and right inferior occipital gyrus) exhibited a discernible attenuation in interaction in ASD, while the synergistic key triad (right cerebellum crus 1 and left postcentral gyrus and left lingual gyrus) indicated a notable decline. The results of classification model further confirmed the potential of the key triads as diagnostic biomarkers.

Transcriptional programs mediating neuronal toxicity and altered glial-neuronal signaling in a Drosophila knock-in tauopathy model.

Missense mutations in the gene encoding the microtubule-associated protein TAU (current and approved symbol is MAPT) cause autosomal dominant forms of frontotemporal dementia. Multiple models of frontotemporal dementia based on transgenic expression of human TAU in experimental model organisms, including Drosophila, have been described. These models replicate key features of the human disease but do not faithfully recreate the genetic context of the human disorder. Here we use CRISPR-Cas-mediated gene editing to model frontotemporal dementia caused by the TAU P301L mutation by creating the orthologous mutation, P251L, in the endogenous Drosophila tau gene. Flies heterozygous or homozygous for Tau P251L display age-dependent neurodegeneration, display metabolic defects, and accumulate DNA damage in affected neurons. To understand the molecular events promoting neuronal dysfunction and death in knock-in flies, we performed single-cell RNA sequencing on approximately 130,000 cells from brains of Tau P251L mutant and control flies. We found that expression of disease-associated mutant tau altered gene expression cell autonomously in all neuronal cell types identified. Gene expression was also altered in glial cells, suggestive of non-cell-autonomous regulation. Cell signaling pathways, including glial-neuronal signaling, were broadly dysregulated as were brain region and cell type-specific protein interaction networks and gene regulatory programs. In summary, we present here a genetic model of tauopathy that faithfully recapitulates the genetic context and phenotypic features of the human disease, and use the results of comprehensive single-cell sequencing analysis to outline pathways of neurotoxicity and highlight the potential role of non-cell-autonomous changes in glia.

Towards a more robust non-invasive assessment of functional connectivity

Abstract Non-invasive evaluation of functional connectivity, based on source-reconstructed estimates of phase-difference-based metrics, is notoriously non-robust. This is due to a combination of factors, ranging from a misspecification of seed regions to suboptimal baseline assumptions, and residual signal leakage. In this work, we propose a new analysis scheme of source-level phase-difference-based connectivity, which is aimed at optimizing the detection of interacting brain regions. Our approach is based on the combined use of sensor subsampling and dual-source beamformer estimation of all-to-all connectivity on a prespecified dipolar grid. First, a pairwise two-dipole model, to account for reciprocal leakage in the estimation of the localized signals, allows for a usable approximation of the pairwise bias in connectivity due to residual leakage of “third party” noise. Secondly, using sensor array subsampling, the recreation of multiple connectivity maps using different subsets of sensors allows for the identification of consistent spatially localized peaks in the 6-dimensional connectivity maps, indicative of true brain region interactions. These steps are combined with the subtraction of null coherence estimates to obtain the final coherence maps. With extensive simulations, we compared different analysis schemes for their detection rate of connected dipoles, as a function of signal-to-noise ratio, phase difference, and connection strength. We demonstrate superiority of the proposed analysis scheme in comparison to single-dipole models, or an approach that discards the zero phase difference component of the connectivity. We conclude that the proposed pipeline allows for a more robust identification of functional connectivity in experimental data, opening up new possibilities to study brain networks with mechanistically inspired connectivity measures in cognition and in the clinic.

Decomposing Neural Circuit Function into Information Processing Primitives.

It is challenging to measure how specific aspects of coordinated neural dynamics translate into operations of information processing and, ultimately, cognitive functions. An obstacle is that simple circuit mechanisms-such as self-sustained or propagating activity and nonlinear summation of inputs-do not directly give rise to high-level functions. Nevertheless, they already implement simple the information carried by neural activity. Here, we propose that distinct functions, such as stimulus representation, working memory, or selective attention, stem from different combinations and types of low-level manipulations of information or information processing primitives. To test this hypothesis, we combine approaches from information theory with simulations of multi-scale neural circuits involving interacting brain regions that emulate well-defined cognitive functions. Specifically, we track the information dynamics emergent from patterns of neural dynamics, using quantitative metrics to detect where and when information is actively buffered, transferred or nonlinearly merged, as possible modes of low-level processing (storage, transfer and modification). We find that neuronal subsets maintaining representations in working memory or performing attentional gain modulation are signaled by their boosted involvement in operations of information storage or modification, respectively. Thus, information dynamic metrics, beyond detecting which network units participate in cognitive processing, also promise to specify how and when they do it, that is, through which type of primitive computation, a capability that may be exploited for the analysis of experimental recordings.

Modeling the Functional Network for Spatial Navigation in the Human Brain.

Spatial navigation is a complex function involving the integration and manipulation of multisensory information. Using different navigation tasks, many promising results have been achieved on the specific functions of various brain regions (e.g., hippocampus, entorhinal cortex, and parahippocampal place area). Recently, it has been suggested that a non-aggregate network process involving multiple interacting brain regions may better characterize the neural basis of this complex function. This paper presents an integrative approach for constructing and analyzing the functionally-specific network for spatial navigation in the human brain. Briefly, this integrative approach consists of three major steps: 1) to identify brain regions important for spatial navigation (nodes definition); 2) to estimate functional connectivity between each pair of these regions and construct the connectivity matrix (network construction); 3) to investigate the topological properties (e.g., modularity and small worldness) of the resulting network (network analysis). The presented approach, from a network perspective, could help us better understand how our brain supports flexible navigation in complex and dynamic environments, and the revealed topological properties of the network can also provide important biomarkers for guiding early identification and diagnosis of Alzheimer's disease in clinical practice.

Individualized Seizure Cluster Prediction Using Machine Learning and Chronic Ambulatory Intracranial EEG.

Epilepsy patients often experience acute repetitive seizures, known as seizure clusters, which can progress to prolonged seizures or status epilepticus if left untreated. Predicting the onset of seizure clusters is crucial to enable patients to receive preventative treatments. Additionally, studying the patterns of seizure clusters can help predict the seizure type (isolated or cluster) after observing a just occurred seizure. This paper presents machine learning models that use bivariate intracranial EEG (iEEG) features to predict seizure clustering. Specifically, we utilized relative entropy (REN) as a bivariate feature to capture potential differences in brain region interactions underlying isolated and cluster seizures. We analyzed a large ambulatory iEEG dataset collected from 15 patients and spanned up to 2 years of recordings for each patient, consisting of 3341 cluster seizures (from 427 clusters) and 369 isolated seizures. The dataset's substantial number of seizures per patient enabled individualized analyses and predictions. We observed that REN was significantly different between isolated and cluster seizures in majority of the patients. Machine learning models based on REN: 1) predicted whether a seizure will occur soon after a given seizure with up to 69.5% Area under the ROC Curve (AUC), 2) predicted if a seizure is the first one in a cluster with up to 55.3% AUC, outperforming baseline techniques. Overall, our findings could be beneficial in addressing the clinical burden associated with seizure clusters, enabling patients to receive timely treatments and improving their quality of life.

Pairing explainable deep learning classification with clustering to uncover effects of schizophrenia upon whole brain functional network connectivity dynamics

Many studies have analyzed resting state functional magnetic resonance imaging (rs-fMRI) dynamic functional network connectivity (dFNC) data to elucidate the effects of neurological and neuropsychiatric disorders upon the interactions of brain regions over time. Existing studies often use either machine learning classification or clustering algorithms. Additionally, several studies have used clustering algorithms to extract features related to brain states trajectories that can be used to train interpretable classifiers. However, the combination of explainable dFNC classifiers followed by clustering algorithms is highly underutilized. In this study, we show how such an approach can be used to study the effects of schizophrenia (SZ) upon brain activity. Specifically, we train an explainable deep learning model to classify between individuals with SZ and healthy controls. We then cluster the resulting explanations, identifying discriminatory states of dFNC. We lastly apply several novel measures to quantify aspects of the classifier explanations and obtain additional insights into the effects of SZ upon brain network dynamics. Specifically, we uncover effects of schizophrenia upon subcortical, sensory, and cerebellar network interactions. We also find that individuals with SZ likely have reduced variability in overall brain activity and that the effects of SZ may be temporally localized. In addition to uncovering effects of SZ upon brain network dynamics, our approach could provide novel insights into a variety of neurological and neuropsychiatric disorders in future dFNC studies.

Combining temporal and spatial attention for seizure prediction.

Approximately 1% of the world population is currently suffering from epilepsy. Successful seizure prediction is necessary for those patients. Influenced by neurons in their own and surrounding locations, the electroencephalogram (EEG) signals collected by scalp electrodes carry information of spatiotemporal interactions. Therefore, it is a great challenge to exploit the spatiotemporal information of EEG signals fully. In this paper, a new seizure prediction model called Gatformer is proposed by fusing the graph attention network (GAT) and the Transformer. The temporal and spatial attention are combined to extract EEG information from the perspective of spatiotemporal interactions. The model aims to explore the temporal dependence of single-channel EEG signals and the spatial correlations among multi-channel EEG signals. It can automatically identify the most noteworthy interaction in brain regions and achieve accurate seizure prediction. Compared with the baseline models, the performance of our model is significantly improved. The false prediction rate (FPR) on the private dataset is 0.0064/h. The average accuracy, specificity and sensitivity are 98.25%, 99.36% and 97.65%. The proposed model is comparable to the state of the arts. Experiments on different datasets show that it has good robustness and generalization performance. The high sensitivity and low FPR prove that this model has great potential to realize clinical assistance for diagnosis and treatment.

Combined behavioral and EEG evidence for the 70 Hz frequency selection of short-term spinal cord stimulation in disorders of consciousness.

This study investigated the prognostic effect of electroencephalography (EEG) instant effects of single spinal cord stimulation (SCS) on clinical outcome in disorders of consciousness (DOC) and the time-dependent brain response during the recovery of consciousness prompted by SCS. Twenty three patients with DOC underwent short-term SCS (stSCS) implantation operation. Then, all patients received the postoperative EEG test including EEG record before (T1) and after (T2) single SCS session. Subsequently, 2 weeks stSCS treatment was performed and revised coma recovery scale (CRS-R) and EEG data were collected. Finally, they were classified into effective and ineffective groups at 3-month follow-up (T6). The parietal-occipital (PO) connectivity and clustering coefficients (CC) in the beta band of the effective group at the 1 week after the treatment (T5) were found to be higher than preoperative assessment (T0). Correlation analysis showed that the change in beta CC at T1/T2was correlated with the change in CRS-R at T0/T6. In addition, the change in PO connectivity and CC in the beta at T0/T5 were also correlated with the change in CRS-R at T0/T5. SCS may facilitate the recovery of consciousness by enhancing local information interaction in posterior brain regions. And the recovery can be predicted by beta CC in the EEG test.

Estimation of Discriminative Multimodal Brain Network Connectivity Using Message-Passing-Based Nonlinear Network Fusion.

Effective estimation of brain network connectivity enables better unraveling of the extraordinary complexity interactions of brain regions and helps in auxiliary diagnosis of psychiatric disorders. Considering different modalities can provide comprehensive characterizations of brain connectivity, we propose the message-passing-based nonlinear network fusion (MP-NNF) algorithm to estimate multimodal brain network connectivity. In the proposed method, the initial functional and structural networks were computed from fMRI and DTI separately. Then, we update every unimodal network iteratively, making it more similar to the others in every iteration, and finally converge to one unified network. The estimated brain connectivities integrate complementary information from multiple modalities while preserving their original structure, by adding the strong connectivities present in unimodal brain networks and eliminating the weak connectivities. The effectiveness of the method was evaluated by applying the learned brain connectivity for the classification of major depressive disorder (MDD). Specifically, 82.18% classification accuracy was achieved even with the simple feature selection and classification pipeline, which significantly outperforms the competing methods. Exploration of brain connectivity contributed to MDD identification suggests that the proposed method not only improves the classification performance but also was sensitive to critical disease-related neuroimaging biomarkers.