Evaluation of Mean Shift, ComBat, and CycleGAN for Harmonizing Brain Connectivity Matrices Across Sites.

Connectivity matrices derived from diffusion MRI (dMRI) provide an interpretable and generalizable way of understanding the human brain connectome. However, dMRI suffers from inter-site and between-scanner variation, which impedes analysis across datasets to improve robustness and reproducibility of results. To evaluate different harmonization approaches on connectivity matrices, we compared graph measures derived from these matrices before and after applying three harmonization techniques: mean shift, ComBat, and CycleGAN. The sample comprises 168 age-matched, sex-matched normal subjects from two studies: the Vanderbilt Memory and Aging Project (VMAP) and the Biomarkers of Cognitive Decline Among Normal Individuals (BIOCARD). First, we plotted the graph measures and used coefficient of variation (CoV) and the Mann-Whitney U test to evaluate different methods' effectiveness in removing site effects on the matrices and the derived graph measures. ComBat effectively eliminated site effects for global efficiency and modularity and outperformed the other two methods. However, all methods exhibited poor performance when harmonizing average betweenness centrality. Second, we tested whether our harmonization methods preserved correlations between age and graph measures. All methods except for CycleGAN in one direction improved correlations between age and global efficiency and between age and modularity from insignificant to significant with p-values less than 0.05.

Evaluation of Mean Shift, ComBat, and CycleGAN for Harmonizing Brain Connectivity Matrices Across Sites.

Connectivity matrices derived from diffusion MRI (dMRI) provide an interpretable and generalizable way of understanding the human brain connectome. However, dMRI suffers from inter-site and between-scanner variation, which impedes analysis across datasets to improve robustness and reproducibility of results. To evaluate different harmonization approaches on connectivity matrices, we compared graph measures derived from these matrices before and after applying three harmonization techniques: mean shift, ComBat, and CycleGAN. The sample comprises 168 age-matched, sex-matched normal subjects from two studies: the Vanderbilt Memory and Aging Project (VMAP) and the Biomarkers of Cognitive Decline Among Normal Individuals (BIOCARD). First, we plotted the graph measures and used coefficient of variation (CoV) and the Mann-Whitney U test to evaluate different methods' effectiveness in removing site effects on the matrices and the derived graph measures. ComBat effectively eliminated site effects for global efficiency and modularity and outperformed the other two methods. However, all methods exhibited poor performance when harmonizing average betweenness centrality. Second, we tested whether our harmonization methods preserved correlations between age and graph measures. All methods except for CycleGAN in one direction improved correlations between age and global efficiency and between age and modularity from insignificant to significant with p-values less than 0.05.

Laminar and dorsoventral organization of layer 1 interneuronal microcircuitry in superficial layers of the medial entorhinal cortex.

Layer 1 (L1) interneurons (INs) participate in various brain functions by gating information flow in the neocortex, but their role in the medial entorhinal cortex (MEC) is still unknown, largely due to scant knowledge of MEC L1 microcircuitry. Using simultaneous triple-octuple whole-cell recordings and morphological reconstructions, we comprehensively depict L1IN networks in the MEC. We identify three morphologically distinct types of L1INs with characteristic electrophysiological properties. We dissect intra- and inter-laminar cell-type-specific microcircuits of L1INs, showing connectivity patterns different from those in the neocortex. Remarkably, motif analysis reveals transitive and clustered features of L1 networks, as well as over-represented trans-laminar motifs. Finally, we demonstrate the dorsoventral gradient of L1IN microcircuits, with dorsal L1 neurogliaform cells receiving fewer intra-laminar inputs but exerting more inhibition on L2 principal neurons. These results thus present a more comprehensive picture of L1IN microcircuitry, which is indispensable for deciphering the function of L1INs in the MEC.

A Proposed Human Structural Brain Connectivity Matrix in the Center for Morphometric Analysis Harvard-Oxford Atlas Framework: A Historical Perspective and Future Direction for Enhancing the Precision of Human Structural Connectivity with a Novel Neuroanatomical Typology.

A complete structural definition of the human nervous system must include delineation of its wiring diagram (e.g., Swanson LW. Brain architecture: understanding the basic plan, 2012). The complete formulation of the human brain circuit diagram (BCD [Front Neuroanat. 2020;14:18]) has been hampered by an inability to determine connections in their entirety (i.e., not only pathway stems but also origins and terminations). From a structural point of view, a neuroanatomic formulation of the BCD should include the origins and terminations of each fiber tract as well as the topographic course of the fiber tract in three dimensions. Classic neuroanatomical studies have provided trajectory information for pathway stems and their speculative origins and terminations [Dejerine J and Dejerine-Klumpke A. Anatomie des Centres Nerveux, 1901; Dejerine J and Dejerine-Klumpke A. Anatomie des Centres Nerveux: Méthodes générales d'étude-embryologie-histogénèse et histologie. Anatomie du cerveau, 1895; Ludwig E and Klingler J. Atlas cerebri humani, 1956; Makris N. Delineation of human association fiber pathways using histologic and magnetic resonance methodologies; 1999; Neuroimage. 1999 Jan;9(1):18-45]. We have summarized these studies previously [Neuroimage. 1999 Jan;9(1):18-45] and present them here in a macroscale-level human cerebral structural connectivity matrix. A matrix in the present context is an organizational construct that embodies anatomical knowledge about cortical areas and their connections. This is represented in relation to parcellation units according to the Harvard-Oxford Atlas neuroanatomical framework established by the Center for Morphometric Analysis at Massachusetts General Hospital in the early 2000s, which is based on the MRI volumetrics paradigm of Dr. Verne Caviness and colleagues [Brain Dev. 1999 Jul;21(5):289-95]. This is a classic connectional matrix based mainly on data predating the advent of DTI tractography, which we refer to as the "pre-DTI era" human structural connectivity matrix. In addition, we present representative examples that incorporate validated structural connectivity information from nonhuman primates and more recent information on human structural connectivity emerging from DTI tractography studies. We refer to this as the "DTI era" human structural connectivity matrix. This newer matrix represents a work in progress and is necessarily incomplete due to the lack of validated human connectivity findings on origins and terminations as well as pathway stems. Importantly, we use a neuroanatomical typology to characterize different types of connections in the human brain, which is critical for organizing the matrices and the prospective database. Although substantial in detail, the present matrices may be assumed to be only partially complete because the sources of data relating to human fiber system organization are limited largely to inferences from gross dissections of anatomic specimens or extrapolations of pathway tracing information from nonhuman primate experiments [Front Neuroanat. 2020;14:18, Front Neuroanat. 2022;16:1035420, and Brain Imaging Behav. 2021;15(3):1589-1621]. These matrices, which embody a systematic description of cerebral connectivity, can be used in cognitive and clinical studies in neuroscience and, importantly, to guide research efforts for further elucidating, validating, and completing the human BCD [Front Neuroanat. 2020;14:18].

Deep Learning Classifies Low- and High-Grade Glioma Patients with High Accuracy, Sensitivity, and Specificity Based on Their Brain White Matter Networks Derived from Diffusion Tensor Imaging.

Classifying low-grade glioma (LGG) patients from high-grade glioma (HGG) is one of the most challenging tasks in planning treatment strategies for brain tumor patients. Previous studies derived several handcrafted features based on the tumor's texture and volume from magnetic resonance images (MRI) to classify LGG and HGG patients. The accuracy of classification was moderate. We aimed to classify LGG from HGG with high accuracy using the brain white matter (WM) network connectivity matrix constructed using diffusion tensor tractography. We obtained diffusion tensor images (DTI) of 44 LGG and 48 HGG patients using routine clinical imaging. Fiber tractography and brain parcellation were performed for each patient to obtain the fractional anisotropy, axial diffusivity, radial diffusivity, and mean diffusivity weighted connectivity matrices. We used a deep convolutional neural network (DNN) for classification and the gradient class activation map (GRAD-CAM) technique to identify the neural connectivity features focused on by the DNN. DNN could classify both LGG and HGG with 98% accuracy. The sensitivity and specificity values were above 0.98. GRAD-CAM analysis revealed a distinct WM network pattern between LGG and HGG patients in the frontal, temporal, and parietal lobes. Our results demonstrate that glioma affects the WM network in LGG and HGG patients differently.

Metapopulation persistence can be inferred from incomplete surveys.

Habitat destruction and fragmentation are principal causes of species loss. While a local population might go extinct, a metapopulation-populations inhabiting habitat patches connected by dispersal-can persist regionally by recolonizing empty patches. To assess metapopulation persistence, two widely adopted indicators in conservation management are metapopulation capacity and patch importance. However, we face a fundamental limitation in that assessing metapopulation persistence requires that we survey or sample all the patches in a landscape: often these surveys are logistically challenging to conduct and repeat, which raises the question whether we can learn enough about the metapopulation persistence from an incomplete survey. Here, we provide a robust statistical approach to infer metapopulation capacity and patch importance by sampling a portion of all patches. We provided analytic arguments on why the metapopulation capacity and patch importance can be well predicted from sub-samples of habitat patches. Full-factorial simulations with more complex models corroborate our analytic predictions. We applied our model to an empirical metapopulation of mangrove hummingbirds (Amazilia boucardi). On the basis of our statistical framework, we provide some sampling suggestion for monitoring metapopulation persistence. Our approach allows for rapid and effective inference of metapopulation persistence from incomplete patch surveys.

Ayu-Characterization of healthy aging from neuroimaging data with deep learning and rsfMRI.

Estimating brain age and establishing functional biomarkers that are prescient of cognitive declines resulting from aging and different neurological diseases are still open research problems. Functional measures such as functional connectivity are gaining interest as potentially more subtle markers of neurodegeneration. However, brain functions are also affected by "normal" brain aging. More information is needed on how functional connectivity relates to aging, particularly in the absence of neurodegenerative disorders. Resting-state fMRI enables us to investigate functional brain networks and can potentially help us understand the processes of development as well as aging in terms of how functional connectivity (FC) matures during the early years and declines during the late years. We propose models for estimation of the chronological age of a healthy person from the resting state brain activation (rsfMRI). In this work, we utilized a dataset (N = 638, age-range 20-88) comprising rsfMRI images from the Cambridge Centre for Aging and Neuroscience (Cam-CAN) repository of a healthy population. We propose an age prediction pipeline Ayu which consists of data preprocessing, feature selection, and an attention-based model for deep learning architecture for brain age assessment. We extracted features from the static functional connectivity (sFC) to predict the subject's age and classified them into different age groups (young, middle, middle, and old ages). To the best of our knowledge, a classification accuracy of 72.619 % and a mean absolute error of 6.797, and an r 2 of 0.754 reported by our Ayu pipeline establish competitive benchmark results as compared to the state-of-the-art-approach. Furthermore, it is vital to identify how different functional regions of the brain are correlated. We also analyzed how functional regions contribute differently across ages by applying attention-based networks and integrated gradients. We obtained well-known resting-state networks using the attention model, which maps to within the default mode network, visual network, ventral attention network, limbic network, frontoparietal network, and somatosensory network connected to aging. Our analysis of fMRI data in healthy elderly Age groups revealed that dynamic FC tends to slow down and becomes less complex and more random with increasing age.

Thresholding Functional Connectivity Matrices to Recover the Topological Properties of Large-Scale Neuronal Networks.

The identification of the organization principles on the basis of the brain connectivity can be performed in terms of structural (i.e., morphological), functional (i.e., statistical), or effective (i.e., causal) connectivity. If structural connectivity is based on the detection of the morphological (synaptically mediated) links among neurons, functional and effective relationships derive from the recording of the patterns of electrophysiological activity (e.g., spikes, local field potentials). Correlation or information theory-based algorithms are typical routes pursued to find statistical dependencies and to build a functional connectivity matrix. As long as the matrix collects the possible associations among the network nodes, each interaction between the neuron i and j is different from zero, even though there was no morphological, statistical or causal connection between them. Hence, it becomes essential to find and identify only the significant functional connections that are predictive of the structural ones. For this reason, a robust, fast, and automatized procedure should be implemented to discard the "noisy" connections. In this work, we present a Double Threshold (DDT) algorithm based on the definition of two statistical thresholds. The main goal is not to lose weak but significant links, whose arbitrary exclusion could generate functional networks with a too small number of connections and altered topological properties. The algorithm allows overcoming the limits of the simplest threshold-based methods in terms of precision and guaranteeing excellent computational performances compared to shuffling-based approaches. The presented DDT algorithm was compared with other methods proposed in the literature by using a benchmarking procedure based on synthetic data coming from the simulations of large-scale neuronal networks with different structural topologies.

Spike signal transmission between modules and the predictability of spike activity in modular neuronal networks.

Modularity is a common feature of the nervous system across species and scales. Although it has been qualitatively investigated in network science, very little is known about how it affects spike signal transmission in neuronal networks at the mesoscopic level. Here, a neuronal network model is built to simulate dynamic interactions among different modules of neuronal networks. This neuronal network model follows the organizational principle of modular structure. The neurons can generate spikes like biological neurons, and changes in the strength of synaptic connections conform to the STDP learning rule. Based on this neuronal network model, we first quantitatively studied whether and to what extent the connectivity within and between modules can affect spike signal transmission, and found that spike signal transmission heavily depends on the connectivity between modules, but has little to do with the connectivity within modules. More importantly, we further found that the spike activity of a module can be predicted according to the spike activities of its adjacent modules through building a resting-state functional connectivity matrix.

Understanding connectivity of pearl oyster populations within Tuamotu atoll semi-closed lagoons: Cumulative insight from genetics and biophysical modelling approaches.

Connectivity affects species demography, (meta)population dynamics, evolution, phylogeny and biogeography. Various methodological approaches are applied to measure connectivity. Biophysical modelling can explore systematically the influence of atmospheric, oceanic and ecological forcing, while genetics measures connectivity patterns within the sampling strategy limit. In the Pacific Ocean pearl farming lagoons, the activity relies on spat collecting of the black lipped pearl oyster Pinctada margaritifera occurring after the larval dispersal phase, which follows spawning from wild or farmed populations. Biophysical 3D modelling and genomic studies have both separately brought insights on within-lagoon connectivity and on the origin of spats. Here, we combined previous genetics results with new realistic biophysical modelling scenarios to elucidate connectivity in Ahe Atoll lagoon. When combined, we identified the weather sequence likely explaining the realized connectivity observations. We discuss the strengths, weaknesses, opportunities and threats of combining these two approaches considering specific pearl farming demographic connectivity questions.

Robust intra-individual estimation of structural connectivity by Principal Component Analysis.

Fiber tractography based on diffusion-weighted MRI provides a non-invasive characterization of the structural connectivity of the human brain at the macroscopic level. Quantification of structural connectivity strength is challenging and mainly reduced to "streamline counting" methods. These are however highly dependent on the topology of the connectome and the particular specifications for seeding and filtering, which limits their intra-subject reproducibility across repeated measurements and, in consequence, also confines their validity. Here we propose a novel method for increasing the intra-subject reproducibility of quantitative estimates of structural connectivity strength. To this end, the connectome is described by a large matrix in positional-orientational space and reduced by Principal Component Analysis to obtain the main connectivity "modes". It was found that the proposed method is quite robust to structural variability of the data.

Dual stream neural networks for brain signal classification.

Objective. The primary objective of this work is to develop a neural nework classifier for arbitrary collections of functional neuroimaging signals to be used in brain-computer interfaces (BCIs).Approach. We propose a dual stream neural network (DSNN) for the classification problem. The first stream is an end-to-end classifier taking raw time-dependent signals as input and generating feature identification signatures from them. The second stream enhances the identified features from the first stream by adjoining a dynamic functional connectivity matrix aimed at incorporating nuanced multi-channel information during specified BCI tasks.Main results. The proposed DSNN classifier is benchmarked against three publicly available datasets, where the classifier demonstrates performance comparable to, or better than the state-of-art in each instance. An information theoretic examination of the trained network is also performed, utilizing various tools, to demonstrate how to glean interpretive insight into how the hidden layers of the network parse the underlying biological signals.Significance.The resulting DSNN is a subject-independent classifier that works for any collection of 1D functional neuroimaging signals, with the option of integrating domain specific information in the design.

A Three-Stage Teacher, Student Neural Networks and Sequential Feed Forward Selection-Based Feature Selection Approach for the Classification of Autism Spectrum Disorder.

Autism disorder, generally known as Autism Spectrum Disorder (ASD) is a brain disorder characterized by lack of communication skills, social aloofness and repetitions in the actions in the patients, which is affecting millions of the people across the globe. Accurate identification of autistic patients is considered a challenging task in the domain of brain disorder science. To address this problem, we have proposed a three-stage feature selection approach for the classification of ASD on the preprocessed Autism Brain Imaging Data Exchange (ABIDE) rs-fMRI Dataset. In the first stage, a large neural network which we call a "Teacher " was trained on the correlation-based connectivity matrix to learn the latent representation of the input. In the second stage an autoencoder which we call a "Student" autoencoder was given the task to learn those trained "Teacher" embeddings using the connectivity matrix input. Lastly, an SFFS-based algorithm was employed to select the subset of most discriminating features between the autistic and healthy controls. On the combined site data across 17 sites, we achieved the maximum 10-fold accuracy of 82% and for the individual site-wise data, based on 5-fold accuracy, our results outperformed other state of the art methods in 13 out of the total 17 site-wise comparisons.

Structural Network Analysis Using Diffusion MRI Tractography in Parkinson's Disease and Correlations With Motor Impairment.

Functional impairment of spatially distributed brain regions in Parkinson's disease (PD) suggests changes in integrative and segregative network characteristics, for which novel analysis methods are available. To assess underlying structural network differences between PD patients and controls, we employed MRI T1 gray matter segmentation and diffusion MRI tractography to construct connectivity matrices to compare patients and controls with data originating from two different centers. In the Dutch dataset (Data-NL), 14 PD patients, and 15 healthy controls were analyzed, while 19 patients and 18 controls were included in the Canadian dataset (Data-CA). All subjects underwent T1 and diffusion-weighted MRI. Patients were assessed with Part 3 of the Unified Parkinson's Disease Rating Scale (UPDRS). T1 images were segmented using FreeSurfer, while tractography was performed using ExploreDTI. The regions of interest from the FreeSurfer segmentation were combined with the white matter streamline sets resulting from the tractography, to construct connectivity matrices. From these matrices, both global and local efficiencies were calculated, which were compared between the PD and control groups and related to the UPDRS motor scores. The connectivity matrices showed consistent patterns among the four groups, without significant differences between PD patients and control subjects, either in Data-NL or in Data-CA. In Data-NL, however, global and local efficiencies correlated negatively with UPDRS scores at both the whole-brain and the nodal levels [false discovery rate (FDR) 0.05]. At the nodal level, particularly, the posterior parietal cortex showed a negative correlation between UPDRS and local efficiency, while global efficiency correlated negatively with the UPDRS in the sensorimotor cortex. The spatial patterns of negative correlations between UPDRS and parameters for network efficiency seen in Data-NL suggest subtle structural differences in PD that were below sensitivity thresholds in Data-CA. These correlations are in line with previously described functional differences. The methodological approaches to detect such differences are discussed.

Asymptotic profiles of the steady states for an SIS epidemic patch model with asymmetric connectivity matrix.

The dynamics of an SIS epidemic patch model with asymmetric connectivity matrix is analyzed. It is shown that the basic reproduction number [Formula: see text] is strictly decreasing with respect to the dispersal rate of the infected individuals. When [Formula: see text], the model admits a unique endemic equilibrium, and its asymptotic profiles are characterized for small dispersal rates. Specifically, the endemic equilibrium converges to a limiting disease-free equilibrium as the dispersal rate of susceptible individuals tends to zero, and the limiting disease-free equilibrium has a positive number of susceptible individuals on each low-risk patch. Furthermore, a sufficient and necessary condition is provided to characterize that the limiting disease-free equilibrium has no positive number of susceptible individuals on each high-risk patch. Our results extend earlier results for symmetric connectivity matrix, providing a positive answer to an open problem in Allen et al. (SIAM J Appl Math 67(5):1283-1309, 2007).

A Scalable FPGA Architecture for Randomly Connected Networks of Hodgkin-Huxley Neurons.

Human intelligence relies on the vast number of neurons and their interconnections that form a parallel computing engine. If we tend to design a brain-like machine, we will have no choice but to employ many spiking neurons, each one has a large number of synapses. Such a neuronal network is not only compute-intensive but also memory-intensive. The performance and the configurability of the modern FPGAs make them suitable hardware solutions to deal with these challenges. This paper presents a scalable architecture to simulate a randomly connected network of Hodgkin-Huxley neurons. To demonstrate that our architecture eliminates the need to use a high-end device, we employ the XC7A200T, a member of the mid-range Xilinx Artix®-7 family, as our target device. A set of techniques are proposed to reduce the memory usage and computational requirements. Here we introduce a multi-core architecture in which each core can update the states of a group of neurons stored in its corresponding memory bank. The proposed system uses a novel method to generate the connectivity vectors on the fly instead of storing them in a huge memory. This technique is based on a cyclic permutation of a single prestored connectivity vector per core. Moreover, to reduce both the resource usage and the computational latency even more, a novel approximate two-level counter is introduced to count the number of the spikes at the synapse for the sparse network. The first level is a low cost saturated counter implemented on FPGA lookup tables that reduces the number of inputs to the second level exact adder tree. It, therefore, results in much lower hardware cost for the counter circuit. These techniques along with pipelining make it possible to have a high-performance, scalable architecture, which could be configured for either a real-time simulation of up to 5120 neurons or a large-scale simulation of up to 65536 neurons in an appropriate execution time on a cost-optimized FPGA.

Some bounds on the distance-sum-connectivity matrix.

The distance-sum-connectivity matrix of a graph G is expressed by δ(i) and δ(j) such that i,j∈V . δ(i) and δ(j) are represented by a sum of the distance matrices for i

Diffusion tensor imaging using multiple coils for mouse brain connectomics.

The correlation between brain connectivity and psychiatric or neurological diseases has intensified efforts to develop brain connectivity mapping techniques on mouse models of human disease. The neural architecture of mouse brain specimens can be shown non-destructively and three-dimensionally by diffusion tensor imaging, which enables tractography, the establishment of a connectivity matrix and connectomics. However, experiments on cohorts of animals can be prohibitively long. To improve throughput in a 7-T preclinical scanner, we present a novel two-coil system in which each coil is shielded, placed off-isocenter along the axis of the magnet and connected to a receiver circuit of the scanner. Preservation of the quality factor of each coil is essential to signal-to-noise ratio (SNR) performance and throughput, because mouse brain specimen imaging at 7 T takes place in the coil-dominated noise regime. In that regime, we show a shielding configuration causing no SNR degradation in the two-coil system. To acquire data from several coils simultaneously, the coils are placed in the magnet bore, around the isocenter, in which gradient field distortions can bias diffusion tensor imaging metrics, affect tractography and contaminate measurements of the connectivity matrix. We quantified the experimental alterations in fractional anisotropy and eigenvector direction occurring in each coil. We showed that, when the coils were placed 12 mm away from the isocenter, measurements of the brain connectivity matrix appeared to be minimally altered by gradient field distortions. Simultaneous measurements on two mouse brain specimens demonstrated a full doubling of the diffusion tensor imaging throughput in practice. Each coil produced images devoid of shading or artifact. To further improve the throughput of mouse brain connectomics, we suggested a future expansion of the system to four coils. To better understand acceptable trade-offs between imaging throughput and connectivity matrix integrity, studies may seek to clarify how measurement variability, post-processing techniques and biological variability impact mouse brain connectomics.

Brain connectivity and novel network measures for Alzheimer's disease classification.

We compare a variety of different anatomic connectivity measures, including several novel ones, that may help in distinguishing Alzheimer's disease (AD) patients from controls. We studied diffusion-weighted magnetic resonance imaging from 200 subjects scanned as part of the Alzheimer's Disease Neuroimaging Initiative. We first evaluated measures derived from connectivity matrices based on whole-brain tractography; next, we studied additional network measures based on a novel flow-based measure of brain connectivity, computed on a dense 3-dimensional lattice. Based on these 2 kinds of connectivity matrices, we computed a variety of network measures. We evaluated the measures' ability to discriminate disease with a repeated, stratified 10-fold cross-validated classifier, using support vector machines, a supervised learning algorithm. We tested the relative importance of different combinations of features based on the accuracy, sensitivity, specificity, and feature ranking of the classification of 200 people into normal healthy controls and people with early or late mild cognitive impairment or AD.

OPTIMIZING BRAIN CONNECTIVITY NETWORKS FOR DISEASE CLASSIFICATION USING EPIC.

We propose a method to adaptively select an optimal cortical segmentation for brain connectivity analysis that maximizes feature-based disease classification performance. In standard structural connectivity analysis, the cortex is typically subdivided (parcellated) into N anatomical regions. White matter fiber pathways from tractography are used to compute an N × N matrix, which represents the pairwise connectivity between those regions. We optimize this representation by sampling over the space of possible region combinations and represent each configuration as a set partition of the N anatomical regions. Each partition is assigned a score using accuracy from a support vector machine (SVM) classifier of connectivity matrices in a group of patients and controls. We then define a high-dimensional optimization problem using simulated annealing to identify an optimal partition for maximum classification accuracy. We evaluate the results separately on test data using cross-validation. Specifically, we demonstrate results on the ADNI-2 dataset, where we optimally parcellate the cortex to yield an 85% classification accuracy using connectivity information alone. We refer to our method as evolving partitions to improve connectomics (EPIC).