Skip to main content

Über dieses Buch

This book constitutes the refereed proceedings of the Second International Workshop on Connectomics in NeuroImaging, CNI 2018, held in conjunction with MICCAI 2018 in Granada, Spain, in September 2018.

The 15 full papers presented were carefully reviewed and selected from 20 submissions. The papers deal with new advancements in network construction, analysis, and visualization techniques in connectomics and their use in clinical diagnosis and group comparison studies as well as in various neuroimaging applications.



Towards Ultra-High Resolution 3D Reconstruction of a Whole Rat Brain from 3D-PLI Data

3D reconstruction of the fiber connectivity of the rat brain at microscopic scale enables gaining detailed insight about the complex structural organization of the brain. We introduce a new method for registration and 3D reconstruction of high- and ultra-high resolution ( 64 \(\upmu \)m and 1.3 \(\upmu \)m pixel size) histological images of a Wistar rat brain acquired by 3D polarized light imaging (3D-PLI). Our method exploits multi-scale and multi-modal 3D-PLI data up to cellular resolution. We propose a new feature transform-based similarity measure and a weighted regularization scheme for accurate and robust non-rigid registration. To transform the 1.3 \(\upmu \)m ultra-high resolution data to the reference blockface images a feature-based registration method followed by a non-rigid registration is proposed. Our approach has been successfully applied to 278 histological sections of a rat brain and the performance has been quantitatively evaluated using manually placed landmarks by an expert.
Sharib Ali, Martin Schober, Philipp Schlömer, Katrin Amunts, Markus Axer, Karl Rohr

FOD-Based Registration for Susceptibility Distortion Correction in Connectome Imaging

Multi-shell, high resolution diffusion MRI (dMRI) data from the Human Connectome Project (HCP) provides an unprecedented opportunity for the in vivo mapping of human brain pathways. It was recently noted, however, that significant distortions remain present in the data of most subjects preprocessed by the HCP-Pipeline, which have been widely distributed and used extensively in connectomics research. Fundamentally this is caused by the reliance of the HCP tools on the B0 images for registering data from different phase encodings (PEs). In this work, we develop an improved framework to remove the residual distortion in data generated by the HCP-Pipeline. Our method is based on more advanced registration of fiber orientation distribution (FOD) images, which represent information of dMRI scans from all gradient directions and thus provide more reliable contrast to align data from different PEs. In our experiments, we focus on the brainstem area and compare our method with the preprocessing steps in the HCP-Pipeline. We show that our method can provide much improved distortion correction and generate FOD images with more faithful representation of brain pathways.
Yuchuan Qiao, Wei Sun, Yonggang Shi

GIFE: Efficient and Robust Group-Wise Isometric Fiber Embedding

Tractography is a prevalent technique for in vivo imaging of the white matter fibers (a.k.a. the tractograms), but it is also known to be error-prone. We previously propose the Group-wise Tractogram Analysis (GiTA) framework for identifying anatomically valid fibers across subjects according to cross-subject consistency. However, the original framework is based on computationally expensive brute-force KNN search. In this work, we propose a more general and efficient extension of GiTA. Our main idea is to find the finite dimensional vector-space representation of the fiber tracts of varied lengths across different subjects, and we call it the group-wise isometric fiber embedding (GIFE). This novel GIFE framework enables the application of the powerful and efficient vector space data analysis methods, such as the k-d tree KNN search, to GiTA. However, the conventional isometric embedding frameworks are not suitable for GIFE due to the massive fiber tracts and the registration errors in the original GiTA framework. To address these issues, we propose a novel method called multidimensional extrapolating (MDE) to achieve GIFE. In our experiment, simulation results show quantitatively that our method outperforms the other methods in terms of computational efficiency/tractability and robustness to errors in distance measurements for real fiber embedding. In addition, real experiment for group-wise optic radiation bundle reconstruction also shows clear improvement in anatomical validity of the results from our MDE method for 47 different subjects from the Human Connectome Project, compared to the results of other fiber embedding methods.
Junyan Wang, Yonggang Shi

Multi-modal Brain Tensor Factorization: Preliminary Results with AD Patients

Global brain network parameters suffer from low classification performance and fail to provide an insight into the neurodegenerative diseases. Besides, the variability in connectivity definitions poses a challenge. We propose to represent multi-modal brain networks over a population with a single 4D brain tensor (B) and factorize B to get a lower dimensional representation per case and per modality. We used 7 known functional networks as the canonical network space to get a 7D representation. In a preliminary study over a group of 20 cases, we assessed this representation for classification. We used 6 different connectivity definitions (modalities). Linear discriminant analysis results in 90–95% accuracy in binary classification. The assessment of the canonical coordinates reveals Salience subnetwork to be the most powerful in classification consistently over all connectivity definitions. The method can be extended to include functional networks and further be used to search for discriminating subnetworks.
Göktekin Durusoy, Abdullah Karaaslanlı, Demet Yüksel Dal, Zerrin Yıldırım, Burak Acar

Intact Connectional Morphometricity Learning Using Multi-view Morphological Brain Networks with Application to Autism Spectrum Disorder

The morphology of anatomical brain regions can be affected by neurological disorders, including dementia and schizophrenia, to various degrees. Hence, identifying the morphological signature of a specific brain disorder can improve diagnosis and better explain how neuroanatomical changes associate with function and cognition. To capture this signature, a landmark study introduced, brain morphometricity, a global metric defined as the proportion of phenotypic variation that can be explained by brain morphology derived from structural brain MRI scans. However, this metric is limited to investigating morphological changes using low-order measurements (e.g., regional volumes) and overlooks how these changes can be related to each other (i.e., how morphological changes in region A are influenced by changes in region B). Furthermore, it is derived from a pre-defined anatomical similarity matrix using a Gaussian function, which might not be robust to outliers and constrains the locality of data to a fixed bandwidth. To address these limitations, we propose the intact connectional brain morphometricity (ICBM), a metric that captures the variation of connectional changes in brain morphology. In particular, we use multi-view morphological brain networks estimated from multiple cortical attributes (e.g., cortical thickness) to learn an intact space that first integrates the morphological network views into a unified space. Next, we learn a multi-view morphological similarity matrix in the intact space by adaptively assigning neighbors for each data sample based on local connectivity. The learned similarity capturing the shared traits across morphological brain network views is then used to derive our ICBM via a linear mixed effect model. Our framework shows the potential of the proposed ICBM in capturing the connectional neuroanatomical signature of brain disorders such as Autism Spectrum Disorder.
Alaa Bessadok, Islem Rekik

Neonatal Morphometric Similarity Networks Predict Atypical Brain Development Associated with Preterm Birth

Morphometric similarity networks (MSNs) have been recently proposed as a novel, robust, and biologically plausible approach to generate structural connectomes from neuroimaging data. In this work, we apply this method to multi-centre neonatal data (postmenstrual age range: 37–45 weeks) to predict brain dysmaturation in preterm infants. To achieve this goal, we combined different imaging sequences (diffusion and structural MRI) to extract a set of metrics from cortical and subcortical brain regions (e.g. regional volumes, diffusion tensor metrics, neurite orientation dispersion and density imaging features) which were used to construct a similarity network. A regression model was then trained to predict postmenstrual age at the time of scanning from inter-regional connections. Finally, to quantify brain maturation, the Relative Brain Network Maturation Index (RBNMI) was computed as the difference between predicted and actual age. The model predicted chronological age with a mean absolute error of 0.88 (±0.63) weeks, and it consistently predicted preterm infants to have a lower RBNMI than term infants. We conclude that MSNs derived from multimodal imaging predict chronological brain development accurately, and provide a data-driven approach for defining cerebral dysmaturation associated with preterm birth.
Paola Galdi, Manuel Blesa, Gemma Sullivan, Gillian J. Lamb, David Q. Stoye, Alan J. Quigley, Michael J. Thrippleton, Mark E. Bastin, James P. Boardman

Heritability Estimation of Reliable Connectomic Features

Brain imaging genetics is an emerging research field to explore the underlying genetic architecture of brain structure and function measured by different imaging modalities. However, not all the changes in the brain are a consequential result of genetic effect and it is usually unknown which imaging phenotypes are promising for genetic analyses. In this paper, we focus on identifying highly heritable measures of structural brain networks derived from diffusion weighted imaging data. Using the twin data from the Human Connectome Project (HCP), we evaluated the reliability of fractional anisotropy measure, fiber length and fiber number of each edge in the structural connectome and seven network level measures using intraclass correlation coefficients. We then estimated the heritability of those reliable network measures using SOLAR-Eclipse software. Across all 64,620 network edges between 360 brain regions in the Glasser parcellation, we observed \(\sim \)5% of them with significantly high heritability in fractional anisotropy, fiber length or fiber number. All the tested network level measures, capturing the network integrality, segregation or resilience, are highly heritable, with variance explained by the additive genetic effect ranging from 59% to 77%.
Linhui Xie, Enrico Amico, Paul Salama, Yu-chien Wu, Shiaofen Fang, Olaf Sporns, Andrew J. Saykin, Joaquín Goñi, Jingwen Yan, Li Shen

Topological Data Analysis of Functional MRI Connectivity in Time and Space Domains

The functional architecture of the brain can be described as a dynamical system where components interact in flexible ways, constrained by physical connections between regions. Using correlation, either in time or in space, as an abstraction of functional connectivity, we analyzed resting state fMRI data from 1003 subjects. We compared several data preprocessing strategies and found that independent component-based nuisance regression outperformed other strategies, with the poorest reproducibility in strategies that include global signal regression. We also found that temporal vs. spatial functional connectivity can encode different aspects of cognition and personality. Topological analyses using persistent homology show that persistence barcodes are significantly correlated to individual differences in cognition and personality, with high reproducibility. Topological data analyses, including approaches to model connectivity in the time domain, are promising tools for representing high-level aspects of cognition, development, and neuropathology.
Keri L. Anderson, Jeffrey S. Anderson, Sourabh Palande, Bei Wang

Riemannian Regression and Classification Models of Brain Networks Applied to Autism

Functional connectivity from resting-state functional MRI (rsfMRI) is typically represented as a symmetric positive definite (SPD) matrix. Analysis methods that exploit the Riemannian geometry of SPD matrices appropriately adhere to the positive definite constraint, unlike Euclidean methods. Recently proposed approaches for rsfMRI analysis have achieved high accuracy on public datasets, but are computationally intensive and difficult to interpret. In this paper, we show that we can get comparable results using connectivity matrices under the log-Euclidean and affine-invariant Riemannian metrics with relatively simple and interpretable models. On ABIDE Preprocessed dataset, our methods classify autism versus control subjects with 71.1% accuracy. We also show that Riemannian methods beat baseline in regressing connectome features to subject autism severity scores.
Eleanor Wong, Jeffrey S. Anderson, Brandon A. Zielinski, P. Thomas Fletcher

Defining Patient Specific Functional Parcellations in Lesional Cohorts via Markov Random Fields

We propose a hierarchical Bayesian model that refines a population-based atlas using resting-state fMRI (rs-fMRI) coherence. Our method starts from an initial parcellation and then iteratively reassigns the voxel memberships at the subject level. Our algorithm uses a maximum a posteriori inference strategy based on the neighboring voxel assignments and the Pearson correlation coefficients between the voxel time series and the parcel reference signals. Our method is generalizable to different initial atlases, ensures spatial and temporal contiguity in the final network organization, and can handle subjects with brain lesions, whose rs-fMRI data varies tremendously from that of a healthy cohort. We validate our method by comparing the intra-network cohesion and the motor network identification against two baselines: a standard functional parcellation with no reassignment and a recently published method with a purely data-driven reassignment procedure. Our method outperforms the original functional parcellation in intra-network cohesion and both methods in motor network identification.
Naresh Nandakumar, Niharika S. D’Souza, Jeff Craley, Komal Manzoor, Jay J. Pillai, Sachin K. Gujar, Haris I. Sair, Archana Venkataraman

Data-Specific Feature Selection Method Identification for Most Reproducible Connectomic Feature Discovery Fingerprinting Brain States

Machine learning methods present unprecedented opportunities to advance our understanding of the connectomics of brain disorders. With the proliferation of extremely high-dimensional connectomic data drawn from multiple neuroimaging sources (e.g., functional and structural MRIs), effective feature selection (FS) methods have become indispensable components for (i) disentangling brain states (e.g., early vs late mild cognitive impairment) and (ii) identifying connectional features that might serve as biomarkers for treatment. Strangely, despite the extensive work on identifying stable discriminative features using a particular FS method, the challenge of choosing the best one from a large pool of existing FS techniques for optimally achieving (i) and (ii) using a dataset of interest remains unexplored. In essence, the question that we aim to address in this work is: “Given a set of feature selection methods \(\{FS_1, \dots , FS_K \}\), and a dataset of interest, which FS method might produce the most reproducible and ‘trustworthy’ connectomic features that accurately differentiate between two brain states?” This paper is an attempt to address this question by evaluating the performance of a particular feature selection for a specific data type in fulfilling criteria (i) and (ii). To this aim, we propose to model the relationships between a set of FS methods using a multi-graph architecture, where each graph quantifies the feature reproducibility power between graph nodes at a fixed number of top ranked features. Next, we integrate the reproducibility graphs with a discrepancy graph which captures the difference in classification performance between FS methods. This allows to identify, for a dataset of interest, the ‘central’ node with the highest degree, which reveals the most reliable and reproducible FS method for the target brain state classification task along with the most discriminative features fingerprinting these brain states. We evaluated our method on multi-view brain connectomic data for late mild cognitive impairment vs Alzheimer’s disease classification. Our experiments give insights into reproducible connectional features fingerprinting late dementia brain states.
Nicolas Georges, Islem Rekik

Towards Effective Functional Connectome Fingerprinting

The ability to uniquely characterize individual subjects based on their functional connectome (FC) is a key requirement for progress towards precision neuroscience. The recent availability of dense scans from individuals has enabled the neuroscience community to investigate the possibility of individual characterization. FC fingerprinting is a new and emerging problem where the goal is to uniquely characterize individual subjects based on FC. Recent studies reported near 100% accuracy suggesting that unique characterization of individuals is an accomplished task. However, there are multiple key aspects of the problem that are yet to be investigated. Specifically, (i) the impact of the number of subjects on fingerprinting performance needs to be studied, (ii) the impact of granularity of parcellation used to construct FC needs to be quantified, (iii) approaches to separate subject-specific information from generic information in the FC are yet to be explored. In this study, we investigated these three directions using publicly available resting-state functional magnetic resonance imaging data from the Human Connectome Project. Our results suggest that fingerprinting performance deteriorates with increase in the number of subjects and with the decrease in the granularity of parcellation. We also found that FC profiles of a small number of regions at high granularity capture subject-specific information needed for effective fingerprinting.
Kendrick Li, Gowtham Atluri

Connectivity-Driven Brain Parcellation via Consensus Clustering

We present two related methods for deriving connectivity-based brain atlases from individual connectomes. The proposed methods exploit a previously proposed dense connectivity representation, termed continuous connectivity, by first performing graph-based hierarchical clustering of individual brains, and subsequently aggregating the individual parcellations into a consensus parcellation. The search for consensus minimizes the sum of cluster membership distances, effectively estimating a pseudo-Karcher mean of individual parcellations. We assess the quality of our parcellations using (1) Kullback-Liebler and Jensen-Shannon divergence with respect to the dense connectome representation, (2) inter-hemispheric symmetry, and (3) performance of the simplified connectome in a biological sex classification task. We find that the parcellation based-atlas computed using a greedy search at a hierarchical depth 3 outperforms all other parcellation-based atlases as well as the standard Dessikan-Killiany anatomical atlas in all three assessments.
Anvar Kurmukov, Ayagoz Musabaeva, Yulia Denisova, Daniel Moyer, Boris Gutman

GRAND: Unbiased Connectome Atlas of Brain Network by Groupwise Graph Shrinkage and Network Diffusion

Network science is enhancing our understanding of how the human brain works at a systems level. A complete population-wise mapping of region-to-region connections, called connectome atlas, is the key to gaining a more full understanding for network-related brain disorders and for discovering biomarkers for early diagnosis. Since a brain network is commonly encoded in an adjacency matrix, it is difficult to apply the state-of-the-art atlas construction approaches by normalizing and averaging the individual adjacency matrices into a common space. In this paper, we propose a novel data-driven approach to construct an unbiased connectome atlas to capture both shared and complementary network topologies across individual brain networks, offering insight into the full spectrum of brain connectivity. Specifically, we employ a hypergraph to model the manifold of a population of brain networks. In this hypergraph, each node represents the individual participant’s brain network, and the edge weight captures the distance between two participants’ brain networks. The construction of a connectome atlas can be achieved using a hierarchical process of graph shrinkage toward the latent common space where the network topologies of all individual brain networks gradually become similar to each other. During the graph shrinkage, the adjacency matrix of each brain network is transformed to the common space by a series of diffusion matrices which exchange the connectome information with respect to the adjacency matrices on the neighboring hypergraph nodes such that the most representative characteristics of network topology are eventually propagated to the final connectome atlas. We have validated our connectome atlas construction method on the simulated brain network data and DTI data of 111 twin pairs in determining the genetic contribution of the structural connectivity.
Guorong Wu, Brent Munsell, Paul Laurienti, Moo K. Chung

Structural Subnetwork Evolution Across the Life-Span: Rich-Club, Feeder, Seeder

The impact of developmental and aging processes on brain connectivity and the connectome has been widely studied. Network theoretical measures and certain topological principles are computed from the entire brain, however there is a need to separate and understand the underlying subnetworks which contribute towards these observed holistic connectomic alterations. One organizational principle is the rich-club - a core subnetwork of brain regions that are strongly connected, forming a high-cost, high-capacity backbone that is critical for effective communication in the network. Investigations primarily focus on its alterations with disease and age. Here, we present a systematic analysis of not only the rich-club, but also other subnetworks derived from this backbone - namely feeder and seeder subnetworks. Our analysis is applied to structural connectomes in a normal cohort from a large, publicly available life-span study. We demonstrate changes in rich-club membership with age alongside a shift in importance from ’peripheral’ seeder to feeder subnetworks. Our results show a refinement within the rich-club structure (increase in transitivity and betweenness centrality), as well as increased efficiency in the feeder subnetwork and decreased measures of network integration and segregation in the seeder subnetwork. These results demonstrate the different developmental patterns when analyzing the connectome stratified according to its rich-club and the potential of utilizing this subnetwork analysis to reveal the evolution of brain architectural alterations across the life-span.
Markus D. Schirmer, Ai Wern Chung

Correction to: GRAND: Unbiased Connectome Atlas of Brain Network by Groupwise Graph Shrinkage and Network Diffusion

The original version of the chapter “GRAND: Unbiased Connectome Atlas of Brain Network by Groupwise Graph Shrinkage and Network Diffusion”, starting on p.127 was revised. An acknowledgment has been added. The original chapter was corrected.
Guorong Wu, Brent Munsell, Paul Laurienti, Moo K. Chung


Weitere Informationen

Premium Partner