Unsupervised Manifold Learning Using High-Order Morphological Brain Networks Derived From T1-w MRI for Autism Diagnosis

Abstract

Brain disorders, such as Autism Spectrum Disorder (ASD), alter brain functional (from fMRI) and structural (from diffusion MRI) connectivities at multiple levels and in varying degrees. While unraveling such alterations have been the focus of a large number of studies, morphological brain connectivity has been out of the research scope. In particular, shape-to-shape relationships across brain regions of interest (ROIs) were rarely investigated. As such, the use of networks based on morphological brain data in neurological disorder diagnosis, while leveraging the advent of machine learning, could complement our knowledge on brain wiring alterations in unprecedented ways. In this paper, we use conventional T1-weighted MRI to define morphological brain networks (MBNs), each quantifying shape relationship between different cortical regions for a specific cortical attribute at both low-order and high-order levels. While typical brain connectomes investigate the relationship between two ROIs, we propose high-order MBN which better captures brain complex interactions by modeling the morphological relationship between pairs of ROIs. For ASD identification, we present a connectomic manifold learning framework, which learns multiple kernels to estimate a similarity measure between ASD and normal controls (NC) connectional features, to perform dimensionality reduction for clustering ASD and NC subjects. We benchmark our ASD identification method against both supervised and unsupervised state-of-the-art methods, while depicting the most discriminative high- and low-order relationships between morphological regions in the left and right hemispheres.

Predicting full-scale and verbal intelligence scores from functional Connectomic data in individuals with autism Spectrum disorder

“…ASD subjects had been diagnosed based on the autism criteria sets in the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition (DSM-IV-TR) (Zwaigenbaum et al 2015). Each brain image was parcellated into 116 regions of interest (ROIs) using automatic labelling atlas (AAL) template (Tzourio-Mazoyer et al 2002). Then, functional connectomes, represented by 116 × 116 symmetric matrices, were generated for each subject.…”

Diagnosis of Autism Spectrum Disorder Using Central-Moment Features From Low- and High-Order Dynamic Resting-State Functional Connectivity Networks

“…Recently, emerging connectomic studies have demonstrated that examining more complex interactions involving multiple ROIs can provide more valuable insights into brain disease fingerprinting and diagnosis (Chen et al, 2016;Zhang et al, 2016Zhang et al, , 2017aGuo et al, 2017; FIGURE 1 | Flow chart of constructing low-(Lo-D-FCNs) and high-order dynamic functional connectivity networks (Ho-D-FCNs), where (A1) denotes the resting-state functional MRI (rs-fMRI) time series associated with each region of interest (ROI), (A2) denotes the second rs-fMRI subseries based on a sliding window, (B1) is the Lo-D-FCNs, (B2) is the second subnetwork of Lo-D-FCNs, (C1) is the Ho-D-FCNs, and (C2) denotes the second subnetwork from Ho-D-FCNs. Morris and Rekik, 2017;Soussia and Rekik, 2018;Zhao et al, 2018). Correspondingly, those FCNs, reflecting complex interactions across multiple ROIs, are referred as the highorder FCN (Ho-FCN).…”

Estimation of gender-specific connectional brain templates using joint multi-view cortical morphological network integration

“…Although fMRI and dMRI neuroimaging modalities allowed the discovery of predictive brain connections fingerprinting gender differences, they may have a few limitations. On the one hand, functional MRI can produce spurious and noisy connectomes due to the low signal-to-noise ratio induced by non-neural noise (Soussia and Rekik 2018). On the other hand, diffusion MRI can produce biased and largely variable structural connectomes depending on the employed fiber tractography method; a fact supported by a recent study (Petrov et al 2017) which evaluated 35 methods to generate structural connectomes and showed that how variations in diffusion MRI pre-processing steps affect network reliability and its ability to classify subjects remains opaque.…”

“…Specifically, CMNs model the relationship in morphology between different cortical regions quantified using specific cortical measurements. For instance, CMNs were investigated in neurodegenerative disorders (Mahjoub et al 2018;Lisowska et al 2019) as well as in neuropsychiatric disorders (Soussia and Rekik 2018;Georges et al 2020). (Nebli and Rekik 2019) presented the first study on gender differences using CMNs of healthy subjects.…”