Support Vector Machines, Multidimensional Scaling and Magnetic Resonance Imaging Reveal Structural Brain Abnormalities Associated With the Interaction Between Autism Spectrum Disorder and Sex

Abstract

Despite substantial efforts, it remains difficult to identify reliable neuroanatomic biomarkers of autism spectrum disorder (ASD) based on magnetic resonance imaging (MRI) and diffusion tensor imaging (DTI). Studies which use standard statistical methods to approach this task have been hampered by numerous challenges, many of which are innate to the mathematical formulation and assumptions of general linear models (GLM). Although the potential of alternative approaches such as machine learning (ML) to identify robust neuroanatomic correlates of psychiatric disease has long been acknowledged, few studies have attempted to evaluate the abilities of ML to identify structural brain abnormalities associated with ASD. Here we use a sample of 110 ASD patients and 83 typically developing (TD) volunteers (95 females) to assess the suitability of support vector machines (SVMs, a robust type of ML) as an alternative to standard statistical inference for identifying structural brain features which can reliably distinguish ASD patients from TD subjects of either sex, thereby facilitating the study of the interaction between ASD diagnosis and sex. We find that SVMs can perform these tasks with high accuracy and that the neuroanatomic correlates of ASD identified using SVMs overlap substantially with those found using conventional statistical methods. Our results confirm and establish SVMs as powerful ML tools for the study of ASD-related structural brain abnormalities. Additionally, they provide novel insights into the volumetric, morphometric, and connectomic correlates of this epidemiologically significant disorder.

Automatic classification of autism spectrum disorder in children using cortical thickness and support vector machine

“…Such findings are consistent with current literature evidence (Cogill & Wang, 2016; Crippa et al., 2015; Zoccante et al., 2010), which highlighted a putative involvement of the aforementioned areas in the pathogenesis of ASD symptoms, therefore confirming the reliability of SVM approaches in order to identify early diagnostic marker in ASD. The best accuracy obtained in our study was reached by the concatenation of gray matter thickness values of the following ROIs: right superior frontal gyrus, medial orbital frontal cortex, postcentral gyrus and left frontal pole, middle temporal gyrus, entorhinal cortex, insula, right hemisphere, and temporal gyrus, consistently with previous findings, which identified medial, ventromedial, and dorsolateral regions of the frontal lobe as being associated with ASD‐related neuroanatomical abnormalities (Irimia et al., 2018). More specifically, the right medial orbital sulcus has been postulated to be implicated in ASD‐related social impairment (Amaral et al., 2008), selectively in repetitive or stereotyped behaviors, in the abnormal actions of obsessive‐compulsive disorder (Atmaca et al., 2007; Whiteside et al., 2004), in decision making and in expectation rewarding (Kringelbach, 2005), all of which are impaired in ASD.…”

“…P-A = postero-anterior. This image was obtained with the tool available at https://developmentalimagingmcri.github.io/freesurfer_statsurf_display/ (Irimia et al, 2018). More specifically, the right medial orbital sulcus has been postulated to be implicated in ASD-related social impairment (Amaral et al, 2008), selectively in repetitive or stereotyped behaviors, in the abnormal actions of obsessive-compulsive disorder (Atmaca et al, 2007;Whiteside et al, 2004), in decision making and in expectation rewarding (Kringelbach, 2005), all of which are impaired in ASD.…”

Machine learning for autism spectrum disorder diagnosis using structural magnetic resonance imaging: Promising but challenging

“…The SA growth rate increased mainly in the left and right middle occipital gyri, right lingual gyrus area, and right cuneus ( Hazlett et al, 2017 ). Differences in the right occipital lobe are consistent with other studies ( Irimia et al, 2018 ; Landhuis, 2020 ) that explain visual perception differences between ASD patients and HCs. This appears to correlate with a report by Irimia et al (2018) that discovered the ASD group had higher areas and connectivity densities in the cuneus, occipital lobes, and the superior and transverse occipital sulci than the HC group.…”

“…Like most previous research (Irimia et al, 2018), it suffers from the over-aggregation of features on insufficient sample size. Using MRI and DTI data, the study evaluated the applicability of SVMs for studying the relationships between an ASD diagnosis and gender.…”

Neural manifold analysis of brain circuit dynamics in health and disease

“…Multidimensional scaling has been used to study changes in neural population dynamics in both large-scale neural simulations (Phoka et al, 2012) and in neural population recordings (Luczak et al, 2009). It has been applied for characterising glomerular activity across the olfactory bulb in predicting odorant quality perceptions (Youngentob et al, 2006), integrative cerebral cortical mechanisms during viewing (Tzagarakis et al, 2009), neuroplasticity in the processing of pitch dimensions (Chandrasekaran et al, 2007), emotional responses to music in patients with temporal lobe lesions (Dellacherie et al, 2011), and structural brain abnormalities associated with autism spectrum disorder (Irimia et al, 2018).…”