Multivariate models of brain volume for identification of children and adolescents with fetal alcohol spectrum disorder

Abstract

Magnetic resonance imaging (MRI) studies of fetal alcohol spectrum disorder (FASD) have shown reductions of brain volume associated with prenatal exposure to alcohol. Previous studies consider regional brain volumes independently but ignore potential relationships across numerous structures. This study aims to (a) identify a multivariate model based on regional brain volume that discriminates children/adolescents with FASD versus healthy controls, and (b) determine if FASD classification performance can be increased by building classification models separately for each sex. Three‐dimensional T1‐weighted MRI from two independent childhood/adolescent datasets were used for training (79 FASD, aged 5.7–18.9 years, 35 males; 81 controls, aged 5.8–18.5 years, 32 males) and testing (67 FASD, aged 6.0–19.6 years, 38 males; 74 controls, aged 5.2–19.5 years, 42 males) a classification model. Using FreeSurfer, 87 regional brain volumes were extracted for each subject and were used as input into a support vector machine generating a classification model from the training data. The model performed moderately well on the test data with accuracy 77%, sensitivity 64%, and specificity 88%. Regions that contributed heavily to prediction in this model included temporal lobe and subcortical gray matter. Further investigation of two separate models for males and females showed slightly decreased accuracy compared to the model including all subjects (male accuracy 70%; female accuracy 67%), but had different regional contributions suggesting sex differences. This work demonstrates the potential of multivariate analysis of brain volumes for discriminating children/adolescents with FASD and provides indication of the most affected regions.

High‐resolution diffusion tensor imaging identifies hippocampal volume loss without diffusion changes in individuals with prenatal alcohol exposure

“…Ethnicity differed between the control and PAE groups in this study; however, bilateral whole‐hippocampus volume reductions in the current PAE group are similar to those reported in an ethnically diverse sample (Astley et al, 2009). Moreover, a machine‐learning study that demonstrated discrimination of children and adolescents with FASD based on brain volume reductions, including the hippocampus, suggests that volume differences in PAE do not differ based on ethnicity (Little & Beaulieu, 2020).…”

“…One study demonstrated that expected sex differences (female hippocampal volumes smaller than males) were present in controls but absent in individuals with PAE (McLachlan et al, 2020). Moreover, machine learning of multivariate models for the prediction of PAE diagnosis based solely on brain volumes identified right hippocampus volume in the top five features for a male‐specific model that was not as highly ranked in a female‐specific model (Little & Beaulieu, 2020).…”

Current Socioeconomic Status Correlates With Brain Volumes in Healthy Children and Adolescents but Not in Children With Prenatal Alcohol Exposure

“…From the overall NeuroDevNet sample, 177 participants underwent brain MRI including 3D T1-weighted MPRAGE (1 × 1 × 1 mm 3 in ~5–6 min; for acquisition details see Little and Beaulieu, 2020 ) at four imaging sites in Canada (University of British Columbia, UBC, 3T Philips Intera; University of Alberta, UofA, 1.5T Siemens Sonata; University of Manitoba, UofM, 3T Siemens Trio; and Queen’s University, 3T Siemens Trio). In total, 20 participants were excluded after visual inspection for motion artifacts and quality control from the CIVET quality control program for segmentation and surface extraction, including eight controls and nine PAE from motion artifacts, and two controls and one PAE from segmentation errors in local areas.…”

The cingulum in very preterm infants relates to language and social-emotional impairment at 2 years of term-equivalent age

“…While previous reports of altered WM in preterm children have been obtained using univariate analytical methods with high exploratory power – e.g., region of interest (ROI)-based analysis ( Caldinelli et al, 2017 , Dodson et al, 2017 , Murray et al, 2016 ), tract-based spatial statistics (TBSS) ( Coker-Bolt et al, 2016 , Collins et al, 2019 , Hollund et al, 2018 , Jurcoane et al, 2016 , Murner-Lavanchy et al, 2018 ), and tensor-based morphometry (TBM) ( Rajagopalan et al, 2017 ) – these methods may be too conservative to detect subtle, spatially distributed differences because they require corrections for multiple comparisons to control the expected false discovery rate (FDR) ( Ecker et al, 2010b ). By contrast, a multivariate pattern analysis (MVPA) (e.g., support vector machine (SVM) and logistic regression models) accounts for interregional correlations and features increased sensitivity to abnormalities in neural systems ( Ecker et al, 2010 , Li et al, 2014 , Little and Beaulieu, 2019 , Schadl et al, 2018 ). MVPA uses multivariate features from neuroimaging data to classify individual observations into different groups and thus reveals the contributing spatial and/or temporal patterns associated with the categories ( Lao et al, 2004 ).…”