Table 6.

Summary of recent and representative studies aiming to distinguish individuals with ASD from TD individuals using multivariate analysis of ABIDE imaging data. Reported sample sizes are the numbers used for classification and do not necessarily reflect the study’s total sample size.

Reference	Study Participants	Experimental Methods	Key Features	Multivariate Technique	Key Results
Zhou et al. (2014)204	127 children with ASD and 153 TD children	Obtained rs-fMRI data from ABIDE repository and used a multi-parametric analytic approach, including network analysis to study connectivity	4 of 22 quantitative imaging features	Random decision tree	Classified with 98% accuracy for the full data set and 68% accuracy when using ten-fold cross validation
lidaka (2015)205	312 children and adolescents with ASD and 328 TD controls	Examined rs-fMRI data taken from ABIDE to analyze functional connectivity through correlation matrices	632 cells from the correlation matrix	Probabilistic neural network	Achieved 89% accuracy, 92% sensitivity, and 87% specificity using leave-one-out cross-validation
Kam et al. (2017)206	61 individuals with ASD and 72 TD individuals, all under 20 years old	Acquired rs-fMRI data from ABIDE data site to distinguish functional networks through hierarchical clustering	Connectivity features from five clusters	Discriminative restricted Boltzmann machine	Using ten-fold cross-validation, classified with 75% sensitivity and 85% specificity
Sadeghi et al. (2017)207	29 adolescents and adults with ASD and 31 TD controls	Analyzed properties of functional networks constructed from MRI images in the ABIDE data set	17 features from nodal metrics	SVM	Averaged 92% classification accuracy with five-fold cross-validation; 68% accuracy in independent set
Syed et al. (2017)208	392 individuals with ASD and 407 age- and sex-matched TD controls	Identified reproducible independent components of functional networks from ABIDE rs-fMRI data	Regions from the default mode network	k-means clustering	Clustering yielded 89% sensitivity and 90% specificity
Bi et al. (2018)209	45 individuals with ASD and 39 TD individuals	Evaluated connectivity from ABIDE rs-fMRI data through application of graph theory	272 graph metrics	Random SVM cluster	Obtained accuracies as high as 96% on the testing subset (26 samples, or 30% of total)
Heinsfeld et al. (2018)210	505 individuals with ASD and 530 TD individuals	Constructed connectivity matrices using correlations for regions’ time series averages using ABIDE rs-fMRI data	19900 functional connectivity features	Deep neural network	Achieved 70% accuracy, 74% sensitivity and 63% specificity with ten-fold cross-validation
Kong et al. (2019)211	78 individuals with ASD and 104 TD individuals	Analyzed brain connectivity through networks based on cortical regions constructed from ABIDE MRI data	3000 of the top cortical grey matter volume features	Deep neural network	Classified with up to 90% accuracy, 84% sensitivity, and 96% specificity using ten-fold cross-validation