Table 7.

Summary of recent and representative studies aiming to distinguish individuals with ASD from TD individuals using multivariate analysis of EEG and MEG 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
Bosl et al. (2011)212	46 infants at high risk for ASD, and 33 low-risk controls	Collected EEG data and computed modified multiscale entropy as an indicator of normal brain development	Low, high, and mean multiscale entropy values for each of 64 channels	k-nearest neighbors, SVM, naive Bayes	Classified with accuracies between 72% and 77% in 9-month-olds using ten-fold cross-validation
Duffy and Als (2012)213	430 children with ASD and 554 TD children	Calculated spectral coherence variables from EEG measurements	40 spectral coherence factors	Discriminant analysis	Averaged 86% sensitivity and 89% specificity across ten split-half analyses and including all age groups
Khan et al. (2013)214	17 adolescents and young adults with ASD and 20 TD controls	Measured task-related local and long-range functional connectivity from MEG data	Four functional connectivity metrics	Quadratic discriminant analysis	Classified with 90% accuracy, 87% sensitivity, and 95% specificity
Jamal et al. (2014)215	12 children with ASD and 12 TD children	Extracted brain connectivity features from EEG measurements	4 of 36 brain connectivity features	Polynomial kernel SVM	With leave-one-out cross-validation, achieved 95% accuracy, 86% sensitivity, and 100% specificity
Khan et al. (2015)216	15 children and adolescents with ASD and 20 TD controls	Evaluated functional connectivity using tactile and resting state MEG recordings	Local functional connectivity index, Granger causality	Discriminant analysis	Achieved 87% sensitivity and 90% specificity using ten-fold cross-validation
Khan et al. (2016)217	15 children and adolescents with ASD and 20 TD controls	Used MEG and structural MRI to investigate abnormal functional connectivity	Three neurophysiological measures	Discriminant analysis	Averaged 90% sensitivity and 95% specificity with ten-fold cross-validation
Bosl et al. (2018)171	35 infants later diagnosed with ASD and 153 infants with no ASD diagnosis	Collected EEG measurements from 3 to 36 months of age to predict ASD diagnosis by 36 months of age	Subset of nonlinear invariant signal features selected from 1026 total	Radial basis function kernel SVM	Predicted ASD with 82–100% sensitivity and 88–99% specificity, depending on age, using leave-one-out cross-validation