Machine learning identification of EEG features predicting working memory performance in schizophrenia and healthy adults

Abstract

BackgroundWith millisecond-level resolution, electroencephalographic (EEG) recording provides a sensitive tool to assay neural dynamics of human cognition. However, selection of EEG features used to answer experimental questions is typically determined a priori. The utility of machine learning was investigated as a computational framework for extracting the most relevant features from EEG data empirically.MethodsSchizophrenia (SZ; n = 40) and healthy community (HC; n = 12) subjects completed a Sternberg Working Memory Task (SWMT) during EEG recording. EEG was analyzed to extract 5 frequency components (theta1, theta2, alpha, beta, gamma) at 4 processing stages (baseline, encoding, retention, retrieval) and 3 scalp sites (frontal-Fz, central-Cz, occipital-Oz) separately for correctly and incorrectly answered trials. The 1-norm support vector machine (SVM) method was used to build EEG classifiers of SWMT trial accuracy (correct vs. incorrect; Model 1) and diagnosis (HC vs. SZ; Model 2). External validity of SVM models was examined in relation to neuropsychological test performance and diagnostic classification using conventional regression-based analyses.ResultsSWMT performance was significantly reduced in SZ (p < .001). Model 1 correctly classified trial accuracy at 84 % in HC, and at 74 % when cross-validated in SZ data. Frontal gamma at encoding and central theta at retention provided highest weightings, accounting for 76 % of variance in SWMT scores and 42 % variance in neuropsychological test performance across samples. Model 2 identified frontal theta at baseline and frontal alpha during retrieval as primary classifiers of diagnosis, providing 87 % classification accuracy as a discriminant function.ConclusionsEEG features derived by SVM are consistent with literature reports of gamma’s role in memory encoding, engagement of theta during memory retention, and elevated resting low-frequency activity in schizophrenia. Tests of model performance and cross-validation support the stability and generalizability of results, and utility of SVM as an analytic approach for EEG feature selection.

Remembered or Forgotten?—An EEG-Based Computational Prediction Approach

“…CWT + SVM [21]: this approach used continuous wavelet transform to extract time-frequency features and then used 1-norm SVM to predict memory performance. Since the data were band-pass filtered between 0.05 and 15 Hz according to the structure of ConvEEGNN, the frequency bands extracted for 1-norm SVM are Theta 1 (centered at 4.00 Hz), Theta 2 (centered at 6.42 Hz) and Alpha (centered at 11.26 Hz).…”

“…The prediction accuracy and significance for ConvEEGNN are compared to: (1) LDA (2) ANN-1 (3) ANN-2 (4) SVM (5) SVM + LDA [20] (6) CWT + SVM [21]. The red bold line represents average prediction accuracy for each approach.…”

“…Methods from pattern recognition have been used for this prediction problem [19–21]. In a recent fMRI study, MVPA has been used to predict subsequent memory performance for 19 participants according to the period of encoding phonogram stimuli [19].…”

“…The accuracy using EEG signals might be relative lower than that using fMRI signals because of the higher spatial resolution of fMRI. In a recent EEG study with Sternberg Working Memory Task (SWMT), SVM has been used to identify signal features associated with working memory performance for 40 schizophrenia adults and 12 healthy adults [21]. Using continuous wavelet transform (CWT), EEG of each trial was analyzed to extract time-frequency and spatial features, including 5 frequency bands at 4 processing stages and 3 scalp sites.…”

“…This approach predicted SWMT trial performance with 84% accuracy in healthy adults and 74% accuracy in schizophrenia adults. Overall, the related work mentioned in [19–21] all used methods from pattern recognition to extract information from data and predict memory performance for each stimulus.…”