Machine learning identification of EEG features predicting working memory performance in schizophrenia and healthy adults.
ABSTRACT: BACKGROUND:With 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. METHODS:Schizophrenia (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. RESULTS:SWMT 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. CONCLUSIONS:EEG 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.
Project description:BACKGROUND:Impairment in episodic memory is one of the most robust findings in schizophrenia. Disruptions of fronto-temporal functional connectivity that could explain some aspects of these deficits have been reported. Recent work has identified abnormal hippocampal function in unmedicated patients with schizophrenia (SZ), such as increased metabolism and glutamate content that are not always seen in medicated SZ. For these reasons, we hypothesized that altered fronto-temporal connectivity might originate from the hippocampus and might be partially restored by antipsychotic medication. METHODS:Granger causality methods were used to evaluate the effective connectivity between frontal and temporal regions in 21 unmedicated SZ and 20 matched healthy controls (HC) during performance of an episodic memory retrieval task. In 16 SZ, effective connectivity between these regions was evaluated before and after 1-week of antipsychotic treatment. RESULTS:In HC, significant effective connectivity originating from the right hippocampus to frontal regions was identified. Compared to HC, unmedicated SZ showed significant altered fronto-temporal effective connectivity, including reduced right hippocampal to right medial frontal connectivity. After 1-week of antipsychotic treatment, connectivity more closely resembled the patterns observed in HC, including increased effective connectivity from the right hippocampus to frontal regions. CONCLUSIONS:These results support the notion that memory disruption in schizophrenia might originate from hippocampal dysfunction and that medication restores some aspects of fronto-temporal dysconnectivity. Patterns of fronto-temporal connectivity could provide valuable biomarkers to identify new treatments for the symptoms of schizophrenia, including memory deficits.
Project description:Working memory (WM) processes depend on our momentary mental state and therefore exhibit considerable fluctuations. Here, we investigate the interplay of task-preparatory and task-related brain activity as represented by pre-stimulus BOLD-fluctuations and spectral EEG from the retention periods of a visual WM task. Visual WM is used to maintain sensory information in the brain enabling the performance of cognitive operations and is associated with mental health. We tested 22 subjects simultaneously with EEG and fMRI while performing a visuo-verbal Sternberg task with two different loads, allowing for the temporal separation of preparation, encoding, retention and retrieval periods. Four temporally coherent networks (TCNs)-the default mode network (DMN), the dorsal attention, the right and the left WM network-were extracted from the continuous BOLD data by means of a group ICA. Subsequently, the modulatory effect of these networks' pre-stimulus activation upon retention-related EEG activity in the theta, alpha, and beta frequencies was analyzed. The obtained results are informative in the context of state-dependent information processing. We were able to replicate two well-known load-dependent effects: the frontal-midline theta increase during the task and the decrease of pre-stimulus DMN activity. As our main finding, these two measures seem to depend on each other as the significant negative correlations at frontal-midline channels suggested. Thus, suppressed pre-stimulus DMN levels facilitated later task related frontal midline theta increases. In general, based on previous findings that neuronal coupling in different frequency bands may underlie distinct functions in WM retention, our results suggest that processes reflected by spectral oscillations during retention seem not only to be "online" synchronized with activity in different attention-related networks but are also modulated by activity in these networks during preparation intervals.
Project description:Patients with schizophrenia show abnormal dynamics and structure of temporally -coherent networks (TCNs) assessed using fMRI, which undergo adaptive shifts in preparation for a cognitively demanding task. During working memory (WM) tasks, patients with schizophrenia show persistent deficits in TCNs as well as EEG indices of WM. Studying their temporal relationship during WM tasks might provide novel insights into WM performance deficits seen in schizophrenia. Simultaneous EEG-fMRI data were acquired during the performance of a verbal Sternberg WM task with two load levels (load 2 and load 5) in 17 patients with schizophrenia and 17 matched healthy controls. Using covariance mapping, we investigated the relationship of the activity in the TCNs before the memoranda were encoded and EEG spectral power during the retention interval. We assessed four TCNs - default mode network (DMN), dorsal attention network (dAN), left and right working memory networks (WMNs) - and three EEG bands - theta, alpha, and beta. In healthy controls, there was a load-dependent inverse relation between DMN and frontal midline theta power and an anti-correlation between DMN and dAN. Both effects were not significantly detectable in patients. In addition, healthy controls showed a left-lateralized load-dependent recruitment of the WMNs. Activation of the WMNs was bilateral in patients, suggesting more resources were recruited for successful performance on the WM task. Our findings support the notion of schizophrenia patients showing deviations in their neurophysiological responses before the retention of relevant information in a verbal WM task. Thus, treatment strategies as neurofeedback -targeting prestates could be beneficial as task performance relies on the preparatory state of the brain.
Project description:Studies have shown that individuals with schizophrenia suffer from memory impairments. In this study, we combined proton magnetic resonance spectroscopy (¹H-MRS) and functional magnetic resonance imaging (fMRI) to clarify the neurobiology of memory deficits in schizophrenia.We used single-voxel MRS acquired in the left hippocampus and fMRI during performance of a memory task to obtain measures of neurochemistry and functional response in 28 stable, medicated participants with schizophrenia (SZ) and 28 matched healthy controls (HC).The SZ group had significantly decreased blood oxygen level-dependent (BOLD) signal in left inferior frontal gyrus (IFG) during encoding and in the anterior cingulate cortex (ACC) and superior temporal gyrus (STG) during retrieval. We did not find significant differences in N-acetylaspartate/creatine (NAA/Cr) or glutamate+glutamine (Glx/Cr) levels between the groups, but did find a significant positive correlation between NAA/Cr and Glx/Cr in the HC group that was absent in the SZ group. There were no significant correlations between BOLD and MRS measured in the hippocampus. Further analyses revealed a negative correlation between left IFG BOLD and task performance in the SZ group. Finally, in the HC group, the left IFG BOLD was positively correlated with Glx/Cr.We replicated findings of reduced BOLD signal in left IFG and of an altered relationship between IFG BOLD response and task performance in the SZ. The absence of correlation between NAA/Cr and Glx/Cr levels in patients might suggest underlying pathologies of the glutamate-glutamine cycle and/or mitochondria.
Project description:Visual working memory (VWM) distinguishes between representations relevant for imminent versus future perceptual goals. We investigated how the brain sequentially prioritizes visual working memory representations that serve consecutive tasks. Observers remembered two targets for a sequence of two visual search tasks, thus making one target currently relevant, and the other prospectively relevant. We show that during the retention interval prior to the first search, lateralized parieto-occipital EEG alpha (8-14 Hz) suppression is stronger for current compared with prospective search targets. Crucially, between the first and second search task, this difference in posterior alpha lateralization reverses, reflecting the change in priority states of the two target representations. Connectivity analyses indicate that this switch in posterior alpha lateralization is driven by frontal delta/low-theta (2-6 Hz) activity. Moreover, this frontal low-frequency signal also predicts task performance after the switch. We thus obtained evidence for large-scale network interactions underlying the flexible shifting between the priority states of multiple memory representations in VWM.
Project description:Using a wireless single channel EEG device, we investigated the feasibility of using short-term frontal EEG as a means to evaluate the dynamic changes of mental workload. Frontal EEG signals were recorded from twenty healthy subjects performing four cognitive and motor tasks, including arithmetic operation, finger tapping, mental rotation and lexical decision task. Our findings revealed that theta activity is the common EEG feature that increases with difficulty across four tasks. Meanwhile, with a short-time analysis window, the level of mental workload could be classified from EEG features with 65%-75% accuracy across subjects using a SVM model. These findings suggest that frontal EEG could be used for evaluating the dynamic changes of mental workload.
Project description:Schizophrenia (SZ) is associated with increased somatic morbidity and mortality, in addition to cognitive impairments similar to those seen in normal aging, which may suggest that pathological accelerated aging occurs in SZ. Therefore, we aim to evaluate the relationships of age, telomere length (TL), and CCL11 (aging and inflammatory biomarkers, respectively), gray matter (GM) volume and episodic memory performance in individuals with SZ compared to healthy controls (HC). One hundred twelve participants (48 SZ and 64 HC) underwent clinical and memory assessments, structural MRI, and had their peripheral blood drawn for biomarkers analysis. Comparisons of group means and correlations were performed. Participants with SZ had decreased TL and GM volume, increased CCL11, and worse memory performance compared to HC. In SZ, shorter TL was related to increased CCL11, and both biomarkers were related to reduced GM volume, all of which were related to worse memory performance. Older age was only associated with reduced GM, but longer duration of illness was related with all the aforementioned variables. Younger age of disease onset was associated with increased CCL11 levels and worse memory performance. In HC, there were no significant correlations except between memory and GM. Our results are consistent with the hypothesis of accelerated aging in SZ. These results may indicate that it is not age itself, but the impact of the disease associated with a pathological accelerated aging that leads to impaired outcomes in SZ.
Project description:There is significant current interest in decoding mental states from electroencephalography (EEG) recordings. EEG signals are subject-specific, are sensitive to disturbances, and have a low signal-to-noise ratio, which has been mitigated by the use of laboratory-grade EEG acquisition equipment under highly controlled conditions. In the present study, we investigate single-trial decoding of natural, complex stimuli based on scalp EEG acquired with a portable, 32 dry-electrode sensor system in a typical office setting. We probe generalizability by a leave-one-subject-out cross-validation approach. We demonstrate that support vector machine (SVM) classifiers trained on a relatively small set of denoised (averaged) pseudotrials perform on par with classifiers trained on a large set of noisy single-trial samples. We propose a novel method for computing sensitivity maps of EEG-based SVM classifiers for visualization of EEG signatures exploited by the SVM classifiers. Moreover, we apply an NPAIRS resampling framework for estimation of map uncertainty, and thus show that effect sizes of sensitivity maps for classifiers trained on small samples of denoised data and large samples of noisy data are similar. Finally, we demonstrate that the average pseudotrial classifier can successfully predict the class of single trials from withheld subjects, which allows for fast classifier training, parameter optimization, and unbiased performance evaluation in machine learning approaches for brain decoding.
Project description:Functional connectivity (FC) patterns obtained from resting-state functional magnetic resonance imaging data are commonly employed to study neuropsychiatric conditions by using pattern classifiers such as the support vector machine (SVM). Meanwhile, a deep neural network (DNN) with multiple hidden layers has shown its ability to systematically extract lower-to-higher level information of image and speech data from lower-to-higher hidden layers, markedly enhancing classification accuracy. The objective of this study was to adopt the DNN for whole-brain resting-state FC pattern classification of schizophrenia (SZ) patients vs. healthy controls (HCs) and identification of aberrant FC patterns associated with SZ. We hypothesized that the lower-to-higher level features learned via the DNN would significantly enhance the classification accuracy, and proposed an adaptive learning algorithm to explicitly control the weight sparsity in each hidden layer via L1-norm regularization. Furthermore, the weights were initialized via stacked autoencoder based pre-training to further improve the classification performance. Classification accuracy was systematically evaluated as a function of (1) the number of hidden layers/nodes, (2) the use of L1-norm regularization, (3) the use of the pre-training, (4) the use of framewise displacement (FD) removal, and (5) the use of anatomical/functional parcellation. Using FC patterns from anatomically parcellated regions without FD removal, an error rate of 14.2% was achieved by employing three hidden layers and 50 hidden nodes with both L1-norm regularization and pre-training, which was substantially lower than the error rate from the SVM (22.3%). Moreover, the trained DNN weights (i.e., the learned features) were found to represent the hierarchical organization of aberrant FC patterns in SZ compared with HC. Specifically, pairs of nodes extracted from the lower hidden layer represented sparse FC patterns implicated in SZ, which was quantified by using kurtosis/modularity measures and features from the higher hidden layer showed holistic/global FC patterns differentiating SZ from HC. Our proposed schemes and reported findings attained by using the DNN classifier and whole-brain FC data suggest that such approaches show improved ability to learn hidden patterns in brain imaging data, which may be useful for developing diagnostic tools for SZ and other neuropsychiatric disorders and identifying associated aberrant FC patterns.
Project description:Color is a perceptual stimulus that has a significant impact on improving human emotion and memory. Studies have revealed that colored multimedia learning materials (MLMs) have a positive effect on learner's emotion and learning where it was assessed by subjective/objective measurements. This study aimed to quantitatively assess the influence of colored MLMs on emotion, cognitive processes during learning, and long-term memory (LTM) retention using electroencephalography (EEG). The dataset consisted of 45 healthy participants, and MLMs were designed in colored or achromatic illustrations to elicit emotion and that to assess its impact on LTM retention after 30-min and 1-month delay. The EEG signal analysis was first started to estimate the effective connectivity network (ECN) using the phase slope index and expand it to characterize the ECN pattern using graph theoretical analysis. EEG results showed that colored MLMs had influences on theta and alpha networks, including (1) an increased frontal-parietal connectivity (top-down processing), (2) a larger number of brain hubs, (3) a lower clustering coefficient, and (4) a higher local efficiency, indicating that color influences information processing in the brain, as reflected by ECN, together with a significant improvement in learner's emotion and memory performance. This is evidenced by a more positive emotional valence and higher recall accuracy for groups who learned with colored MLMs than that of achromatic MLMs. In conclusion, this paper demonstrated how the EEG ECN parameters could help quantify the influences of colored MLMs on emotion and cognitive processes during learning.