Javier Gomez-Pilar

Adaptive semi-supervised classification to reduce intersession non-stationarity in multiclass motor imagery-based brain-computer interfaces

The intersession non-stationarity in electroencephalogram (EEG) data is a major issue to robust operation of brain-computer interfaces (BCIs). The aim of this paper is to propose a semi-supervised classification algorithm whereby the model is gradually enhanced with unlabeled data collected online. Additionally, a processing stage is introduced before classification to adaptively reduce the small fluctuations between the features from training and evaluation sessions. The key element of the classification algorithm is an optimized version of kernel discriminant analysis called spectral regression kernel discriminant analysis (SRKDA) in order to meet the low computational cost requirement for online BCI applications. Four different approaches, SRKDA and sequential updating semi-supervised SRKDA (SUSS-SRKDA) with or without adaptive processing stage are considered to quantify the advantages of semi-supervised learning and adaptive stage. The session-to-session performance for each of them is evaluated on the multiclass problem (four motor imagery tasks: the imagination of movement of the left hand, right hand, both feet, and tongue) posed in the BCI Competition IV dataset 2a. The results agree with previous studies reporting semi-supervised learning enhances the adaptability of BCIs to non-stationary EEG data. Moreover, we show that reducing the inter-session non-stationarity before classification further boosts its performance. The classification method combining adaptive processing and semi-supervised learning is found to yield the highest session-to session transfer results presented so far for this multiclass dataset: accuracy (77%) and Cohen?s kappa coefficient (0.70). Thus, the proposed methodology could be of great interest for real-life BCIs.

Neural Network Reorganization Analysis During an Auditory Oddball Task in Schizophrenia Using Wavelet Entropy

The aim of the present study was to characterize the neural network reorganization during a cognitive task in schizophrenia (SCH) by means of wavelet entropy (WE). Previous studies suggest that the cognitive impairment in patients with SCH could be related to the disrupted integrative functions of neural circuits. Nevertheless, further characterization of this effect is needed, especially in the time-frequency domain. This characterization is sensitive to fast neuronal dynamics and their synchronization that may be an important component of distributed neuronal interactions; especially in light of the disconnection hypothesis for SCH and its electrophysiological correlates. In this work, the irregularity dynamics elicited by an auditory oddball paradigm were analyzed through synchronized-averaging (SA) and single-trial (ST) analyses. They provide complementary information on the spatial patterns involved in the neural network reorganization. Our results from 20 healthy controls and 20 SCH patients showed a WE decrease from baseline to response both in controls and SCH subjects. These changes were significantly more pronounced for healthy controls after ST analysis, mainly in central and frontopolar areas. On the other hand, SA analysis showed more widespread spatial differences than ST results. These findings suggest that the activation response is weakly phase-locked to stimulus onset in SCH and related to the default mode and salience networks. Furthermore, the less pronounced changes in WE from baseline to response for SCH patients suggest an impaired ability to reorganize neural dynamics during an oddball task.

Adaptive Stacked Generalization for Multiclass Motor Imagery-Based Brain Computer Interfaces

Practical motor imagery-based brain computer interface (MI-BCI) applications are limited by the difficult to decode brain signals in a reliable way. In this paper, we propose a processing framework to address non-stationarity, as well as handle spectral, temporal, and spatial characteristics associated with execution of motor tasks. Stacked generalization is used to exploit the power of classifier ensembles for combining information coming from multiple sources and reducing the existing uncertainty in EEG signals. The outputs of several regularized linear discriminant analysis (RLDA) models are combined to account for temporal, spatial, and spectral information. The resultant algorithm is called stacked RLDA (SRLDA). Additionally, an adaptive processing stage is introduced before classification to reduce the harmful effect of intersession non-stationarity. The benefits of the proposed method are evaluated on the BCI Competition IV dataset 2a. We demonstrate its effectiveness in binary and multiclass settings with four different motor imagery tasks: left-hand, right-hand, both feet, and tongue movements. The results show that adaptive SRLDA outperforms the winner of the competition and other approaches tested on this multiclass dataset.

Assessment of Time and Frequency Domain Entropies to Detect Sleep Apnoea in Heart Rate Variability Recordings from Men and Women

Heart rate variability (HRV) provides useful information about heart dynamics both under healthy and pathological conditions. Entropy measures have shown their utility to characterize these dynamics. In this paper, we assess the ability of spectral entropy (SE) and multiscale entropy (MsE) to characterize the sleep apnoea-hypopnea syndrome (SAHS) in HRV recordings from 188 subjects. Additionally, we evaluate eventual differences in these analyses depending on the gender. We found that the SE computed from the very low frequency band and the low frequency band showed ability to characterize SAHS regardless the gender; and that MsE features may be able to distinguish gender specificities. SE and MsE showed complementarity to detect SAHS, since several features from both analyses were automatically selected by the forward-selection backward-elimination algorithm. Finally, SAHS was modelled through logistic regression (LR) by using optimum sets of selected features. Modelling SAHS by genders reached significant higher performance than doing it in a jointly way. The highest diagnostic ability was reached by modelling SAHS in women. The LR classifier achieved 85.2% accuracy (Acc) and 0.951 area under the ROC curve (AROC). LR for men reached 77.6% Acc and 0.895 AROC, whereas LR for the whole set reached 72.3% Acc and 0.885 AROC. Our results show the usefulness of the SE and MsE analyses of HRV to detect SAHS, as well as suggest that, when using HRV, SAHS may be more accurately modelled if data are separated by gender.

An Asynchronous P300-Based Brain-Computer Interface Web Browser for Severely Disabled People

This paper presents an electroencephalographic (EEG) P300-based brain–computer interface (BCI) Internet browser. The system uses the “odd-ball” row-col paradigm for generating the P300 evoked potentials on the scalp of the user, which are immediately processed and translated into web browser commands. There were previous approaches for controlling a BCI web browser. However, to the best of our knowledge, none of them was focused on an assistive context, failing to test their applications with a suitable number of end users. In addition, all of them were synchronous applications, where it was necessary to introduce a “read-mode” command in order to avoid a continuous command selection. Thus, the aim of this study is twofold: 1) to test our web browser with a population of multiple sclerosis (MS) patients in order to assess the usefulness of our proposal to meet their daily communication needs; and 2) to overcome the aforementioned limitation by adding a threshold that discerns between control and non-control states, allowing the user to calmly read the web page without undesirable selections. The browser was tested with sixteen MS patients and five healthy volunteers. Both quantitative and qualitative metrics were obtained. MS participants reached an average accuracy of 84.14%, whereas 95.75% was achieved by control subjects. Results show that MS patients can successfully control the BCI web browser, improving their personal autonomy.

Assessment of Neurofeedback Training by means of Motor Imagery based-BCI for Cognitive Rehabilitation

The age-related impairment is an increasing problem due to the aging suffered by the population, especially in developed countries. It is usual to use electroencephalogram (EEG)-based Brain Computer Interface (BCI) systems by means of the signal in order to assist and to improve the quality of life of people with disabilities. However, a parallel research line addresses the problem by the use of BCI systems as a way to train cognitive areas to achieve a deceleration of cognitive impairment or even an improvement. In this regard, a neurofeedback training (NFT) tool using motor imagery-based BCI, was developed. Training consists on imagery motor exercises combined with memory and logical relation tasks. In order to assess the effectiveness of the application 40 subjects, older than 59 years old, took part in this study. Our NFT application was tested by 20 subjects and their scores of a neuropsychological test were compared with the remaining 20 subjects who did not perform the NFT. Results show a significant improvement of three cognitive features after performing the NFT: visual perception, expressive speech, and immediate memory. Therefore, evidences show that the performance of a NFT tool based on motor imagery tasks could be a positive activity for slow down the aging effects.

Functional EEG network analysis in schizophrenia: Evidence of larger segregation and deficit of modulation

Objective: Higher mental functions depend on global cerebral functional coordination. Our aim was to study fast modulation of functional networks in schizophrenia that has not been previously assessed. Methods: Graph‐theory was used to analyze the electroencephalographic (EEG) activity during an odd‐ball task in 57 schizophrenia patients (18 first episode patients, FEPs) and 59 healthy controls. Clustering coefficient (CLC), characteristic path length (PL) and small‐worldness (SW) were computed at baseline ([−300 0] ms prior to stimulus delivery) and response ([150 450] ms post‐stimulus) windows. Clinical and cognitive assessments were performed. Results: CLC, PL and SW showed a significant modulation between baseline and response in controls but not in patients. Patients obtained higher CLC and SW at baseline, lower CLC and higher PL at response, and diminished modulation of CLC and SW as compared to controls. In patients, CLC and SW modulation were inversely associated to cognitive performance in executive tasks and directly associated to working memory. Similar patterns were observed in FEPs. CLC and SW during the baseline were inversely associated to their respective modulation magnitudes. Conclusions: Our results are coherent with a hyper‐segregated network at baseline (higher CLC) and a decreased modulation of the functional connectivity during cognition in schizophrenia. HighlightsA deficit in fast modulation of functional network properties during an odd‐ball task was found in schizophrenia patientsThis deficit was also found in first‐episode patientsThere was a significant association between network modulation deficits and cognition in the patients

Exploring non-stationarity patterns in schizophrenia: neural reorganization abnormalities in the alpha band

OBJECTIVE The aim of this paper was to characterize brain non-stationarity during an auditory oddball task in schizophrenia (SCH). The level of non-stationarity was measured in the baseline and response windows of relevant tones in SCH patients and healthy controls. APPROACH Event-related potentials were recorded from 28 SCH patients and 51 controls. Non-stationarity was estimated in the conventional electroencephalography frequency bands by means of Kullback-Leibler divergence (KLD). Relative power (RP) was also computed to assess a possible complementarity with KLD. MAIN RESULTS Results showed a widespread statistically significant increase in the level of non-stationarity from baseline to response in all frequency bands for both groups. Statistically significant differences in non-stationarity were found between SCH patients and controls in beta-2 and in the alpha band. SCH patients showed more non-stationarity in the left parieto-occipital region during the baseline window in the beta-2 band. A leave-one-out cross validation classification study with feature selection based on binary stepwise logistic regression to discriminate between SCH patients and controls provided a positive predictive value of 72.73% and negative predictive value of 78.95%. SIGNIFICANCE KLD can characterize transient neural reorganization during an attentional task in response to novelty and relevance. Our findings suggest anomalous reorganization of neural dynamics in SCH during an oddball task. The abnormal frequency-dependent modulation found in SCH patients during relevant tones is in agreement with the hypothesis of aberrant salience detection in SCH. The increase in non-stationarity in the alpha band during the active task supports the notion that this band is involved in top-down processing. The baseline differences in the beta-2 band suggest that hyperactivation of the default mode network during attention tasks may be related to SCH symptoms. Furthermore, the classification improved when features from both KLD and RP were used, supporting the idea that these measures can be complementary.

Alterations of Effective Connectivity Patterns in Mild Cognitive Impairment: An MEG Study

Neuroimaging techniques have demonstrated over the years their ability to characterize the brain abnormalities associated with different neurodegenerative diseases. Among all these techniques, magnetoencephalography (MEG) stands out by its high temporal resolution and noninvasiveness. The aim of the present study is to explore the coupling patterns of resting-state MEG activity in subjects with mild cognitive impairment (MCI). To achieve this goal, five minutes of spontaneous MEG activity were acquired with a 148-channel whole-head magnetometer from 18 MCI patients and 26 healthy controls. Inter-channel relationships were investigated by means of two complementary coupling measures: coherence and Granger causality. Coherence is a classical method of functional connectivity, while Granger causality quantifies effective (or causal) connectivity. Both measures were calculated in the five conventional frequency bands: delta (δ, 1-4 Hz), theta (θ, 4-8 Hz), alpha (α, 8-13 Hz), beta (β, 13-30 Hz), and gamma (γ, 30-45 Hz). Our results showed that connectivity values were lower for MCI patients than for controls in all frequency bands. However, only Granger causality revealed statistically significant differences between groups (p-values < 0.05, FDR corrected Mann-Whitney U-test), mainly in the beta band. Our results support the role of MCI as a disconnection syndrome, which elicits early alterations in effective connectivity patterns. These findings can be helpful to identify the neural substrates involved in prodromal stages of dementia.

Modulation of brain network parameters associated with subclinical psychotic symptoms

OBJECTIVE Static deficits in small-world properties of brain networks have been described in clinical psychosis, but task-related modulation of network properties has been scarcely studied. Our aim was to assess the modulation of those properties and its association with subclinical psychosis and cognition in the general population. METHOD Closeness centrality and small-worldness were compared between pre-stimulus baseline and response windows of an odd-ball task in 200 healthy individuals. The correlation between modulation of network parameters and clinical (scores in the Community Assessment of Psychological Experiences) and cognitive measures (performance in the dimensions included in the Brief Assessment of Cognition in Schizophrenia battery) was analyzed, as well as between these measures and the corresponding network parameters during baseline and response windows during task performance. RESULTS In the theta band, closeness centrality decreased and small-worldness increased in the response window. Centrality and small-worldness modulation were, respectively, directly and inversely associated with subclinical symptoms. CONCLUSIONS A widespread modulation of network properties in theta band was observed, with a transient increase of small-worldness during the response window, compatible with a transiently more integrated cortical activity associated to cognition. This supports the relevance of electroencephalography to study of normal and altered cognition and its substrates. A relative deficit in the ability to reorganize brain networks may contribute to subclinical psychotic symptoms.