A Discriminative Multi-Output Gaussian Processes Scheme for Brain Electrical Activity Analysis

Abstract

The study of brain electrical activity (BEA) from different cognitive conditions has attracted a lot of interest in the last decade due to the high number of possible applications that could be generated from it. In this work, a discriminative framework for BEA via electroencephalography (EEG) is proposed based on multi-output Gaussian Processes (MOGPs) with a specialized spectral kernel. First, a signal segmentation stage is executed, and the channels from the EEG are used as the model outputs. Then, a novel covariance function within the MOGP known as the multispectral mixture kernel (MOSM) allows us to find and quantify the relationships between different channels. Several MOGPs are trained from different conditions grouped in bi-class problems, and the discrimination is performed based on the likelihood score of the test signals against all the models. Finally, the mean likelihood is computed to predict the correspondence of new inputs with each class’s existing models. Results show that this framework allows us to model the EEG signals adequately using generative models and allows analyzing the relationships between channels of the EEG for a particular condition. At the same time, the set of trained MOGPs is well suited to discriminate new input data.

Modeling Neonatal EEG Using Multi-Output Gaussian Processes

“…To the best of our knowledge, the work presented here is the first attempt at modelling and classifying neonatal EEG with MOGPs and, therefore, the first validation of the MOSM kernel on such setting. In the same fashion as those of [35], our results are auspicious and promising as a jumping off point for reliable MOGP-based seizure detection mechanisms deployed at real-world clinical environments. Although in this work we only considered seizure/nonseizure classification, the proposed method can be easily extended to seizure type classification with suitable datasets by means of multilabel classifiers.…”

“…Methods in the classifier include SVM, NN, RNN, which are fed with 98 features including amplitude, complexity and oscillatory behaviour of EEG. [35] MOGP-based EEG binary classification (not neonatal). The MOSM kernel is used and the discrimination is performed based on the likelihood score of the test signals only, no other feature-based classifiers are considered.…”

“…Also, [39] proposed a hierarchical GP for spatio-temporal representation of EEG and source localization. Recently, [35] proposed a MOSM-based discriminative approach to binary classification of EEG with emotional content and hand movement prediction.…”

Multiple Kernel Stein Spatial Patterns for the Multiclass Discrimination of Motor Imagery Tasks

“…In the same approach, the Filter-Bank CSP (FBCSP) exploited the potential correlation between CSP characteristics extracted from different bands to improve the signal discrimination [8]. In general, filter-banked feature extraction approaches outperform conventional multichannel time-series representations, including traditional CSP, in supervised learning schemes [12]. Nonetheless, spectral variants of CSP hardly decode MI tasks that activate spatially close regions, no matter the frequency differences [13].…”