Electroencephalogram-based decoding cognitive states using convolutional neural network and likelihood ratio based score fusion

Abstract

Electroencephalogram (EEG)-based decoding human brain activity is challenging, owing to the low spatial resolution of EEG. However, EEG is an important technique, especially for brain–computer interface applications. In this study, a novel algorithm is proposed to decode brain activity associated with different types of images. In this hybrid algorithm, convolutional neural network is modified for the extraction of features, a t-test is used for the selection of significant features and likelihood ratio-based score fusion is used for the prediction of brain activity. The proposed algorithm takes input data from multichannel EEG time-series, which is also known as multivariate pattern analysis. Comprehensive analysis was conducted using data from 30 participants. The results from the proposed method are compared with current recognized feature extraction and classification/prediction techniques. The wavelet transform-support vector machine method is the most popular currently used feature extraction and prediction method. This method showed an accuracy of 65.7%. However, the proposed method predicts the novel data with improved accuracy of 79.9%. In conclusion, the proposed algorithm outperformed the current feature extraction and prediction method.

A framework for closed-loop neurofeedback for real-time EEG decoding

“…More automated approaches include e.g. iterative variable selection based on spectral and ERP features for each participant [113] or neural networks for extraction of EEG features with subsequent classification [114,115]. Generally, these methods are computationally expensive, but hold great potential for optimization of feature extraction in real-time neurofeedback systems.…”

Single-Trial Decoding of Scalp EEG Under Natural Conditions

“…In case of decoding of scalp EEG, the research area is still progressing, and relatively few studies document detection of brain states in regards to semantic categories (often discrimination between two high-level categories) [Simanova et al, 2010, Murphy et al, 2011, Wang et al, 2012, Taghizadeh-Sarabi et al, 2014, Stewart et al, 2014, Kaneshiro et al, 2015, Zafar et al, 2017. EEG-based decoding of human brain activity has significant potential due to excellent time resolution and the possibility of real-life acquisition, however, the signal is extremely diverse, subject-specific, sensitive to disturbances, and has a low signal-to-noise ratio, hence, posing a major challenge for both signal processing and machine learning [Nicolas-Alonso and Gomez-Gil, 2012].…”

“…Due to before-mentioned challenges, previous studies have been performed in controlled laboratory settings with high-grade EEG acquisition equipment [Simanova et al, 2010, Murphy et al, 2011, Wang et al, 2012, Stewart et al, 2014, Kaneshiro et al, 2015, Zafar et al, 2017. Visual stimuli paradigms can often not be described as naturalistic, due to 1) repeated presentation of identical experimental trials, and 2) iconic views of objects and lack of complexity of semantic context [Simanova et al, 2010, Murphy et al, 2011, Wang et al, 2012, Taghizadeh-Sarabi et al, 2014, Stewart et al, 2014, Kaneshiro et al, 2015.…”

“…Visual stimuli paradigms can often not be described as naturalistic, due to 1) repeated presentation of identical experimental trials, and 2) iconic views of objects and lack of complexity of semantic context [Simanova et al, 2010, Murphy et al, 2011, Wang et al, 2012, Taghizadeh-Sarabi et al, 2014, Stewart et al, 2014, Kaneshiro et al, 2015. Generalizability of decoding classifier models to novel participants is rare, due to subjectspecific modelling approaches [Simanova et al, 2010, Wang et al, 2012, Stewart et al, 2014, Taghizadeh-Sarabi et al, 2014, Kaneshiro et al, 2015, Zafar et al, 2017. Moreover, a number of participants are occasionally excluded from analysis due to artifacts and low classification accuracy [Taghizadeh-Sarabi et al, 2014, Zafar et al, 2017.…”

“…Generalizability of decoding classifier models to novel participants is rare, due to subjectspecific modelling approaches [Simanova et al, 2010, Wang et al, 2012, Stewart et al, 2014, Taghizadeh-Sarabi et al, 2014, Kaneshiro et al, 2015, Zafar et al, 2017. Moreover, a number of participants are occasionally excluded from analysis due to artifacts and low classification accuracy [Taghizadeh-Sarabi et al, 2014, Zafar et al, 2017.…”