A Generic Transferable EEG Decoder for Online Detection of Error Potential in Target Selection

Abstract

Reliable detection of error from electroencephalography (EEG) signals as feedback while performing a discrete target selection task across sessions and subjects has a huge scope in real-time rehabilitative application of Brain-computer Interfacing (BCI). Error Related Potentials (ErrP) are EEG signals which occur when the participant observes an erroneous feedback from the system. ErrP holds significance in such closed-loop system, as BCI is prone to error and we need an effective method of systematic error detection as feedback for correction. In this paper, we have proposed a novel scheme for online detection of error feedback directly from the EEG signal in a transferable environment (i.e., across sessions and across subjects). For this purpose, we have used a P300-speller dataset available on a BCI competition website. The task involves the subject to select a letter of a word which is followed by a feedback period. The feedback period displays the letter selected and, if the selection is wrong, the subject perceives it by the generation of ErrP signal. Our proposed system is designed to detect ErrP present in the EEG from new independent datasets, not involved in its training. Thus, the decoder is trained using EEG features of 16 subjects for single-trial classification and tested on 10 independent subjects. The decoder designed for this task is an ensemble of linear discriminant analysis, quadratic discriminant analysis, and logistic regression classifier. The performance of the decoder is evaluated using accuracy, F1-score, and Area Under the Curve metric and the results obtained is 73.97, 83.53, and 73.18%, respectively.

Detection of Error-Related Potentials in Stroke Patients from EEG Using an Artificial Neural Network

“…The classification accuracies for the different calibration schemes were significantly higher than chance level (calculated with a significance level of 5% [ 53 ]). The accuracies above chance level for the between-day and across-participant calibration schemes are in agreement with several other studies that have reported that ErrPs can be detected using these approaches [ 17 , 22 , 30 , 31 , 32 , 33 , 35 , 36 , 54 ]. The classification accuracies obtained for LDA in between-day and across-participant calibration are similar to what has been reported previously, but lower for ANN.…”

“…Moreover, it may not be needed to use as many ErrP and NonErrP epochs as were utilized in this study; it has been reported that a steep increase in classification accuracies was observed after ~50–100 ErrPs were used for calibration [ 33 , 36 , 40 ]. The classification accuracies based on within-day calibration are in agreement with findings in several other studies which have reported a detection performance in the range of roughly 70–90% [ 12 , 13 , 15 , 16 , 17 , 22 , 28 , 30 , 32 , 33 , 56 ]; it should be noted that different classification metrics have been reported. This was the case when using either features or the entire epoch as input for the ANN classifier.…”

“…It has been shown in several studies that ErrPs can be decoded (see [ 9 , 10 ] for a review about decoding ErrPs) and how the performance of the decoder can be optimized using different types of features [ 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 ], classifiers [ 11 , 12 , 17 , 18 , 22 , 23 ], time windows [ 24 , 25 , 26 ], and channels [ 17 , 25 , 27 , 28 ]. Moreover, it has been shown that generalized ErrP detectors can be transferred across different tasks and types of ErrPs such as observation and interaction ErrPs, across participants, and time which eliminates the need for calibration (see, e.g., [ 11 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 ]). This transfer may be successful due to the stability of the ErrP over time as good test/retest reliability has been reported for evoking the ErrP [ 37 , 38 , 39 ].…”

The Effect of Static and Dynamic Visual Stimulations on Error Detection Based on Error-Evoked Brain Responses

“…Moreover, the error recognition based on ErrPs of the late time window was better in dynamic stimulation mode. Static and dynamic visual stimuli in ErrPs-based BCI systems have been widely used, e.g., P300-BCIs with ErrPs correction to improve accuracies [9,23], human-robot co-adaptation using ErrPs [16], and implicit cursor control based on ErrPs [15]. This paper studied the difference between the two stimulation modes and the selection of time window, to improve the BCIs' error recognition accuracy.…”

“…Observation ErrPs are the key point of error recognition in BCI control. They are generated when a person observes an error in the task performed by a second subject (an agent or other human) [8,9]. The observation ErrPs are usually evoked by visual stimulations, which can be roughly divided into static and dynamic ones.…”