Combining Statistical Analysis and Machine Learning for EEG Scalp Topograms Classification.

Alexander Kuc,Sergey Korchagin,Alexander E Hramov,Natalia Shusharina,Vladimir A Maksimenko

doi:10.3389/fnsys.2021.716897

Abstract

Incorporating brain-computer interfaces (BCIs) into daily life requires reducing the reliance of decoding algorithms on the calibration or enabling calibration with the minimal burden on the user. A potential solution could be a pre-trained decoder demonstrating a reasonable accuracy on the naive operators. Addressing this issue, we considered ambiguous stimuli classification tasks and trained an artificial neural network to classify brain responses to the stimuli of low and high ambiguity. We built a pre-trained classifier utilizing time-frequency features corresponding to the fundamental neurophysiological processes shared between subjects. To extract these features, we statistically contrasted electroencephalographic (EEG) spectral power between the classes in the representative group of subjects. As a result, the pre-trained classifier achieved 74% accuracy on the data of newly recruited subjects. Analysis of the literature suggested that a pre-trained classifier could help naive users to start using BCI bypassing training and further increased accuracy during the feedback session. Thus, our results contribute to using BCI during paralysis or limb amputation when there is no explicit user-generated kinematic output to properly train a decoder. In machine learning, our approach may facilitate the development of transfer learning (TL) methods for addressing the cross-subject problem. It allows extracting the interpretable feature subspace from the source data (the representative group of subjects) related to the target data (a naive user), preventing the negative transfer in the cross-subject tasks.

Highlights

Machine learning (ML) has become a new standard in brain signals analysis (Hramov et al, 2021)
Similar to our recent study (Kuc et al, 2021), we reduced the number of experimental conditions considering a = 30%, 50% as the low ambiguity (LA) stimuli and a = 80%, 90% as high ambiguity (HA) stimuli
We found that t2 (β = −0.462, p = 0.037) (Figure 6B) and t3 (β = −0.524, p = 0.016) (Figure 6C) significantly predicted convolutional neural network (CNN) accuracy

Summary

Introduction

Machine learning (ML) has become a new standard in brain signals analysis (Hramov et al, 2021). ML is a model-free approach that successfully operates with data without prior knowledge of its origin. When the mathematical model of the time series is unknown, ML can build this model based on training data. Being trained on a representative amount of data, ML enables the classification, detection, and prediction of the newly acquired data. These aspects of ML meet the fundamental requirements for brain-computer interfaces (BCIs). BCI often utilizes brain activity biomarkers that barely have an exact mathematical model.

Methods

Results

Discussion

Conclusion