Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Robustness of Asynchronous EEG-Based Movement Prediction using Classifier Ensembles

Niklas Kueper, Kartik Chari, Elsa Andrea Kirchner

TL;DR

It is demonstrated that classifier ensembles and appropriate postprocessing methods effectively enhance the asynchronous detection of movement intentions from EEG signals and yields greater improvements in online classification than in offline classification, and reduces false detections.

Abstract

Objective: Stroke is one of the leading causes of disabilities. One promising approach is to extend the rehabilitation with self-initiated robot-assisted movement therapy. To enable this, it is required to detect the patient's intention to move to trigger the assistance of a robotic device. This intention to move can be detected from human surface electroencephalography (EEG) signals; however, it is particularly challenging to decode when classifications are performed online and asynchronously. In this work, the effectiveness of classifier ensembles and a sliding-window postprocessing technique was investigated to enhance the robustness of such asynchronous classification. Approach: To investigate the effectiveness of classifier ensembles and a sliding-window postprocessing, two EEG datasets with 14 healthy subjects who performed self-initiated arm movements were analyzed. Offline and pseudo-online evaluations were conducted to compare ensemble combinations of the support vector machine (SVM), multilayer perceptron (MLP), and EEGNet classification models. Results: The results of the pseudo-online evaluation show that the two model ensembles significantly outperformed the best single model for the optimal number of postprocessing windows. In particular, for single models, an increased number of postprocessing windows significantly improved classification performances. Interestingly, we found no significant improvements between performances of the best single model and classifier ensembles in the offline evaluation. Significance: We demonstrated that classifier ensembles and appropriate postprocessing methods effectively enhance the asynchronous detection of movement intentions from EEG signals. In particular, the classifier ensemble approach yields greater improvements in online classification than in offline classification, and reduces false detections, i.e., early false positives.

Enhancing Robustness of Asynchronous EEG-Based Movement Prediction using Classifier Ensembles

TL;DR

It is demonstrated that classifier ensembles and appropriate postprocessing methods effectively enhance the asynchronous detection of movement intentions from EEG signals and yields greater improvements in online classification than in offline classification, and reduces false detections.

Abstract

Objective: Stroke is one of the leading causes of disabilities. One promising approach is to extend the rehabilitation with self-initiated robot-assisted movement therapy. To enable this, it is required to detect the patient's intention to move to trigger the assistance of a robotic device. This intention to move can be detected from human surface electroencephalography (EEG) signals; however, it is particularly challenging to decode when classifications are performed online and asynchronously. In this work, the effectiveness of classifier ensembles and a sliding-window postprocessing technique was investigated to enhance the robustness of such asynchronous classification. Approach: To investigate the effectiveness of classifier ensembles and a sliding-window postprocessing, two EEG datasets with 14 healthy subjects who performed self-initiated arm movements were analyzed. Offline and pseudo-online evaluations were conducted to compare ensemble combinations of the support vector machine (SVM), multilayer perceptron (MLP), and EEGNet classification models. Results: The results of the pseudo-online evaluation show that the two model ensembles significantly outperformed the best single model for the optimal number of postprocessing windows. In particular, for single models, an increased number of postprocessing windows significantly improved classification performances. Interestingly, we found no significant improvements between performances of the best single model and classifier ensembles in the offline evaluation. Significance: We demonstrated that classifier ensembles and appropriate postprocessing methods effectively enhance the asynchronous detection of movement intentions from EEG signals. In particular, the classifier ensemble approach yields greater improvements in online classification than in offline classification, and reduces false detections, i.e., early false positives.
Paper Structure (29 sections, 5 figures)

This paper contains 29 sections, 5 figures.

Table of Contents

  1. Introduction
  2. Methods
  3. Results
  4. Discussion
  5. Conclusion and Outlook

Figures (5)

  • Figure 1: Pseudo-online evaluation scheme: It was distinguished between A) correct trials (top), B) early detection trials (middle), and C) no detection trials (bottom). In red, the deadtime period (-5s to -4s) is shown, where possible predictions are ignored by the decision logic, and in orange, the resting period is shown (-4s to -0.75s), where no movement onsets should be predicted. Finally, in green, the target period (-0.75s to 0.15s) is shown, during which a detection of movement onset is expected.
  • Figure 2: Offline performance comparison of all classification methods: Besides the dummy classifier (turquoise), the results and statistical comparison of single models (in blue), two models (in red), as well as the results of three models (in gold) are illustrated. In green, the comparisons of the best-performing models (E-SE-SME and E-ME-SME) are shown. The significance levels are shown in the top-left corner of the figure. Only significant results are indicated; if not indicated, the comparison was not significant.
  • Figure 3: Pseudo-online performance comparison for different numbers of postprocessing windows: Besides the dummy classifier, the results of all single models, two model ensembles, and three model ensembles are shown. The coloring indicates the number of postprocessing windows used. The results of the statistical analysis of the model-wise number of postprocessing windows are indicated with black U-shaped brackets. The significance levels are located on the top left of the figure. Only significant results are indicated, if not indicated, the comparison was not significant.
  • Figure 4: Pseudo-online performances for the selected number of postprocessing windows: The results and the comparisons of the statistical analysis of single models (in blue), two models (in red), as well as the results of three models (in gold) are illustrated. In green, the comparisons of the best-performing models (comparisons E3-SE2-SME2 and E3-ME2-SME2) are shown. The significance levels are located on the top left of the figure. Only significant results are indicated, if not indicated, the comparison was not significant.
  • Figure 5: Pseudo-online early detections for all classification methods: The results and the comparisons of the statistical analysis of single models (in blue), two models (in red), as well as the results of three models (in gold) are illustrated. In green, the comparisons of the best-performing models (comparisons E3-SE2-SME2 and E3-ME2-SME2) are shown. The significance levels are located on the top left of the figure. Only significant results are indicated, if not indicated, the comparison was not significant.