Influential Factors of an Asynchronous BCI for Movement Intention Detection

Abstract

In recent years, asynchronous brain computer interface (BCI) systems have been utilized in many domains such as robot controlling, assistive technology, and rehabilitation. In such BCI systems, movement intention detection algorithms are used to detect movement desires. In recent years, movement-related cortical potential (MRCP), an electroencephalogram (EEG) pattern representing voluntary movement intention, attracts wide attention in movement intention detection. Unfortunately, low MRCP detection accuracy makes the asynchronous BCI system impractical for real usage. In order to develop an effective MRCP detection algorithm, EEG data have to be properly preprocessed. In this work, we investigate the relationship and effects of three factors including frequency bands, spatial filters, and classifiers on MRCP classification performance to determine best settings. In particular, we performed a systematic performance investigation on combinations of five frequency bands, five spatial filters, and six classifiers. The EEG data were acquired from subjects performing series of self-paced ankle dorsiflexions. Analysis of variance (ANOVA) statistical test was performed on F1 scores to investigate effects of these three factors. The results show that frequency bands and spatial filters depend on each other. The combinations directly affect the F1 scores, so they have to be chosen carefully. The results can be used as guidelines for BCI researchers to effectively design a preprocessing method for an advanced asynchronous BCI system, which can assist the stroke rehabilitation.

Figure 1

The experimental protocol shows the task of ballistic ankle dorsiflexion sequence followed by idle states. Idle states are as long as the user desires. These idle states normally take about 3 to 7 seconds.

Figure 2

Multiple comparison results of the three-way interactions among different spatial filter, frequency, and classifier. The best combination is when classifier = SVM, spatial filter = SL, and frequency = [0.01–1] Hz. (a) Estimated marginal means of F1 score with classifier = LDA. (b) Estimated marginal means of F1 score with classifier = SVM. (c) Estimated marginal means of F1 score with classifier = 1-NN-ED. (d) Estimated marginal means of F1 score with classifier = 1-NN-DTW. (e) Estimated marginal means of F1 score with classifier = MF. (f) Estimated marginal means of F1 score with classifier = TM.

Figure 3

Comparison of the estimated marginal means of F1 scores in various classifiers, showing SVM as the dominant classifier.

Figure 4

Comparison of the estimated marginal means of F1 scores in various frequency bands on different spatial filters. SL spatial filter at [0.01-1] Hz frequency clearly provides the best result.

Figure 5

The topoplot demonstrates the F1 score of each channel for every subject when frequency was set with [0.01-1] Hz using SL spatial filter. The channels with higher discriminant power appear brighter. (a) Subject 1. (b) Subject 2. (c) Subject 3. (d) Subject 4. (e) Subject 5. (f) Subject 6. (g) Subject 7. (h) Subject 8. (i) Subject 9.