Connectionists: Spatio-temporal beamforming for fast time-locked EEG component detection

[2016 IEEE International Workshop on Machine Learning fo Signal Processing]

We developed a new algorithm for time-locked EEG component detection and 
applied to it even to single-trial P300, N400 and even to synchronous, 
amplitude-based SSVEP detection. The algorithm is based on a 
spatiotemporal extension of the original spatial linearly-contrained 
minimum-variance (LCMV) beamformer.

All papers can be retrieved from 
https://kuleuven.box.com/s/1ff5qozg68jok7dddxror0ev45fr6ola

N400: The N400 event-related potential (ERP) is mainly employed to study 
language processing, and is traditionally extracted by averaging over 
multiple trials. Being able to analyse individual trials (epochs) has 
clear statistical advantages, and allows for more detailed hypotheses. 
The study by Van Vliet et al. 2016 has shown that the spatiotemporal 
beamformer can be successfully employed for the analysis of each 
individual spatiotemporal N400 trial, and that with this technique more 
detailed analysis can be achieved.

- M. van Vliet, N. Chumerin, S. De Deyne, J. R. Wiersema, W. Fias, G. 
Storms, and M. M. Van Hulle. Single-trial erp component analysis using a 
spatiotemporal lcmv beamformer. IEEE Transactions on Biomedical 
Engineering, 63(1):5566, Jan 2016.

- Code available at https://github.com/wmvanvliet/ERP-beamformer

P300: The P300 ERP is often used for building a brain-computer interface 
(BCI), that allows (re-)establishing a communication channel for people 
suffering from motor-and communication disabilities (ALS, LIS, stroke, 
etc.). Limiting the number of trials (i.e., how many times all targets 
are flashed) increases the communication speeed of the BCI. A study by 
Wittevrongel and Van Hulle 2016, has employed the spatiotemporal 
beamformer towards this purpose, and has shown state-of-the-art 
performance in P300 detection, with the advantage of faster training of 
the classifier.

- B. Wittevrongel and M. M. Van Hulle. Faster p300 classifier training 
using spatiotemporal beamforming. International Journal of Neural 
Systems, 26(03):1650014, May 2016.

SSVEP: The steady-state visual evoked potential (SSVEP) is another 
paradigm used for building BCIs. The goal is to detect which frequency 
is present in the (mainly occipital) EEG. While traditionally analyses 
investigate the frequency domain, a study from Wittevrongel and Van 
Hulle 2016 has employed a combination of a time domain techinque and the 
spatiotemporal beamformer. The study has shown an improvement over 
existing state-of-the-art algorithms, especially for short signals (~ 1 
second).

- B. Wittevrongel and M. M. Van Hulle. Frequency- and phase encoded 
ssvep using spatiotemporal beamforming. PLOS ONE, 11(8):e0159988, Aug 
2016. DOI:10.1371/journal.pone.0159988

- B. Wittevrongel and M. M. Van Hulle. Hierarchical online ssvep 
spelling achieved with spatiotemporal beamforming. 2016 IEEE Statistical 
Signal Processing Workshop (SSP), Jun 2016.