Independent component analysis on saccade-related EEG - Comparing the ensemble averaging method and independent component analysis

Arao Funase, Allan K. Barros, Sigeru Okuma, Andrzej Cichocki, Tohru Yagi

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)


We have researched saccade-related EEG to investigate the brain function in saccadic eye movement. In previous studies, we analyzed saccade-related EEG using an ensemble averaging method. However, the ensemble averaging method has some problems. First, EEG signals must be recorded repetitively under the same conditions. In addition, the ensemble averaging method can possibly extinguish EEG signals which have unstable latency between a trigger and the peak point of an EEG signal. Therefore, we applied saccade-related EEG to an independent component analysis (ICA). ICA can extract a signal from mixed signals through the assumption that source signals are mutually independent and recorded signals are mixed lineally from source signals. ICA can analyze non-processed signals. Therefore, ICA overcomes the problems of the ensemble averaging method. In this field, many algorithms were proposed. "Fast ICA with Reference signal (FICAR)" was selected, however, because this algorithm can extract a desired signal close to a reference signal. In this experiment, there were two tasks. The first (eye movement task) was that subjects move their eyes toward a visual target (LED) located on the right and left sides. The second (non-eye movement task) was that subjects do not move their eyes toward the visual target when it is turned on. EEG signals were recorded through electrodes arranged according to the international 10-20 method and were analyzed with the FICAR. Reference signals had one peak and each reference signal had a different peak time. EEG signals were analyzed with each reference signal. As a result, in the case of the eye movement task, the FICAR extracted independent components (ICs) which peaked just before saccade. Few ICs were extracted under the non-eye movement task. The features for peak timing, peak amplitude, and dependence on existence of saccade are the same as the features for the ensemble averaging results. The ICs extraction rate is 71. 6% under the eye movement task and 25. 2% under the non-eye movement task. The ICs were extracted by reference signals which peak just before saccade. Therefore the ICs are related to saccade-related EEG.

Original languageEnglish
Pages (from-to)342-351
Number of pages10
JournalJapanese Journal of Medical Electronics and Biological Engineering
Issue number4
Publication statusPublished - 2003
Externally publishedYes


Dive into the research topics of 'Independent component analysis on saccade-related EEG - Comparing the ensemble averaging method and independent component analysis'. Together they form a unique fingerprint.

Cite this