Extracting task-related activation components from optical topography measurement using independent components analysis

Abstract

Optical topography (OT) signals measured during an experiment that used activation tasks for certain brain functions contain neuronal-activation induced blood oxygenation changes and also physiological changes. We used independent component analysis to separate the signals and extracted components related to brain activation without using any hemodynamic models. The analysis procedure had three stages: first, OT signals were separated into independent components (ICs) by using a time-delayed decorrelation algorithm; second, task-related ICs (TR-ICs) were selected from the separated ICs based on their mean intertrial cross-correlations; and third, the TR-ICs were categorized by k-means clustering into TR activation-related ICs (TR-AICs) and TR noise ICs (TR-NICs). We applied this analysis procedure to the OT signals obtained from experiments using one-handed finger-tapping tasks. In the averaged waveform of the TR-AICs, a small overshoot can be seen for a few seconds after the onset of each task and a few seconds after it ends, and the averaged waveforms of the TR-NICs have an N-shaped pattern.