The proposed ICA method: The original hemodynamic response is reconstructed using independent components (ICs).
ICA concept: (a) examples of 5 hemodynamic signals (i.e., a hemodynamic response due to neuronal activation, low frequency oscillation #1, respiratory, motion artifact, and low frequency oscillation #2), (b) the power-spectral densities of (a), (c) three signals used in the ICA of http://www.cis.hut.fi/projects/ica/fastica, (d) three identified ICs from (c), and (e) the power-spectral densities of (d) that can be compared with (b), where IC 1, IC 2, and IC 3 are claimed to be the estimated hemodynamic responses , , and , respectively.
Experimental design: (a) optodes configuration (Fp1 and Fp2 in the International 10-20 System are used as reference points), (b) experimental procedure for arithmetic task.
A typical example of raw HbO, HbR, and HbT data (Subject 1, Ch. 12).
PSDs of typical 4 ICs chosen from Figure 5 (Subject 8).
HbO comparison (Subject 8): the low-pass-filtered (LPF) signal h LPF(k) at Ch. 15 (dotted), the reconstructed hemodynamic response h ICA(k) using Eq. (7) (solid), and the modeled hemodynamic signal h M (k) (the t-value between h LPF(k) and h M (k) is 37 and that between h ICA(k) and h M (k) is 51).
Comparison of h ICA(k) using (7) (solid) and h IĆA(k) using only four ICs with high t-values (dotted): The drifting behavior of h LPF(k) during 360-450 s has been removed.
Comparison of the highest t-values of h LPF(k) and h ICA(k), respectively, for all 8 subjects and their grand averages.
Article metrics loading...
Full text loading...