In this study we used NIRS, which can cover frontal and superior temporal surface areas, to determine the patterns of cortical activation of healthy subjects performing the Shiritori task with overt speech. We found significant hemodynamic responses bilaterally in 42 (80.8Â %) of the 52 channels examined. The wide distribution of significantly activated channels across the frontal and superior temporal areas found in this study is in line with the results of previous studies, which determined cortical activation patterns of healthy adults during the Shiritori task without overt speech by using MEG and fMRI, respectively [14, 27].
Using cluster analysis of the changes in [oxy-Hb] per 0.1 s, we divided the 52 channels into 4 clusters, each comprised of channels showing similar changes in [oxy-Hb] and distributional region symmetrically across hemispheres (Fig. 3). It is speculated that the representative cortical regions in each cluster could be the inferior frontal gyrus and superior temporal gyrus in Cluster I, the middle frontal gyrus in Cluster II, the superior frontal gyrus in Cluster III, and the supplementary motor area in Cluster IV according to [19, 26], the former showed the probabilistic correspondence relation between anatomical cortical surface areas and the international 10–20 system for EEG, and the latter used the same NIRS machine (ETG-4000 with probe of 52 channels) and method for setting the probe as our study, and demonstrates a figure to clarify correspondence between the channels and the anatomical areas.
Next, we determined the characteristics of the changes in [oxy-Hb] (waveforms) throughout the task by analyzing an averaged waveform (representative waveform) across the channels included in each cluster. The differences among the representative waveforms were determined from the three successive task periods set: the steep slope of the initial rise (Fig. 4a), the duration of this rise (Fig. 4b), and the subsequent gentle slope up until the end of the task (Fig. 4c). The most highly-activated representative waveform (out of the four waveforms) was that of Cluster I, which was independent from the other three clusters according to the dendrogram derived from cluster analysis (Fig. 3). Its initial rise was steeper and lasted longer than that of the others during the first task period, and almost plateaued during the last period. Since Cluster I would mainly include the inferior frontal and post-superior temporal gyri encompassing Broca’ s and Wernicke’ s areas respectively which constitute language-related networks in the brain [27], the variations seen between the cluster waveforms might reflect strong functional connectivity between the channels in the same cluster [11]. However it is currently not known whether differences between clusters depend on different cognitive components involved in tasks, or whether they reflect specific traits of the neurons comprising the different cortical regions. Applying the present analysis to some cognitive tasks and resting states may help to clarify the situation.
With regard to laterality, it was confirmed that the channels showing significant cortical activation during the Shiritori task were distributed bilaterally, and that this bilateral distribution was found in 60Â % of subjects. This result is opposed to the left-sided dominance during the Shiritori tasks reported in MEG and fMRI studies [14, 27]. Although further research is required to determine the cortical laterality during Shiritori, it is important to point out that the duration of the Shiritori task in the present study was longer than that of the other studies in this field. This would increase the cognitive loads of the task, leading to activation of more cortical areas, including regions in the right hemisphere.
Next, with regard to the wide frontal and superior temporal activation found in this study, we attempt to speculate the reason for this activation associated with Shiritori task performance. As with conventional WGL, executive functions such as word retrieval, storing previously generated word in working memory, maintaining cognitive effort, inhibiting inappropriate responses and sufficient processing speed aiming at better performance would be demanded in order to perform the Shiritori task. These functions are primarily subserved by the frontal lobe [1, 10, 20]. The Shiritori task, however, appears to have greater executive demands than the conventional WGL. This is due to the fact that the cognitive loads of set-shifting and inhibition may be higher because of the initial syllable changes for each word and avoiding saying a noun ending with /n/, respectively. In addition, playing the Shiritori task alone would demand self-monitoring and planning because the subject must think back to the last noun said as well as think ahead when choosing the next noun. Moreover, the Shiritori task is thought to place greater cognitive loads on internal phonological processing in order to repeat and listen to the last syllable of the previous noun in mind until generating the next noun, and the superior temporal activity is considered to be responsible for this function [20, 27]. Thus, Shiritori might require further activation of the frontal and superior temporal areas than conventional WGL.
This study has several limitations. First, the cortical region/channel association in the present study differs depending on subject’s head shape as the channels depart from Fpz (the international 10/20 system for EEG) at which we set the most ventral center photodiode. Thus, especially in superior temporal areas, a perfect cortical region/channel association would not be ensured. In the future study for the purposes of accurate identification of localization, it is required to investigate the cortical region/channel association in each subject or to register the channel positions onto the standard template brain when group statistics is performed [25, 28]. Second, changes in [oxy-Hb] are expressed as relative values compared to a baseline task, and it’s assumed that optical path length is constant across subjects or measurement points. Thus, evaluations using this type of NIRS (continuous wave-based) machine which is unable to measure absolute values should be considered estimates. Nevertheless, it was argued that this kind of NIRS machine is able to measure quite precise changes in [oxy-Hb] as well as [deoxy-Hb] since the path length did not change more than 10 % among channels [6, 7]. Third, it was recently reported that NIRS signals would include considerable components from skin blood flow, a fact that is attracting the interest of researchers using NIRS [16, 23]. Although this has been confirmed only at the frontal pole (forehead) thus far, clarifying the hemodynamic responses across the whole scalp and seeking methods to minimize the influence of skin blood flow is urgently needed to justify the results determined by NIRS and to ensure NIRS studies continue. However, since a strong correlation between NIRS and fMRI signals has also been reported [3, 21], the activity seen in the present study could also include considerable signals from neural activity. Therefore, at the very least, it appears we succeeded in outlining the cortical activity to perform the Shiritori task. The development of techniques that can separate systemic and brain activity components from NIRS signals is awaited.