本学科４年生の徳田竜也くんが、Neuroscience2017（第４０回日本神経科学会大会）にて、Junior Investigator Poster Award(ジュニア研究者ポスター賞）を受賞いたしました。
【学会名】Neuroscience 2017 （第40回日本神経科学大会）
Bottom-up algorithm for removing motion artifacts in fNIRS data of children
Tatsuya Tokuda 1：Syuuhei Yamamoto 1：Minako Uga 1,4：Masako Nagashima 2,：Takahiro Ikeda 2：Yukifumi Monden1,2,3：Ippeita Dan 1
1Applied Cognitive Neuroscience Laboratory, Faculty of Science and Technology, Chuo University
2Department of Pediatrics, Jichi Medical University
3Department of Pediatrics, International University of Health and Welfare
4Center for Development of Advanced Medical Technology, Jichi Medical University
Functional near-infrared spectroscopy (fNIRS) non-invasively measures cortical hemodynamic changes hemoglobin species (oxy-Hb and deoxy-Hb) by using two or more near infrared spectra. fNIRS has gained increasing attention lately in neuropsychiatric and cognitive studies of children. In comparison to other authentic neuroimaging modalities such as fMRI, fNIRS has large advantage: it offers less degree of head fixation is less prone to motion artifacts. However, although these merits are generally applicable to adult subjects, it is still difficult to obtain child data without motion artifacts.
The complications of the motion artifacts have been investigated for ages. Nevertheless, it is not easy at all to detect and remove the artifacts by utilizing automatic methods especially for children data. Consequently, such motions artifacts have been detected by raters visually. This further evokes an additional complication: detection and removal of the artifacts is time-consuming and raters’ evaluations may not always match to each other. In order to avoid the complications, we have developed a bottom-up algorithm so as to remove the artifacts based on numerical assessment of child fNIRS data in a similar manner as experienced raters.
We set three criteria for the algorithm. They included positive correlation between oxy-Hb and deoxy-Hb, transition detected per 1 second, and sequential artifacts disrupting correct measurement of the brain activation. By utilizing this algorithm, we reanalyzed the previous research data monitoring the oxy-Hb signal changes of thirty children performing a go/no-go task.
We evaluated activation of right prefrontal cortex during go/no-go task. There was a significant brain activation (t(30)=3.52, p<0.05), which is a comparable degree of activation when artifact removal was performed by experienced raters. Inter-rater agreement between manual detection and the current algorithm was also considerably large (kappa coefficient > 0.7).
We propose that the current algorism can be used for validating previous fNIRS child data in an objective way. However, for its application to novel fNIRS child data, we may need further optimization for parameters used in the algorithm.