Comparing different pre-processing routines for infant fNIRS data

Abstract

Highlights Five pre-processing pipelines were compared on synthetic and real infant NIRS data. Strict inclusion criteria limit the number of trials, but recover HRF accurately. Artifact correction retains larger amounts of data, but may lower HRF amplitude. No difference between applying pre-processing to optical density or concentration data.

Oxygenated hemoglobin signal provides greater predictive performance of experimental condition than de-oxygenated

“…The copyright holder for this preprint this version posted November 20, 2021. ; https://doi.org/10. 1101/2021 The oxygenated hemoglobin signal provided the best decoding performance (Table 1), outperforming both the de-oxyhemoglobin and total hemoglobin signals. The mean ROC-AUC was greatest for the HbO data in all four datasets and was larger than the HbR data (M=7.7, SD=2.3), t(3)=5.6, p=.011.…”

“…The copyright holder for this preprint this version posted November 20, 2021. ; https://doi.org/10. 1101/2021 In this passive experimental task, six participants were asked to focus their vision on a computer monitor positioned one meter in front of them. The screen displayed either an oscillating black-and-white checkerboard pattern (condition) or a static grey cross (control) in the center of a black screen.…”

“…The copyright holder for this preprint this version posted November 20, 2021. ; https://doi.org/10. 1101/2021…”

“…Conversely, the de-oxygenated signal is less affected by extracranial physiological noise (Dravida et al, 2017;Heinzel et al, 2013;Kirilina et al, 2012;Obrig et al, 2000) and provides greater spatial selectivity (Cannestra et al, 2003;Plichta et al, 2007). The cortical contribution within total hemoglobin signals, however, may be significantly higher the mean or slope of the oxygenated signal is a more robust classifier than de-oxygenated or total hemoglobin for task-evoked neural activity (Gemignani and Gervain, 2021;Mihara et al, 2012;Naseer and Hong, 2013;Naseer et al, 2014). However, the manual selection of features based on the time course of the fNIRS signal may bias the results towards one chromophore, as the time courses of these signals differ in amplitude and latency (Scholkmann et al, 2014;Tam and Zouridakis, 2014).…”

The neural correlates of inhibitory control in 10-month-old infants: a functional near-infrared spectroscopy study

“…This would not only result in dissipation of the necessary response prepotency built up during the task, it would also present more opportunities for distraction and boredom resulting in additional data attrition. Indeed, a considerable strength of this study is the high number of blocks completed by infants (6 -14 blocks, equivalent to 36 -84 individual trials), which also meant that we could apply stringent block inclusion criteria to ensure that data included in analyses had a high signal-to-noise ratio and was a reliable index of neural activation (Gemignani & Gervain, 2021).…”