Comparing fNIRS signal qualities between approaches with and without short channels

Xin Zhou; Gabriel Sobczak; Colette M McKay; Ruth Y Litovsky

doi:10.1371/journal.pone.0244186

. 2020 Dec 23;15(12):e0244186. doi: 10.1371/journal.pone.0244186

Comparing fNIRS signal qualities between approaches with and without short channels

Xin Zhou ^1,^*, Gabriel Sobczak ², Colette M McKay ^3,⁴, Ruth Y Litovsky ^1,^5,⁶

Editor: Manabu Sakakibara⁷

PMCID: PMC7757903 PMID: 33362260

Abstract

Functional near-infrared spectroscopy (fNIRS) is a non-invasive technique used to measure changes in oxygenated (HbO) and deoxygenated (HbR) hemoglobin, related to neuronal activity. fNIRS signals are contaminated by the systemic responses in the extracerebral tissue (superficial layer) of the head, as fNIRS uses a back-reflection measurement. Using shorter channels that are only sensitive to responses in the extracerebral tissue but not in the deeper layers where target neuronal activity occurs has been a ‘gold standard’ to reduce the systemic responses in the fNIRS data from adults. When shorter channels are not available or feasible for implementation, an alternative, i.e., anti-correlation (Anti-Corr) method has been adopted. To date, there has not been a study that directly assesses the outcomes from the two approaches. In this study, we compared the Anti-Corr method with the ‘gold standard’ in reducing systemic responses to improve fNIRS neural signal qualities. We used eight short channels (8-mm) in a group of adults, and conducted a principal component analysis (PCA) to extract two components that contributed the most to responses in the 8 short channels, which were assumed to contain the global components in the extracerebral tissue. We then used a general linear model (GLM), with and without including event-related regressors, to regress out the 2 principal components from regular fNIRS channels (30 mm), i.e., two GLM-PCA methods. Our results found that, the two GLM-PCA methods showed similar performance, both GLM-PCA methods and the Anti-Corr method improved fNIRS signal qualities, and the two GLM-PCA methods had better performance than the Anti-Corr method.

Introduction

Functional near-infrared spectroscopy (fNIRS) is a non-invasive technique that uses near-infrared light to measure the concentration changes in the oxygenated (HbO) and deoxygenated (HbR) hemoglobin in brain tissue. Since changes in hemoglobin in the cerebral blood flow are associated with the changes in local neural activity through neurovascular coupling [1], fNIRS has been implemented in human brain imaging research [see reviews by 2,3]. Currently, fNIRS faces a major challenge due to confounding signals in the data causing poor signal-to-noise ratios (SNRs). The confounding signals include motion artifacts [4], systemic activity such as respiration, heart rate, low-frequency fluctuations in blood pressure [5,6] and Mayer waves [7] in both the extracerebral and cerebral tissue [8], and spontaneous (non-evoked) neural activity in the cerebral tissue [9]. The current study compared two approaches that have been widely used to improve fNIRS signal quality by reducing systemic activity and noise from extracerebral tissue.

fNIRS uses a back-reflection measurement from a light source to an associated detector. As near-infrared light travels in and out of the extracerebral tissue, it passes through highly vascularized regions, including the scalp and skull, which have an average total thickness of about 13 mm [10]. fNIRS is very sensitive to the responses in extracerebral tissues [11] and physiological signals and comparatively less sensitive to changes in hemoglobin in the deeper cerebral tissue [12]. Therefore, removing the extracerebral components and physiological signals in the fNIRS signals is crucial for improving the fidelity of measures of neuronal activity. Several methods have been used in the literature to reduce the systemic activity in fNIRS signals. The first method [13] applies a low-pass filter, e.g., with a cut-off frequency of 0.1 Hz, to remove respiration and cardiac signals. However, this method fails to remove the low-frequency (below 0.1 Hz) cardiovascular oscillations in the blood flow such as Mayer waves that overlap in the frequency domain with neuronal responses recorded using block design [7,14].

A second method [15–18] performs a principal component analysis (PCA) on the fNIRS signals to remove motion artifacts and extracerebral responses. This method assumes that noise sources including extracerebral responses and motion artifacts that are much larger in amplitude compared to physiological signals, are homogenous across channels thus contribute the most to signals in all the channels. In this way, regressing out the first few principal components (PCs) in fNIRS signals would reduce the motion artifacts and the global component in the extracerebral tissue. This method is also referred to as a spatial filter. Another method that is also based on spatial separation, independent component analysis (ICA), has been used in some studies to reduce systemic noise without short channels [19–21]. The study reported in [21] compared the efficiency of PCA and ICA methods in reducing the systemic responses in fNIRS signals. Their results found that both methods were able to reduce the systemic responses, but the ICA method was inferior to the PCA method. However, when motion artifacts are rare in the fNIRS data, the assumption that motion artifacts are present in the first few PCA components would no longer be valid and instead lead to a risk of reducing the wanted neuronal signals in the fNIRS data. Further, the spatial separation methods assume that the extracerebral responses are not correlated with or independent from neuronal signals in the cerebral tissue, which might not be valid. In the cases when the extracerebral responses and cerebral responses are correlated, the PCA and ICA methods can not separate them to further remove the extracerebral responses in the fNIRS signals.

Besides the spatial separation methods (PCA and ICA) as explained above, an anti-correlation method (Anti-Corr) was proposed [22,23]. This method assumes that the changes in neuronal HbO and HbR responses are anti-correlated, whereas the systemic-activity or noise-related changes in HbO and HbR are not [23]. For example, motion artifacts cause positive correlations between changes in the HbO and HbR responses [22]. Due to a relatively high efficacy in reducing the systemic noise in fNIRS signals, the Anti-Corr method has been widely used to reduce systemic activity in fNIRS signals [24–27]. However, the fundamental assumption of the Anti-Corr method may not be entirely valid, as changes in neuronal HbO and HbR responses are not perfectly anti-correlated, with HbR changes lagging in time by ~1-2s or so behind HbO changes [28]. Hence, the anti-correlation method is likely to yield errors when used to quantify changes in neuronal activity.

Inspired by the PCA method, a more refined method known as the short-channel subtraction method has been introduced to capture and reduce systemic responses in the fNIRS signals due to extracerebral tissue activity [29–31]. The short-channel subtraction method is based on two assumptions. First, the short channels are configured on the scalp such that only responses from extracerebral tissue are measured, without contribution from the relatively deeper cerebral tissue [32]. Secondly, systemic responses such as respiration, heartbeat, Mayer waves, changes in blood pressure and cutaneous blood flow in the extracerebral tissue are partially homogeneous. Therefore, the global component measured by the short channels also exists in signals from the regular fNIRS channels [33]. Using a general linear model (GLM) regression analysis a GLM-PCA method can be applied to remove the short-channel component from the regular fNIRS channels [30,31,34]. The GLM-PCA method with short channels compared to PCA and ICA methods with no short channels mentioned above are not suited to handle the issue of failing to separate the extracerebral and neuronal responses, if they are correlated. Besides the GLM-PCA method, some studies used a linear regression [32,35] or an adaptive filtering [36,37] to subtract responses recorded using shorter channels from signals in the regular fNIRS channels. The linear regression and adaptive filtering assume that the local scalp responses measured by short channels also exist in the paired regular fNIRS channel. Hence, these ‘direct’ subtraction methods were often adopted when dual optodes were used, i.e., with a light source connected with one detector at closer (less than 10 mm) and one at regular distances (30 mm). The direct subtraction method was also adopted when short channels were within 15-mm distance from regular channels, center-to-center [34]. However, in studies that use many regular channels, pairing each regular fNIRS channel with one or more shorter channels is technically impractical and costly.

The short-channel subtraction has been considered as the ‘gold standard’ to reduce the responses from the extracerebral tissue in fNIRS data. However, there does not appear to be agreement in the literature on the ideal range for source-detector distances. Brigadoi and Cooper [38] used Monte Carlo simulations to propose that source-detector distances for adults should be no longer than 8.4 mm to optimize extraction of responses from extracerebral tissue without the inclusion of cerebral tissue. However, published studies have used short channels with distances of 13–15 mm [30,31,33,34,39,40], 20 mm [41] and even up to 25 mm [42]. The potential difficulty with using larger source-detector distances for “short channels” is that statistical regression to data from the regular fNIRS channels can introduce false negatives into the results. That is, by regressing out the short-channel components, which possibly included cerebral responses, the cerebral responses in the regular fNIRS channels would also be reduced. The study reported in [43] found that source-detector distances of 13-mm yielded poorer outcomes in improving fNIRS signal qualities compared to 6-mm distances.

Thus far, we have argued that shorter source-detector distances are generally better for lowering false-negative rates, and channels as small as ~5 mm have been introduced [32]. However, very short distances between light sources and detectors can also introduce challenges, because with shorter distances, the detectors can be saturated with photons [38]. To date, manufactures have been able to produce 8-mm short channels [7] that seem to successfully mitigate light saturation problems and minimize the amount of photons in short channel photon paths entering into cerebral tissue layers.

Despite the current ‘gold standard’ of including short channel in fNIRS studies, there is a paucity of research on the functioning of short channels, including data quality at the individual level and responses that have been measured. Hence, the current study provided detailed information on the functioning of eight 8-mm short channels for references when future studies consider whether to include short channels. We employed the GLM-PCA method by including the first two principal components (PCs) from the short channels to represent the event-related and non-related systemic response in the extracerebral tissue. Two GLM-PCA methods were conducted, with and without including stimulus-related regressors, to identify whether the event-related systemic response dominated in the extracerebral tissue. We further compared the efficacy of reducing systemic responses hence improving fNIRS signal quality between using the GLM-PCA methods and the Anti-Corr method without short channels to evaluate the necessity of including short channels. To evaluate fNIRS measures of neuronal activity evoked by an auditory task versus noise, we calculated the contrast-to-noise ratios (CNRs) from the HbO responses, which has been widely used to assess fNIRS signal quality [18,20,44]. We expected that the GLM-PCA methods with short channels would perform better at improving fNIRS signal quality, i.e., showing greater CNR values compared to the Anti-Corr method.

Methods

Participants

Twenty-three adults (10 males, mean and standard deviation (SD) of age: 22.70 ± 3.07 years) participated. This study was approved by the University of Wisconsin–Madison’s Human Subjects Institutional Review board and all the participants gave written consent. All participants were native English speakers, students at the University of Wisconsin-Madison. All participants passed the pure-tone audiometry using air conduction, with hearing thresholds less than 20 dB HL at octave frequencies between 125 Hz and 8000 Hz.

fNIRS data collection and short channels

The current study used a continuous-wave NIRS instrument (NIRScout system, NIRx medical technologies, LLC) with 16 LED light sources (Fig 1A, red arrows) and 16 avalanche photodiode (APD) detectors (Fig 1A, blue arrows) for data collection, with a sampling frequency of 3.91 Hz. Each LED light source emitted near-infrared light with wavelengths of 760 nm and 850 nm. A light source paired with detectors located at about 30-mm distance provided fNIRS channels that collected signals. In total, forty-four 30-mm (regular) channels were included and grouped into twelve regions of interest (ROIs) based on their anatomical positions, with ROI1 –ROI6 on the left and ROI7 –ROI12 on the right side (Fig 1C). These ROIs covered the ventral-lateral pathways of cortical regions for auditory speech processing, including the frontal cortex, inferior prefrontal, anterior temporal cortex, and the superior temporal areas (see reviews with meta-analyses in [45,46]). As shown in Fig 1C, each ROI consisted of 4 adjacent or crossing channels (labeled circles), which were aimed to measure signals from neighboring local brain regions. Except that ROI2 and ROI6 only consisted of two crossing channels. A NIRScap (NIRX medical technologies, LLC) matched to each participant’s head circumference was used to hold the light sources and detectors on the head. The optodes were located based on the standardized 10–10 system [47]. As the distances between some source-detector pairs on the NIRScap were above the optimal distance (30 mm) for fNIRS data recording, plastic spacers were used to reposition the two optodes and keep the distances between them at 30 mm.

Eight 8-mm channels were also used, whereby an extra silicon photodiode detector was split into 8 groups of dual-detectors (Fig 1A, green arrows), with each group surrounding a light source at 8-mm distance (Fig 1C, green dots with green lines). Panel (B) shows a schematic diagram of the connection between a light source and a short-channel detector, or between a light source and a regular detector. The eight short-channel detectors were located on two hemispheres, with 2 on the inferior frontal and 2 on the temporal lobes of each side.

The study was conducted in a standard acoustics double-walled sound booth designed by the company Industrial Acoustics Company (IAC), while participants sat in an armchair wearing a NIRSCap of suitable size. During preparation, to center the cap and correctly position the optodes on the head, the Cz was positioned halfway the distance of Nasion to Inion and halfway the distance between the two pre-auricular points. Further, the frontal electrodes Fp1 was positioned at 10% of the Nasion-Inion distance (a few centimeters above the eyebrows). The cap was attached to a chest wrap for fixation. Then the gains of light intensity at the APD detectors were checked to ensure that all the channels had at least ‘acceptable’ light intensity. In the event that a channel did not show good intensity, poor contact between the optodes with the skin usually accounted for this error, i.e., optodes in a poor angle and hair artifacts. To correct this issue, optodes were first removed from the NIRSCap to push away the hair from the contact zone and ensure perpendicular contact with the scalp, and then replaced. The procedure was repeated until most of the optodes receive at least ‘acceptable’ light intensity shown on the data collection software interface.

Stimuli and experimental design

Auditory speech stimuli consisted of sentences with a uniform structure (name, verb, number, adjective, and noun), for instance, ‘Bob has six green hats’. Sentences, unprocessed or with a noise-vocoder [48], were presented to the left ear, in the presence of background four-talker speech babble that was presented to the left ear or both ears simultaneously. The babble was also filtered with a noise-vocoder, and presented at a speech-to-noise ratio (StNR) of -10 dB or -15 dB. In total, eight hearing conditions were involved during data collection.

fNIRS data were collected using a pseudo-random block design in three separate sessions. In each session, stimuli in four different conditions at the same StNR (-10 dB or -15 dB) were presented in a random order. Each session started with a 30-s silent baseline; after each block of 13.6-s stimulation consisting of 5 sentences, there was a silent period ranging from 25 to 35 s. Nine blocks of stimuli per condition were presented in total. Throughout the test, there was a white cross on a black monitor 1.5 m in front of the participants. Participants were given verbal and written instructions to look at the cross and avoid dramatic head movement during data collection. They were also required to attend to the speech sentences and count the number of color words in each block and to push mouse buttons to indicate an even or odd number of colors at the end of each block. fNIRS data reported in this study, was part of a project that examined listening effort in normal hearing listeners under a total of 8 hearing conditions. As the aim of the current study was to examine the efficacy of the short-channel subtraction method but not about listening effort, we presented fNIRS data in one of easy hearing conditions only, i.e., unprocessed speech presented with noise in both ears at -10 dB StNR. The experiment was run in the neurobehavioral Presentation software (https://www.neurobs.com/).

fNIRS data analysis

The fNIRS signals recorded by the NIRScout system were exported to MATLAB (MATLAB R2017a) for further analysis, with a combination of customized software written by the authors and scripts adopted from Homer2 [13]. The steps taken in the signal processing are listed below and are also illustrated in Fig 2.

Fig 2 — SCI is for scalp coupling index. ‘d’ is for the distance between source and detector in an fNIRS channel.

Remove step-like noise. Step-like noise can be caused by sudden loss of contact between optodes and the skin, or interposition of hair, during data collection. If not removed, the step-like noise would skew fNIRS signals. To remove step-like artifacts in the data of each channel (y), the derivative of y was first estimated as X. Any absolute values in X, i.e., abs(X_i), which were two standard deviations or more above the mean of the absolute values of X, were set as zeros. Response y (with step-like artifacts removed) was then recovered by calculating the cumulative sum of the updated X.
Exclude channels of poor quality. A third-order Butterworth bandpass filter with cut-off frequencies of 0.5 and 1.5 Hz was applied to raw fNIRS intensity data to extract heartbeat signals in individual channels. Channels that showed poor correlations of heartbeat signals between the two different NIR wavelengths, which has been called the scalp coupling index (SCI) in Pollonini et al. [42], were excluded from further analysis. Because the heartbeat is a characteristic component of the fNIRS signal, channels that fail to record the heartbeat signals, i.e., poor SCI, are unlikely to record other physiological or neural responses. In the study by Pollonini et al. [42], channels with SCI less than 0.75 were excluded. In the current study, the 8-mm short channels had relatively poor success in quite a few subjects (Fig 3A) likely due to hair artifacts. Hence, a lower cut-off frequency (SCI > = 0.15) was set compared to that in [35] to ensure for each participant there were at least 4 shorter channels remaining for further analysis of the data from each session. The number ‘4’ was decided as Sato et al. [31] reported that a GLM-PCA method with 4 short channels provided a robust estimation of cerebral activity. Further, keeping the threshold of SCI > = 0.15 ensured that short channels from both the frontal cortex and the temporal cortex were included in the PCA, to better represent the common components in the extracerebral tissue across a wide brain region.
Convert light intensity to optical density. See Huppert et al. [13].
Correct motion artifacts. The motion artifacts might be caused by the physical displacement of the optodes from the surface of the participant’s head. The wavelet decomposition method proposed by Molavi and Dumont [49] was then performed to correct motion artifacts. With wavelet decomposition, motion artifacts appear as abrupt breaks in the wavelet domain, whereas hemodynamic responses to stimuli have less variable coefficients. To remove the motion artifacts, wavelet coefficients out of the interquartile range of 0.1 were set to zero, the same setting as in Zhou et al. [25].
Calculate the concentration changes of HbO and HbR. To adjust the effect of age and wavelengths of near-infrared light on the calculation of ΔHbO and ΔHbR, the method proposed by Scholkmann and Wolf [50] was used to calculate wavelength-dependent differential pathlength factor. The optical density data were then converted into the ΔHbO and ΔHbR using the modified Beer-Lambert law [51].
Calculate the block-average of ΔHbO and ΔHbR without or with reducing systemic responses. For all the 8-mm and 30-mm channels with SCI above 0.15, block-average fNIRS responses were first calculated without using the GLM-PCA or the Anti-Corr method. A third-order Butterworth band-pass filter (cut-off frequency at 0.01–0.09 Hz) was applied to remove high-frequency physiological signals, such as respiration and heartbeats (Fig 2, red box). Block-average fNIRS responses were then calculated, with baseline averages of each block, i.e., 5 s before stimulus onset, being subtracted. Responses in each ROI of individual participants were calculated as the grand mean of block-average fNIRS responses across channels clustered in the ROI.
Reduce systemic responses. The Anti-Corr method proposed by Cui et al. [22] was used to reduce the systemic responses (Fig 2, purple box) without short channels. The Anti-Corr method was applied to the calculated concentration changes of HbO and HbR for the regular channels (step 5). Then the same band-pass filter and block-average as explained in step 6) were conducted.
Reduce extracerebral components. The GLM-PCA method was applied to reduce the extracerebral components in the HbO and HbR responses, separately (Fig 2, yellow box) in the regular channels. For fNIRS signals in both short and regular channels, a third-order Butterworth band-pass filter (0.01–1.5 Hz) was applied to remove the low- and high-frequency physical noise. A PCA was then performed on all short channels with SCI > = 0.15 for each session of data per participant. Fig 3B shows the number of short channels (yellow) included in PCA. The first two principal components (PCs) that contributed the most to the short-channel responses were assumed to be the ‘global’ components across channels and needed to be removed. The mean ± SD contributions of the first two PCs extracted from HbO and HbR data to the short-channel responses explained 71.88% ± 7.65% and 58.66% ± 6.60%, respectively, of the variance in the signals across participants for three sessions of data collection. These two PCs were used as two regressors for a GLM with another four stimulus-specific (event-related) regressors, each for one hearing condition. Each stimulus-specific regressor was built by a box-car function for the stimulation convolved with two-gamma hemodynamic response functions (HRF), adopted from Kamran et al. [52]. Though previous studies suggested that including a derivative of the HRF would result in a more accurate estimation of the neuronal activity by accounting for shifts of time-to-peak and response window [24,53], concerns about over-parameterizing the GLM led to exclusions of derivatives in the current study. In a second method (GLM-PCA2), only the two PCs were used as regressors for the GLM, without stimulus-specific regressors. The GLM-PCA and GLM-PCA2 methods were performed on the HbO and HbR data from each of the 30-mm channels, separately, and the product of the two PCs and the corresponding coefficients from GLM was then subtracted from each channel. The same band-pass filter and block-average were then conducted on the signals in the 30-mm channels, based on which responses in ROIs were calculated.
Calculate the CNR values. To evaluate the signal quality, the CNR values of ΔHbO responses were calculated for individual channels, by first identifying the peak of block-averaged ΔHbO responses within 5–16 s after stimulus onset. The CNR was calculated as the mean ΔHbO within 10 s centered at the peak (HbO_sig), divided by the root-mean-square of the sum of the variance of HbO_sig and the variance of 5s baseline noise (HbO_noise), i.e., CNR = mean(HbO_sig)/sqrt(std(HbO_sig)^{^2} + std(HbO_noise)^{^2}). The mean CNR across channels clustered in each ROI was then calculated. We expected greater than zero CNRs in all ROIs, which covered the ventral pathways of the auditory speech processing [45,46]. When individuals were listening to and processing sentences, neuronal activity from cerebral tissues in the ventral speech pathway would increase, which was associated with increased ΔHbO responses through neurovascular coupling [54].

Fig 3 — Panel A plots the mean (circles) and standard deviation (bars) of SCI values across three sessions for the 8 short channels, i.e., S1—S8. The magenta horizontal dash line denotes the cut-off threshold (SCI = 0.15); channels that showed SCI less than the threshold were excluded from further analysis. Panel B shows the histogram of numbers of regular (right) and short channels (left) that were included in the GLM-PCA methods for each session of data collection among participants.

Statistical analysis

To evaluate the efficacy of Anti-Corr and two GLM-PCA methods in reducing the systemic noise in fNIRS signals, versus the results with no correction, a multilevel linear model was conducted on the CNR values calculated from different methods [55], with subjects being a random factor and ROIs (N = 12) being the within-subjects factor. We did not conduct a repeated-measures analysis of variance (ANOVA) as performing F-tests requires independence (lack of sphericity) between measures, whereas fNIRS responses in the current study from the 12 ROIs were dependent. Benjamini-Hochberg post-doc tests were conducted to reveal the differences between different methods.

Results

fNIRS data quality in the shorter and regular channels

The data quality in the short (8-mm) channels was calculated as SCI and the mean ± SD SCI values among the three sessions of data collection for all the 8 short channels in 23 participants are shown in Fig 3A. As shown in Fig 3A, the data quality in short channels varied among participants. While some participants (1–7) showed SCI close to 1 (perfect correlation) in all the 8 channels (8 dots in each column), other participants (14–23) showed SCIs near 0 or in negative range in one or more short channels. The numbers of short channels included for PCA analysis (SCI > = 0.15) in three sessions among participants are shown in Fig 3B (left). The mean and standard error of the mean (SEM) of the exclusion ratio of the short channels was 10.69% ± 3.17%. The number of regular (30-mm) channels included for further analysis for each session are shown in Fig 3B (right). The mean ± SEM exclusion ratio of regular channels was 2.80% ± 0.97%.

Block-average responses in the 8-mm channels

After excluding the channels of poor quality (SCI < 0.15), the group average of ΔHbO and ΔHbR of the eight short (8-mm) channels (S1 –S8) were examined. Fig 4 plots the block-averaged fNIRS responses in the eight short channels. Channels S1 –S4 were on the left hemisphere and S5 –S8 were located at the symmetric positions on the right side. Group means (solid lines) and SEMs (shaded areas) are plotted for ΔHbO (red) and ΔHbR responses (blue). Each epoch consists of 5-s baseline and 32-s responses after stimulus onset, with stimulus duration (marked by vertical lines) of 13.6 s. As shown in Fig 4, all the short channels had suppressive responses, i.e., decreased ΔHbO and increased ΔHbR after stimulus onset, opposite to expected event-evoked neuronal ΔHbO and ΔHbR—increase in ΔHbO and decrease in ΔHbR.

fNIRS responses with and without short-channel subtraction

To reduce the systemic noise in fNIRS signals, the Anti-Corr method without short channels and the two GLM-PCA short-channel subtraction methods were applied (see Fig 2). fNIRS responses in all ROIs before and after using Anti-Corr and GLM-PCA to reduce the systemic noise were compared. Fig 5 shows the group mean and SEM ΔHbO responses without reducing the systemic components (original, red), with ΔHbO responses after applying the Anti-Corr (purple) or the GLM-PCA method (with stimulus-related and two PC regressors, yellow), or the GLM-PCA2 (with only the two PC regressors, brown). As shown in Fig 5, the original block-average ΔHbO responses (red) were generally suppressive, i.e., decrease in responses after stimulus onset and below baseline (zero). This response pattern is similar to the globally suppressive responses in the short channels (Fig 4), suggesting that the suppressive responses in the extracerebral tissue largely contributed to the signals in the regular fNIRS channels. After applying the Anti-Corr (purple) or the two GLM-PCA methods (yellow and brown), the suppressive components were reduced. The block-average ΔHbO responses across most of the 12 ROIs showed increased response after stimulus onset and gradually went back to baseline after stimulus offset, resembling the expected patterns of neuronal activity related to ΔHbO. Further, the two GLM-PCA methods showed almost identical block-averaged results, and both reduced more of the suppressive components in the fNIRS signals in all the ROIs compared to the Anti-Corr method.

Fig 5 — The block-average results, i.e., group means (solid lines) and standard error of means (SEM, shaded areas) of ΔHbO responses without further reducing the systemic responses (Original, red) and after applying the Anti-Corr (purple), the GLM-PCA (with stimulus-related regressors and 2 PCs, yellow), and the GLM-PCA2 (with the 2 PCs as regressors, brown) methods are plotted. The two vertical black dash lines plot the onset and offset of stimuli, with a duration of 13.6 s. Responses of zero are relative to the average of 5-s baseline.

Fig 6 plots the CNR values with the Original values (red) and after applying the Anti-Corr (purple) or the GLM-PCA (yellow) or the GLM-PCA2 (brown) method for the 12 ROIs. As shown in Fig 6, applying the two GLM-PCA methods increased the CNR values compared to the Original values (red). Further, the CNRs values from GLM-PCA and GLM-PCA2 methods were almost identical. In eleven of the 12 ROIs, the CNR values from the GLM-PCA method were also greater than those from the Anti-Corr method (purple). Results from the multilevel linear model analysis showed significant differences between methods (χ²(3) = 33.01, p < 0.001) and between ROIs (χ²(11) = 200.75, p < 0.001). Results from the Benjamini-Hochberg post-hoc tests revealed significant differences in CNR values between Anti-Corr and the Original method (p = 0.014), between the two GLM-PCA methods and the Original (p < 0.001), between the two GLM-PCA and Anti-Corr (p < 0.001), but no differences between GLM-PCA and GLM-PCA2 (p = 0.992). As the choice of SCI = 0.15 was low, we also provided block-averaged results and CNR values with a cut-off of SCI = 0.75 in supplementary materials. Results from the statistical analyses demonstrated that, with SCI = 0.75 the two GLM-PCA methods significantly improved the CNR values compared to the original signals and compared to the Anti-Corr method, similar to the results with SCI = 0.15 reported above.

Discussion

The present study compared the performance of two approaches to reducing the systemic noise in the fNIRS signals, i.e., the short-channel subtraction (GLM-PCA) method and an anti-correlation (Anti-Corr) method, which can be used when shorter channels are not available or feasible for implementation. We implemented the GLM-PCA methods with the two PCs extracted from short-channels responses, with and without stimulus-related regressors, i.e., GLM-PCA and GLM-PCA2, respectively. fNIRS data were collected from 23 young normal-hearing adults when listening to speech sentences in the left ear with bilateral noise at -10 dB StNR. fNIRS measures were examined from 12 ROIs from two hemispheres (Fig 1). To assess whether any of the three methods improved fNIRS signal qualities, we calculated the CNR values from ΔHbO responses before and after applying the methods. Our results found that both the Anti-Corr and GLM-PCA methods improved the fNIRS signal qualities; there were no significant differences between the two GLM-PCA methods and they both performed better than the Anti-Corr method.

The current study used eight 8-mm channels, which were positioned on the bilateral inferior frontal and temporal lobes (Fig 1), to measure responses from the extracerebral tissues but not the deeper cerebral tissues. Our results demonstrated that responses in all the 8 short channels (Fig 4) were generally suppressive, i.e., ΔHbO responses decreased and ΔHbR increased after stimulus onset. Further, the ΔHbO responses in all 12 ROIs that consisted of regular channels showed suppressive patterns (red, in Fig 5). Similar suppressive responses were also found by Kirilina et al. [6], where a time-domain NIRS system was used with diode lasers, providing one 1-mm short channel and four 30-mm regular channels. In that study, four regular channels were positioned on the forehead with 2 on each side, and the short channel was positioned on the right forehead. Kirilina et al. [6] found that all channels showed suppressive responses when participants were performing a visual semantic categorization task. In addition to fNIRS, Kirilina et al. [6] also scanned 2 participants with fMRI and participants performed the same tasks as in fNIRS sessions. They reported that the suppressive fNIRS (ΔHbO) responses had similar patterns to the extracranial (or extracerebral) responses recorded from fMRI. Kirilina et al. [6] proposed that the suppressive responses were due to vasoconstriction in the superficial skin tissue, which might occur when participants were under mental stress or performing arithmetic tasks, resulting in a decreased blood flow, thus decreased ΔHbO responses.

Since fNIRS responses recorded from each channel were a combination of homogeneous global responses and region-specific responses from local areas [33], it is possible that the suppressive patterns found in the short channels in both our study and Kirilina et al. [6] have similar origins that are globally homogeneous, and not dependent on the channel locations or experimental tasks. In the current study, after applying the two GLM-PCA methods, the suppressive components were reduced and ΔHbO responses in most of the 12 ROIs showed increased patterns after auditory stimulation in line with our expectations. These results suggest that the short-channel components largely contributed to fNIRS signals and that the short-channel subtraction method improved fNIRS signal qualities. However, the group average ΔHbO responses in the ROI6 and ROI10 were still suppressive after applying GLM-PCA. This result might be due to the locations of the two ROIs on the head, i.e., the temporal lobes on both sides (Fig 1C). Optodes that were involved in these regions and held by the NIRScap, were located around or above the auricles, where there was a lot of hair that could contaminate the data collection. Further, with a length of 30 mm (regular channels), near-infrared light has a limited penetration depth (about 15 mm) in adults, thus we were not able to record cerebral activity in the deeper tissues such as temporal sulcus underneath some of the channels in the two ROIs. Therefore, the fNIRS auditory-evoked signals in channels above these regions on the head might mostly consist of extracerebral components and noise, with few or no neuronal-activity related changes in hemodynamic responses.

Our results showed that the GLM-PCA (short-channel subtraction) method performed better at improving fNIRS signal qualities compared to the Anti-Corr (anti-correlation) method used previously [23–27]. This difference between methods could be contributed to by assumptions in the Anti-Corr method that may not be entirely valid. For example, the Anti-Corr method assumes that neuronal-activity-related ΔHbO and ΔHbR are anti-correlated, which is not entirely valid as ΔHbR lags behind ΔHbO. The two GLM-PCA methods using data from 8-mm channels assume that only responses from extracerebral tissue are measured from the shorter channels [38]. Further, PCA was applied to the signals from at least 4 out of 8 short channels that were spread over the bilateral inferior prefrontal and temporal cortex. The first two of the principal components were assumed to represent the systemic responses and/or residual motion artifacts that globally existed in the extracerebral tissue. Subtracting the components from the regular fNIRS channels led to greater CNR values from ΔHbO response, i.e., larger ratios of cerebral neuronal activity versus noise. Interestingly, two GLM-PCA methods, with one only using the two PCs (GLM-PCA2) and the other also including the stimulus-related regressors and (GLM-PCA), showed almost identical block averaged results and CNRs values for 12 ROIs. The results suggest that the two PCs were not correlated with the stimulus-related regressors, hence, the two GLM returned similar results for the two PCs. The finding that the two PCs in the short channels were not correlated with the event-related regressors suggests the dominance of the non-related systemic responses in the extracerebral tissue compared to the event-related systemic responses.

Despite the benefits of using GLM-PCA methods to reduce noise in fNIRS data, several limitations must be addressed. First, to retain at least 4 short channels for PCA [31], we only excluded channels with SCI less than 0.15. With this cut-off threshold, in 3 out of 23 subjects nearly 40% of short channels (Fig 3A; i.e., average of 3 to 4 out of 8 short channels) per session were excluded from data analysis. However, the criterion of SCI < 0.15 might be too low; including short channels of poor quality for PCA might introduce artifacts to fNIRS signals when performing GLM-PCA methods rather than reducing the systemic noise. Second, the ratio of short channels rejected was larger than that of the regular channels. This difference was likely due to the procedure of pushing aside the hair underneath optodes to improve the light intensity for the regular channels, but not for the short channels, since the dual-detectors were underneath the same grommets that held the light sources onto the head (see Fig 1). Third, the efficacy of the short-channel subtraction method might vary among participants and between testing sessions. In the subjects where the SCI values were above threshold but poor in most of the channels, suggesting the difficulty in recording fNIRS responses, the short-channel subtraction method might not work as well as in the subjects where fNIRS data generally had more acceptable signal qualities. Further, the data quality in short channels from some participants also varied across sessions (Fig 3A), which could be due to the shifts of optodes on the head hence having poor contact with the skin. Though, participants were instructed to make dramatic movements to avoid the optodes shifting. In the more complicated tasks that involved a lot of head movements, it is important to better fix the optodes are on the head, to collect data of good quality. Otherwise, short-channel subtraction methods based on short channels that mostly measured motion artifacts may not be able to efficiently reduce the systemic responses in the extracerebral tissue.

Conclusion

This study compared fNIRS signal qualities by using the current ‘gold standard’ method, i.e., short-channel subtraction, with an anti-correlation method that has been applied in studies when short-channels were not available or infeasible for implementation. Our results found that fNIRS responses in all the shorter channels and regular (30-mm) channels before short-channel subtraction were overall suppressive (opposite to the expected neuronal activity). In the regular channels, after the short-channel components being subtracted, fNIRS signals showed patterns similar to neuronal activity. Comparisons of results from the two approaches found that both methods improved the CNR values from ΔHbO, but the CNRs for the short-channel subtraction methods exceeded those from the anti-correlation method, indicating that short channels should be used in preference to the anti-correlation method if feasible.

Supporting information

S1 Fig. Block-average ΔHbO responses in 12 regions of interest (ROIs).

Results in channels with SCI> = 0.75 were included. The block-average results, i.e., group means (solid lines) and standard error of means (SEM, shaded areas) of ΔHbO responses without further reducing the systemic responses (Original, red) and after applying the Anti-Corr (purple), the GLM-PCA (with stimulus-related regressors and 2 PCs, yellow), and the GLM-PCA2 (with the 2 PCs as regressors, brown) methods are plotted. The two vertical black dash lines plot the onset and offset of stimuli, with a duration of 13.6 s. Responses of zero are relative to the average of 5-s baseline.

(TIF)

Click here for additional data file.^{(325.7KB, tif)}

S2 Fig. Contrast-to-noise ratios (CNRs) from different methods.

Results from channels with SCI> = 0.75 were included. Violin plots show the CNRs from HbO responses before (Original, red) and after applying Anti-Corr (purple) or GLM-PCA method (yellow), or GLM-PCA2 (brown) in individual participants in 12 regions of interest (ROIs). Results from the multilevel linear model analysis on the CNR values found a significant interaction between methods and ROIs (χ²(33) = 58.74, p = 0.004), and significant differences between methods (χ²(3) = 239.92, p < 0.001), and between ROIs (χ²(11) = 89.39, p < 0.001). Results from the Benjamini-Hochberg post-hoc tests revealed no significant differences in CNR values between Anti-Corr and the Original method (p = 0.26), but demonstrated significant differences between the two GLM-PCA methods and the Original (p < 0.001), between the two GLM-PCA and Anti-Corr (p < 0.001), but no differences between GLM-PCA and GLM-PCA2 (p = 0.77).

(TIF)

Click here for additional data file.^{(294.4KB, tif)}

S1 Dataset

(XLSX)

Click here for additional data file.^{(107.9KB, xlsx)}

Acknowledgments

The authors appreciate the time and support from all the research participants. We also thank various colleagues from the binaural hearing and speech lab for suggestions about experimental design and participant recruitment, including Alan Kan, Z. Ellen Peng, Emily Burg, Shelly Godar, and Tanvi Thakkar.

Data Availability

All relevant data are within the manuscript and its Supporting information files.

Funding Statement

The work was supported by NIH-NIDCD (https://www.nidcd.nih.gov/, R01DC003083 to RL), UW-Madison’s Office of the Vice Chancellor for Research, and a Core grant from NIH-NICHD (U54HD090256 to Waisman Center). The sponsors and funders were not involved in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

1.Jobsis FF. Noninvasive, infrared monitoring of cerebral and myocardial oxygen sufficiency and circulatory parameters. Science. 1977;198(4323):1264–7. 10.1126/science.929199 [DOI] [PubMed] [Google Scholar]
2.Ferrari M, Quaresima V. A brief review on the history of human functional near-infrared spectroscopy (fNIRS) development and fields of application. Neuroimage. 2012;63(2):921–35. 10.1016/j.neuroimage.2012.03.049 [DOI] [PubMed] [Google Scholar]
3.Scholkmann F, Kleiser S, Metz AJ, Zimmermann R, Mata Pavia J, Wolf U, et al. A review on continuous wave functional near-infrared spectroscopy and imaging instrumentation and methodology. Neuroimage. 2014;85 Pt 1:6–27. 10.1016/j.neuroimage.2013.05.004 [DOI] [PubMed] [Google Scholar]
4.Scholkmann F, Spichtig S, Muehlemann T, Wolf M. How to detect and reduce movement artifacts in near-infrared imaging using moving standard deviation and spline interpolation. Physiological measurement. 2010;31(5):649–62. 10.1088/0967-3334/31/5/004 [DOI] [PubMed] [Google Scholar]
5.Kirilina E, Yu N, Jelzow A, Wabnitz H, Jacobs AM, Tachtsidis I. Identifying and quantifying main components of physiological noise in functional near infrared spectroscopy on the prefrontal cortex. Frontiers in human neuroscience. 2013;7:864 10.3389/fnhum.2013.00864 [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Kirilina E, Jelzow A, Heine A, Niessing M, Wabnitz H, Bruhl R, et al. The physiological origin of task-evoked systemic artefacts in functional near infrared spectroscopy. Neuroimage. 2012;61(1):70–81. 10.1016/j.neuroimage.2012.02.074 [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Yücel MA, Selb J, Aasted CM, Lin P-Y, Borsook D, Becerra L, et al. Mayer waves reduce the accuracy of estimated hemodynamic response functions in functional near-infrared spectroscopy. Biomedical optics express. 2016;7(8):3078–88. 10.1364/BOE.7.003078 [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Habermehl C, Schmitz C, Koch SP, Mehnert J, Steinbrink J, editors. Investigating hemodynamics in scalp and brain using high-resolution diffuse optical tomography in humans. Biomedical Optics; 2012: Optical Society of America. [Google Scholar]
9.Tachtsidis I, Scholkmann F. Erratum: Publisher’s note: False positives and false negatives in functional near-infrared spectroscopy: issues, challenges, and the way forward. Neurophotonics. 2016;3(3):039801 10.1117/1.NPh.3.3.039801 [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Strangman GE, Zhang Q, Li Z. Scalp and skull influence on near infrared photon propagation in the Colin27 brain template. Neuroimage. 2014;85 Pt 1:136–49. 10.1016/j.neuroimage.2013.04.090 [DOI] [PubMed] [Google Scholar]
11.Hirasawa A, Yanagisawa S, Tanaka N, Funane T, Kiguchi M, Sorensen H, et al. Influence of skin blood flow and source-detector distance on near-infrared spectroscopy-determined cerebral oxygenation in humans. Clin Physiol Funct I. 2015;35(3):237–44. 10.1111/cpf.12156 [DOI] [PubMed] [Google Scholar]
12.Al-Rawi PG, Smielewski P, Kirkpatrick PJ. Evaluation of a near-infrared spectrometer (NIRO 300) for the detection of intracranial oxygenation changes in the adult head. Stroke. 2001;32(11):2492–500. 10.1161/hs1101.098356 [DOI] [PubMed] [Google Scholar]
13.Huppert TJ, Diamond SG, Franceschini MA, Boas DA. HomER: a review of time-series analysis methods for near-infrared spectroscopy of the brain. Appl Opt. 2009;48(10):D280–98. 10.1364/ao.48.00d280 [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Erdogan SB, Yucel MA, Akin A. Analysis of task-evoked systemic interference in fNIRS measurements: insights from fMRI. Neuroimage. 2014;87:490–504. 10.1016/j.neuroimage.2013.10.024 [DOI] [PubMed] [Google Scholar]
15.Zhang YH, Brooks DH, Franceschini MA, Boas DA. Eigenvector-based spatial filtering for reduction of physiological interference in diffuse optical imaging. Journal of biomedical optics. 2005;10(1). 10.1117/1.1852552 [DOI] [PubMed] [Google Scholar]
16.Zhang X, Noah JA, Dravida S, Hirsch J. Signal processing of functional NIRS data acquired during overt speaking. Neurophotonics. 2017;4(4):041409 10.1117/1.NPh.4.4.041409 [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Zhang X, Noah JA, Hirsch J. Separation of the global and local components in functional near-infrared spectroscopy signals using principal component spatial filtering. Neurophotonics. 2016;3(1). 10.1117/1.NPh.3.1.015004 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Cooper RJ, Seib J, Gagnon L, Phillip D, Schytz HW, Iversen HK, et al. A systematic comparison of motion artifact correction techniques for functional near-infrared spectroscopy. Frontiers in neuroscience. 2012;6 10.3389/fnins.2012.00147 [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Santosa H, Hong MJ, Kim SP, Hong KS. Noise reduction in functional near-infrared spectroscopy signals by independent component analysis. The Review of scientific instruments. 2013;84(7):073106 10.1063/1.4812785 [DOI] [PubMed] [Google Scholar]
20.Bauernfeind G, Wriessnegger SC, Daly I, Muller-Putz GR. Separating heart and brain: on the reduction of physiological noise from multichannel functional near-infrared spectroscopy (fNIRS) signals. Journal of neural engineering. 2014;11(5):056010 10.1088/1741-2560/11/5/056010 [DOI] [PubMed] [Google Scholar]
21.Virtanen J, Noponen T, Merilainen P. Comparison of principal and independent component analysis in removing extracerebral interference from near-infrared spectroscopy signals. Journal of biomedical optics. 2009;14(5):054032 10.1117/1.3253323 [DOI] [PubMed] [Google Scholar]
22.Cui X, Bray S, Reiss AL. Functional near infrared spectroscopy (NIRS) signal improvement based on negative correlation between oxygenated and deoxygenated hemoglobin dynamics. Neuroimage. 2010;49(4):3039–46. 10.1016/j.neuroimage.2009.11.050 [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Yamada T, Umeyama S, Matsuda K. Separation of fNIRS Signals into Functional and Systemic Components Based on Differences in Hemodynamic Modalities. PloS one. 2012;7(11). 10.1371/journal.pone.0050271 [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Wijayasiri P, Hartley DEH, Wiggins IM. Brain activity underlying the recovery of meaning from degraded speech: A functional near-infrared spectroscopy (fNIRS) study. Hearing research. 2017;351:55–67. 10.1016/j.heares.2017.05.010 [DOI] [PubMed] [Google Scholar]
25.Zhou X, Seghouane AK, Shah A, Innes-Brown H, Cross W, Litovsky RY, et al. Cortical Speech Processing in Postlingually Deaf Adult Cochlear Implant Users, as Revealed by Functional Near-Infrared Spectroscopy. Trends in hearing. 2018;22:2331216518786850 10.1177/2331216518786850 [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Wiggins IM, Hartley DEH. A Synchrony-Dependent Influence of Sounds on Activity in Visual Cortex Measured Using Functional Near-Infrared Spectroscopy (fNIRS). PloS one. 2015;10(3). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Lawrence RJ, Wiggins IM, Anderson CA, Davies-Thompson J, Hartley DE. Cortical correlates of speech intelligibility measured using functional near-infrared spectroscopy (fNIRS). Hearing research. 2018;370:53–64. 10.1016/j.heares.2018.09.005 [DOI] [PubMed] [Google Scholar]
28.Jasdzewski G, Strangman G, Wagner J, Kwong KK, Poldrack RA, Boas DA. Differences in the hemodynamic response to event-related motor and visual paradigms as measured by near-infrared spectroscopy. Neuroimage. 2003;20(1):479–88. 10.1016/s1053-8119(03)00311-2 [DOI] [PubMed] [Google Scholar]
29.Gagnon L, Perdue K, Greve DN, Goldenholz D, Kaskhedikar G, Boas DA. Improved recovery of the hemodynamic response in diffuse optical imaging using short optode separations and state-space modeling. Neuroimage. 2011;56(3):1362–71. 10.1016/j.neuroimage.2011.03.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Gagnon L, Yücel MA, Boas DA, Cooper RJ. Further improvement in reducing superficial contamination in NIRS using double short separation measurements. Neuroimage. 2014;85:127–35. 10.1016/j.neuroimage.2013.01.073 [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Sato T, Nambu I, Takeda K, Aihara T, Yamashita O, Isogaya Y, et al. Reduction of global interference of scalp-hemodynamics in functional near-infrared spectroscopy using short distance probes. NeuroImage. 2016;141:120–32. 10.1016/j.neuroimage.2016.06.054 [DOI] [PubMed] [Google Scholar]
32.Saager RB, Berger AJ. Direct characterization and removal of interfering absorption trends in two-layer turbid media. JOSA A. 2005;22(9):1874–82. 10.1364/josaa.22.001874 [DOI] [PubMed] [Google Scholar]
33.Funane T, Atsumori H, Katura T, Obata AN, Sato H, Tanikawa Y, et al. Quantitative evaluation of deep and shallow tissue layers’ contribution to fNIRS signal using multi-distance optodes and independent component analysis. Neuroimage. 2014;85 Pt 1:150–65. 10.1016/j.neuroimage.2013.02.026 [DOI] [PubMed] [Google Scholar]
34.Gagnon L, Cooper RJ, Yücel MA, Perdue KL, Greve DN, Boas DA. Short separation channel location impacts the performance of short channel regression in NIRS. Neuroimage. 2012;59(3):2518–28. 10.1016/j.neuroimage.2011.08.095 [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Scholkmann F, Metz AJ, Wolf M. Measuring tissue hemodynamics and oxygenation by continuous-wave functional near-infrared spectroscopy—how robust are the different calculation methods against movement artifacts? Physiological measurement. 2014;35(4):717 10.1088/0967-3334/35/4/717 [DOI] [PubMed] [Google Scholar]
36.Zhang Q, Brown EN, Strangman GE. Adaptive filtering for global interference cancellation and real-time recovery of evoked brain activity: a Monte Carlo simulation study. Journal of biomedical optics. 2007;12(4):044014–12. 10.1117/1.2754714 [DOI] [PubMed] [Google Scholar]
37.Zhang Q, Strangman GE, Ganis G. Adaptive filtering to reduce global interference in non-invasive NIRS measures of brain activation: how well and when does it work? Neuroimage. 2009;45(3):788–94. 10.1016/j.neuroimage.2008.12.048 [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Brigadoi S, Cooper RJ. How short is short? Optimum source–detector distance for short-separation channels in functional near-infrared spectroscopy. Neurophotonics. 2015;2(2):025005-. 10.1117/1.NPh.2.2.025005 [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Weder S, Zhou X, Shoushtarian M, Innes-Brown H, McKay C. Cortical Processing Related to Intensity of a Modulated Noise Stimulus—a Functional Near-Infrared Study. Journal of the Association for Research in Otolaryngology. 2018;19(3):273–86. 10.1007/s10162-018-0661-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Gregg NM, White BR, Zeff BW, Berger AJ, Culver JP. Brain specificity of diffuse optical imaging: improvements from superficial signal regression and tomography. Frontiers in neuroenergetics. 2010;2. [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Yamada T, Umeyama S, Matsuda K. Multidistance probe arrangement to eliminate artifacts in functional near-infrared spectroscopy. Journal of biomedical optics. 2009;14(6):064034–12. 10.1117/1.3275469 [DOI] [PubMed] [Google Scholar]
42.Pollonini L, Olds C, Abaya H, Bortfeld H, Beauchamp MS, Oghalai JS. Auditory cortex activation to natural speech and simulated cochlear implant speech measured with functional near-infrared spectroscopy. Hearing research. 2014;309:84–93. 10.1016/j.heares.2013.11.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Goodwin JR, Gaudet CR, Berger AJ. Short-channel functional near-infrared spectroscopy regressions improve when source-detector separation is reduced. Neurophotonics. 2014;1(1). 10.1117/1.NPh.1.1.015002 [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Scarpa F, Cutini S, Scatturin P, Dell’Acqua R, Sparacino G. Bayesian filtering of human brain hemodynamic activity elicited by visual short-term maintenance recorded through functional near-infrared spectroscopy (fNIRS). Optics express. 2010;18(25):26550–68. 10.1364/OE.18.026550 [DOI] [PubMed] [Google Scholar]
45.Vigneau M, Beaucousin V, Herve PY, Duffau H, Crivello F, Houde O, et al. Meta-analyzing left hemisphere language areas: phonology, semantics, and sentence processing. Neuroimage. 2006;30(4):1414–32. 10.1016/j.neuroimage.2005.11.002 [DOI] [PubMed] [Google Scholar]
46.Vigneau M, Beaucousin V, Herve PY, Jobard G, Petit L, Crivello F, et al. What is right-hemisphere contribution to phonological, lexico-semantic, and sentence processing? Insights from a meta-analysis. Neuroimage. 2011;54(1):577–93. 10.1016/j.neuroimage.2010.07.036 [DOI] [PubMed] [Google Scholar]
47.Acharya JN, Hani A, Cheek J, Thirumala P, Tsuchida TN. American Clinical Neurophysiology Society Guideline 2: Guidelines for Standard Electrode Position Nomenclature. Journal of clinical neurophysiology: official publication of the American Electroencephalographic Society. 2016;33(4):308–11. [DOI] [PubMed] [Google Scholar]
48.Shannon RV, Zeng F-G, Kamath V, Wygonski J, Ekelid M. Speech recognition with primarily temporal cues. Science. 1995;270(5234):303–4. 10.1126/science.270.5234.303 [DOI] [PubMed] [Google Scholar]
49.Molavi B, Dumont GA. Wavelet based motion artifact removal for Functional Near Infrared Spectroscopy. Conference proceedings: Annual International Conference of the IEEE Engineering in Medicine and Biology Society IEEE Engineering in Medicine and Biology Society Annual Conference. 2010;2010:5–8. [DOI] [PubMed]
50.Scholkmann F, Wolf M. General equation for the differential pathlength factor of the frontal human head depending on wavelength and age. Journal of biomedical optics. 2013;18(10):105004 10.1117/1.JBO.18.10.105004 [DOI] [PubMed] [Google Scholar]
51.Delpy DT, Cope M, van der Zee P, Arridge S, Wray S, Wyatt J. Estimation of optical pathlength through tissue from direct time of flight measurement. Physics in medicine and biology. 1988;33(12):1433 10.1088/0031-9155/33/12/008 [DOI] [PubMed] [Google Scholar]
52.Kamran MA, Jeong MY, Mannan MM. Optimal hemodynamic response model for functional near-infrared spectroscopy. Frontiers in behavioral neuroscience. 2015;9 10.3389/fnbeh.2015.00151 [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Friston KJ, Fletcher P, Josephs O, Holmes A, Rugg M, Turner R. Event-related fMRI: characterizing differential responses. Neuroimage. 1998;7(1):30–40. 10.1006/nimg.1997.0306 [DOI] [PubMed] [Google Scholar]
54.Ferrari M, Mottola L, Quaresima V. Principles, techniques, and limitations of near infrared spectroscopy. Canadian journal of applied physiology = Revue canadienne de physiologie appliquee. 2004;29(4):463–87. 10.1139/h04-031 [DOI] [PubMed] [Google Scholar]
55.Bliese PD. Within-group agreement, non-independence, and reliability: Implications for data aggregation and analysis. 2000. [Google Scholar]

PLoS One. doi: 10.1371/journal.pone.0244186.r001

Decision Letter 0

Chelsea Dobbins

24 Jul 2020

PONE-D-20-04747

Comparing fNIRS signal qualities between approaches with and without short channels

PLOS ONE

Dear Dr. Zhou,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The paper has been written well on an important area of fNIRS research. While praising many aspects of the article, the majority of the reviewers suggest a major revision is required to improve and deepen the introduction, method and discussion and include more details of the PCA.

Please work through and address the reviewer's comments in your revised version.

Please submit your revised manuscript by Sep 07 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Chelsea Dobbins

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Thank you for stating the following in the Competing Interests section:

"Dr. Litovsky discloses that she is a consultant for Frequency Therapeutics. The other authors have

all certified that they have NO affiliations with or involvement in any organization or entity with

any financial interest or non-financial interest in the subject matter or materials discussed in this

manuscript."

Please confirm that this does not alter your adherence to all PLOS ONE policies on sharing data and materials, by including the following statement: "This does not alter our adherence to PLOS ONE policies on sharing data and materials.” (as detailed online in our guide for authors http://journals.plos.org/plosone/s/competing-interests). If there are restrictions on sharing of data and/or materials, please state these. Please note that we cannot proceed with consideration of your article until this information has been declared.

Please include your updated Competing Interests statement in your cover letter; we will change the online submission form on your behalf.

Please know it is PLOS ONE policy for corresponding authors to declare, on behalf of all authors, all potential competing interests for the purposes of transparency. PLOS defines a competing interest as anything that interferes with, or could reasonably be perceived as interfering with, the full and objective presentation, peer review, editorial decision-making, or publication of research or non-research articles submitted to one of the journals. Competing interests can be financial or non-financial, professional, or personal. Competing interests can arise in relationship to an organization or another person. Please follow this link to our website for more details on competing interests: http://journals.plos.org/plosone/s/competing-interests

3. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Partly

Reviewer #3: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: I Don't Know

Reviewer #3: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The paper entitled “comparing fNIRS signal qualities betrween approaches with and without short channel” Is a well written paper that provides interesting information about methodological choices for the NIRS data processing. I found the paper easy to read and hypotheses and results are easy to understand. The final contribution would certainly be valuable for future functional NIRS studies. This being said I think the paper would greatly benefits from some clarifications of methodological choices and reorganization of some parts. You will find below a list of questions and recommendations related to it.

• Introduction

- A description of the different systemic components (HR, Mayer waves…) and how they interfere with the neural signal could be useful for the naive or NIRS beginners. For instance how the HR could be summed on block averaging (e.g., Kirilina 2012) and how Mayer waves’ frequency overlap with the neurovascular response.

- Why did you choose to use GLM-PCA and what's the difference with other regression like the almost conventional Ac= AL-α.AS (Scholkmann 20014 and Nirx short-channel ressources)?

- You choose PCA over ICA. Please justify this choice.

• Method

- I would suggest justifying more the limitation of the study in the comparison of only two methods. Indeed, you mentioned 5 different technics in the introduction and, if the choice of the Gold standard short channel is obvious, the choice of Anti-Corr only is not justified enough.

- I’m aware that fNIRS processing pipeline has not reached a consensus yet, however I have few concerns about your choices that you may need to clarify/justify.

- Filtering: You did not mention (or I missed it) the first band-pass filter you applied on the ΔHb [0.1 1.5] Hz in the text, hence no justification for this range is available. Since there seem to be some agreement for a [0.01 0.09] frequency band (see Pinti et al., 2019), why did you choose these values?

- Short-channel regression: You chose to apply the short-channel correction over the averaged concentration data whereas numerous researchers have stated that it is better to do it on the OD. Even the code provided by Nirx for Homer2 states that the regression should be applied on the OD. Could you justify this choice and state how this could affect the preprocessing results.

- ROIs: A figure with the ROIs would really help visualizing your results and discussion. In addition a brief justification for this choice should be written in the Method section.

- Statistics: It’s a bit annoying to discover the statistical analyses in the results section. I would prefer an a priori description of them in the Method section.

- What was the device acquisition sampling frequency?

- Participants: Please detail the material used for hearing assessments.

L120. Normal channel distance average is certainly around 30 mm but according to the 10-20 positioning system you should have variations around this distance (it can be roughly from 25 up to 40 mm according to Nirsite). You need to add more information about this issue and how it can be partially corrected using Nirx spacers (only to prevent channel from exceeding 30mm).

L122. Please provide more information about cap positioning over the head to ensure homogeneity across participants.

L133. Define IAC the first time you use it.

L169. Justify such difference compared to Pollonini. If you have clearly described why you need to have at least 4 short channels of correct quality, it seems that a SCI threshold of 0.15 is very low. Looking at your figure 3, if you choose a criterion of SCI >0.5 for instance, you'll be closer to Pollonini's and will only lose participant N° 19 while still keeping an average of 6.9 short channels for the remaining 22 participants whit a better quality.

From L189 to L211: this part could be simplified and clearer if you state only once that all three data types (Not corrected, Anti-Corr and Short-channel) where finally filtered and block-averaged in the same way.

• Results

- L353. As a reader I would be curious to have a look at the 2 principal components.

• Discussion

- As APD are known to have a better light sensitivity, I’d like to know whether your conclusion would remain the same for normal/standard photodiode.

- I’m a little bit surprised regarding the results for the auditory cortex. Since several studies have succeeded in probing this region with conventional NIRS setup (see references below), I’m not convinced about your argument. At this point, I recommend doing a Monte Carlo simulation of photon travel with your setup in AtlasViewer to discuss your result with actual depth penetration values. I’m currently doing research with a comparable montage and have quiet descent sensitivity in the superior temporal sulcus, precisely in the primary auditory cortex. Here again a picture of the ROIs would help in the results interpretation.

Refs:

Santosa, H., Hong, M. J., & Hong, K. S. (2014). Lateralization of music processing with noises in the auditory cortex: an fNIRS study. Frontiers in Behavioral Neuroscience, 8, 418.

Y. Minagawa-Kawai, K. Mori, I. Furuya, R. Hayashi, Y. Sato. Assessing cerebral representations of short and long vowel categories by NIRS. Neuroreport, 13 (2002), pp. 581-584

Plichta, M. M., Gerdes, A. B., Alpers, G. W., Harnisch, W., Brill, S., Wieser, M. J., & Fallgatter, A. J. (2011). Auditory cortex activation is modulated by emotion: a functional near-infrared spectroscopy (fNIRS) study. Neuroimage, 55(3), 1200-1207.

• Figures

Please cite image sources (NIRX/Nisite for instance)

Reviewer #2: Overall, I think that this paper still lacks strong and clear results.

The manuscript requires some additional work to better present the results and, possibly, also the data analysis.

I have two main concerns about this paper.

[1] The first is methodological.

The GLM-PCA method includes a GLM model that uses as regressors, in addition to the two principal components of the PCA on the short channels, also information about the temporal locations of the stimuli, i.e. the “stimulus-regressors”.

So the GLM is “informed” that there are the stimuli.

After, the comparison between the three pipelines is performed on the capacity of recovering the effect of the stimuli.

I think that this is not completely fair, as the results from the pipeline with short-channels come from a step that uses the same information that is “searched” to process the signals.

- What would be the result of using a GLM only based on the two main components?

- Or, conversely, what would be the results on the two pipelines not based on short-channels when a GLM with “stimulus-regressors” is used?

[2] The second pertains to the organization of the study.

This is a methodological paper with the aim of comparing signal processing pipelines.

However, there is not a strong result on the comparison here.

Instead, there are many qualitative considerations (many figures, maybe too many) but not a clear quantitative assessment.

I think this is partially due to the organization of the manuscript.

First, the Methods section completely lacks any information about how the pipelines will be compared, and on what basis the comparison methods have been defined. These methods are mentioned in the Results but without any strong rationale (see also below).

Second, the Results section presents many findings that are not strictly “results”. They are more qualitative descriptions that do not add any real contribution to the study. I think the Results should focus only on the comparison of the pipelines, which is the aim of the manuscript.

Partially, it is also due to the fact that the comparison methods seem quite arbitrary: there is not a motivation about how and why the Authors decided to use the correlation with the HRF. Is this choice supported by some functional hypothesis? Are there other papers that adopt the same methods?

On one side, I understand that it is hard to objectively compare the results of the three pipelines, having not a ground truth about what an ideal signal would look like.

However, the methods adopted rely on many - explicit or implicit - assumptions and so they appear quite arbitrary. For instance: why using the ROI? Is it correct to expect the same (temporally and spatially) brain activity response on all the ROIs? Are all ROIs involved in the processing of the stimulus? What makes the correlation with the expected HRF an appropriate quality measure?

Finally, it is partially because it is not completely clear what is the object of the evaluation.

Is it the use of short-channels, or the GLM-PCA algorithm? Or both?

If the short-channels: what about using other preprocessing algorithms?

If the algorithm: what about using the GLM-PCA on the “usual” channels?

If both: why both? And why was the GLM-PCA chosen?

- The Authors should try to better organize the manuscript and adequately motivate the methods adopted.

- The purpose of the study should be better identified and the Results section should focus only on the presentation of the quantitative results.

Then there are some major and minor issues.

Major

- The preprocessing procedure involved a step in which some channels were removed. As this step operates on each subject independently, it is important to state how many channels have been used in the downstream analysis (see Bizzego et al. 2020);

- Short channels are used to different sources of noise from the standard channel, as mentioned in the Introduction. However, in this study the Authors adopt a number of countermeasures to avoid and remove noise: they ask the subjects to remain still and they filter the signals to remove noise. The Authors should comment on that, as the adopted countermeasures could make useless the use of short-channels and other processing procedures.

- Why only the first two principal components have been used? Why two? This choice should be better motivated.

- Figure 2: it seems that the standard channels in the short-channel “track” undergo two bandpass filters: the first with 0.1 - 1.5 Hz, and the second 0.01-0.1 Hz. But this is not mentioned in the manuscript. Was the first filter really applied? If so, then the two filters would actually filter al the frequency components, as the bands do not overlap.

- Why were the channels merged into ROIs? How were these ROIs created? Do they group channels according to the Brodmann Areas? How many channels have been used for each ROI and participant?

- L. 291-293: why the multilevel linear model? What is the hypothesis that is tested? Any reference from the literature? This sentence is also unclear.

- L. 298-300: what is this follow up contrast analysis? Contrast between what? If reported then it should be described and motivated in the Methods section.

Minor

- L. 94: I think that what is compared should be the efficacy of the pipelines and not the “quality” of the signals. The quality of the signal is something that is given, the pipelines can be more or less robust to low quality signals.

- L. 119-120: “A single source …” This sentence is not clear

- L. 134-135: I think “optimization” is not an appropriate term. Optimization usually indicates a procedure where the parameters of an algorithm are tuned to maximise the efficacy.

- L. 156-162: the description of this step could be more accurate. Maybe “derivative” is more appropriate than “deviation”; I think the Authors also used the absolute values of Xi

- L. 171-173: This sentence should motivate the use of ‘4’ as the minimum number of short-channels. But it is quite unclear. What do the Authors mean by ‘low cost budget”? Why is budget important here?

The following comments are on parts of the text that I think could be removed. In case the Authors decide to put them somewhere else (e.g. Supplementary Information)

- L. 219-221: How the Authors verified this? I think this should be presented as an hypothetical explanation.

- L. 225-228: This is not a finding as the threshold was specifically set to obtain these results

- L. 237-245: the Methods section does not mention the group-averages on the short-channels; this part comes from nowhere. Why is the readed introduced to these results? What is the purpose of showing them? If reported, they should be better motivated and logically included in the procedure.

References

Bizzego et al. (2020). Commentary: Current status and issues regarding pre-processing of fNIRS neuroimaging data: an investigation of diverse signal filtering methods within a general linear model framework. Frontiers in Human Neuroscience

doi: 10.3389/fnhum.2020.00247

Reviewer #3: This paper is concerned with a comparison between two methods of removing systemic influence from the fNIRS signal. The authors have compared the anti-correlation approach with a PCA method based on short channels. This is a useful paper for researchers working with fNIRS, especially as some aspects of signal treatment and fNIRS quantification remain in development.

In this particular paper, 23 participants listened to sentences as the primary experimental task. There is no description of the sentences/stimulus material. I can see how long each lasted for and the temporal characteristics of stimulus presentation, but I would like some concrete examples of the material and confirmation that no motor response was required from the participant.

Figure 2 provided a clear presentation of the different treatments of the fNIRS data that form the basis for this paper. However, there was one aspect that I did not understand and I'm concerned that it may confound the comparison that the authors wish to make. The original data and anti-correlation data are subjected to the same signal processing as the GLM_PCA route except for a bandpass filter of 0.1-1.5Hz that was only applied to the GLM-PCA paper. What was the purpose of this filter? Was it only applied to the 8mm (short channels)? It was not clear to me from reading the paper, why this filter had been applied to that particular data set and not the other two.

The authors are very diligent with respect to filtering the signal quality utilising the SCI method developed by Pollonini et al and showing the results in Figure 3. On page 10, they state that the original method used <0.75 as an exclusion criterion, which they felt was too strict, so instead they opt for a cut-off frequency of 0.15, which is even lower. This part of the text does not make sense and should be corrected, in addition, the authors should justify the basis of their adjustment to the original exclusion criterion - not simply say that it was too strict, i.e. if they had used it, how much data would have been lost?

I would also like more details of the PCA described on page 11. My reading of the text suggests that the PCA is based on all short-channel data, or was the PCA specific to location or hemisphere? This was not clear to me. I also did not quite understand how regressors were calculated, did the PCA represent HbO or Hbb or the HRF function? Finally, the authors extracted the largest two PCs as regressors for the GLM, but I would like to know what proportion of the total variance was accounted for by these 2 PCs.

I was also very unclear about how the ROIs had been derived from the montage and the text on this topic at the bottom of page 13 should be expanded to make this clear. Perhaps the authors could indicate the ROI on the montage diagram shown in Figure 1?

When it comes to the comparison between the three data sets, I felt that the authors relied too heavily on data visualisation. The authors make the claim that GLM-PCA method reduced the suppressive component and we can see that in the plots, or at least we can for HbO, the deoxygenated signal is more difficult to see. Their method of testing correlation with HRF is OK but a little indirect. Other authors, e.g. Fairclough et al (2018) who have compared anti-correlation methods with raw data have done by comparing effect sizes in response to independent variables across different ANOVA. I would prefer the authors to adopt a similar approach by comparing activation during pre-stimulus period with activation during the presentation of stimuli via ANOVA. If the authors are concerned about sphericity, they can test each data set individually. I feel that this is important because most researchers want to know if these methods will make their data more sensitive to experimental effects - and this aspect is not really addressed in the current manuscript.

As a final point, I would like the authors to consider in the Discussion whether their results would generalise to more complex tasks that involve more head movement and a motor component. They use a task that is very controlled (no response, minimal head movement) and they should consider whether their results would generalise to different types of task.

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Attachment

Submitted filename: Reviews.pdf

Click here for additional data file.^{(292KB, pdf)}

PLoS One. 2020 Dec 23;15(12):e0244186. doi: 10.1371/journal.pone.0244186.r002

Author response to Decision Letter 0

5 Sep 2020

Please see our point-to-point responses to reviewers, submitted as a separate word document (Response to Reviewers.docx), attached at the end of the submission/revision.

Attachment

Submitted filename: Response to Reviewers.docx

Click here for additional data file.^{(40KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0244186.r003

Decision Letter 1

Manabu Sakakibara

10 Nov 2020

PONE-D-20-04747R1

Comparing fNIRS signal qualities between approaches with and without short channels

PLOS ONE

Dear Dr. Zhou,

<o:p></o:p>

The original three experts in the field have carefully reviewed the manuscript entitled, "Comparing fNIRS signal qualities between approaches with and without short channels", their comments are appended below. <o:p></o:p>

<o:p></o:p>

All the reviewers acknowledged that the revised manuscript has been greatly improved; The first and the third reviewer satisfied the revision while the second one pointed out some concerns regarding description and presentation of methods and results. <o:p></o:p>

<o:p></o:p>

I will make the final judgement after receipt of your responses and the appropriate revision. <o:p></o:p>

Please submit your revised manuscript by Dec 25 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

We look forward to receiving your revised manuscript.

Kind regards,

Manabu Sakakibara, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (if provided):

I will make the final judgement after receipt of your response and the appropriate revision.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: (No Response)

Reviewer #3: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: (No Response)

**********

6. Review Comments to the Author

Reviewer #1: The authors have addressed all my comments and recommendations. The paper gained in strength and is clearer now.

Reviewer #2: I think the analysis, results and manuscript have been improved: now the flow from the aims, methods and hypotheses and results is more logical and easier to follow.

I still have some methodological concerns:

- The authors should comment on an implicit assumption that I think it is critical: that an increase of HbO is expected during the stimulation. The CNR measure relies heavily on this assumption and would be pointless if no response is observed. The Authors should provide some references in the literature to support this assumption.

In parallel, instead of using custom ROIs, I would find more reasonable to group channels based on BAs, showing that those associated to the processing of the acoustic stimuli show greater activation.

- In any case, I urge the Authors to describe the rules they followed to compose the ROIs.

The major issue is still about the organization of the contents in the manuscript and readability of the text.

- description of methods in the introduction: I suggest to first describe all methods that do not require short channels, and then those with short channels.

- L98-104 interrupt the logical flow of the description and they raise a point that is never addressed later. Maybe this paragraph can be removed.

- L117-133 this part is very hard to follow, some sentences are also ambiguous or too complicated. I strongly suggest to rephrase and reorganize the structure of this paragraph. I understand that the qualitative description of the brain response in short channels is considered of potential interest for the reader, but it is important to give some structure to this part: try to distinguish the different aims and between qualitative and quantitative results. I would suggest to keep the term "data" only to indicate signals or experimental measures. Here the authors probably mean "information"/"description" of the functioning of the short channels.

- L53 and throughout the text: please ensure that the "Ref [x]" is an acceptable format to cite references;

- The description of the stimuli should state how many different stimuli are used, if they were repeated etc. I feel that the complete description of the experimental settings is needed to understand how the data were collected. Some of these details are reported after in the text, but I think it would be clearer to aggregate these details here. This should also be the place where the report what are the different experimental conditions and, in particular, why the chose this, and why only one.

- L152-153 The rephrased sentence is still unclear: maybe the authors just wanted to define what is considered to be a "channel" in fNIRS data? If so, also take the opportunity to clarify other points: how many total "regular" channels have been collected for each signal (should be 40: this information can be extrapolated but it is better to explicitly state that); what do they mean with "overlapping" channels: I think (hope) this does not mean that they refer to the same anatomical location (i.e.: same detector-source pair), so probably the Authors mean "near" instead of overlapping.

- I find the description of the montage (L157-168) a bit rough: what is a "bundle" in this case? please avoid ambiguous terms. Figure 1A is not very informative, it can be removed; channels in Figure 1C should be highlighted; Figure 1C and 1D can be merged if different colors are used to highlight channels belonging to different ROIs. Credits of Figure 1B are missing - I think. L166-168: I think that what authors mean is that whenever the detector was placed more than 30 mm from the source, the detector is re-positioned at 30 mm from the source, by means of a custom plastic holder.

- L182-190: as previously commented, I would avoid separating the description of the experimental design from the description of the stimuli. Also the description should be complete and authors should motivate why the only focused on one condition.

- Please use mathematical formulae and symbols; first write the equation, then describe the terms in the equation.

- L205: "0.5 - 1.5 Hz signals" does not mean anything. What do the authors mean? Are these the original signals after the 0.5-1.5 Hz bandpass filtering?

- Setting such a low SCI is quite critical. I would prefer to see the analysis on fewer subjects than a pipeline with such extreme parameter values. I would suggest to include a supplementary material with the results with the original SCI=0.75 threshold.

- The pipeline description can be improved. The authors should separately describe the steps that are applied to all methods (1-5; 6 and 9) and those that are particular of each method (7 / 8 / 9). It is hard for the reader to exactly understand what was done.

- The authors should comment on the fact that a decrease in HbO is observed in short channels. Was it expected? Why? Any similar finding in the literature.

- Figure 3B is not optimal to explain how many regular and short channels were actually used: I would try with an histogram.

Reviewer #3: I would like to thank the authors for their full and considered responses to the points raised in my review. I’m happy with their responses and have no further questions with respect to the revised manuscript.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Sébastien Scannella

Reviewer #2: No

Reviewer #3: Yes: Stephen Fairclough

PLoS One. 2020 Dec 23;15(12):e0244186. doi: 10.1371/journal.pone.0244186.r004

Author response to Decision Letter 1

23 Nov 2020

Please see our point-to-point responses in the document 'Response to Reviewers.docx'.

Attachment

Submitted filename: Response to reviewers.docx

Click here for additional data file.^{(26.1KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0244186.r005

Decision Letter 2

Manabu Sakakibara

7 Dec 2020

Comparing fNIRS signal qualities between approaches with and without short channels

PONE-D-20-04747R2

Dear Dr. Zhou,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Manabu Sakakibara, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

Reviewer #2: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #2: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #2: Yes

**********

6. Review Comments to the Author

Reviewer #2: Dear Authors

Thanks for considering my suggestions and editing the manuscript.

I think that now the contents are more readable and understandable.

I think the question about the "suppressive" patterns in the short-channels is still open and, in my opinion, quite substantial.

The constitutive assumption behind the use of short-channels is that the signal should not be associated with the brain response.

Now, building on the interpretation you provided, based on Kirilina et al 2012, it seems that this is not really true.

I am afraid that a consequence of this is that any time you use short channels, since they present a suppressive response, the "corrected" HbO signal will present a "spurious" increased activation.

I never had any experience with short channels, and I thank you that I learned a lot from this paper.

I also understand that this issue goes beyond the scope of this paper; however, this concern remains.

I think that including a critical discussion about this issue would benefit this study (and the reader): even if it is hard to provide answers, it is better to highlight a question than leaving a black hole.

Thank you and congratulations for this study.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

PLoS One. doi: 10.1371/journal.pone.0244186.r006

Acceptance letter

Manabu Sakakibara

10 Dec 2020

PONE-D-20-04747R2

Comparing fNIRS signal qualities between approaches with and without short channels

Dear Dr. Zhou:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Manabu Sakakibara

Academic Editor

PLOS ONE

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 Fig. Block-average ΔHbO responses in 12 regions of interest (ROIs).

(TIF)

Click here for additional data file.^{(325.7KB, tif)}

S2 Fig. Contrast-to-noise ratios (CNRs) from different methods.

(TIF)

Click here for additional data file.^{(294.4KB, tif)}

S1 Dataset

(XLSX)

Click here for additional data file.^{(107.9KB, xlsx)}

Attachment

Submitted filename: Reviews.pdf

Click here for additional data file.^{(292KB, pdf)}

Attachment

Submitted filename: Response to Reviewers.docx

Click here for additional data file.^{(40KB, docx)}

Attachment

Submitted filename: Response to reviewers.docx

Click here for additional data file.^{(26.1KB, docx)}

Data Availability Statement

All relevant data are within the manuscript and its Supporting information files.

[pone.0244186.ref001] 1.Jobsis FF. Noninvasive, infrared monitoring of cerebral and myocardial oxygen sufficiency and circulatory parameters. Science. 1977;198(4323):1264–7. 10.1126/science.929199 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref002] 2.Ferrari M, Quaresima V. A brief review on the history of human functional near-infrared spectroscopy (fNIRS) development and fields of application. Neuroimage. 2012;63(2):921–35. 10.1016/j.neuroimage.2012.03.049 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref003] 3.Scholkmann F, Kleiser S, Metz AJ, Zimmermann R, Mata Pavia J, Wolf U, et al. A review on continuous wave functional near-infrared spectroscopy and imaging instrumentation and methodology. Neuroimage. 2014;85 Pt 1:6–27. 10.1016/j.neuroimage.2013.05.004 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref004] 4.Scholkmann F, Spichtig S, Muehlemann T, Wolf M. How to detect and reduce movement artifacts in near-infrared imaging using moving standard deviation and spline interpolation. Physiological measurement. 2010;31(5):649–62. 10.1088/0967-3334/31/5/004 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref005] 5.Kirilina E, Yu N, Jelzow A, Wabnitz H, Jacobs AM, Tachtsidis I. Identifying and quantifying main components of physiological noise in functional near infrared spectroscopy on the prefrontal cortex. Frontiers in human neuroscience. 2013;7:864 10.3389/fnhum.2013.00864 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref006] 6.Kirilina E, Jelzow A, Heine A, Niessing M, Wabnitz H, Bruhl R, et al. The physiological origin of task-evoked systemic artefacts in functional near infrared spectroscopy. Neuroimage. 2012;61(1):70–81. 10.1016/j.neuroimage.2012.02.074 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref007] 7.Yücel MA, Selb J, Aasted CM, Lin P-Y, Borsook D, Becerra L, et al. Mayer waves reduce the accuracy of estimated hemodynamic response functions in functional near-infrared spectroscopy. Biomedical optics express. 2016;7(8):3078–88. 10.1364/BOE.7.003078 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref008] 8.Habermehl C, Schmitz C, Koch SP, Mehnert J, Steinbrink J, editors. Investigating hemodynamics in scalp and brain using high-resolution diffuse optical tomography in humans. Biomedical Optics; 2012: Optical Society of America. [Google Scholar]

[pone.0244186.ref009] 9.Tachtsidis I, Scholkmann F. Erratum: Publisher’s note: False positives and false negatives in functional near-infrared spectroscopy: issues, challenges, and the way forward. Neurophotonics. 2016;3(3):039801 10.1117/1.NPh.3.3.039801 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref010] 10.Strangman GE, Zhang Q, Li Z. Scalp and skull influence on near infrared photon propagation in the Colin27 brain template. Neuroimage. 2014;85 Pt 1:136–49. 10.1016/j.neuroimage.2013.04.090 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref011] 11.Hirasawa A, Yanagisawa S, Tanaka N, Funane T, Kiguchi M, Sorensen H, et al. Influence of skin blood flow and source-detector distance on near-infrared spectroscopy-determined cerebral oxygenation in humans. Clin Physiol Funct I. 2015;35(3):237–44. 10.1111/cpf.12156 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref012] 12.Al-Rawi PG, Smielewski P, Kirkpatrick PJ. Evaluation of a near-infrared spectrometer (NIRO 300) for the detection of intracranial oxygenation changes in the adult head. Stroke. 2001;32(11):2492–500. 10.1161/hs1101.098356 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref013] 13.Huppert TJ, Diamond SG, Franceschini MA, Boas DA. HomER: a review of time-series analysis methods for near-infrared spectroscopy of the brain. Appl Opt. 2009;48(10):D280–98. 10.1364/ao.48.00d280 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref014] 14.Erdogan SB, Yucel MA, Akin A. Analysis of task-evoked systemic interference in fNIRS measurements: insights from fMRI. Neuroimage. 2014;87:490–504. 10.1016/j.neuroimage.2013.10.024 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref015] 15.Zhang YH, Brooks DH, Franceschini MA, Boas DA. Eigenvector-based spatial filtering for reduction of physiological interference in diffuse optical imaging. Journal of biomedical optics. 2005;10(1). 10.1117/1.1852552 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref016] 16.Zhang X, Noah JA, Dravida S, Hirsch J. Signal processing of functional NIRS data acquired during overt speaking. Neurophotonics. 2017;4(4):041409 10.1117/1.NPh.4.4.041409 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref017] 17.Zhang X, Noah JA, Hirsch J. Separation of the global and local components in functional near-infrared spectroscopy signals using principal component spatial filtering. Neurophotonics. 2016;3(1). 10.1117/1.NPh.3.1.015004 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref018] 18.Cooper RJ, Seib J, Gagnon L, Phillip D, Schytz HW, Iversen HK, et al. A systematic comparison of motion artifact correction techniques for functional near-infrared spectroscopy. Frontiers in neuroscience. 2012;6 10.3389/fnins.2012.00147 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref019] 19.Santosa H, Hong MJ, Kim SP, Hong KS. Noise reduction in functional near-infrared spectroscopy signals by independent component analysis. The Review of scientific instruments. 2013;84(7):073106 10.1063/1.4812785 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref020] 20.Bauernfeind G, Wriessnegger SC, Daly I, Muller-Putz GR. Separating heart and brain: on the reduction of physiological noise from multichannel functional near-infrared spectroscopy (fNIRS) signals. Journal of neural engineering. 2014;11(5):056010 10.1088/1741-2560/11/5/056010 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref021] 21.Virtanen J, Noponen T, Merilainen P. Comparison of principal and independent component analysis in removing extracerebral interference from near-infrared spectroscopy signals. Journal of biomedical optics. 2009;14(5):054032 10.1117/1.3253323 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref022] 22.Cui X, Bray S, Reiss AL. Functional near infrared spectroscopy (NIRS) signal improvement based on negative correlation between oxygenated and deoxygenated hemoglobin dynamics. Neuroimage. 2010;49(4):3039–46. 10.1016/j.neuroimage.2009.11.050 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref023] 23.Yamada T, Umeyama S, Matsuda K. Separation of fNIRS Signals into Functional and Systemic Components Based on Differences in Hemodynamic Modalities. PloS one. 2012;7(11). 10.1371/journal.pone.0050271 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref024] 24.Wijayasiri P, Hartley DEH, Wiggins IM. Brain activity underlying the recovery of meaning from degraded speech: A functional near-infrared spectroscopy (fNIRS) study. Hearing research. 2017;351:55–67. 10.1016/j.heares.2017.05.010 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref025] 25.Zhou X, Seghouane AK, Shah A, Innes-Brown H, Cross W, Litovsky RY, et al. Cortical Speech Processing in Postlingually Deaf Adult Cochlear Implant Users, as Revealed by Functional Near-Infrared Spectroscopy. Trends in hearing. 2018;22:2331216518786850 10.1177/2331216518786850 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref026] 26.Wiggins IM, Hartley DEH. A Synchrony-Dependent Influence of Sounds on Activity in Visual Cortex Measured Using Functional Near-Infrared Spectroscopy (fNIRS). PloS one. 2015;10(3). [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref027] 27.Lawrence RJ, Wiggins IM, Anderson CA, Davies-Thompson J, Hartley DE. Cortical correlates of speech intelligibility measured using functional near-infrared spectroscopy (fNIRS). Hearing research. 2018;370:53–64. 10.1016/j.heares.2018.09.005 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref028] 28.Jasdzewski G, Strangman G, Wagner J, Kwong KK, Poldrack RA, Boas DA. Differences in the hemodynamic response to event-related motor and visual paradigms as measured by near-infrared spectroscopy. Neuroimage. 2003;20(1):479–88. 10.1016/s1053-8119(03)00311-2 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref029] 29.Gagnon L, Perdue K, Greve DN, Goldenholz D, Kaskhedikar G, Boas DA. Improved recovery of the hemodynamic response in diffuse optical imaging using short optode separations and state-space modeling. Neuroimage. 2011;56(3):1362–71. 10.1016/j.neuroimage.2011.03.001 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref030] 30.Gagnon L, Yücel MA, Boas DA, Cooper RJ. Further improvement in reducing superficial contamination in NIRS using double short separation measurements. Neuroimage. 2014;85:127–35. 10.1016/j.neuroimage.2013.01.073 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref031] 31.Sato T, Nambu I, Takeda K, Aihara T, Yamashita O, Isogaya Y, et al. Reduction of global interference of scalp-hemodynamics in functional near-infrared spectroscopy using short distance probes. NeuroImage. 2016;141:120–32. 10.1016/j.neuroimage.2016.06.054 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref032] 32.Saager RB, Berger AJ. Direct characterization and removal of interfering absorption trends in two-layer turbid media. JOSA A. 2005;22(9):1874–82. 10.1364/josaa.22.001874 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref033] 33.Funane T, Atsumori H, Katura T, Obata AN, Sato H, Tanikawa Y, et al. Quantitative evaluation of deep and shallow tissue layers’ contribution to fNIRS signal using multi-distance optodes and independent component analysis. Neuroimage. 2014;85 Pt 1:150–65. 10.1016/j.neuroimage.2013.02.026 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref034] 34.Gagnon L, Cooper RJ, Yücel MA, Perdue KL, Greve DN, Boas DA. Short separation channel location impacts the performance of short channel regression in NIRS. Neuroimage. 2012;59(3):2518–28. 10.1016/j.neuroimage.2011.08.095 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref035] 35.Scholkmann F, Metz AJ, Wolf M. Measuring tissue hemodynamics and oxygenation by continuous-wave functional near-infrared spectroscopy—how robust are the different calculation methods against movement artifacts? Physiological measurement. 2014;35(4):717 10.1088/0967-3334/35/4/717 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref036] 36.Zhang Q, Brown EN, Strangman GE. Adaptive filtering for global interference cancellation and real-time recovery of evoked brain activity: a Monte Carlo simulation study. Journal of biomedical optics. 2007;12(4):044014–12. 10.1117/1.2754714 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref037] 37.Zhang Q, Strangman GE, Ganis G. Adaptive filtering to reduce global interference in non-invasive NIRS measures of brain activation: how well and when does it work? Neuroimage. 2009;45(3):788–94. 10.1016/j.neuroimage.2008.12.048 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref038] 38.Brigadoi S, Cooper RJ. How short is short? Optimum source–detector distance for short-separation channels in functional near-infrared spectroscopy. Neurophotonics. 2015;2(2):025005-. 10.1117/1.NPh.2.2.025005 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref039] 39.Weder S, Zhou X, Shoushtarian M, Innes-Brown H, McKay C. Cortical Processing Related to Intensity of a Modulated Noise Stimulus—a Functional Near-Infrared Study. Journal of the Association for Research in Otolaryngology. 2018;19(3):273–86. 10.1007/s10162-018-0661-0 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref040] 40.Gregg NM, White BR, Zeff BW, Berger AJ, Culver JP. Brain specificity of diffuse optical imaging: improvements from superficial signal regression and tomography. Frontiers in neuroenergetics. 2010;2. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref041] 41.Yamada T, Umeyama S, Matsuda K. Multidistance probe arrangement to eliminate artifacts in functional near-infrared spectroscopy. Journal of biomedical optics. 2009;14(6):064034–12. 10.1117/1.3275469 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref042] 42.Pollonini L, Olds C, Abaya H, Bortfeld H, Beauchamp MS, Oghalai JS. Auditory cortex activation to natural speech and simulated cochlear implant speech measured with functional near-infrared spectroscopy. Hearing research. 2014;309:84–93. 10.1016/j.heares.2013.11.007 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref043] 43.Goodwin JR, Gaudet CR, Berger AJ. Short-channel functional near-infrared spectroscopy regressions improve when source-detector separation is reduced. Neurophotonics. 2014;1(1). 10.1117/1.NPh.1.1.015002 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref044] 44.Scarpa F, Cutini S, Scatturin P, Dell’Acqua R, Sparacino G. Bayesian filtering of human brain hemodynamic activity elicited by visual short-term maintenance recorded through functional near-infrared spectroscopy (fNIRS). Optics express. 2010;18(25):26550–68. 10.1364/OE.18.026550 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref045] 45.Vigneau M, Beaucousin V, Herve PY, Duffau H, Crivello F, Houde O, et al. Meta-analyzing left hemisphere language areas: phonology, semantics, and sentence processing. Neuroimage. 2006;30(4):1414–32. 10.1016/j.neuroimage.2005.11.002 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref046] 46.Vigneau M, Beaucousin V, Herve PY, Jobard G, Petit L, Crivello F, et al. What is right-hemisphere contribution to phonological, lexico-semantic, and sentence processing? Insights from a meta-analysis. Neuroimage. 2011;54(1):577–93. 10.1016/j.neuroimage.2010.07.036 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref047] 47.Acharya JN, Hani A, Cheek J, Thirumala P, Tsuchida TN. American Clinical Neurophysiology Society Guideline 2: Guidelines for Standard Electrode Position Nomenclature. Journal of clinical neurophysiology: official publication of the American Electroencephalographic Society. 2016;33(4):308–11. [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref048] 48.Shannon RV, Zeng F-G, Kamath V, Wygonski J, Ekelid M. Speech recognition with primarily temporal cues. Science. 1995;270(5234):303–4. 10.1126/science.270.5234.303 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref049] 49.Molavi B, Dumont GA. Wavelet based motion artifact removal for Functional Near Infrared Spectroscopy. Conference proceedings: Annual International Conference of the IEEE Engineering in Medicine and Biology Society IEEE Engineering in Medicine and Biology Society Annual Conference. 2010;2010:5–8. [DOI] [PubMed]

[pone.0244186.ref050] 50.Scholkmann F, Wolf M. General equation for the differential pathlength factor of the frontal human head depending on wavelength and age. Journal of biomedical optics. 2013;18(10):105004 10.1117/1.JBO.18.10.105004 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref051] 51.Delpy DT, Cope M, van der Zee P, Arridge S, Wray S, Wyatt J. Estimation of optical pathlength through tissue from direct time of flight measurement. Physics in medicine and biology. 1988;33(12):1433 10.1088/0031-9155/33/12/008 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref052] 52.Kamran MA, Jeong MY, Mannan MM. Optimal hemodynamic response model for functional near-infrared spectroscopy. Frontiers in behavioral neuroscience. 2015;9 10.3389/fnbeh.2015.00151 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0244186.ref053] 53.Friston KJ, Fletcher P, Josephs O, Holmes A, Rugg M, Turner R. Event-related fMRI: characterizing differential responses. Neuroimage. 1998;7(1):30–40. 10.1006/nimg.1997.0306 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref054] 54.Ferrari M, Mottola L, Quaresima V. Principles, techniques, and limitations of near infrared spectroscopy. Canadian journal of applied physiology = Revue canadienne de physiologie appliquee. 2004;29(4):463–87. 10.1139/h04-031 [DOI] [PubMed] [Google Scholar]

[pone.0244186.ref055] 55.Bliese PD. Within-group agreement, non-independence, and reliability: Implications for data aggregation and analysis. 2000. [Google Scholar]

PERMALINK

Comparing fNIRS signal qualities between approaches with and without short channels

Xin Zhou

Gabriel Sobczak

Colette M McKay

Ruth Y Litovsky

Roles

Abstract

Introduction

Methods

Participants

fNIRS data collection and short channels

Fig 1. Short-channel setting and fNIRS montage for data collection.

Stimuli and experimental design

fNIRS data analysis

Fig 2. Diagram of fNIRS signal processing and systemic response reduction.

Fig 3. Scalp coupling index (SCI) in the short channels and the number of channels involved for data analysis.

Statistical analysis

Results

fNIRS data quality in the shorter and regular channels

Block-average responses in the 8-mm channels

Fig 4. Short-channel responses.

fNIRS responses with and without short-channel subtraction

Fig 5. Block-average ΔHbO responses in 12 regions of interest (ROIs).

Fig 6. Contrast-to-noise ratios (CNRs) from different methods.

Discussion

Conclusion

Supporting information

Acknowledgments

Data Availability

Funding Statement

References

Decision Letter 0

Chelsea Dobbins

Roles

Author response to Decision Letter 0

Decision Letter 1

Manabu Sakakibara

Roles

Author response to Decision Letter 1

Decision Letter 2

Manabu Sakakibara

Roles

Acceptance letter

Manabu Sakakibara

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases