Elevation of intracranial pressure affects the relationship between hemoglobin concentration and neuronal activation in human somatosensory cortex

Abstract During neuronal activation, a local decrease of deoxygenated hemoglobin concentration (deoxy‐Hb) occurs which is the basis of functional brain imaging with blood oxygenation level dependent functional magnetic resonance imaging (BOLD‐fMRI). Elevated intracranial pressure (eICP) has been shown to impair functional deoxy‐Hb changes. This study investigated this effect and its relation to the underlying neuronal activity in the human primary somatosensory cortex (SI). Functional near‐infrared spectroscopy (fNIRS) during somatosensory evoked potentials (SEP) monitoring was performed on 75 subjects during conditions of median nerve stimulation (MNS) and resting state, combined with normal breathing (NB) and eICP by escalating breathing maneuvers (breath holding [BH], Valsalva maneuver with 15 mmHg [V15] and 35 mmHg expiratory pressure [V35]). During NB, fNIRS revealed a typical oxygenated hemoglobin concentration (oxy‐Hb) increase with deoxy‐Hb decrease during MNS enabling SI brain mapping. Breathing maneuvers associated eICP produced a known global change of oxy‐Hb and deoxy‐Hb with and without MNS. When subtracting measurements during resting state from measurements during MNS, neither functional oxy‐Hb nor deoxy‐Hb changes could be recovered while SEPs remained unchanged. In conclusion, Valsalva‐induced eICP prevents oxy‐Hb and deoxy‐Hb changes during neuronal activation in SI. This finding raises questions on the validity of oxy‐Hb‐ and deoxy‐Hb‐based brain imaging (e.g., BOLD‐fMRI) during eICP.

Similarly, notwithstanding this limitation, fNIRS is also being adapted to patient applications, for example, in brain-computerinterfacing (Banville, Gupta, & Falk, 2017). A recent review has put forward the problem that on the one hand observed oxygenation changes in fNIRS signals can be falsely attributed to neuronal activity although their origin is systemic or even extra-cerebral, and on the other hand fNIRS can miss neuronal activity because systemic reactions interfere with oxygenation changes (Tachtsidis & Scholkmann, 2016).
Elevated intracranial pressure (eICP), a condition often associated with brain tumors, has been found to relevantly impair NVC. In a rat model, eICP caused by mock cerebrospinal fluid (CSF) infusion led to an amplitude reduction and-during very high intracranial pressure (ICP)-a reversal of deoxy-Hb changes during primary somatosensory cortex (SI) activation . We recently reproduced this finding in human primary motor cortex (MI) when applying Valsalva-induced eICP during MI activation via finger tapping and finding an amplitude decrease of functional deoxy-Hb changes ultimately leading to a complete abolishment of functional brain imaging with fNIRS when based on deoxy-Hb (Knauth, Heldmann, Münte, & Royl, 2017). In order to reproduce this effect in a different cortical area, and at the same time enabling the obtainment of a direct, that is, electrophysiological, measurement of concurrent neuronal activation we transferred this study to human SI by investigating neuronal activation and NVC responses during median nerve stimulation (MNS) with different breathing maneuvers leading to changes in ICP.

| MATERIALS AND METHODS
The study was approved by the ethical committee at the University of Lübeck and conducted in accordance with the Declaration of Helsinki (64. WMA General Assembly, October 2013, Fortaleza, Brazil) and the understanding and informed consent of each subject prior to their inclusion in the study.

| Subjects
Seventy-five healthy volunteers (mean age 23.5 years; range 18-47 years; 47 women; 66 right-handers) with no history of neurological or pneumological disorders participated in the study.

| Electrophysiological measurement
In order to obtain a direct measurement of neuronal activity during MNS, somatosensory evoked potentials (SEP) were recorded as applied in clinical routine. Two EEG electrodes were placed on the subject's scalp at the C3 0 = CP3 (right-hand stimulation) and C4 0 = CP4 (left-hand stimulation) position above the postcentral gyrus, 3 cm occipital and 7 cm lateral of Cz, according to the 10-20 system for EEG (Nuwer et al., 1998) and fixed with 3 × 3 cm white plaster. A reference electrode was placed on the forehead and a ground electrode a few centimeters proximal of the stimulation block above the carpal tunnel. Impedances were kept below 5 kΩ. The electrodes were connected to the measuring unit of the applied ISIS IOM System (OSIRIS Neurostimulator, NeuroExplorer software version 4.4.9.0, INOMED, Emmendingen Germany) that also triggered MNS. SEPs were averaged within each trial and recorded to the hard disk of the connected computer.

| fNIRS measurement
Measurements of hemoglobin (Hb) changes were performed with an optical topography system (ETG-4000, Hitachi Medical Corporation, Tokyo Japan) using continuous-wave fNIRS (Kamran, Mannan, & Jeong, 2016). A total of 30 optodes (16 emitters alternating with 14 detectors) was attached to an elastic head cap and arranged as two 3 × 5 grids with an area of 14 × 19 cm each covering the left and right hemisphere's SI (Figures 1 and 2). The optode grid was adjusted so that the EEG electrodes for SEP measurement always stood in relation to the same optodes (C3 0 between optode 16 and 17, C4 0 F I G U R E 1 Optode grid. A total of 30 optodes (16 emitters alternating with 14 detectors) was attached to an elastic head cap and arranged as two 3 × 5 grids with an area of 14 × 19 cm each covering the left and right hemisphere's SI. From these 2 × 15 optodes 2 × 22 detector channels were generated, located halfway between emitter and detector F I G U R E 2 Experimental setup. A blood pressure cuff was connected to a 10 ml plastic syringe. By blowing into the syringe the test person-produced endexpiratory pressures of 0 mmHg, 10-20 mmHg, or 30-40 mmHg, depending on the current breathing task. Each task was indicated on a computer screen in front of the test person. MNS was applied manually by the experimental examiner when required during protocol (here shown for left hand MNS). A: stimulation electrode, B: ground electrode, C: diverting electrodes C3 0 (right hand MNS) and C4 0 (left hand MNS), D: reference electrode, E: optode bundle, F: Valsalva device, G: water available during resting periods between optode 27 and 27, Figure 1). With this positioning, which was based on previous studies' findings examining the relation between the international 10-20 system and underlying cortical areas, the SI (postcentral gyrus) was covered (Okamoto et al., 2004;Sato et al., 2006;Steinmetz, Fürst, & Meyer, 1989;Towle et al., 1993). From these 2 × 15 optodes 2 × 22 detector channels were generated, located halfway between emitter and detector. The Hb absorption coefficient is most sensitive to measuring the Hb concentration in spectral bands centered at 700 and 960 nm (Strojnik & Paez, 2013). The applied ETG-4000 system, therefore, measured absorption changes at the omitted light at 695 nm (dominated by deoxy-Hb) and at 830 nm (dominated by oxy-Hb). The inter-optode distance was 30 mm. The sampling frequency was 10 Hz. Changes of oxy-Hb and deoxy-Hb were calculated by a built-in software using a modified Lambert Beer algorithm (Cope & Delpy, 1988;Kamran et al., 2016).

| MNS
Participants were instructed to hold the stimulated hand still in their lap during measurement and to move as little as possible (especially avoidance of jaw and head movement) to avert noise from muscular activity. SI was activated by contralateral MNS at the wrist via two stimulation electrodes (anode and cathode) with the applied ISIS IOM System ( Figure 2). Stimulation trains with rectangular pulses (sampling rate 10,000 Hz, time resolution = 0.1 ms, 800 measuring points per 20 s, amplification 100, 30 Hz HP-filter, notch filter from 5 to 600 Hz) were applied for 20 s with a stimulus frequency of 4.3 Hz. The motor threshold (mean 10.8 mA, range 7-25 mA) was individually identified by visible activation of the Musculus abductor pollicis brevis and then increased for another 3-4 mA to achieve a reliable above-threshold stimulation (Buchner et al., 2014). A minimal resting interval of 40 s was maintained between two successive stimulation trains. Before starting the experiment, a test run of MNS including a reproduction with 200 summations each was performed (Vogel, 2006). If the test trials produced noisy curves, minimal modifications of the location of the stimulation electrodes and/or EEG electrodes were made in order to optimize both, MNS and SEP recording. Minimal N20 amplitude of 1.0 μV was ensured.

| Breathing tasks
Previous studies have shown that breathing against resistance temporarily increases ICP (Brimioulle, Moraine, Norrenberg, & Kahn, 1997;Knauth et al., 2017;Prabhakar, Bithal, Suri, Rath, & Dash, 2007). Four different breathing maneuvers were performed by each participant to modify ICP during the experiment: (a) normal breathing (NB), (b) breath holding for 20 s with 0 mmHg forced expiratory pressure (BH), (c) weak Valsalva maneuver (VM) with 15 mmHg forced expiratory pressure against resistance (V15), and (d) strong VM with 35 mmHg forced expiratory pressure against resistance (V35). To control VMs' intensity, the forced expiratory pressure was measured with a self-constructed device (10 ml plastic syringe connected to a blood pressure cuff manometer [Pstras, Thomaseth, Waniewski, Balzani, & Bellavere, 2016]). The subject was asked to take a deep breath, and then blow into the syringe maintaining the target pressure for the task's duration. A range for the two pressure levels (10-20 mmHg = V15, 30-40 mmHg = V35) was tolerated since maintenance of one exact pressure level is difficult to achieve for most people. During resting periods, subjects were allowed to breathe freely. For BH subjects were asked to inhale before and to keep their airways open during the breathhold, in order to not build up a higher expiratory pressure. For all breathing maneuvers task compliance was monitored by the experimental examiner. Before starting the experiment, a training session was inserted to familiarize participants with different breathing tasks.

| Partial pressure of end-tidal carbon dioxide after breathing maneuvers
In order to assess the contribution of hypercapnia to the observed changes, additional experiments with Partial pressure of end-tidal carbon dioxide (PetCO 2 ) measurements were performed. For this purpose, we employed a calibrated Evita XL Respiratory Ventilator from our intensive care unit (Drägerwerk Co.KGaA, Moislinger Allee 53-55, Lübeck, Germany). On five additional subjects (three women; age range 23-46 years) the same experimental protocol was applied with randomized breathing maneuvers analogue to fNIRS experiments (see PetCO 2 and environmental CO 2 , the mouthpiece of the ventilating system was occluded during breath holding. Analogue to the experimental protocol described above, the subjects performed NB, BH, V15, and V35 10 × each. PetCO 2 was written down by the examiner during the 2 s preparation time before and after each task. Due to the ventilating system's reaction time to PetCO 2 changes, values after breathing maneuvers were averaged from the two breaths after the ending of each task.

| Data analysis and statistical testing
Data were analyzed with custom-written scripts using MATLAB ® R2018a (The MathWorks, Inc.). Primarily, the data sets were analyzed for each subject individually. Later, a group analysis of all collected data sets was performed. Since the goal of our study was to investigate changes during eICP with a high signal-to-noise ratio (SNR), only subject data with a relevant and robust NVC response during NB were included in further analysis. Active measuring channels were selected for each participant by displaying the oxy-Hb's, deoxy-Hb's, and total-Hb's mean time course during NB + MNS (in total 10 × 20 s per subject) for all 22 channels. If NB + MNS showed a sufficient hemodynamic response (oxy-Hb increase, deoxy-Hb decrease, total-Hb increase [Obrig, 2002], by at least 1.5 SD), the channel was included into further analysis. In case a subject showed more than one relevant channel, all time-course data of the active channels were averaged before further analysis.

| Hemoglobin concentration-time courses in SI
After subtracting baseline values from 11 to 0 s before the beginning of each condition, concentration-time courses of oxy-Hb, deoxy-Hb, and total-Hb were first averaged for each of the seven tasks across all trials within each subject and then plotted as mean with a ± 95% confidence interval across all 44 subjects. To control for systematic data trends, corresponding concentration changes during four consecutive time intervals of 66 s were taken from the resting period conditions and averaged. It is known that VMs induce large global oxygenation changes (Wu, Bandettini, Harper, & Handwerker, 2015). Therefore, an additional subtraction analysis was performed to extract the functional oxygenation signal from recorded global changes. In detail, within each subject, the averaged Hb time courses for breathing tasks without MNS were linearly subtracted from corresponding averaged Hb time courses for breathing tasks with simultaneous MNS.

| SEP-analysis
SEPs were averaged for each condition across all trials within each subject and then averaged across all 44 displayed data sets. In order to account for SEP differences due to tissue resistance (skin and skull), The main part comprised five blocks with seven tasks each in random order. NB, normal breathing, MNS, median nerve stimulation, VM, Valsalva maneuver, BH, breath holding normalization was performed before averaging across subjects. Averaged within-subject SEP time courses were divided by the subject's maximum SEP amplitude. Following the experimental protocol, every data set aimed at averaging across 3 × 5 SEPs (BH, V15, and V35) and 1 × 10 SEPs (NB) for the respective condition. In 12 of 64 experiments, SEPs from individual stimulation trains (17 out of the totally planned 1,600 trains across all 64 experiments) could not be saved due to a technical failure and were therefore interpolated from the remaining obtained SEPs of the same condition when averaging.

| Statistical analysis of time-course data
To statistically assess Valsalva-induced eICP's influence on functional responses to MNS, the resulting mean amplitude of deoxy-Hb, oxy-Hb, and total-Hb during MNS (9-20 s after initiation of 20 s right hand MNS) and the mean SEP amplitude were calculated for each condition. A paired t test (including Bonferroni correction) was applied for testing differences between baseline conditions and breathing tasks. Results with p < .05 were considered statistically significant, results with p < .0005 were considered highly significant.

| Functional imaging data
While the analysis explained in Section 2.8.3 is suitable to assess the and delivered discernible hemodynamic SI responses contralaterally for both sides. Figure 4 illustrates the number of channels with functional hemoglobin changes during right (a) and left (b) hand MNS. The additional data from left hand MNS were used to assess functional imaging but were not included in the time course analysis to avoid a statistical bias since these data were not obtained from independent samples.

| Cortical hemoglobin concentration time courses during MNS and eICP
Time courses of regional deoxy-Hb, oxy-Hb, and total-Hb changes  During all breathing tasks, the typical NVC response is not discernable. Instead, a known large polyphasic oxy-Hb and total-Hb change can be observed. A first positive peak becomes apparent 2 s after the onset of the breathing maneuver, which increases in amplitude with escalation of the breathing maneuver (0.055 mM*mm during BH, 0.1 mM*mm during V15, 0.17 mM*mm during V35). At 8 s after the onset of the breathing maneuver, a first minimum occurs (−0.18 mM*mm after BH). This minimum shows a trend toward a larger amplitude during V15 (−0.27 mM*mm) and a smaller amplitude during V35 (−0.2 mM*mm). When applying V15 and V35, it is followed by a second maximum immediately after the end of the breathing maneuver which is of negative amplitude after V15 (−0.1 mM*mm) but reaches positive values during V35 (0.2 mM*mm).
A second minimum 2 s after the end of the breathing maneuver is seen after V15 (−0.18 mM*mm) and has a smaller amplitude after V35 (−0.11 mM*mm). Finally, before recovery to baseline values, a third maximum is reached 8 s after the end of the breathing maneuver that increases in amplitude with escalating breathing maneuvers

| Statistical analysis of time course data
To further assess the NVC response during eICP, the mean amplitude of deoxy-Hb, oxy-Hb and total-Hb at the maximum response (9-20s after MNS onset) as well as the mean SEP amplitude (absolute values, that is, positive and negative values) of each breathing condition were analyzed across subjects (n = 44). For BH, V15, and V35 the concentration time courses without MNS were subtracted from the corresponding concentration time courses with MNS (as described for

| Functional imaging data
Functional neuroimaging displays visible contrasts between areas that differ in their functional state and thus brain maps of activated areas can be obtained. To assess Valsalva-induced eICP's influence on functional neuroimaging based on oxy-Hb and deoxy-Hb, an additional imaging analysis was performed in all subjects in which distinct functional oxygenation changes were obtained during right and left hand MNS (n = 16). In order to obtain a complete image rather than irregularly positioned points, all individual measuring channels were converted to an image matrix before further analysis. To this end, the fNIRS data of each hemisphere's grid (2 × 15 channels) were linearly interpolated, resulting in an image matrix of 5 × 9 pixels for each hemisphere. In each resulting pixel, a pooled t test was performed comparing its mean amplitude during MNS (0-20 s) with its mean amplitude before MNS initiation (−7 to −2 s before MNS) in all subjects. For the BH, V15, and V35 paradigms, the above-mentioned subtraction analysis was performed before this test in order to remove global changes associated with the breathing maneuvers. Figure 7 displays the results of this functional imaging analysis. During NB, both oxy-Hb and deoxy-Hb based functional imaging clearly identifies the activated SI (Figure 7a,e,i,m). However, in none of the conditions involving breathing maneuvers, this brain mapping can be reproduced. Corroborating the time course data, this analysis shows that functional imaging based on oxy-Hb and deoxy-Hb concentration time courses is not capable to localize neuronal activation during Valsalva-induced eICP.
F I G U R E 5 Influence of eICP on vascular responses (time courses of regional oxy-, deoxy-, and total-Hb) and neuronal activity in the SI during MNS. Displayed are time courses averaged across all trials of the respective condition within each subject, then depicted as mean across subjects with a ± 95% confidence interval from 44 left SI areas (right hand MNS) from 44 subjects. Light gray box: preparation time (2 s) depicted on a computer screen, dark gray box: time interval per condition (20 s). (a) During MNS, a typical local hyperoxygenation (increase in oxy-and total-Hb, decrease in deoxy-Hb) becomes apparent in contralateral SI. When accompanied by the breathing maneuvers BH, V15, and V35 (b-d), this vascular response cannot be discerned. Instead, a large global multiphasic response known to be associated with the applied breathing maneuvers is seen. Corresponding vascular responses without MNS (f-h) follow a very similar time course. (i) contains the same data as (a) but on a larger scale. When subtracting (f-h) from (b-d) the MNS evoked component of the vascular response should become apparent (j-l). However, although the oxy-Hb time course during MNS + BH suggests a trend toward a small hyperoxygenation, no significant MNS induced local hyperoxygenation can be recovered. At the same time, neuronal activity as measured by SEPs remains largely unchanged during the different breathing maneuvers (m-p). This suggests that oxy-Hb and deoxy-Hb no longer signalize neuronal activity in the SI during eICP 4. Functional imaging with fNIRS correctly localizes the corresponding SI during MNS + NB but fails to do so during breathing maneuver induced eICP. Therefore, functional imaging of the SI based on deoxy-Hb is impaired during eICP.
In the following, we will address methodological issues of our study and discuss possible implications.

| Breathing maneuvers as means of ICP elevation
Experimental invasive ICP elevation by mock CSF infusion has been shown to reduce and even reverse the functional deoxy-Hb response in rat SI . Translated to patients undergoing brain mapping with BOLD-fMRI this has fundamental consequences as activated areas might not be localized correctly in the possibly damaged brain, for example, in brain tumor patients (Lindauer et al., 2010). When attempting to transfer this effect to and assess it in humans, graded VMs are an obvious noninvasive mean for eICP attainment. Depending on the amplitude of the applied intrathoracic pressure, jugular venous outflow is reduced which causes a displacement of CSF into the cranial cavity (Wostyn, Audenaert, & De Deyn, 2009) and elevates ICP (Williams, 1981). It should be noted that the applied BH maneuver, although formally not a VM, would also lead to a small increase of intrathoracic pressure with a probable small increase in ICP (Thomason & Glover, 2008;Wu et al., 2015).
Since breath holding and VMs increase alveolar CO 2 concentration, the effects of relative arterial hypercapnia might interfere or contribute to the observed changes. For this reason, we performed additional F I G U R E 6 Box plots across experiments (44 SI areas from 44 subjects) with mean amplitudes of functional responses during MNS. During NB, deoxy-Hb (a), oxy-Hb (b) and total-Hb (c) show MNS-evoked concentration changes corresponding to a typical hyperoxygenating NVC response. With all three escalating breathing maneuvers (NB, V15, and V35), this response is abolished with significant alterations of the concentration change during MNS (*p < .05, **p < .0005, Paired t test with Bonferroni correction). (d) shows relative SEP amplitudes (absolute values) during NB and during the different breathing maneuvers with no apparent change signifying preserved neuronal activity during eICP PetCO 2 measurements confirming a PetCO 2 increase by about 10% or 4 mmHg evoked by breath holding maneuvers. This raises the question to what extent the effects observed in our study can be attributed to an arterial PetCO 2 increase. With a rise of PetCO 2 by 6 mmHg, a reduction of functional CBF increases by about 30% can be expected (Whittaker, Driver, Bright, & Murphy, 2016). Considering the obtained fNIRS responses to MNS (Figure 5i), a reduction of this magnitude would still be detectable in our data analysis. Thus, the observed changes exceed the magnitude expected from a PetCO 2 increase by 4 mmHg. In addition, the temporal profile of our fNIRS responses to the breathing maneuvers argues against a sole CO 2 effect. On the one hand, the gradual return of oxy-Hb and deoxy-Hb to baseline values after the breathing maneuvers observed in our experiments could suggest a gradual washout of accumulated arterial pCO 2 . On the other hand, the complex multiphasic response observed during the maneuver has a prompt onset too fast to be attributed to blood gas changes.
Furthermore, an accumulation of arterial pCO 2 over the course of the experiment could be put forward as a relevant confounder of our data.
However, this seems unlikely, considering the return of elevated PetCO 2 to baseline values after the next breathing maneuver in our experimental data. Therefore, while breath holding leads to hypercapnia (Lindholm & Lundgren, 2006;Sasse, Berry, Nguyen, Light, & Mahutte, 1996;Seddon, Thacker, Jurd, & Loveman, 2014), eICP effects probably outweigh hypercapnia effects. However, future fNIRS and fMRI studies should include PetCO2 monitoring, as it can change not only by forced breathing maneuvers: a speech task-induced mild PetCO 2 decrease by 4% produced relevant cortical oxygenation changes in a recent fNIRS study that could be falsely attributed to neuronal activation (Scholkmann, Gerber, Wolf, & Wolf, 2013).
Another confounder of our eICP paradigm stems from its transient characteristic, that is, the 20 s breathing maneuver does not allow for a new equilibrium of total-Hb, oxy-Hb, and deoxy-Hb to take place. However, the advantage of this approach with fast recovery is that different eICP levels can be applied in randomized order, which in turn increases the validity of the observed effect. By subtracting trials without parallel MNS from trials with MNS we were able to examine the effect of breathing maneuvers. Confirming our previous study's findings in the human MI (Knauth et al., 2017) and the initial animal study's results  we observed an eICP correlated decrease in the deoxy-Hb response. Taken together, though the results of this study differ in some ways (discussed below), we are confident that Valsalva-induced eICP is an appropriate mean to study the effect of eICP on NVC related changes in oxy-Hb and deoxy-Hb in human SI.
While the alignment of experimental protocols between our previous study in the human MI (Knauth et al., 2017) and our current study on SI allows for comparability, it also leads to a common limitation. A VM inevitably leads to global extracranial blood flow changes, for example, from the scalp, possibly interfering with fNIRS measurements. This problem is discussed in more detail in our previous study on MI (Knauth et al., 2017). Briefly summarized, while superficial scalp veins have been shown to contribute to fNIRS signals (Kirilina et al., 2012), fNIRS measurements with different penetration depths found a rather small contribution from superficial layers (Tsubaki, Kojima, Furusawa, & Onishi, 2013) during VMs. In addition, a combination of NIRS with Laser Doppler Flowmetry did not reveal a relevant influence of surface vasomotor activity on cortical NIRS data (Aletti et al., 2012). Therefore, although we cannot exclude a small contribution of extracranial oxygenation signals the predominant fNIRS signals from our study are most likely of cortical origin. Nevertheless, the used NIRS setup with only one penetration depth constitutes a limitation of our study. Future studies on systemic changes should be done with optimized NIRS systems which are able to discriminate signals from scalp flow and other signals not related to cortical blood flow.

| Dropout rate and SNR
Since we aimed to investigate pathological influences on physiological NVC our experimental setup was dependent on a strong and T A B L E 1 Paired t test (visualized in Figure 6) Cerebral blood volume ≙ total hemoglobin concentration. p-values were determined and Bonferroni corrected (primary p-values multiplied by four) for 12 tests concerning relative changes in deoxy-Hb, oxy-Hb, and total-Hb, of which 5 were significant (*p < .05) and 3 were very significant (**p < .0005). For deoxy-Hb all applied tests (4) except for BH versus V15 were significant or very significant, for oxy-Hb three out of four tests were and for total-Hb two of the applied tests were statistically significant. In detail, the differences between deoxy-Hb and oxy-Hb amplitudes for NB conditions and conditions with eICP (BH, V15, and V35) were all significant. The total-Hb amplitude showed a significant difference between NB and V15 as well as between NB and V35. This underlines the time course analysis' finding that functional deoxy-Hb amplitudes are significantly smaller during V15 and V35 when compared to NB and BH. The oxy-Hb's and total Hb's mean amplitude also gets smaller with the escalation of breathing tasks. These findings support the theory of a deoxy-Hb decoupling mechanism with an increasing positive difference and the passing of the zero line during the breathing maneuvers' progression.
unambiguous cortical MNS oxygenation response during NB. If such a response could not be detected during baseline conditions with NB, an assessment under nonbaseline conditions (= eICP) would not have rendered value to this endeavor. Therefore, the original data from 113 experiments were confined to 64 which were further analyzed.
Forty-nine sets were excluded because no discernable oxygenation change was observed during MNS + NB after averaging across trials post hoc. The dropout rate was therefore even higher than in our previous study on the human MI where 13 from 59 experimental subjects were excluded (Knauth et al., 2017). Such a preselection based on SNR bears the risk of systematic bias. Our paradigm only included 10 repetitions for MNS + NB which might have been too few for many experimental subjects to achieve a sufficient SNR. In theory, additional averaging by adding more experimental blocks could have increased the number of experiments with an unambiguous and clear oxygenation change during MNS. However, due to the exhausting VM paradigm, a corresponding lengthening of the experimental duration was not a practical option. Instead, we increased the number of experimental subjects. We also considered the alternative of applying an event-related design instead of a block design to achieve a higher SNR (Klingner et al., 2011). However, we wanted to ensure sufficient comparability to the analogue study on the MI (Knauth et al., 2017).
The encountered heterogenous SNR is not unexpected. fNIRS studies have described multiple SNR impairments, with assumed explanations including age, cooperation, anatomical characteristics such as skull thickness, inaccurate placements of NIRS optodes, wavelength dependence of sensitivity and CSF layers, physiological phenomena relating to brain activation and systemic changes in blood pressure or heart rate (Moriguchi et al., 2017;Sato et al., 2005;Strangman, Franceschini, & Boas, 2003). In summary, we are convinced that our confining of the study population to subjects with high SNR does not distort the results of our study but instead raises their validity.

| Valsalva-induced eICP and multiphasic oxy-Hb and deoxy-Hb changes in SI
One main goal of this study was to examine whether the effect observed in a previous study employing Valsalva-induced eICP on NVC in the human MI (Knauth et al., 2017) could be reproduced in the human SI. To this end, the experimental protocols were equalized.
This enabled us to compare the effect of Valsalva-induced eICP on oxy-Hb and deoxy-Hb in the SI to the corresponding effect in the MI. The Valsalva-evoked response profile and magnitude of oxy-Hb, total-Hb, and deoxy-Hb were very similar in the two studies, confirming our interpretation of this response as a global stimulus affecting several cortical areas similarly. Oxy-Hb and total Hb followed a multiphasic response profile with three peaks and two troughs.
F I G U R E 7 Effect of ICP elevation via breath holding tasks on fNIRS imaging. Interpolated from the bilateral 3 × 5 optode grid an interpolated bilateral 5 × 9 image matrix with 45 pixels was created for each hemisphere and every condition. Each pixel was subjected to a pooled t test comparing its mean amplitude during MNS (0-20s) with its mean amplitude before MNS (−7 to −2 s before MNS initiation) in all subjects ( (Tsubaki et al., 2013) and BOLD-fMRI (Henderson et al., 2002) (Niederhauser, Rosenbaum, Gore, & Jarquin-Valdivia, 2008) while another fNIRS study reported a missing response in 3 of 10 subjects and inconsistent data in three of the remaining subjects (Tanosaki et al., 2001). Another study did not report any subjects with failed SI activity but showed data of only medium SNR (Takeuchi et al., 2009).
When using fMRI during MNS one study did not obtain functional activity in two of nine subjects, while all nine subjects showed significant activation during a finger opposition task (Spiegel, Tintera, Gawehn, Stoeter, & Treede, 1999). Another fMRI study found MNS activation in the SI to be considerably dependent upon attention, although it is not clear whether this BOLD signal variance was due to differences in neuronal activity since no parallel EEG recording was obtained (Backes, Mess, van Kranen-Mastenbroek, & Reulen, 2000).
Moreover, one study suggests a frequency dependence of the obtained fMRI responses upon MNS (Kampe, Jones, & Auer, 2000). It has been demonstrated that sensory stimuli can evoke EEG responses without concurrent positive BOLD-fMRI response. On the contrary, a decrease in BOLD contrast was recorded and interpreted as a focal deactivation due to a reduced neuronal baseline activity induced by inhibitory interneurons (Blankenburg et al., 2003). While we were able to recover functional oxygen changes in the motor cortex by subtracting responses to BH without responses with finger tapping (Knauth et al., 2017), only a weak trend toward functional oxygenation changes remains when applying BH to MNS and SI. It is therefore tempting to speculate that breathing differences also contributed to the variance in the previously mentioned studies and the general problem of acquiring robust MNS induced functional Hb responses.
Variance might be reduced by training subjects to pace their breathing and-if breath holding is applied-to hold breath after exhalation instead of inspiration (Handwerker, Gazzaley, Inglis, & D'Esposito, 2007).

| Functional brain imaging of SI based on oxy-Hb and deoxy-Hb changes
The analysis of eICP induced changes on oxy-Hb and deoxy-Hb during MNS was accomplished with two different methods. In one approach, averaged time course data of activated fNIRS channels were analyzed with statistical tests on amplitude changes as described and discussed above. In the second approach, we assessed functional imaging data obtained from a subgroup of 16 subjects in which MNSinduced changes of oxy-Hb and deoxy-Hb were obtained for both left-and right-side stimulation. By performing this analysis, we were able to ensure a maximum contrast of functional cortical changes between two conditions (left-side and right-side) and thus investigate the behavior of this contrast as a surrogate of functional imaging capability during Valsalva-induced eICP. In addition, a possible displacement of the activated volume during eICP (e.g., by minor shifts of the cortex or blood redistribution within the different arteriovenous compartments) as the explanation for the encountered amplitude changes could be made visible. However, corroborating the time course results, functional maps of the correct SI were obtained only during NB (Figure 7a,e,i,m). With parallel breathing maneuvers, no functional map could be recovered which makes a displacement of the origin of the Hb signal improbable. This finding applied to both, oxy-Hb and deoxy-Hb, which again is in contrast to our previous study that employed the same eICP paradigm on human MI during finger tapping and found a preserved oxy-Hb functional map during eICP with an abolished functional imaging capability for deoxy-Hb. When interpreting this disparity, the same points apply as in the discussion of the time course data. In summary, with regard to the sustained neuronal activity visible in the SEPs we encountered, we propose that deoxy-Hb and oxy-Hb do not indicate neuronal activation during eICP. This finding questions the validity of brain mapping based on these blood components in patients with eICP, for example, fMRI in brain tumor patients. However, the transfer of this study's results to patients with eICP cannot be done in a straightforward manner for two main reasons. Firstly, hypercapnia is a possible contributing factor in the applied VM but not in patients with eICP due to disease conditions. Secondly, as opposed to patients with disease-related eICP, our paradigm only involved eICP transiently. Nevertheless, if this finding can be reproduced in future studies involving patients with chronic eICP, imaging methods based on deoxy-Hb should be critically questioned toward their validity (Lindauer et al., 2010) and their ability to protect eloquent brain areas during surgery. Alternative imaging methods are, for example, navigated transcranial magnetic stimulation (Picht, Frey, Thieme, Kliesch, & Vajkoczy, 2016) or direct intraoperative electrical mapping (Kuchcinski et al., 2015) that do not depend on NVC. Our results also give reason to critically accompany the recent advance of fNIRS into the field of brain-computer interfaces (Banville et al., 2017) since many applications rely on the presupposition of a measurable and valid NVC.

| Perspective of a noninvasive ICP estimation
Our previous study with a gradual amplitude reduction of finger tapping induced deoxy-Hb in the human MI during eICP gave rise to the outlook that this effect might be used to estimate ICP non-invasively.
Such a neuromonitoring method would be a relevant advance in neuromonitoring of neurocritical care patients. When deciding to do a parallel study during MNS we also wanted to transfer this paradigm to a functional activation method that is not dependent on cooperation and therefore can be applied to unconscious patients. Regarding the final results, we doubt that neuromonitoring by MNS will offer the possibility of ICP estimation. However, future studies should employ different, more robust stimulation paradigms, for example, visual stimulation, to further investigate this clinically relevant approach.

| CONCLUSIONS
When applying eICP by escalating breathing maneuvers to human SI activation by MNS we encountered a significant reduction and even abolishment of both oxy-Hb and deoxy-Hb responses with preserved neuronal activity as measured by EEG. Thus, this study adds evidence to the hypothesis that eICP decouples Hb changes from the neuronal activity and therefore renders methods like BOLD-fMRI and fNIRS blind to neuronal activation.