The assessment of brain activity through the fNIRS technique relies on the neurovascular-coupling phenomenon, similarly to fMRI. The technique relies on the relationship between neural activity and changes in regional cerebral blood flow (rCBF) affecting oxygenation and hemoglobin content in blood vessels. The physiological and physical principles of the technique have been extensively described elsewhere (Obrig and Villringer, 2003; Scholkmann et al., 2014). Both fMRI and fNIRS are grounded on the model that states that an increase in oxygen metabolism due to neural activity is accompanied by a disproportional increase in rCBF causing a local hyperoxygenation (Fox and Raichle, 1986). This cascade of events is reflected in an increase in the concentration of oxy-hemoglobin and a decrease in concentration of deoxyhemoglobin (Chance and Villringer, 1997). fNIRS takes advantage to the fact that these hemoglobin chromophores are selective absorbers of light in the near infrared part of the spectrum, while biological intervening tissues such as bone and skin are transparent. Therefore, the principle of the technique is to non-invasively quantify the light migrating from sources to detectors positioned in the scalp taking into account these different absorption and scattering properties.

The hemodynamic changes for the present study were obtained from the optical changes collected using two different continuous-wave, functional near-infrared spectroscopy systems (NIRScout16x16 and NIRSport8x8, NIRx Medical Technologies, Glen Head, NY). Only one system has been used per experiment and its choice was primarily based on: (i) number of subjects per run (i.e., single subject or hyperscanning); (ii) number and size of cortical area(s) of interest; and (iii) whether movement of the entire body was expected during the measurement.

The NIRScout system, consisting of 16 LED illumination sources (760 and 850 nm) and 16 fiber optic detectors, was used to collect data in the experiment based on hyperscanning (violin duo) and to achieve coverage of frontal and motor cortical areas (piano playing). The NIRSport wearable system with 8 LED sources (760 and 850 nm) and 8 active detectors was used in the table-tennis and in the continuous monitoring in daily activities experiments. The sampling rate for the experiments was 7.81 Hz, with the exception of the piano measurement, which was 3.91 Hz.

Sources and detectors were placed on the region(s) of interest of the measuring cap (10-5 international system, except for the continuous measurement, which used a frontal headband). Their spatial distribution on the cap was chosen to result in channels (i.e., source-detector pairs) with standard inter-optode distances of approximately 30 mm, thus guaranteeing an optimal balance between signal-to-noise ratio and sensitivity to the cortex (Li et al., 2011; Strangman et al., 2013). The setups used in each experiment and the corresponding number of channels are depicted in Figure 1.

In this first illustrative experiment, we show the feasibility of fNIRS recording during physical activities that require moderate levels of motion (Supplementary Material, video Table Tennis). This was a table tennis trial in which the participant was an athlete, male, 30 years old, right-handed and with 16 years of regular practice. The experiment consisted of a block design with three active conditions: forehand, predictable and unpredictable conditions alternated with rest (baseline). The order of active conditions was randomized and each block had the duration of 20 s (30 s for rest) and was repeated 10 times. In the forehand condition, the ball was always directed to the right side of the table of the participant, while in the predictable condition the ball was directed twice to the right side and then twice to the left side (thus inducing a backhand hit). In the unpredictable condition, there were no prior fixed locations (right or left) for the hit. All the hits were returned to the right side of the partner (J.R.S) at the other end of the table. Here we aimed to identify cortical regions which are more active in unpredictable than in predictable condition. In order to do so, the oxy-Hb and deoxy-Hb time courses of each channel were modeled with a general linear model (GLM), with a correction for serial autocorrelations using the hemodynamic response function (HRF) precoloring method (Friston et al., 2000) on the downsampled signal at 1 Hz. We computed the contrast of the HRF betas estimates between unpredictable and predictable conditions, and results for each channel were Bonferroni corrected for multiple comparisons.

In addition, we also evaluated the coefficient-of-variation as a metric for the signal quality across the four conditions. The CV (CV = standard deviation / mean) of each wavelength was calculated at each frame considering a sliding-window of 5 s width (yielding 21 frames, centered at the frame of interest). This is a similar procedure as used and described by Piper et al. (2014) to evaluate the influence of motion artifacts during physical exercise. The motor/premotor setup was used in this experiment (Figure 1, top-left). Our hypothesis was that left motor/premotor regions would be more active in the unpredictable condition.

The results are shown in Figure 2 (top) (see Table S1, Figure S1 in the Supplementary Material for statistical results of all channels). As expected, the statistical parametric maps highlight that these regions were indeed more active during the unpredictable when compared to the predictable condition. Interestingly, the right premotor cortex was also more active, suggesting that location uncertainty required motor preparation in both hemispheres. Moreover, the boxplots suggest that the signal reliability was similar between the two conditions. However, as expected, the coefficient-of-variation was slightly smaller during rest (~1% less).

In this second illustrative experiment, we highlight the feasibility of fNIRS recording during the performance of two musical excerpts in the piano and we compared cortical activation levels in rhythmically simple vs. polyrhythmic conditions (Supplementary Material, video Piano). Both excerpts were 30 s long and corresponded to the first bars of the following compositions by Alexander Scriabin: Etude op. 42, no. 5, in C sharp minor, and Etude op. 8, no. 2, in F sharp minor, for the rhythmically simple vs. polyrhythmic conditions, respectively. We used a block-design with two active conditions (simple vs. poly) alternated with rest (baseline). The order of conditions was randomized. Each cycle had the duration of 20 s and was repeated 10 times. The participant (P.V.) was a professional pianist, 51 years old, female, right-handed with 45 years of regular practice. Statistical parametric maps of activation and the CV metric were computed in a similar fashion as the Table Tennis experiment described above. The setup is depicted in Figure 1 (bottom-left).

Figure 2 (middle) depicts the results suggesting that the right dorsolateral prefrontal cortex (DLPFC) was more active in the polyrhythmic than in the rhythmically simple condition (see Table S2, Figure S2 in the Supplementary Material for statistical results of all channels). This finding points toward a greater cognitive effort and sustained attention in polyrhythmic playing. The coefficients-of-variation shown in the boxplots suggest a satisfactory signal quality during the acquisition of the data.

In this third case study, we investigated the temporal patterns of cortical activation during human interaction when playing a violin duo with simultaneous fNIRS acquisition from both musicians (each of them playing a different musical line) (Supplementary Material, video Violin Duo). The participants (professional level, 41 and 50 years old, males, right-handed with 28 and 40 years of regular practice, respectively) were instructed to play a 30 s excerpt of Antonio Vivaldi's Allegro from the Concerto No 1 in E Major, Op. 8, RV 269, "Spring". We used the setup described in Figure 1 (top-right).

We analyzed the cortical activation synchronization between subjects. Thus, for each channel of the two violinists, we carried out a correlation analysis of the oxyhemoglobin concentration during the performance of this piece. An interbrain correlation map was built based on the Spearman correlation coefficient, as shown in Figure 2 (bottom-left; see Table S3, Figure S3 in the Supplementary Material for statistical results of all channels). Interestingly, higher synchronization between the brains of the two violinists occurred in parietal and frontal regions, in premotor and somatomotor areas.

Finally, we demonstrate the possibility of using fNIRS for long periods of data acquisition. In this concept proof, the participant (G.Z.M.), a 26 years old male (right handed) was monitored during approximately 4 h, from 10 a.m. to 2 p.m., while carrying out daily activities related to his work (mostly software coding and writing technical support emails). We used a prefrontal setup, as described in Figure 1 (bottom-right).

A 1-min sliding window was used to quantify the interhemispheric functional connectivity of the lateral orbitofrontal cortex. The Spearman correlation coefficient between the oxyhemoglobin concentration signals was obtained for each window, providing a time series for the interhemispheric functional connectivity. This time-varying connectivity and its respective estimated power spectrum (smoothed periodogram) are shown in Figure 2 (bottom-right). Interestingly, the spectrum suggests a periodicity with peaking at 0.002 Hz. We prefer to avoid conjecturing on the physiological interpretation of this finding, but this specific frequency seems to not be directly related to a non-functional component (e.g., physiological noise). However, it is important to mention that a previous resting state fMRI study has already reported cyclic fluctuations in functional connectivity (Handwerker et al., 2012).

GZ is employed by NIRx Medizintechnik GmbH. The other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.