A total of 68 participants participated in this study; twenty-five young adults (YA), nineteen older adults (OA) and twenty-four people with PD. Young and older adult participants were recruited via local advertisement (i.e., posters within campus notice boards and e-mails circulated to students and staff members). People with PD were recruited from local Neurology clinics via referrals from movement disorder specialist neurologists. All study procedures were approved by the Oregon Health and Science University Institutional Review Board, with written informed consent obtained before participation.

Participants were included if they were aged 20–40 (young adults) or 50–90 (older adults or PD) years, able to stand or walk for 2 min without assistance. People with PD were included if they had a diagnosis of PD as defined by the UK Brain Bank criteria, Hoehn and Yahr score II-IV and were taking anti-Parkinsonian medication. Exclusion criteria were; musculoskeletal, vestibular, visual, or other medical condition that affected gait or balance.

Participant characteristics of age, sex, height, and weight were recorded. Global cognition was measured with the Montreal Cognitive Assessment (MoCA) (45). All testing in people with PD was performed in the OFF medication state, ~12 h after taking last medication dosage. Disease severity was measured using the Movement Disorders Society (MDS-revised) unified Parkinson's disease rating scale (MDS-UPDRS) (46), freezing-status was measured using the new Freezing of Gait Questionnaire (nFOGQ) (47), and levodopa equivalent daily dosage (LEDD) was calculated (48). Specifically, out of twenty-four participants with PD, twelve had freezing of gait, as reported by the nFOGQ (nFOGQ>1).

The participants completed two different motor tasks at self-selected pace; (1) a 2-min 360° turning-in-place task, alternating 360° turning to the left and 360° turning to the right; and (2) a 2-min walking task. Each condition included a baseline of 20 s of standing at the start and end of the task (with 80 s of performing the turning or walking task in between). The walking condition was conducted over-ground with participants walking back and forth over a 9 m straight path, with a 180° degree turn at each end. Condition order was randomized for the participants, with breaks between tasks if needed. A research assistant walked with the participants and stood by the participants during turning to ensure their safety.

Test re-test reliability of the fNIRS measurement of PFC activity during the continuous turning and walking trials was conducted within the young adult subjects. The young adult group performed the same walking and turning tasks for a second time after having the fNIRS device removed and replaced following a short break (~5–10 min).

A mobile fNIRS system (Oxymon, Artinis, Netherlands) was used to record changes in oxygenated hemoglobin (HbO₂) and deoxygenated hemoglobin (HHb) within the PFC at a sampling rate of 50 Hz. Distance from transmitter to detector was 3.5 cm (38) and data was collected and processed in line with previous studies (32, 33). Additionally, two short-separation reference channels at a distance of 1.5 cm (one left and one right hemisphere) were used to allow for removal of peripheral interference (i.e., from blood flow changes in the extra-cerebral layers of the head) in the long source-detector separation channels (44).

All processing of fNIRS signals followed current recommendations where possible. A digitizer (PATRIOT, Polhemus, VT, USA) was used to provide 3-dimensional morphological locations for cortical regions of interest relative to scalp position and the fNIRS optode measure. Data from the digitizer was entered into the software package NIRS-statistical package metric mapping (NIRS-SPM, http://www.nitrc.org/projects/nirs_spm) (49), which was implemented within MATLAB 2017a (Mathworks, MA, USA). NIRS-SPM allows registration of fNIRS channel data onto the Montreal Neurological Institute (MNI) standard brain space (50). NIRS-SPM used probabilistic registration of the fNIRS co-ordinate data to determine channels that related to ROIs at the group level [described in detail elsewhere (51)]. HbO₂ changes were recorded bilaterally (left and right) within the pre-frontal cortex (PFC). The Brodmann areas (BA) that corresponded to the PFC consisted of BA9 and BA10 for all of the participants.

The fNIRS data were processed within custom-made MATLAB algorithms, which consisted of several steps:

The primary outcome for this study was change in oxygenated hemoglobin (HbO₂) from baseline standing to walking or turning, which is a proxy for cortical activation. The fNIRS system used emitter-detector pairs to emit light into the skull that diffused through brain tissues resulting in scattering of multiple photons (55). These photons were then detected by the fNIRS detector channels when exiting the skull after passing through the cortical brain layers (typically ~1–2 cm deep, with an emitter-detector optode distance of 3.5 cm). Importantly, HbO₂, and HHb have different absorption rates for different wavelengths of near-infrared light, which can be analyzed with Beer-Lamberts law equations (56) within the fNIRS software to calculate the relationship between an exciting photon intensity and incident photon intensity to derive changes in HbO₂ and HHb (57). Therefore, the fNIRS system measured optical density of the raw signal and converted this to HbO₂ and HHb using Beer-Lamberts law (57). HbO₂ rather than HHb was reported as our primary outcome due to its sensitivity to walking and cognitive tasks (58, 59). Additionally, changes in HbO₂ concentration within local brain capillary networks are caused by neuron firings with brain activity, which is commonly referred to as neurovascular coupling (60). Relative changes from baseline standing (initial 20 s) in HbO₂ concentration was reported in an attempt to account for between individual physiological variations (33); see below calculations.

Walking and turning periods;

Statistical analysis was conducted in SPSS (v.24, IBM, Armonk, NY, USA) and Shapiro-Wilks tests determined data normality with parametric analysis used throughout. Mean differences with paired-sample t-tests, intra-class correlation coefficients (Absolute Agreement: ICC_{2, 1}) and Bland-Altman plots with 95% limits of agreement (LoA 95%) were used to determine the test re-test reliability of PFC activity measurement via fNIRS between the first and second data captures in young adults. Acceptance ratings for ICCs were set at excellent (>0.75), moderate (0.40–0.75), and poor (<0.40) agreement (61).

PFC HbO₂ activation during turning and walking in young, older and PD groups was reported via mean and standard deviation. Separate linear mixed-effect models compared groups (YA, OA, PD), Periods (Early vs. Late) and PFC regions (Left vs. Right), with a random intercept for each subject within the models. Post-hoc independent t-tests explored significant differences between specific groups (YA vs. OA, YA vs. PD, OA vs. PD). We also compared between groups differences using Cohen's d effect sizes based on mean (SD) scores. Effect sizes were interpreted as small (0.2), medium (0.5), and large (0.8) (61). Due to the exploratory nature of the analysis, statistical significance was set at p < 0.05.

Demographic characteristics of the participants are shown in Table 1. Groups were significantly different for age (p < 0.001), but age was not significantly different between older adults and PD groups (p = 0.166). Similarly, young adults tended to be taller than the older adults and PD subjects, but the older adults and PD groups did not significantly differ for height (p = 0.146). The groups were well-matched for gender (p = 0.160), weight (p = 0.287), and cognitive ability (p = 0.227).

Following removal of fNIRS data that had poor signal quality, data from 19 young adult subjects were analyzed for walking (n = 6 young adult walking trials were excluded following visual signal inspection: 4 first trial and 2 second trial errors) and 25 subjects were analyzed for turning to determine test re-test reliability of fNIRS measurement.

On average, there was moderate (ICC_{2, 1} = 0.67) reliability of PFC activity measured via mobile fNIRS during turning and walking in young adults (Table 2). There were no significant differences between any of the outcomes over the two sessions, with very low average difference between testing sessions (Mean Difference = 0.03 μm, Figures 1, 2). Reliability across the different time periods (early vs. late) and regions of interest (Left, Right or combined) within the PFC ranged from moderate (ICC_{2, 1} = 0.52) to excellent (ICC_{2, 1} = 0.79). Importantly, the primary outcome over the duration of the motor tasks (combined left and right PFC regions) had moderate reliability (turning; ICC_{2, 1} = 0.67, walking; ICC_{2, 1} 0.71) (Table 2).

Table 3 shows the relative change in HbO₂ during continuous turning and walking in the groups. Overall, the PD group had higher average levels of PFC activation during turning and walking than young or older adults across the majority of the time-points and PFC regions. Older adults also had higher average PFC activation during turning and walking compared to young adults.

During walking the groups were significantly different for overall PFC activation (p = 0.025), left PFC activation (p = 0.012) and for the early period (first 40 s) of walking (p = 0.007) (Table 3). There was also a significant difference in PFC activation between early and late periods of walking across the groups (p = 0.012), with higher PFC activity in the early period. Post-hoc testing indicated that the PD group had significantly higher PFC activation during the early period of walking compared to the young (p = 0.008) and older adults (p = 0.037). The PD group also had significantly higher overall (p = 0.025) and left side (p = 0.017) PFC activation than the young adults. However, the older adult group did not have significantly different PFC activation compared to the young adults, with small effect size differences between groups (Table 4). Overall, the PD group had moderate to large differences (Cohen's d 0.52 to 0.86) in PFC activation during walking compared to young and older adults (Table 4), particularly for the early period and left PFC region.

Tables 3, 4 demonstrate that the effect of turning on PFC activation was similar across groups, as during continuous 360° turning there were no significant group differences for any outcome. However, there was a trending difference between PFC regions during turning (left vs. right, p = 0.059), with higher right PFC activity compared to left (Table 3). Despite the lack of a significant difference, effect sizes showed that the early period of turning differentiated the young group from the older adult and PD groups (Table 4; Cohens d of 0.86 and 1.21, respectively), with young adults having the lowest activation of the three groups (Table 3).

The mobile fNIRS device used in this study is a commercial device that allows access to the raw data that registers HbO₂ concentration and subsequent implementation of our custom-made algorithms for data analysis. Re-test reliability of HbO₂ signal recorded with the mobile fNIRS device was conducted within our young adult group with some variations across the conditions (walking or turning), as well as when the HbO₂ signal was broken into different PFC hemispheres (left or right) or trial times (early or late). When using our fNIRS system with short-separation reference channels and fixed data analysis pipeline we found moderate reliability (ICC_{2, 1} 0.67) of the PFC HbO₂ signal during the continuous walking and turning tasks. These results, together with small LoA between trials, indicated that we could be confident that this signal is reliable, as reported in Figures 1, 2. However, there were changes in reliability when breaking the fNIRS signal down into individual hemispheres or times of the signal (early vs. late), which is similar to previous static fNIRS findings (62). On the basis of our results, it can be stated that HbO₂ outcomes measured using a mobile fNIRS device during continuous (2 min) walking or turning are relatively stable. However, when breaking the signal into specific features, such as hemispheres or time-periods, the stability of the HbO₂ signal can be altered.

This study confirms that PFC activity, measured through mobile fNIRS, can show differences in PFC activation during continuous walking and turning between young, older and subjects with PD. Specifically, we found that people with PD, OFF their medication, had significantly higher PFC activation during walking than young and older adults, in line with previous studies where PD subjects were ON medication (32, 33, 37–40). Higher levels of PFC activation during continuous walking, particularly the early period of walking, in PD compared to the other groups likely reflects the need to use executive-attentional resources even during relatively simple tasks (i.e., usual walking) (21, 26). The use of cognitive resources to compensate for PD related deficits is similar to previously reported theories of PD walking, which hypothesized that to compensate for movement automaticity deficits people with PD increased cognitive control (63), particularly executive control (26). However, executive-attentional deficits are common in PD (64) and these impact the ability to effectively compensate for underlying deficits. Therefore, when tasks become more challenging (such as dual-tasks, obstacles etc.) people with PD may not be able to effectively respond (24), which impacts gait and mobility, with implications for falls risk. Future studies may uncover further age or PD-related deficits with the use of more complex tasks that provide additional cognitive burden to participants.

Interestingly, although older adults had slightly greater PFC activation levels than young adults during walking, our findings were not significantly different which is in agreement to another previous fNIRS study (32). This demonstrated that healthy young and older adults may use their executive-attentional resources in a similar manner when walking, with little need to activate the PFC due to intact lower-level neural structures and activity compared to people with PD. Similarly, there were limited group differences found during the continuous 360° turning-in-place task, which may indicate that this complex task requires cognitive resources regardless of age or disease. Indeed, average PFC activation was higher in young adults, older adults and PD during continuous turning compared to walking, with PD participants having the highest HbO₂ concentrations across groups. Interestingly, although there were no significant group differences, there were large between group effects for PFC activation during the early period of the turning task. This highlighted that young adults had much lower PFC activation during the early period of continuous turning than older adults or people with PD. Aging and PD may therefore lead to greater cognitive control being required to begin complex continuous motor tasks, however less cognitive control is required once the task underway.

PFC activation appears to increase with walking in PD, but not with age. This may represent cortical compensation for sub-cortical dysfunction with PD. Findings suggest that targeting cortical activation, particularly executive-attentional activity within the PFC, with interventions such as pharmaceuticals, cueing strategies (visual, auditory or proprioceptive) or transcranial magnetic or direct current stimulation may help to alleviate the cortical burden of walking in PD. However, future studies are needed to establish findings and examine response to such interventions. Specifically, future studies should assess whether the observed increase in PFC activation is related to PD itself or to freezing of gait. In fact, as freezing of gait could represent a disruption of gait automaticity, people experiencing freezing of gait, even without PD, may show an increased cortical control of gait.

A key strength of this study was the use of short-separation reference channels (1.5 cm apart) to account for the peripheral hemodynamic response that is associated with physical activity (44). Short-separation channels have only been used in one other mobile fNIRS study (65), but they should be used within future work to reduce noise and ensure repeatable findings. Another strength is that we have provided detailed data analysis steps from our fNIRS signal processing, whereas other previous studies often use ‘black-box' data analysis tools such as NIRS-SPM (66) or Homer2 (67).

A limitation of this study was that, despite being the first to asses test re-test reliability of fNIRS HbO₂ measurement during walking and turning, test re-test analysis was only conducted in young adults. Reliability could possibly change in older adults or PD, as previous studies have shown this in patients with TBI (68). However, we may expect that reliability would even be higher in older adults and further in PD as, unlike young adults, they require more cognitive control of movement due to a loss of motor automaticity (1, 28), therefore findings in these groups may be more consistent. However, this is mainly speculative at this point, future studies could examine re-test reliability in older adult and PD using the same protocol developed here. Additionally, we did not examine the influence of disease stage, duration or other clinical factors (e.g., freezing of gait) in PD, which future studies should examine to provide greater understanding of cortical activation during walking and turning in PD.

The 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.