Using functional near‐infrared spectroscopy to measure prefrontal cortex activity during dual‐task walking and navigated walking: A feasibility study

Abstract Introduction While functional near‐infrared spectroscopy (fNIRS) can provide insight into motor‐cognitive deficits during ecologically valid gait conditions, the feasibility of using fNIRS during complex walking remains unknown. We tested the process and scientific feasibility of using an fNIRS device to measure cortical activity during complex walking tasks consisting of straight walking and navigated walking under single and dual‐task (DT) conditions. Methods Nineteen healthy people from 18 to 64 years (mean age: 45.7 years) participated in this study which consisted of three complex walking protocols: (i) straight walking, DT walking (walking while performing an auditory Stroop task) and single‐task auditory Stroop, (ii) straight and navigated walking, and (iii) navigated walking and navigated DT walking. A rest condition (standing still) was also included in each protocol. Process feasibility outcomes included evaluation of the test procedures and participant experience during and after each protocol. Scientific feasibility outcomes included signal quality measures, and the ability to measure changes in concentration of deoxygenated and oxygenated hemoglobin in the prefrontal cortex. Results All participants were able to complete the three protocols with most agreeing that the equipment was comfortable (57.9%) and that the testing duration was adequate (73.7%). Most participants did not feel tired (94.7%) with some experiencing pain (42.1%) during the protocols. The signal qualities were high for each protocol. Compared to the rest condition, there was an increase in oxygenated hemoglobin in the prefrontal cortex when performing dual‐task walking and navigation. Conclusion We showed that our experimental setup was feasible for assessing activity in the prefrontal cortex with fNIRS during complex walking. The experimental setup was deemed acceptable and practicable. Signal quality was good during complex walking conditions and findings suggest that the different tasks elicit a differential brain activity, supporting scientific feasibility.

ing conditions and findings suggest that the different tasks elicit a differential brain activity, supporting scientific feasibility.

dual-task walking, fNIRS, navigation, walking INTRODUCTION
Performing complex tasks involving a combination of motor (e.g., walking) and cognitive skills forms part of everyday living (Kahya et al., 2019), such as walking while talking or navigating in different environments. In healthy individuals, walking is an automated skill requiring minimal cognitive control which is critical for safe ambulation (Bürki et al., 2017;Clark, 2015), but this control is subject to age-related changes (Papegaaij et al., 2014). In recent decades, rising evidence has shown the negative effect of aging and neurological diseases (e.g., Parkinson's disease and multiple sclerosis) on motor-cognitive interaction necessary for an independent lifestyle (Kahya et al., 2019;Roy et al., 2017;Singhal et al., 2020). A growing amount of literature highlights a connection between poor cognitive function, gait variability measures, and falling (Amboni et al., 2013), which bears devastating consequences on health in aging and for people with neurological diseases (Batchelor et al., 2012;Bishnoi et al., 2021). While the neural substrates underlying motor-cognitive difficulties are not fully understood (McIsaac et al., 2018), one first step is to arrive at a method that accurately portrays these deficits in an ecologically valid state.
Recently, walking automaticity has been assessed with functional Magnetic Resonance Imaging (fMRI) in conjunction with imagined walking or foot movements (Bürki et al., 2017) or by behavioral assessment in studies using a dual-task paradigm (Johansson et al., 2021;Kressig et al., 2008). Dual-task (DT) studies involve the simultaneous performance of a primary (e.g., walking) and a secondary task with different goals, the primary task performed alone as a single task (ST) as well as with the secondary task (i.e., DT), for example, walking while performing a cognitive task. The difference in the performance of the primary task under ST and DT conditions provide an indication of the automaticity of the task of interest (Clark, 2015). Although fMRI is considered the gold standard for neuroimaging, the method is sensitive to motion artifacts (Bishnoi et al., 2021;Gramigna et al., 2017) and cannot be used during real walking. Instead, functional near-infrared spectroscopy (fNIRS) has been introduced as a technique to record cortical activity during gait based on the brain's hemodynamics (Bishnoi et al., 2021). The major advantages of fNIRS compared to other neuroimaging techniques are that it is noninvasive, can be used wirelessly, is lightweight, and relatively robust to head movements (Gramigna et al., 2017;Kahya et al., 2019), which makes it a suitable neuroimaging technique to study complex movements in humans.
Some fNIRS studies investigating DT walking and postural control tasks in healthy adults have found increased oxygenation in the prefrontal cortex (PFC) during DT compared to ST (Holtzer et al., 2011;Marusic et al., 2019;Rosso et al., 2017). Overactivation in cortical areas in older adults has been attributed to compensation for agerelated declines in brain structure and function (Rosso et al., 2017), suggesting that the PFC plays an important role in the performance of motor-cognitive dual-tasks.
While fNIRS has been successfully used to measure PFC oxygenation during steady-state DT walking, to the best of our knowledge, no study has tested the feasibility of using fNIRS during complex walking paradigms combining DT and navigation. Complex walking might pose a challenge to participants and could reduce fNIRS signal quality. Therefore, we evaluated the process and scientific feasibility of measuring PFC activity during complex walking protocols involving both navigation and cognitive DT. Process feasibility aspects included acceptability of wearing the measurement systems, adherence to walking protocols and involved tasks, and practicability of simultaneous fNIRS and gait measures. The scientific feasibility was evaluated by investigating signal quality and differences in the fNIRS data between the different task conditions within the walking protocols.

Participants
Nineteen adults (13 females, mean age and range: 45.7, 18-64 years) with the ability to ambulate independently and without any neurological disease or impairment affecting gait were recruited through advertisements. This study was approved by the Regional Board of Ethics in Stockholm (Dnr 2020-05315/2021-01329). All participants received verbal and written information and gave written consent prior to study participation.

Procedure
Prior to the experimental session, participants were asked about health status and to provide their head size (circumference) for the fNIRS measurement. To describe the demographics as well as the physical and cognitive functioning of the study sample, interviews and a neuropsychological test battery were carried out during the experimental session, in addition to the fNIRS measurement. Demographic data consisted of sex, medical history, years of education, living situation and employment status, as well as overall physical activity level quantified using the Frändin and Grimby scale (Frändin et al., 1991).

F I G U R E 1
The three complex walking protocols used in the experiment. (a) Walking with a cognitive dual-task; (b) navigation; and (c) navigation with simultaneous cognitive dual-task.

Complex walking protocols
The fNIRS experiment consisted of three protocols with different complex walking conditions. The protocols were set up according to a block design, with stimuli being walking conditions and cognitive tasks and to walk around the cones alternating between yellow and red cones and to ignore the blue cones. The auditory Stroop task, which has been proven feasible to perform during gait assessments (Johansson et al., 2021), consisted of the Swedish words for high and low in a congruent or incongruent high and low pitch. Words were presented to the participants through wireless headphones. Participants were instructed to respond verbally, as fast and correctly as possible, to the corresponding pitch irrespective of the words presented. The responses were recorded to analyze task accuracy. During each block with an auditory Stroop task, a total of 7-word prompts of high or low were presented in a predetermined randomized order. Participants were instructed to pay equal attention to both tasks when dual-tasking.
Stimulus length was 20 s long, followed by 15 s of rest period to allow for a baseline measure (Amaro & Barker, 2006). Each block condition was performed 6 times in each protocol (e.g., 6 blocks of walking straight) with the time being approximately 12, 8, and 8 min, for protocol 1, protocol 2, and protocol 3, respectively. The conditions within the protocols were randomized to negate the learning effect due to repetition as much as possible.
All test instructions and auditory Stroop were provided through headphones and the test leaders had no interaction with participants during data recording. Participants were given opportunity to practice the auditory Stroop alone and as a DT before starting each protocol.
Participants were also given the opportunity to rest for a few minutes in between the protocols.

fNIRS and complex walking measurement
The measurement of changes in concentration of oxygenated (HbO) and deoxygenated (HHb) hemoglobin in the PFC was assessed using a NIRSPORT 2 (NIRx Medizintechnik, Berlin, Germany) continuous wave fNIRS device. During testing, the participants were fitted with a

F I G U R E 2
The walking tasks used in the complex walking protocols showing (a) straight walking and (b) navigated walking. In protocols 1 and 3, we combined straight walking and navigated walking with simultaneous performance of the auditory Stroop task. Cones are illustrated in red, yellow, and blue along with cones for turning around in grey. Walking path is illustrated in black.

Feasibility assessment
The process and scientific feasibility of measuring PFC activity during complex walking with fNIRS were evaluated. Participants were asked questions directly after each protocol regarding their experience of wearing the fNIRS system and their ability to concentrate on the protocol tasks. After completing all protocols, participants filled out a feasibility questionnaire regarding the acceptability of the whole experiment, with questions pertaining to discomfort, dizziness, pain, fatigue, attention, task prioritization and task difficulty (see Appendix A and B for the full questionnaire).
To assess acceptability of the experiment, questionnaires were analyzed to evaluate if the experiment was performed in an acceptable timeframe and if participants felt tired, dizzy, or experienced pain during or after the experiment. The ability of participants to concentrate during the experiment was also evaluated. For adherence to the walking protocols and involved tasks, completion rates, interruptions during the walking protocols and task performance were evaluated. Practicability was assessed by evaluating if data from all measurement systems could be obtained simultaneously and if data about task performance (i.e., audio recordings and video recordings) was sufficient to be able to score the tasks. Assessing scientific feasibility involved investigating whether the fNIRS data could reveal changes in PFC activity during protocol tasks and if these differences were in line with previous DT studies, where DT walking is generally associated with an increase in prefrontal activity compared to rest (Vitorio et al., 2017). Signal quality was evaluated by calculating scalp coupling index (SCI) values based on the photoplethysmographic cardiac waveform's presence in the fNIRS signal (Pollonini et al., 2016). SCI values range from 0 to 1, where 1 is considered an ideal coupling value and a threshold for acceptable coupling is around 0.7 to 0.8 (Hernandez et al., 2020;Pollonini et al., 2014).
Preprocessing included resampling the raw fNIRS signal to 4 Hz, to account for high autocorrelation in the fNIRS signal, and converting the raw intensity to changes in optical density. Optical density was converted to changes in hemoglobin concentration using the modified Beer-Lambert law (Delpy et al., 1988). Partial path-length factor (PPF) was set to 0.1 (Toyoda et al., 2008).

Process feasibility adherence to walking protocols and involved tasks
All participants were able to complete the three walking protocols successfully without any interruptions, except one participant who had to restart protocol 2 because the wireless headphones turned off.

Protocol 1 (Standing ST and Walking ST/DT)
A majority of the participants did not perceive the Standing ST (90%) and the Walking DT (85%) to be challenging. During dual-tasking, 55% of the participants reported that they were equally focused on both tasks while 40% stated that they were more focused on the auditory Stroop task. Almost all participants (n = 17) scored 100% correct on the auditory Stroop (range 89%−100%).

Protocol 2 (Walking ST and Navigation ST) and Protocol 3 (Navigation ST/DT)
None of the participants perceived that the Navigation ST was challenging while half of the participants reported that the Navigation DT was challenging. During dual-tasking, most of the participants (65%) felt that they focused on both tasks equally, with 90% of the participants reporting that they perceived both tasks to be of equal importance. Most participants (n = 14) responded 100% correctly on the auditory Stroop (range 95%−100%). Visual inspection of navigation indicated that participants understood the task and could perform the navigation with high compliance.

Acceptability of wearing the measurement systems
A majority of participants reported that the fNIRS device was comfortable to wear (57.9%). While most participants did not experience any pain during or after testing (57.9%), some did experience pain during testing (42.1%) and of these, some also after testing (10%). Most pain was reported as some form of pressure on the forehead, while two participants reported pain on the head in general. Wearing the tight fNIRS cap while walking did not make the participants dizzy during (95%) or after (95%) the walking protocols. Participants reported that testing time was adequate (73.7%) and that they did not feel tired during (94.7%) or after (78.9%) testing. All participants (100%) reported that they were able to concentrate during the testing. From putting on equipment to completing the last protocol, the fNIRS measurement took approximately 45 min.

Practicability of simultaneous measurement
Accelerometer data from the APDM system could successfully be obtained for each walking protocol. GoPro videos of each walking TA B L E 1 Demographic and clinical characteristics

Scientific feasibility
Average SCI values calculated for each condition were good, around 0.96, throughout each condition (Table 2). Participants generally had few channels with poor coupling.

DISCUSSION
This study explored the feasibility of using fNIRS to observe changes in the PFC while performing three ecologically valid complex walking protocols in healthy adults. A feasibility questionnaire of each participant's experience during each protocol showed that participants did not report any major acceptability issues wearing the fNIRS device or performing the experiment. Analysis of performed tasks showed that participants were able to complete the protocols without problems and with high task performance. Practicability could be increased by automatic synchronization of measurement systems. Scientific feasibility was evaluated via signal quality, which was high, and by exploring the activation in the PFC during the different conditions in the three TA B L E 2 Average SCI values over all source and detector pairs (excluding short-separation channels) for all participants for different protocol conditions, and average number of channels with poor coupling (SCI < 0.7) per participant The blue to red color bar denotes the t-statistic value, which indicates an increase or decrease in HbO concentration from baseline to the task condition. One participant was not included in any protocol due to issues with block separation triggers, and another participant was excluded for the same reason in protocol 3. protocols, where a group-level analysis of the fNIRS data showed activation in the PFC when performing DT walking and navigation compared to rest.
The present results are consistent with previous studies such as Nieuwhof et al. (2016) who showed that participants experienced a low burden of wearing a wireless fNIRS device during DT walking, and were able to perform walking protocols with little effort. Although some participants reported experiencing pain in our experiment, this was generally reported as pressure from the optodes on the forehead, and every participant was able to complete all protocols without any issues, supporting acceptability of the experiment. Steps could be taken to further increase comfort, such as adding rubber pads to the forehead optodes.
It has been documented that DT walking generally results in PFC activation in healthy adults (Vitorio et al., 2017). Results for protocol 1 (Figure 5a-c) where we observed an increase in HbO during DT in channel AF8-F6, roughly corresponding to BA46/dorsolateral PFC (dlPFC), are comparable to Fraser et al. (2016). Their study found significant increases in HbO levels in the dlPFC during DT walking with an n-back task for both older and young adults. The dlPFC is associated with exec-utive functions (Yogev-Seligmann et al., 2008), which could indicate the increased need for executive resources during DT. While protocol 2 elicited no significant increase in HbO, protocol 3 (Figure 5f-g) led to increases in medial brain regions (roughly BA8), which could indicate that the navigational DT requires different resources. These results also support the scientific feasibility of distinguishing between DT and ST conditions. Contrasts between active conditions were reinforced by considering both HbO and HHb, indicating the importance of combined measures, which has been repeatedly argued for (Hakim et al., 2022).
A novelty of this feasibility study is to measure activation in PFC during three different complex walking protocols, not limiting experiments to steady-state straight walking. The complex walking did not introduce lower SCI values than straight walking (Table 2). Normal walking, a relatively simple motor task, has been documented to not induce much blood oxygenation in the PFC, at least in young healthy adults (Mirelman et al., 2014). While straight walking (Figure 5b and d) did not elicit significant increase compared to baseline, neither did navigated walking in protocol 2 (Figure 5e). This could suggest that our navigation task might not have been challenging enough for a healthy population. However, navigation caused a significant HbO increase in protocol 3 (Figure 5f) but not in protocol 2 (Figure 5e). The reasons for this need further investigation.

Limitations
Although the sample size was sufficient to determine feasibility of fNIRS to measure PFC during complex walking, it might be too small to draw any neurophysiological conclusions. The sample was also highly physically active, something which might not be true for other populations, for example, neurodegenerative disease. It has been documented that oxygenation levels during DT walking can be affected by age (Mirelman et al., 2014); therefore, a limitation of our study could be the large variability in age (18-64 years) in the present study sample.
Our future studies will separate these groups with more participants in each.
To account for extracerebellar blood flow, we included short channels. Nevertheless, it is not completely possible to disentangle cerebral and extracerebral signal components since we did not measure blood pressure or muscle movement. The signal could be biased by jaw movements during auditory Stroop and the oxygenation drop due to exhalation (Scholkmann et al., 2013). Finally, exact anatomical registration of optodes was not performed, for example, with structured-light 3D scanning (Homölle & Oostenveld, 2019), which could make comparisons between subjects in our study less precise.

CONCLUSION
Our study supports both the process and scientific feasibility of fNIRS measurement of PFC activity during complex walking. The measurement systems were deemed acceptable by the participants, and adherence to the protocols was maintained. The complex walking tasks elicited differential brain activity in healthy participants and signal quality was high. This underlines the scientific feasibility of our setup, the usability of fNIRS, and can further be used for development of walking protocols, enabling reproducibility and a common test protocol when using this type of complex walking task.

ACKNOWLEDGMENTS
We thank all participants who took part in this feasibility study. We also thank the uMOVE core facility at Karolinska Institutet.

CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT
Code is available via: https://osf.io/f4p3d/?view_only = c339fe 6a5fce44408765a40788b06763. With respect to the Swedish and EU personal data legislation (GDPR), the data are not freely accessible due to regulations regarding personal integrity in research, public access and privacy. The data are available from the principal investigator of the project: Erika Franzén (erika.franzen@ki.se), on a reasonable request. Any sharing of data will be regulated via a data transfer and user agreement with the recipient.

ETHICS STATEMENT
This study was approved by the Regional Board of Ethics in Stockholm (Dnr 2020-05315 and 2021-01329).