Altered prefrontal cortex responses in older adults with subjective memory complaints and dementia during dual‐task gait: An fNIRS study

People with cognitive impairments show deficits during physical performances such as gait, in particular during cognitively challenging conditions (i.e. dual‐task gait [DTG]). However, it is unclear if people at risk of dementia, such as those with subjective memory complaints (SMC), also display gait and central deficits associated with DTG. In this study, we investigated the effects of single‐ and dual‐task gait (STG and DTG), on left prefrontal cortex (PFC) activation in elderly people with subjective memory complaints (SMC) and Dementia. A total of 58 older adults (aged 65–94 years; 26 Healthy; 23 SMC; 9 Dementia) were recruited. Gait spatiotemporal characteristics (i.e. stride velocity and length) were assessed using an instrumented walkway during STG and DTG. Single‐channel functional near‐infrared spectroscopy over the left PFC was used to measure changes in oxyhaemoglobin (O2Hb) during gait. Stride velocity and length during STG (all p < .05) and DTG (all p < .000) were significantly impaired in people with Dementia compared to Healthy and SMC individuals. No differences were observed between Healthy and SMC. For STG, a greater increase in O2Hb (p < .05) was observed in those with Dementia compared to the Healthy and SMC, while no differences were observed between Healthy and SMC. A significant increase and decline in O2Hb was observed during DTG in the SMC and Dementia groups, respectively, compared to Healthy. Our findings indicate an altered pattern of cerebral haemodynamic response of the left PFC in DTG in people with SMC and Dementia, which may suggest that central changes precede functional impairments in people with SMC.


| INTRODUCTION
Ageing is associated with declines in cognitive functioning and is a significant risk factor for neurodegenerative conditions such as dementia. In the normal ageing process, it is expected that some cognitive declines would be apparent, typically exemplified by increased reaction time and reduced abilities related to attention and executive functioning (Salthouse, 2004). However, in dementia the trajectory of cognitive decline is magnified, significantly impairing activities of daily living and resulting in poorer quality of life. Considering the rapid increase in population ageing worldwide, there have been increased efforts to raise awareness, with an emphasis on lifestyle and dietary modifications, to mitigate the risks-associated age-related cognitive declines and dementia (Simons et al., 2006;Solfrizzi et al., 2008;Di Marco et al., 2014). However, identifying older individuals at greater risk of cognitive declines and dementia remains a challenge particularly in the prodromal stages due to the absence of a clear clinical biomarker that would allow for early detection (Ahmed et al., 2014).
The assessment of physical movement, such as gait, has gained interest as a simple method to test for cognitive-motor functioning particularly in older adults (Ijmker & Lamoth, 2012;Morris et al., 2016;Tian et al., 2017). The control of gait is typically thought to be autonomous that stems from both spinal and supraspinal centres of the central nervous system to maintain the continuous gait cycle during movement, and balance and postural control in relation to the environment in which the individual is walking (e.g. going up or down a hill; Takakusaki, 2017). However, there is a cognitive element to gait that involves attention and executive functioning resources to produce optimal gait .
To investigate the cognitive-motor processes associated with gait, single-and dual-task gait (STG and DTG) paradigms have been previously used. DTG uses a concurrent cognitive task (e.g. word association or counting backwards) during gait, to determine the level of deterioration of gait performance compared to STG (i.e. normal walking) (Howell et al., 2016;Smith et al., 2016;Li et al., 2018). The reduction, or cost, in gait performance during DTG has been hypothesised to be caused by a reduction in attentional resources associated with performing two tasks concurrently (Boisgontier et al., 2013;Nascimbeni et al., 2015).
In healthy older individuals and those with cognitive deficits, impaired DTG performance has been consistently reported (Hausdorff et al., 2008;Taylor et al., 2013). Portable functional neuroimaging modalities such as functional near-infrared spectroscopy (fNIRS) have demonstrated that prefrontal cortex O 2 Hb is increased, even in simple walking conditions, in older compared to younger adults (Mirelman et al., 2017). During DT paradigms, increased cortical activity, particularly in the PFC (Beurskens et al., 2014;Meester et al., 2014), has been associated with increased attentional processing in older adults compared to young adults. Given the role of the PFC in attentional allocation, this evidence lends support to the hypothesis that DT activities which involve a walking component, lead to an increase in cognitive demand where more attentional resources or demands are necessary to maintain DT activities.
Preliminary evidence have suggested that DTG paradigms may be sensitive to detect changes in cognitive functioning, particularly when cognitive impairments are not yet apparent (Beauchet et al., 2017). In this case, people with subjective memory complaints (SMC), those with reported memory impairments, but no clinical indication of memory deficits, may serve as an ideal population to determine if indeed DTG performance differ from age-matched controls with no memory complaints and individuals with a known memory impairment. Individuals with mild cognitive impairment have demonstrated increased PFC activation during DTG, associated with poorer executive function performance compared to age-matched controls (Doi et al., 2013), but to date-limited evidence on DTG performance exists in people with SMC. To the best of our knowledge, no neuroimaging evidence has been reported in relation to DTG performance in people with SMC.
In this pilot study, we compared spatiotemporal gait characteristics (i.e. stride length and velocity) of people with SMC, Dementia and healthy age-matched controls during STG and DTG conditions. We additionally compared the haemodynamic response of the left PFC during STG and DTG of all three groups. We hypothesized that people with SMC and Dementia would have impaired performance in gait outcomes compared to healthy age-matched controls during STG and DTG. We further hypothesized that the haemodynamic response of the left PFC would be greater in the SMC and Dementia groups compared to healthy controls during both gait conditions.

| Participants
In total 58 older participants (aged 65-94 years; 26 Healthy; 23 SMC; 9 Dementia) were recruited from the community and assisted living facilities within the Melbourne, Australia metropolitan area and surrounding regions. To be included in this study, all participants needed to be physically healthy and could walk at least 10 m without assistance. Individuals with history of stroke, head trauma, alcohol or drug dependency, current clinical diagnosis of severe depression or anxiety were excluded. Participants with SMC were determined via verbal confirmation to the question "do you feel like your memory is becoming worse?" and had to provide three examples of day-to-day issues that have occurred regarding their memory. Additionally participants within the SMC group needed to score greater than 24 points on the Montreal Cognitive Assessment (MoCA) to rule out mild cognitive impairments (MCI). Participants with Dementia were required to have been provided with a diagnosis prior to inclusion in the study and have a MoCA score of less than 24. All participants or their carers gave written informed consent prior to participating in this study. This study was approved by the Deakin University Human Research Ethics Committee (DUHREC 2017-054) and conducted in accordance with the Helsinki Declaration.

| Cognitive assessment
The MoCA is a validated 30-point test system designed to screen for cognitive impairments (Larner, 2012). It assesses several cognitive domains that include memory recall, visuospatial memory, language, attention, concentration and working memory. In this study, all participants assigned to the Dementia group attained a combined score for all cognitive domains of lower than 24 points.

| Gait performance
Both STG and DTG performance was assessed on a 4.87m instrumented walkway (ZenoMetrics LLC, Peekskill, NY, USA, sampled at 120 Hz with a 0.5 cm spatial resolution). The instrumented walkway measured spatiotemporal gait characteristics, in particular, stride length and velocity that was used as indices of gait performance in this study. All participants completed a total of 8 trials (8 passes away and back to the starting point) over the walkway using their preferred walking speed. All participants first completed four of the trials of STG, followed by four trials of DTG (counting backwards taking off seven from a pre-randomised list of three-digit numbers). The four STG trials were completed first to avoid any influence on gait and haemodynamic responses of the DTG trials on STG trials. Prior to the STG, all participants were instructed to "walk as you would normally walk" across the instrumented walkway. Prior to the DTG, all participants were instructed to "do your best to maintain your normal walking speed" and "to continue counting if you think you made a mistake". The number of counting responses during DTG was recorded. A schematic diagram of the setup is shown in Figure 1. All steps recorded under each condition were pooled for each participant and the mean of the pooled steps are reported as outcomes.

| Left PFC measures
A portable single-channel fNIRS device (Portalite, Artinis Medical Systems, The Netherlands) was placed over the left PFC that corresponded approximately to the F3 region (based on 10-20 EEG system). The portable fNIRS device emits NIR light at two wavelengths (760 and 850 nm) to detect changes in oxygenated (O 2 Hb) and deoxygenated (HHb) haemoglobin separately. Based on the assumption that NIR light is permeable to bodily tissue, the modified Beer-Lambert law (MBLL) was used to determine the attenuation of NIR light in proportion to the regional change in cerebral O 2 Hb and HHb. The MBLL describes the attenuation and scattering of NIR light as it passes through biological tissue, which underpins the concept of fNIRS (Delpy et al., 1988). Prior to each trial, all participants stood quietly in an unassisted upright position with hands by the side and looking straight ahead for 30s to establish a baseline haemodynamic response. After 30s of baseline measurement, participants were instructed to walk towards the "X" on the other end of the instrumented walkway, walk around the "X" and back towards the start line. In total four trials were performed for each gait condition, with each participant randomly assigned to start with either STG or DTG.

| Data processing and statistical analysis
For all fNIRS measures, raw O 2 Hb and HHb signals were collected using the proprietary software provided with the F I G U R E 1 A diagram of the study setup for the STG and DTG. A trial consists of a participant beginning at the start line, walking towards an "X" marked on the floor (1st pass), walking around the "X" and walking back towards the start line (2nd pass). All participants performed eight trials (4x STG; 4x DTG) that consisted of a total of 16 passes one way Portalite (Oxysoft 3.2.51.4 x64, Artinis Medical Systems, The Netherlands) and processed using HOMER2 (MATLABbased optical imaging toolbox). Prior to pre-processing, all raw fNIRS data were visually inspected for motion artefacts between 10 and 40s time-window follow commencement of the gait tasks. This time window corresponds to the peak fNIRS response which was used for further analysis. Following visual inspection, the raw data were pre-processed using a motion artefact detection and correction algorithm (i.e. principal component analysis [PCA]) (Brigadoi et al., 2014), and the averaged peak change in O 2 Hb and HHb (as defined by highest O 2 Hb and lowest HHb value for each trial less pre-gait baseline values) over a 30s time window was used to compare between GROUPS (Controls vs. SMC vs. Dementia) and GAIT conditions (STG vs. DTG). The pre-processing pipeline in HOMER2 for fNIRS signals is shown in Figure 2.
A one-way analysis of variance (ANOVA) was used to determine significant differences in participant demographic data and total number of counting responses during DTG. Additionally, a repeated measures ANOVA was used to compare within-group (GAIT -STG vs. DTG) and between-group (GROUP -Controls vs. SMC vs. Dementia) factors in step length and velocity and cerebral haemodynamic responses (O 2 Hb and HHb). Post hoc analysis was done using Tukey's honest significant difference (Tukey's HSD). An alpha level of p < .05 was set as the level of significance between comparisons. All data analyses were conducted using Statistical Package for the Social Sciences v25 (SPSS, IBM Inc, USA). All results are presented as Mean ± Standard Deviation (SD) and scatter plot of individual data points.

| Participant demographics
All participant's demographic details are shown in Table 1. One-way ANOVA showed that participants in the Dementia group were significantly older (p < .001) and had significantly lower MoCA scores (p < .001) compared to both Controls and SMC groups. No significant differences were observed between Dementia and Controls or SMC groups for height, weight and education level.

| Step length and velocity
The comparisons of step length and velocity during STG and DTG, and number of counting responses between groups F I G U R E 2 The processing pipeline used to process O 2 Hb and HHb signals. Firstly, all raw signals were converted into changes in Optical Density (OD). A motion artefact detection algorithm was applied to identify potential motion artefacts, by identifying parts of the signal within each trial that exceeded the pre-specified thresholds. After identification of a motion artefact, a principal component analysis (PCA) filter was used to correct any potential motion artefacts identified. A bandpass filter was then applied to filter out any low-and high-frequency noise. Once filtered, the OD signal was converted into concentration changes using the MBLL and the concentration change was averaged over the four trials in each gait condition. *Note that as each participant had a different completion time for each trial, the average time of the four trials for each gait condition was used to set the end time range are shown in Figure 3. Repeated measures ANOVA showed significant main effects for GAIT conditions (F 1, 55 = 10.13, p = .002) and GROUPS (F 2, 55 = 63.53, p < .001) for step length, and similarly for step velocity (GAIT -F 1, 55 = 19.00, p < .001; GROUP -F 2, 55 = 44.36, p < .001).

| DISCUSSION
This study aimed to determine if (a) people with SMC and Dementia would have impaired STG and DTG performance as measured by changes in step length and velocity, and (b) cerebral haemodynamic responses associated with STG and DTG would differ between SMC and Dementia compared to healthy age-matched controls. In partial support of our first hypothesis that gait kinematics would be impaired in people with SMC and Dementia, our results showed impaired step length and velocity in people with Dementia for STG and DTG compared to healthy control and SMC groups. In terms of our second hypothesis, individuals with SMC and F I G U R E 4 Comparisons of (a) O 2 Hb and (b) HHb response to STG and DTG among Controls, SMC and Dementia groups. The peak O 2 Hb response for STG for each group occurred at 21.3 ± 6.5 s (Control), 25.3 ± 8.2 s (SMC) and 23.6 ± 6.7 s (Dementia), while for DTG the peak O 2 Hb response occurred at 17.6 ± 5.5 s (Control), 18.9 ± 4.6 s (SMC) and 20.4 ± 5.5 s (Dementia). Within-group comparisons (STG vs. DTG) indicate a significant increase (p < .001) in O 2 Hb of the left PFC during DTG in both Control and SMC groups, but a significant decrease (p < .001) in O 2 Hb in the Dementia group. Between-group comparisons (Control vs. SMC vs. Dementia) showed a significant increase (p < .001) in O 2 Hb in the left PFC during STG in Dementia compared to Controls and SMC. However, during DTG, a significant increase in O 2 Hb in the left PFC was observed in the SMC group, while a significant decrease (p < .001) in O 2 Hb in the Dementia group was observed. Only the DTG condition performed by the SMC group elicited a significant reduction (p < .001) in HHb (as indicated by a greater negative value) when compared between and within groups. (#) Indicates within-group (STG vs. DTG) significance of p < .001, while (*) indicates between-group significance of p < .001 with other two groups Dementia displayed a differential response in left PFC to DTG. Those with SMC demonstrated a significant increase in left PFC activation during DTG, whist people with dementia showed an increase and decrease in left PFC activation during STG and DTG respectively.

| Gait performance during STG and DTG and total counting responses
Dual-tasking paradigms such as DTG have increasingly been used to investigate the cognitive-motor relationship in various neurodegenerative conditions such as Huntington's disease (Purcell et al., 2020;Radovanovic et al., 2020), Parkinson's disease (Fok et al., 2010;Rochester et al., 2014) and Dementia (Muir et al., 2012;Montero-Odasso et al., 2017). The premise of DTG is such that doing two tasks simultaneously (i.e. a cognitive and gait task) will result in greater utilization of cognitive resources than either tasks performed alone (Ebersbach et al., 1995). In line with previous studies, we showed that the Dementia group had poorer gait performances in the STG and DTG conditions compared to healthy control and SMC groups (Muir et al., 2012;Montero-Odasso et al., 2017). It is now fairly well-established that changes in gait parameters such as stride length, frequency and variability are key motor changes that occur alongside changes with cognitive deficits. Additionally, longitudinal studies have suggested that these changes in gait parameters are predictive of cognitive declines that may have clinical utility in early diagnosis of dementia (Cedervall et al., 2014;Montero-Odasso et al., 2017). Although participants with dementia performed worse compared to the other groups, it should be noted that there were no differences in gait spatiotemporal parameters between the STG and DTG tasks in participants with dementia. While it is unclear as to why this may be the case, a likely reason was that the participants with dementia strategized performing only the gait task, rather than devoting equal attention to both cognitive and gait tasks simultaneously.
Our results further indicated that participants with SMC did not show any difference in gait performance of the STG and DTG tasks compared to the healthy controls. As SMC has been previously shown to predict conversion from normal cognitive functioning to dementia (St John & Montgomery, 2002;Wang et al., 2004), the ability to use functional or behavioural tests that is strongly associated with SMC remains inconclusive. While there is some evidence to show associations between gait parameters (i.e. gait variability) and SMC (Beauchet et al., 2017), other studies showed no associations between gait measures, but rather an association between the fear of falling and SMC (Sakurai et al., 2017). It, therefore, remains to be seen if more robust associations can be made based on functional or behavioural tests in people with SMC. At least in our sample, spatiotemporal gait parameters and MoCA scores of the SMC group were comparable with healthy controls that precludes us from making any conclusions of gait deficiencies in people with SMC.

| Left PFC activation during STG and DTG
While we did not find any differences in gait measures of STG and DTG between SMC and healthy controls, we did observe a significant increase in left PFC activation during the DTG but not the STG task between SMC and healthy controls. To the best of our knowledge, this is the first study to demonstrate such a novel finding. In people with Dementia, there is strong evidence to suggest that atrophy and reduction in left PFC function leads to memory decline and increased caregiver burden (Maillet & Rajah, 2013;Matsuoka et al., 2018). Additionally gait studies further implicate the role of the left PFC in gait control and balance (Harada et al., 2009;Meester et al., 2014). In our age-matched healthy controls, an increase in left PFC activation during DTG compared to STG was expected and indicates an increased utilization of attentional resources to cope with the demands of performing two tasks simultaneously. A recent study by Wagshul et al. (2019) suggested that the increase in PFC activation is likely to stem from neural inefficiency (i.e. increase PFC activation with no concomitant increase in functional performance) that was associated with reduced grey matter volume of the PFC with ageing. This effect has been consistently shown in healthy older adults during DTG (Doi et al., 2013;Ohsugi et al., 2013;Mirelman et al., 2017) or balance (Lin et al., 2017;Teo et al., 2018) tasks that require increased attentional demands. What was, therefore, surprising was that people with SMC displayed a higher level of left PFC activation, significantly greater than that of healthy controls, despite no differences in age or MoCA scores between both groups. We postulate that the increase in left PFC activity in people with SMC may indicate some form of neural compensation, as proposed by the compensation-related utilization of neural circuits hypothesis (CRUNCH; Reuter-Lorenz & Cappell, 2008), where additional cognitive resources are needed to maintain DTG performance. If this is indeed the case, then perhaps this pattern of increased neural activity of the left PFC in people with SMC may represent very early central changes in cognitive processing that precedes any observable change in motor function or behaviour.
In the Dementia group, we observed two key findings that lend further support to our observations in the SMC group. Firstly, during STG performance, left PFC activation was significantly greater, concomitant with poorer gait performance in the Dementia group compared to healthy control and SMC groups. Similarly, this is in line with the CRUNCH framework whereby compensational and/or additional resources may be required to perform basic tasks, when cognitive abilities are compromised. Secondly, we further observed that under DTG conditions, left PFC activity in the Dementia group was reduced compared to healthy controls or SMC. This likely indicates a failure to utilize attentional resources in situations whereby cognitive demands of the task become increasingly overwhelming. A recent systematic review of fNIRS studies in people with mild cognitive impairment or dementia showed decreased resting-state and task-related oxyhaemoglobin response (Yeung & Chan, 2020), supporting our hypothesis of reduced prefrontal cortex activation when executive function is challenged.

| Limitations and future directions
To this end, there are several limitations that we need to acknowledge. Firstly, as we only used a single-channel fNIRS system, our study lacked the spatial resolution to measure from other key brain regions, such as the right PFC, to validate other known ageing models such as the HAROLD (hemispheric asymmetry reduction in older adults) model, which could help explain the pattern of left PFC activity that we observed. Secondly, our study may have lacked the statistical power to detect differences in STG and DTG gait performances between SMC and healthy control groups. Thirdly, our study did not include single task performance for the cognitive task or did we record the number of correct responses during DTG, which may provide an estimate of cognitive interference. However, considering that the focus was in people with SMC, and that both STG and DTG performance and total counting responses were similar between Controls and SMC, we suggest that cognitive interference (if any) is likely to be similar. Thus, the increase in fNIRS responses during the DTG in SMC is therefore likely to represent higher dualtask demand associated with SMC. Fourthly, our Dementia group has significantly less participants, and were older in age compared to the SMC and healthy control groups. It is also likely that the secondary task alone, during DTG, may have been perceived as more difficult to perform in the dementia group, which may be additional to the need to concurrently perform a motor and cognitive task. Finally, the lack of differences in gait parameters between Controls and SMC could be due to the lack in sensitivity of stride length and velocity in detecting specific cognitive changes. Recent evidence from a longitudinal population-based study suggests that double support time was predictive of memory decline in a large sample of older adults compared to other gait parameters (Jayakody et al., 2019). We acknowledge that these limitations may potentially bias our findings. Future studies should consider using a montage sufficient to measure from both left and right PFC and with other brain regions (e.g. sensorimotor and premotor areas) to provide a more comprehensive understanding of brain activation during STG and DTG tasks.

| Conclusion
In conclusion, our results indicate that while participants with SMC do not exhibit functional gait deficits in STG and DTG tasks, an increase in activity of the left PFC during DTG may be indicative of additional recruitment of attentional resources to maintain DTG performance that is comparable to healthy controls. In addition, a reduction in left PFC activation during DTG observed with the Dementia group may be indicative of failure to utilize attentional resources during tasks that have a higher cognitive load. Overall, our findings suggest that central changes may precede functional impairments in SMC, and that fNIRS during DTG may be able to detect early attentional changes, which may lead to detriments in gait seen in people with Dementia.