Experiment protocols for brain-body imaging of locomotion: A systematic review

Introduction Human locomotion is affected by several factors, such as growth and aging, health conditions, and physical activity levels for maintaining overall health and well-being. Notably, impaired locomotion is a prevalent cause of disability, significantly impacting the quality of life of individuals. The uniqueness and high prevalence of human locomotion have led to a surge of research to develop experimental protocols for studying the brain substrates, muscle responses, and motion signatures associated with locomotion. However, from a technical perspective, reproducing locomotion experiments has been challenging due to the lack of standardized protocols and benchmarking tools, which impairs the evaluation of research quality and the validation of previous findings. Methods This paper addresses the challenges by conducting a systematic review of existing neuroimaging studies on human locomotion, focusing on the settings of experimental protocols, such as locomotion intensity, duration, distance, adopted brain imaging technologies, and corresponding brain activation patterns. Also, this study provides practical recommendations for future experiment protocols. Results The findings indicate that EEG is the preferred neuroimaging sensor for detecting brain activity patterns, compared to fMRI, fNIRS, and PET. Walking is the most studied human locomotion task, likely due to its fundamental nature and status as a reference task. In contrast, running has received little attention in research. Additionally, cycling on an ergometer at a speed of 60 rpm using fNIRS has provided some research basis. Dual-task walking tasks are typically used to observe changes in cognitive function. Moreover, research on locomotion has primarily focused on healthy individuals, as this is the scenario most closely resembling free-living activity in real-world environments. Discussion Finally, the paper outlines the standards and recommendations for setting up future experiment protocols based on the review findings. It discusses the impact of neurological and musculoskeletal factors, as well as the cognitive and locomotive demands, on the experiment design. It also considers the limitations imposed by the sensing techniques used, including the acceptable level of motion artifacts in brain-body imaging experiments and the effects of spatial and temporal resolutions on brain sensor performance. Additionally, various experiment protocol constraints that need to be addressed and analyzed are explained.


. Introduction
About 28% of American adults older than 50 in the general community presented with impaired locomotion and its prevalence increased with age (p < 0.001) (Mahlknecht et al., 2013). The causes of impaired locomotion are divided into neurological (e.g., Parkinson's disease, stroke, Multiple Sclerosis, and dementia) (Allali et al., 2018;Buckley et al., 2019) and/or musculoskeletal drivers, such as arthritis and cardiovascular conditions (Blyth et al., 2019;Andonian and Huffman, 2020;Minetto et al., 2020). Among these neurological and musculoskeletal impacts, previous research has identified different patterns of gait disorders, such as parkinsonian (De Bartolo et al., 2020;Guayacán and Martínez, 2021), frontal (Hülser et al., 2022), or spastic gait (Muñoz-Lasa et al., 2019;Norbye et al., 2020). Furthermore, researchers have argued that locomotion should be defined as a syndrome for pre-clinical outcomes, such as motoric cognitive risk syndrome for pre-dementia (Xiang et al., 2021;Li et al., 2022a). Because of the neurological and musculoskeletal correlates of locomotion, rehabilitation treatments have been explored to utilize locomotion training, such as treadmill gait exercise, to improve patients' musculoskeletal ability and further induce neurological benefits (Smania et al., 2011;Hornby et al., 2016;Bassiri et al., 2022). Therefore, each human locomotion contains unique features of neural and musculoskeletal drivers, clinical conditions, the complexities of human development and aging, and signifiers of physical activity for health and wellness (Runge and Hunter, 2006;Pons et al., 2013;Kerkman et al., 2018).
Understanding the neurological and musculoskeletal correlates of locomotion is a pivotal need to pave the way for successful rehabilitation, improve performing locomotion tasks, especially in older adults, or create a digital twin of humans while doing a locomotion task (Dai et al., 2022). Various techniques, such as electroencephalography (EEG), functional near-infrared spectroscopy (fNIRS), and magnetic resonance imaging (MRI), have been developed and deployed to study age-related changes and specific diseases in the neurological and musculoskeletal correlates (Lewis et al., 2019). These various techniques enable us access to the movement phenotypes, such as brain structures (Chenausky and Tager-Flusberg, 2022), the functional substrates (Magrinelli et al., 2021), and motion signatures (Klibaite et al., 2022), in locomotion control during different experimental settings. However, these neuroimaging techniques present advantages and limitations, such as sensitivity to motion artifacts (Abtahi et al., 2020;Bonnal et al., 2022), portability (Sejdić et al., 2019), and spatial and temporal resolutions (Martini et al., 2020;Kumar and Michmizos, 2022). These limitations set up constraints on the design of the experimental protocols. For instance, MRI studies mainly investigated imagined locomotion rather than real-world locomotion (Stolbkov et al., 2019;Skinner et al., 2022). Although the sensitivity of fNIRS and EEG to artifacts has been improved, still most existing studies focus on low-intensity movement (e.g., walking) rather than high intensity, such as running.
To overcome these challenges and limitations of traditional brain imaging techniques, most recent research argued that brainbody imaging techniques that enable simultaneous measurements of dynamics of brain activities and body movements could be a promising method to reveal the profound relationship between brain, body, and behavior (Makeig et al., 2009;Gramann et al., 2010;Gwin et al., 2010;Wagner et al., 2012). The brain/body imaging techniques integrated portable devices, reliable sensing methods against motion artifacts, and capable of monitoring brain activities and locomotion with appropriate temporal and spatial resolutions, and sophisticated data analysis approaches for multimodal data preprocessing and curation. Thus far, some specialized centers and clinics such as GE HealthCare (HealthCare, 2023), SCL Health (Group, 2023), Stratus (Stratus, 2023), Zeto (Zeto, 2023), CMS (CMS, 2023), CNS (Calyx, 2023), NordicNeuroLab (NordicNeuroLab, 2023), and NIRx (NIRx, 2023) have adopted these brain/body imaging tools. With multiple-channel EEG (more than 64) capturing brain activities, these tools track spatial information of brain correlates of human motion and provide highprecision data at a sampling rate of 200-5,00 Hz (Cortney Bradford et al., 2019). Simultaneously, they utilized standard locomotion tools, such as optical motion capture systems (Divya and Peter, 2022), inertial measurement units (Khaksar et al., 2021), and EMG sensors (Hallett et al., 2021) to capture the locomotion data at the same sampling rate as brain activity data. As a result, sophisticated data analysis approaches have been developed to study the relationship between brain, body, and behavior, such as cognitive-motor interference and coherence (Zhu et al., 2022).
However, before these brain-body imaging tools can truly be adopted for clinical use, their effectiveness must be carefully assessed. From a technical point of view, reproducing human locomotion experiments has been problematic due to the lack of protocols, standardization, and benchmarking tools, which ultimately impairs the evaluation of previous research quality and validation of prior understandings (Parmentier et al., 2020;Kameli et al., 2021). Notably, only a few studies focused on developing standard protocols and related design methodology as an open research issue. Therefore, this paper focuses on reviewing the settings of existing experiment protocols, such as locomotion intensity, duration, distance, brain sensor technologies, and corresponding brain activation expressions. First, we develop a conceptual framework to identify the design methodology of experiment protocols. Limitations of each brain-body imaging technique and the corresponding constraints on protocol design are then characterized and mapped into the scope of the systematic review approach. Next, we review existing studies that implement various types of experiment protocols, demonstrating the current gaps in design methodology. Finally, metrics for evaluating the research quality and implications for reproducing prior knowledge are proposed.
Overall, we aim to provide a roadmap for the future development of locomotion analysis methods based on brainbody imaging techniques, including highlighting current progress, identifying various constraints, and suggesting potential research directions. The main contributions of this review paper are to: • Establish intrinsic links between neurological and musculoskeletal correlates of locomotion characteristics and quantifiable measures that brain-body imaging tools can capture; • Review existing experimental protocols for studying neurological and musculoskeletal correlates of locomotion; • Examine the feasibility of replicating the experiments in the laboratory systems, and finally, • Identify gaps and lay out a roadmap for the design methodology of experiment protocols.

. Methodology
The design methodology of experimental protocols for studies on neurological and musculoskeletal correlates of human locomotion is based on careful considerations of cognitive and locomotive demands, observation constrained by the sensing technology, the pathology derived from the research interests, and energetic costs invested by the human subjects. Ideally, the cognitive and locomotive demands are designed in specific thresholds to stimulate certain intertwined relationships between neurological and/or musculoskeletal activities. However, the sensing technology's thresholds of these demands are constrained, including sensitivity to motion artifacts, portability in different environments, and spatial and temporal resolutions for capturing dynamics. For instance, current sensing techniques, such as EEG and fNIRS, claim they have lowered their sensitivity to motion artifacts. Still, most researchers design low-intensity locomotion (e.g., slow walking) to avoid difficulty in denoising efforts. Sometimes, in high-intensity locomotion experiments (e.g., running), the data become useless due to a high amount of movement artifacts (Gwin et al., 2011). Figure 1 illustrates the conceptual framework of the design methodology of experimental protocols.
We developed a systematic search of existing studies based on the conceptual framework. To find the papers, these keywords were searched for in Google Scholar: human brain locomotion/locomotion, OR EEG, OR nirs, OR MRI/ Brain-Body Imaging. Then, all the papers were examined, and the related papers were added to the tables of this paper. The criteria for a paper to be reviewed are: 1) it must examine the brain's signals while the participants perform a locomotion task.
2) The experiment participants should be healthy; the results of this systematic search are shown in Figure 2. Although exploring walking disorders is highly valuable, there are thorough reviews for each locomotion disease. Then, the critical information on protocol design has been extracted and inserted into the tables. In this respect, these parameters in each protocol have been extracted: the type of the surface or the device used for experimenting (e.g., overground and treadmill), the speed of performing the task, the distance that participants have moved, the duration of the task, the type of the sensors used in the experiment (it is more focused on the sensors to read the brain's data), number and age of the participants, special conditions of each research, and the contribution of the research.
All the results were converted to the same units to make the protocols comparable. Accordingly, the distance is stated in meters (m), the speed is in kilometers per hour (km/h), and the duration of the task is in minutes (min); otherwise, the unit is stated.

. Locomotive and cognitive demands
To examine neural activities during human locomotion, single tasks without cognitive demands (e.g., walking) and dual tasks with cognitive demands (e.g., talking while walking) are examined. In this paper, the tasks are divided into walking, running, cycling, and .

FIGURE
Mobile body imaging study selection: Preferred items for systematic reviews and meta-analyzes (PRISMA) flow diagram. Dual-Task-Walking (DTW). The abbreviations used in this paper are explained in Table 1.
• In the participant section, M shows the number of male participants and the rest are Females.
• a ± b is used to describe the participants' age statistics in which a is mean and b is standard deviation. • If participants are older than 60, they are considered as OA.

. . Type : Walking
Walking has been examined in different forms: actual walking, imaginary walking, and simulated walking. Nonetheless, the common point between these studies is the activation of the brain and involvement of higher cognition control areas (Al-Yahya et al., 2011). The studies that have investigated the walking as the human locomotion are shown in Table 2.

. . Type : Running
Only one paper was found that merely focused on examining running as a locomotion task and studying brain activities. This paper's information is described in Table 3.

. . Type : Cycling
Cycling is another locomotive task to study the brain correlates of locomotion. The locomotive demands span from low to high intensity, with various cycling speeds, while few studies gave self-paced instructions to participants. Most studies expected to utilize the locomotive demands to stimulate the dynamics of brain activities for identifying neurological and musculoskeletal benefits. The research studies related to the cycling are demonstrated in Table 4. . . Type -: Walking-running In this section, three research were found that have investigated both walking and running their studies, which meet the defined criteria for research selection in this review. These studies are investigated in Table 5. . . Type -: Walking-cycling Only one paper shown in Table 6 was found that has examined both walking and cycling.
. . Type : Dual walking task-spontaneous locomotive and cognitive demands By switching from single tasks to dual tasks, the agerelated gait changes are more distinguished (Beurskens and Bock, 2012) the reason roots in the fact that the cognitive resources should compensate for the motor impairments (Mirelman et al., 2017). Because of this reason, in this type of locomotion task, usually both young adults (YA) and old adults (OA) are examined. The Dual task walking studies are presented in Table 7. . Results

. . Task analysis
The filtered papers based on the mentioned criteria are 102 papers. Most of the papers, 65, conducted research on low-intensity locomotion, walking, only 4 papers examined running, 17 papers examined cycling, and 18 papers conducted the experiment on dual task, which considers the locomotive and cognitive demands simultaneously. Figure 3 shows the visualization of the tasks, the surface that the experiment has been conducted on, and the brain sensor that has been employed for the experiment. In this regard, the most outer layer of the circle shows the investigated locomotion tasks (walking, running, cycling, and dual-task walking). The middle circle demonstrates the type of the surface (Tr: treadmill, IM: imaginary, OG: overground, Erg: Ergometer bicycle, B: on a board). The type of brain sensor used in the research are shown in the most inner circle (EEG, NIRS, MRI, and PET).

. . Analysis of locomotive intensity
Among the researches that have declared the intensity (e.g., speed) that locomotion has been performed, the papers that have stated their speed (or an average of speed) have been compared with each other. However, the researches that the speed was not defined quantitatively and constant (e.g., the speed was dependent on the length of the participants' feet length Wagner et al., 2012Wagner et al., , 2014Seeber et al., 2014) were excluded from the visualization. Accordingly, the rotational velocities when the brain-body imaging experiment is using an ergometer or a pedaling board are visualized in Figure 4. The experiment using 60 rpm in cycling has been considered a reference speed for most researchers. Besides the rotational velocities of Figure 4, the linear speed for SW and DTW on different surfaces (i.e., Tr and OG) for YA and OA are shown in Figure 5. The bigger the bulb, the more repetition on reported speed. For example, walking at the speed of 2 Km/h is the main standard for researchers on the treadmill when the participants are YA. In addition, in each experiment surface, a trend of using a higher speed for YA compared to the speed used for OA can also be observed in the reviewed papers in this figure. For instance, in DTW OG, the maximum speed reported for YA is 4.6 Km/h, although the maximum speed reported for OA in DTW OG is about 3.8 Km/h.   To avoid the effect of fatigue on the experiment, participants could walk whenever they wanted. Accordingly, the time of experiment could last between 60 to 120 min.
Walking involves baseline theta band activity that significantly increases with loss of balance. Also, in the right-handed and footed participants, the left SMC plays a larger role in sensing loss of balance during walking than the right SMC. Subjects walked for a total of 1 h at 0.75 m/s. Subjects alternately walked in 7.5-min blocks of time at 0% grade and at 15% grade, for a total of 30 min at each condition Comparison of walking with 0% and 15% grades shows greater gamma power during level walking in the left sensorimotor and anterior cingulate clusters. Also, comparison of frequency activation of the artifacts during walking conditions shows that the differences between walking conditions were cortically driven rather than a residual artifact of the experiment.
Treadmill -2 12 min combination of different reactions to the appeared obstacles.
EEG (32 channel), 7 IMU 5M, range: 24-29 To create obstacles, in one scenario a line laser is projected over the treadmill to simulate the appearance of the obstacle. In the second scenario, a screen placed in front of the treadmill changes its color to simulate the appearance of the obstacle EEG role can be developed to detect the unexpected obstacles as EEG potential over the fronto-central area of the subjects' brain change. An accuracy of 79.5% was reported for obstacle detection.   To divide the gait cycle, 3D markers in five positions were employed Activation of cerebral cortex during gait phases is examined. During pre-swing and terminal-stance, cerebral cortex is more actively involved in the control of eight examined muscles.

Nordin et al. (2019a)
Treadmill -1.8, 3.6, 5.4, 7.2 3 min for each speed and considering rest between each experiment Dual layer EEG and 8-channel EMG 9 (6M), 27 ± 4 As the speed is close to running, the subjects were asked to walk even in high speeds. A standing baseline trial was recorded prior to changing speed experiments.
Dual-layer EEG isolates the changes in sensorimotor electrocortical dynamics across walking speeds. Also, dual-layer EEG is beneficial to remove residual artifacts while gait speeds change. In addition, a correlation between different walking phases and alpha/beta spectral power is drawn. Goni-ctrl: the avatar was driven by a goniometer. goniometer sensor placed at hip, knee, and ankle joint angles. BCI-ctrl: the avatar was controlled by BCI. The participants could see the avatar on the 52-inch TV.
In closed-loop walking using BCI and avatar, cortical involvement during walking increases as α/µ are subdued in the posterior parietal cortex and inferior parietal lobe. Low γ modulations in the anterior cingulate cortex and superior temporal gurus may show the increasing voluntary control of human gait.  Each subject participated in four sessions (standing and three speeds). Each session repeated a run 10 times analysis.
The centro-parietal was not covered with the sparse 14-channel electrodes and no spectral changes were observed in SMC. steady-state visual-evoked potential-based BCI can be used to mimic natural walking using consumer-level EEG. the walking was synchronized with a series of cue pacing cue tones and thus the step rate and length were changed based on the pacing cue tempo Analyzing beta band power in the right dorsolateral prefrontal cortex shows two recognizable patterns. One pattern may help in starting and executing the movement and the other one has control and inhibition functions (Wagner et al., 2016). Also, in the posterior medial frontal cortex, an EEG step-cue delay negatively is generated with a peak at 250 ms after anomalous cue tone onsets (Wagner et al., 2019b).
De Sanctis et al.  In conditions that require adjusting the steps based on the visual input, µ, β, and lower γ frequencies in pre-motor and parietal cortices are reduced, which shows these brain areas' activation increase. This activation is higher compared to mirror feedback and a visual attention task, which may indicate additional motor planning and visuomotor processing.
Alchalabi et al.  Although walking is automatic, brain signals, especially the cortical area, can be classified to walking and non-walking signals reliably with high speed. Also, actual waking classification has a higher accuracy compared to imaginary walking.
Nojiri and Iwane Brain scanning while participants imagine Standing, initiating gait, walking, walking with obstacles. The results of these conditions were compared to a rest (control) condition When the cognition demand of the task and the need for processing sensory information increase, the higher brain centers become more engaged. The experiment has two modes: 1organized: sequential activation of muscles related to walking 2-chaotic: non-gait-like pattern activation of muscles. Each experiment repeated 9 times.
There is no difference between activation of chaotic and organized patterns of stimulation. Activation pattern of mental imagery and gait-like plantar stimulation are similar especially in SMA-proper bilaterally and right pre-SMA. The experiment includes ten fMRI sessions consisting of three blocks. Walking and stand positions were randomly shown.
In major gait-related task especially at initiation of a gait, SMA is activated. During termination and stepping over an obstacle, a significant visuomotor network is required.  The visual stimulus was generated by Daz 3D. Before starting experiment, high-resolution structural brain scan obtained for anatomical structure of fMRI data The executed task is coupled with more activation in M1 and the medial cerebellum while imagined task has higher activation in somatosensory cortex, M1, and lateral cerebellum.  The experiments had three conditions: 1-normal walking 2-cognitive interface task: press a button based when a high-pitch sound is heard and ignore the low-pitch sound 3-motor interface task: preventing connection of the rings placed on a stick Doing tasks that involve the brain's motor interface reduces gait velocity and stride length and increases the stride time and temporal-spatial variability. These changes don't occur in tasks require cognitive interface involvement.

. . Analysis of participants' age
To analyze the age of the participants, when the age of the participant is unknown, the research is excluded in the visualization of Figure 6. In this regard, there is a significant gap in the ages between 40 and 60 years old as only studies had participants in this range of age.

. . Analysis of locomotion duration
The duration of locomotion task varies from seconds to hours (Figure 7). In this regard, when it comes to single walking, the experiment on the treadmill with a duration of 12 min is the primary source for experiments. To make the results of studies with different duration comparable, the effect of fatigue on the muscles needs to be studied, and one solution could be recording the data after the locomotion task for a specific time.

. . Analysis of locomotion distance
Besides the duration, another parameter for assessing the locomotion demands is the distance when the participants were asked to walk overground. The walk distance falls in a range of 1 to 150 meters. For better demonstration, only the distances up to 30 meters are shown in Figure 8.

. . Brain activation in brain-body imaging experiments
In this section, the corresponding brain activation to the locomotion is presented. Comparison of the brain activation areas for the locomotion tasks are shown in Figure 9. In this figure, we have only shown the results of the reviewed papers that have specifically mentioned the brain regions' activation sources and their effects. However, if a study generally describes the cortex area is excluded from this visualization. In this figure, ∝ is the proportional symbol, ↑ shows increasing, and ↓ shows reduction of an item. For instance, when the OA's walk speed increases, oxyHB in the Supplementary Motor Area increases.

. Discussion
In this section, the configurations of locomotive and cognitive demands, the research interests regarding neurological and musculoskeletal drivers, and the observational constraints from the sensing techniques are discussed concerning the impact of brainbody imaging sensors on the design methodology of experimental protocols for measuring dynamics of brain, body, and behavior. Since few papers in the field have adopted a conceptual framework to evaluate the quality of the neurological and musculoskeletal correlates of human locomotion extracted using various methods, it is not easy to compare these state-of-the-art methods. Therefore, in this section, we focus on establishing the elements of the proposed  Attentional focus as assessed three times during performing exercise by indicating a point in an analog range of completely on task to completely off task.
Based on oxyHb in right dorsolateral PFC and right medial frontal cortex, the brain's region associated with mental effort is disengaged with the brain's region linked to resting activity in order to keep mental resources for the maintenance of exercise.  The participants had at least 4 h cycling training per week within the last 6 months before the experiment. Improvement in cycling training is closely related to brain cortical activity. Also, the higher cadence, the greater brain functional response. A standardized warm-up (i.e., 5 min at 1 W/kg, 2 min at 3 W/kg, and 1 min at 5 W/kg) and a 5-min recovery period was used before doing the task.
Besides showing the possibility of localizing brain cortical activity during pedaling activity, it is shown that motor cortex activity increases with the increasing of power level and significantly mirrored muscle activity.

3.0
The speed adjustment is based on prior instruction.
cortical activation during bicycling and walking are compared. During movement, while bicycling is associated with stronger decrease in beta power, walking is associated with alpha power reduction.
conceptual framework and discuss how to assess the neurological and musculoskeletal measures extracted from brain-body imaging sensors for further clinical use.

. . Neurological and musculoskeletal drivers
According to our systematic review results, most studies followed the philosophy of medical diagnosis that is rooted in conducting statistical analysis between different groups (e.g., young vs. old, control vs. patient). These groups were asked to perform specific behavioral protocols under which their performances were supposed to show neurological and musculoskeletal differences reflected by the brain-body imaging data. Therefore, various experimental protocols were explored to capture the differences in statistical norms among groups (Kashuba et al., 2020;Warmerdam et al., 2020). The typical statistical analysis approaches include comparisons of statistical norms (e.g., p-value, effect size) and classifications based on machine learning (e.g., accuracy, precision) (Figueiredo et al., 2018;Hatami et al., 2019;Patil et al., 2019;Hausmann et al., 2021). Few studies adopted representation learning methods based on deep neural networks to explore patterns and features of brain and motion signals to show the group differences (Vásquez-Correa et al., 2018;Talo et al., 2019;Song et al., 2021). Furthermore, specific quantitative assessments, such as clinical outcomes or symptoms, were identified as ground truth so that the same reference could examine both groups. However, no consensus on experimental protocols caused challenges in determining reliable ground truth or references. Otherwise, few studies provided explanations and rationales for the design methodology of behavioral experiments.
Although it lacks standardization of experimental protocols, existing studies still generated consistent conclusions on neurological and musculoskeletal correlates of human locomotion. For example, most studies revealed that the active brain regions during various walking protocols include the motor, sensory, and prefrontal cortexes. The brain areas involved in walking behavior span SMA, premotor cortex, sensorimotor, M1, and left and right prefrontal cortex. Notably, the age-related changes showed that the involvement of PFC increased among old adults, especially during high-speed walking. When the subjects need to adjust their postures for cycling, the involvement of PFC decreases in order to keep mental resources for the maintenance of these additional requirements during exercise. The PFC has been implicated in planning complex cognitive behavior, especially in the resting status. Few studies explored the running protocols; therefore, there is little consistent knowledge regarding neurological and musculoskeletal correlates of human running. It is noteworthy that when subjects are required to conduct spontaneous cognitive tasks, the PFC activation increases in all age populations, and its increase is more significant in older adults with and without cognitive impairment. These conclusions have motivated many hypothesis-driven research projects on disease-related changes in neurological and musculoskeletal correlates of human locomotion.
Besides the consistent understanding of neurological correlates of human locomotion, mixed results and conclusions exist due  Each round started and ended with 20 s of standing quietly, with the instruction to refrain from talking and moving the head. More complex walking: negotiating with two physical obstacles during walking. DTW: walking while talking (subtracting 3 s from a 3 digit, predefined number) Needed cognition increases for both young and older people but for older people is more significant. Gait variability in older adults increases with the increase in pre-frontal activation. pre-frontal activation in older people is higher indicating older people rely more on their cognitive resources during walking. Neural activation in the PFC increases with task complexity, similarly, in both younger and older adults.  Walking has lower functional connectivity between SMC areas than standing.
Treadmill -1.5, 3, and 4.5 6.5 min 32-channel EEG 7, Range: 25-33 The dual task was counting the green letters appearing during 3 s on a screen in front of the participants. Task: 0.2 s flashing light, 0.1 s between two flashes and 1 s interval and repeating that for 12 times for 25 target letters.
Feasibility of suing P300 during walking while recording EEG signals from parietal and occipital areas is shown, which can be beneficial for ambulatory conditions.  OA performed five blocks of the response inhibition task while sitting, 9 or 10 blocks while walking and two blocks only walking. YA completed three or four blocks sitting, a minimum of four blocks walking slowly (range: 4-8 blocks), at least four blocks walking quickly (range: 4-8 blocks) and two blocks of each speed walking without the task. Task: speeded visual Go/No-Go task By examining the variability and time of stride in different configurations, only the OA's accuracy drops significantly when performing inhibitory task while walking. Also, the brain's performance in YA is more modulated than OA according to the EEG data of cortical activities. The reason might be an age-associated loss in flexible resource allocation across multiple tasks.
De Sanctis et al. When walking while doing another task, stride time in walking grows by increasing the cognition load of the task. Also, by increasing the age, the cortical motor behavior shifts from automatic to more controlled process.
Mazurek et al. Based on studying visual cortex in visual oddball discrimination during standing and walking, Weighted Phase Lag Index introduced as a potential method for recovering cognitive brain dynamics in the presence of gait-related artifacts.

FIGURE
Task-Surface-Brain sensor visualization: the most outer layer shows the type of the locomotion task; walking, running, cycling, and dual-task walking. The middle layer shows the type of the surface used for the body-brain imaging experiment, which are Tr: treadmill, IM: imaginary task, B: Board with pedals, OG: overground, Erg: Ergometer cycling. The most inner layer shows the type of the brain sensor used in the brain-body imaging.

FIGURE
Comparing locomotive intensity in two types of brain-body imaging tasks performed with di erent rotational speeds: ) cycling ) walking on a board with pedals. Only younger adults have been employed to perform these tasks.
to the heterogeneity of participants' neurological expressions, musculoskeletal variance under imprecise experimental protocols, and observation constraints by the sensing techniques. For instance, some studies concluded that the cerebral cortex controls multiple muscles hierarchically through a few synergies during walking. In contrast, few studies argued that the role of cortical control during walking might not be valid due to the motion artifacts of EEG. Furthermore. Subjects that conducted actual walking were more accurately classified than subjects with imaginary walking. Subjects that walked on a treadmill showed different band activities captured by EEG compared to subjects who walked over the ground. These inconsistent results require further investment in establishing protocols, standardization, and benchmarking tools, which also motivated this research work.

. . Locomotive and cognitive demands
The primary assumption underlying the experimental protocols is that the locomotive and cognitive demands could stimulate changes and patterns in neural activity, which the brain-body sensors could pick up (Ladouce et al., 2019). Thus, the knowledge of the simulation mechanisms further guides the treatment and rehabilitation strategies for potential clinical .
outcomes. However, most experimental protocols involve lowintensity locomotion tasks, such as walking and cycling, while few studies conducted high-intensity locomotion tasks, such as running. In addition, little research explained the design methodology of the experimental protocol, especially the research hypothesis of which types of neural activities might be expected under the experimental configurations, such as duration, intervals, and frequency. Also, conducting behavioral studies on human subjects is always challenging, even more, if considering psychological factors such as state of mind, concentration, and technical dexterity. Therefore, some studies integrated cognitive demands into locomotive tasks, such as avoiding obstacles, determining walking directions, or following an avatar. Other studies developed a dual-task paradigm to examine the involvement of cognition in human locomotion. As illustrated in Figure 1, the conceptual framework argued that more work is needed to examine the simulation mechanisms, thus developing a better design methodology for experimental protocols. The systematic review results showed that several challenging questions remain in the research field. First, benchmarking the locomotive and cognitive demands will be needed. Locomotive tasks have several configurable parameters, such as duration, intervals, and frequency, which need to divide into several levels or intensities. Cognitive tasks also have configurable parameters, such as type and complexity, depending on working memory and executive functions. For instance, for a specific subpopulation's demographic, locomotion and cognition capacity, researchers in designing experimental protocol need a basic understanding of which levels of demands might stimulate the anticipated brain activity. Knowledge of locomotion disorder patterns and corresponding brain areas has helped establish experimental protocols for several neurological diseases, such as parkinsonian gait for Parkinson disease (Ghai et al., 2018) and NIH cognitive toolbox for cognitive impairment (Gershon et al., 2010). Second, manipulation of the tasks needs more research investment. Existing studies rarely consider how the sequential configuration of the tasks stimulates neural activity. Most protocols conducted a heuristic-designed sequence of locomotive tasks and anticipated the brain-body sensors could capture the subtle changes or patterns of neural activities. However, the loop from demands to brain and musculoskeletal activities to sensors, in Figure 1, shows that manipulation of the demands could generate richer information than the heuristic-designed protocol. Third, the experimental protocols should be easily administered to avoid confusion and distraction for participants. Few studies gave a clear description of how the protocols are being instructed. Cueing the participants toward specific tasks could have influenced the expected neural activity and cognitive performance. Therefore, most researches argued that when the experiment protocol is administered and instructed by assistance of a computer is less likely that neural activity and cognitive performance get adversely affected compared to the case that the experiment is controlled and instructed by interference of a human. Hence, replication of previous experiment protocols that were controlled by a human as examiner with a computer-assisted administration provides a high-quality data with removal of the human interference affects (Vrana and Vrana, 2017;Dror, 2020;Young et al., 2022). Moreover, computer assistance in the cognitive load assessment could make it feasible to conduct  the experiment at places out of the clinics and laboratories. These places (e.g., participants' home or a local clinic) are accessible to the participants and individuals are comfortable to perform the experiment with assistance of a computer without interference of an examiner. As positive side effects of the computer-assisted experiment administration, high-quality and cost-effective patient care could be provided (Porrselvi, 2022;Young et al., 2022).

. . Observation constraints by the sensing techniques
Another dimension of our systematic review results is illustrating the impacts of the observation constraints by the brainbody sensing techniques. Previous work has reviewed and discussed the advantages and disadvantages of these techniques, including tolerance of motion artifacts and spatial and temporal resolutions. However, most studies reviewed in this work did not explain the rationale for selecting specific sensing techniques in experimental protocol design. According to our review results, it is obvious to see the impacts of sensing techniques on the research results and generated knowledge. Therefore, we summarize these impacts for enabling guidelines for future researchers to design experimental standards and benchmarks.

. . . Tolerance of motion artifacts
Prior knowledge of sensing techniques has concluded that among three of them, MRI has the lowest tolerance to motion artifacts, EEG less, and fNIRS the highest. Despite efforts in signal processing and denoising to improve each technique's tolerance to motion artifacts, most studies still followed the knowledge and showed different experimental results. Studies that adopted MRI has considered its limited tolerance of motion artifacts and mainly designed imaginary locomotion or simulated surrogate tasks rather than actual locomotion in their experiments (Stolbkov et al., 2019;Amemiya et al., 2021). The central assumption of . /fnins. . these studies is that neurological correlates of these imaginary or surrogate tasks are closely related to neurological activation during actual locomotion. However, our summarized results in locomotion tables and Figure 9 showed that these imaginary or surrogate tasks could not simulate the complex dynamics the brain must execute to adjust and maintain the musculoskeletal patterns during actual locomotion. Especially studies that utilized EEG have shown that actual walking stimulated brain activation patterns that could be classified with higher precision than imagery walking. Therefore, studies that adopted EEG sensors designed a more comprehensive range of locomotion tasks, from imaginary walking and cycling to running. Furthermore, the fNIRS studies are preferred in high-intensity locomotion experiments. More than 50% of studies reviewed in this work that conducted cycling and running tasks adopted fNIRS sensors. Sensitivity of EEG data to motion artifacts is a research concern and the results of some previous published research due to not considering an extensive removal of motion artifacts has been questioned. For instance, the reported EEG results indicating the changes in high-gamma frequency band during walking (Gwin et al., 2011) could be caused by motion artifacts as well (Castermans et al., 2014). To address the considerable drawback, different denoising methods to employ during or after data collection have been developed. During data collection, artifact removal is associated with hardware modifications. In this respect, one effective way has been introduced to separate electrophysiological signals from non-neural signals. To this end, in one approach, two layers are used below the EEG cap. A silicone layer is used on top of the scalp to block electrophysiological signal. Then, a simulated conductive scalp with similar impedance to human scalp is used to measure the voltage differences generated by gait dynamics (Snyder et al., 2015). In another approach, well-known as a EEG dual electrode design, simultaneously EEG data and isolated motion artifacts are recorded by pairs of the electrodes that are electrically independent and mechanically coupled Clark et al., 2020). After the data collection, denoising is coupled with software data processing. In this regard, Independent Component Analysis (ICA), low, high, and band band pass filter are routinely applied. Besides these remedies, ICA-based methods such as adaptive ICA mixture model algorithm (Palmer et al., 2006), extended infomax ICA (Lee et al., 1999) and multiple mixture ICA (Allen et al., 2000) approaches are utilized . Moreover, a developed conductive head phantom and robotic motion platform has emerged as a powerful tool to analyze the artifact removal methods through generating a ground truth for EEG signal. This device is used to evaluate artifact removal methods such as dual-layer EEG and Artifact Subspace Reconstruction (Richer et al., 2020). Also, this device is used to show that electrodes with larger surface reduces the electrodes vulnerability to motion artifacts (Symeonidou et al., 2018). In addition, this device is employed to assess the effect of the motion artifacts parameters such as frequency and amplitude (Oliveira et al., 2016).

. . . Spatial and temporal resolutions
MRI has the highest spatial resolution among the three sensing techniques while fNIRS has the lowest (only centimeters under the skull) (Li et al., 2022b). Accordingly, MRI studies discussed their results with a detailed description of the brain cortex, such as SMA and the dorsal premotor cortex. Nonetheless, lacking the portability feature has significantly limited the application of the MRI in the brain-body imaging of locomotion studies mostly to imaginary task and thus, this scope of study has been deprived from the high spatial resolution of the MRI imaging technique. On the other hand, although EEG has lower spatial resolution than MRI, appropriate source localization approaches could improve its spatial resolution at a cortical level. When source localization is used, the spatial resolution of the fNIRS and EEG are comparable. However, EEG has the highest temporal resolution among the three methods and permits the highest and the most precise data brain investigation compared to the fNIRS and MRI methods. Thus, Looking into the results summarized in our work, we can see that studies that adopted EEG sensors provided the neurological and musculoskeletal correlates on a fine-grained temporal scale, such as neural activation patterns during different movement periods (e.g., swing, stance). These features has made the EEG the most popular technique in the explorations of the brain-body imaging of human locomotion.

. . . Miscellaneous constraints
We also observed other constraints in the design methodology of experimental protocols, such as imbalanced ages of participants. For example, most studies recruited participants aged below 40, while fewer recruited participants older than 60. A significant age gap (40-60) among the participants involved in these studies existed. However, the adults within this age range are baby boomers in the U.S., a large sector of the population, a group deemed by . /fnins. . the Centers for Disease Control and Prevention (CDC) as mid-life stage adults. This age population is facing an increase in clinical and other preventive services to support maintaining good health into older age (Pearson-Stuttard et al., 2019;Doubeni et al., 2021), which means this age group should be a critical population for studies on neurological and musculoskeletal correlates of human locomotion.
Although there might exist real-world challenges to recruiting this specific age population, additional investment in balancing the age distribution of the participants is needed to fulfill a crosslifespan understanding of human locomotion and its neurological and musculoskeletal correlates.

. Conclusion
This paper explores the topic of the design methodology of experimental protocols that aim to study neurological and musculoskeletal correlates of human locomotion using brainbody sensing techniques. The review of many types of neural activities stimulated by human locomotion demonstrates the importance of quantitative analysis using brain-body sensors in potential healthcare applications. By reviewing the current design methodology of experimental protocols, this paper illustrates that the protocol design significantly impacts the experimental results due to the heterogeneity of participants' neurological expressions, musculoskeletal variance under the imprecise locomotive and cognitive demands, and observation constraints by the sensing techniques. Finally, the impacts of the experimental protocols are discussed by reviewing the practical issues to provide implications and guidelines for future researchers to design experimental standards and benchmarks.
Brain-body imaging of human locomotion is a vast area of research. This paper focused on a significant research issue: how to reproduce human locomotion experiments. Therefore, we conducted a systematic review of existing experiment protocols to examine various settings and conditions, such as locomotion intensity, locomotion duration, locomotion distance, brain sensing technologies, and corresponding brain activation expressions. In future work, technologies for locomotion sensing and their advantages and disadvantages will be further examined and discussed. Also, upper-limb locomotion, such as shoulder, elbow, wrist, and finger movement, is a broad study that will be examined in another work. Finally, as the participants of this systematic review are healthy, similar brain-body imaging experiment exploration for neuromechanical disorders would be a valuable work to extend.

Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

Author contributions
JG designed the study, created the research question, and finalized the manuscript. SK implemented the study process, collected and analyzed the data, and initiated the manuscript. NJ developed the study motivation and significance, guided the data analysis, and revised the manuscript. All authors contributed to the article and approved the submitted version.