Brain markers of cognitive-motor control of gait function in Parkinson’s disease: links with physical activity and aerobic fitness

Abstract

[Purpose]

This narrative review synthesizes magnetic resonance imaging (MRI) evidence of (i) the associations between physical activity and/or aerobic fitness and brain markers related to cognitive function in Parkinson’s disease (PD); and (ii) MRI-derived brain markers linked to gait impairment in PD. We also propose a conceptual framework describing how physical activity or fitness may support the cognitive-motor networks associated with gait.

[Methods]

PubMed and Web of Science were searched for MRI studies examining the relationships between physical activity/fitness and cognition-relevant brain markers as well as MRI studies linking brain markers to gait outcomes (including cognition-related gait measures and freezing of gait).

[Results]

Fourteen observational studies met our inclusion criteria. Higher physical activity and/or fitness were seen to be generally associated with better cognitive performance and MRI features suggestive of preserved brain integrity and connectivity within the cognitive-motor networks. Regarding gait, the reviewed evidence suggests a shift toward greater reliance on cognitive-motor control to compensate for impaired automatic motor control, which may occur when this compensatory cognitive control becomes insufficient under increased task demands.

[Conclusion]

Physical activity and fitness may contribute to gait function in patients with Parkinson’s disease by supporting the cognitive-motor integration required for walking. Longitudinal and interventional studies using ecologically valid paradigms that approximate real-world (including dual-task) walking, ideally combined with wearable neuroimaging approaches, are needed to clarify how changes in physical activity or fitness are related to neural markers of gait.

INTRODUCTION

Parkinson’s disease (PD) is the second most common neurodegenerative disorder after Alzheimer’s disease [1]. It is characterized by progressive dopaminergic degeneration in the substantia nigra, leading to cardinal motor symptoms such as tremor, rigidity, and bradykinesia. Among these, gait impairment is a particularly disabling feature, as it directly increases the risk of falls and compromises functional independence. Increasing evidence indicates that gait dysfunction in patients with PD cannot be explained solely by motor deficits. Cognitive impairment is closely linked to impaired gait control, particularly under complex or attention-demanding conditions, such as dual-task walking, in which walking is performed simultaneously with another cognitive or motor task [2-4]. These observations have prompted a conceptual shift toward viewing gait dysfunction in PD as a manifestation of disrupted cognitive-motor integration rather than a purely motor phenomenon.

In daily life, gait is not an automated motor behavior but a cognitively demanding task that requires continuous monitoring of the environment, goal management, and adaptive motor planning [5]. In individuals with PD, impairments in executive function, attention, and processing speed further exacerbate gait instability, particularly when cognitive demands are increased. This vulnerability is most evident during dual-task conditions, where competition for limited cognitive resources leads to reduced gait speed, increased variability, and an increased risk of falls [6]. Neuroimaging studies increasingly suggest that these deficits reflect alterations in distributed neural networks involving the prefrontal, parietal, and subcortical regions that jointly support cognitive control and motor execution. However, although the behavioral association between cognition and gait is well established, the underlying neurobiological substrates linking these domains remain incompletely understood.

Physical exercise and habitual physical activity have been widely used as non-pharmacological approaches to improve motor and cognitive outcomes in patients with PD [7]. Accumulating evidence suggests that higher levels of physical activity, reflecting behavioral engagement in bodily movement, and aerobic fitness, reflecting physiological capacity commonly assessed by ${\dot{\mathrm{VO}}}_{2\max }$ or ${\dot{\mathrm{VO}}}_{2\mathrm{peak}}$, are associated with more efficient functional organization of cognitive-motor neural networks, including reduced prefrontal cortical activation during walking and enhanced functional network integration, potentially reflecting compensatory or reserve-related mechanisms [8,9]. Although these variables may have partially overlapping neural relevance, they reflect different behavioral and physiological dimensions. Consistent with this notion, individuals with PD who engage in higher levels of physical activity demonstrate superior gait performance under both single- and dual-task conditions compared to their less active counterparts [10]. Importantly, emerging neuroimaging studies indicate that these behavioral benefits may be supported by structural and functional brain markers—such as preserved subcortical volumes and enhanced connectivity within cognitive and sensorimotor networks—suggesting that physical activity and fitness may modulate the neural systems supporting cognitive-motor integration in PD.

Despite growing interest in the relationships among physical activity, cognitive function, and gait performance in PD, the neurobiological mechanisms underlying these domains remain insufficiently characterized. In particular, evidence regarding magnetic resonance imaging (MRI) brain markers associated with physical activity or fitness, and their relevance to cognitive-motor interactions, has accumulated in a fragmented manner. Therefore, the purpose of this narrative review is to synthesize observational evidence on MRI-based brain markers associated with physical activity and aerobic fitness in individuals with PD, and to examine how these brain markers relate to cognitive function and gait performance. We propose a conceptual framework linking physical activity and aerobic fitness to brain MRI markers relevant to cognitive-motor integration and gait function in patients with PD. We first review MRI-based brain markers associated with physical activity and aerobic fitness and then consider how these markers may relate to cognitive function and gait performance.

METHODS

This narrative review aimed to explore the relationship between exercise capacity and cognitive function in individuals with PD, with a particular focus on neurobiological mechanisms assessed using MRI. To identify relevant literature, a comprehensive search of two major databases, PubMed and Web of Science, was conducted, covering studies published within the last 15 years (2010-2024). The search strategy included combinations of the following keywords: “Parkinson’s disease,” “exercise,” “cognition,” and “MRI.”

To capture the range of research examining the relationship between physical activity or fitness levels and brain structure and function in patients with PD, observational studies meeting the following criteria: (1) included individuals diagnosed with Parkinson’s disease; (2) assessed motor or non-motor (especially cognitive) outcomes in relation to physical activity or exercise; (3) included MRI-based assessments of brain structure or function; and (4) were peer-reviewed and published in English.

The initial database search identified 95 relevant records (PubMed, 52; Web of Science, 43). After excluding 20 duplicates, 75 articles remained. During the screening phase, 32 records were excluded, including review articles (n = 9), case or intervention studies (n = 22), and non-English language publications (n = 1). Subsequently, 43 full-text articles were assessed for eligibility. Based on the exclusion criteria, studies that did not report relevant outcomes (n = 3), did not include patients with PD (n = 18), or were unrelated to the research questions (n = 8) were excluded, leaving 14 studies in the final review (Figure 1). The full texts of eligible articles were reviewed in detail, and key information was extracted, including study design, sample characteristics, outcome measures (motor, cognitive, or neuroimaging), and major findings. The included articles are summarized in Supplementary Table 1.

Figure 1. Flow Diagram of Study Selection.

RESULTS

Physical Activity, Cognitive Function, and Brain MRI Markers in PD

Cognitive impairment is a common non-motor symptom of PD that affects attention, executive function, memory, and visuospatial ability [11]. These deficits may occur even in the early stages of the disease and are associated with progression to Parkinson’s disease dementia (PDD), which significantly affects quality of life [12,13]. Although dopaminergic degeneration in the nigrostriatal pathway plays a role in cognitive decline, additional mechanisms, including cholinergic system deficits, alpha-synuclein accumulation, and structural brain changes, also contribute to the pathophysiology of cognitive impairment [14-16].

Recent observational studies have examined whether physical activity or aerobic fitness is associated with MRI brain markers relevant to cognitive function. In a study by Petkus et al. (2021), 45 participants with PD completed a graded exercise test to estimate ${\dot{\mathrm{VO}}}_{2\max }$ and underwent structural MRI. The results demonstrated that higher aerobic fitness was significantly associated with better episodic memory, and that this relationship was mediated by greater thalamic volume. These findings suggest that aerobic fitness may preserve subcortical structural integrity, which in turn may support cognitive performance. Similarly, Jarrahi et al. (2021) investigated the association between aerobic fitness and intrinsic functional connectivity using resting-state functional MRI in 24 individuals with PD. Estimated ${\dot{\mathrm{VO}}}_{2\max }$ was positively associated with connectivity in the sensorimotor network (SMN) and cognitive networks subserving executive and memory function in the posterior default mode network (DMN) and salience network (SN). These results imply that aerobic fitness may contribute to the maintenance of integrative network communication in patients with PD.

The multimodal neuroimaging study of Sacheli et al. (2018) used positron emission tomography and task-based functional MRI to compare 17 patients with PD classified as either habitual exercisers or sedentary. The habitual exercisers exhibited greater dopamine release in the caudate nucleus and increased ventral striatum activation during a reward-processing task. These neurobiological changes were accompanied by lower apathy and bradykinesia scores, suggesting that regular physical activity may enhance both motivational and motor systems via dopaminergic modulation. Donahue et al. (2022) conducted a study with 96 PD participants using 7-day wrist-worn actigraphy to quantify physical activity. They reported that moderate-intensity physical activity (MIPA), defined as a weighted weekly total of > 150 min of activity at 3.00-5.99 metabolic equivalents (METs) plus twice the time spent in activity at ≥ 6.0 METs, but not light-intensity physical activity, was significantly associated with better global cognition, executive function, visuospatial abilities, and memory. Resting-state fMRI analyses in a subsample of 50 participants revealed that individuals meeting the World Health Organization’s MIPA guidelines (≥ 150 min/week) demonstrated higher functional connectivity in regions spanning the hippocampus, brainstem, cingulate, and parietal cortex. Moreover, functional connectivity significantly moderated the relationship between physical activity levels and cognitive performance, suggesting that neural network integration may mediate the cognitive benefits of physical activity.

Taken together, these findings suggest that both physical activity and aerobic fitness are positively associated with structural (e.g., thalamic volume) and functional (e.g., network-level connectivity) brain MRI markers related to cognitive performance in PD. Moreover, evidence of dopaminergic modulation suggests additional neurochemical mechanisms through which regular physical activity may exert its effects. Although the cross-sectional designs of these studies preclude definitive conclusions regarding causality, the overall pattern supports the hypothesis that MIPA may contribute to the preservation of cognitive function in PD. This effect may be mediated by multiple neurobiological pathways, including the maintenance of subcortical structures, enhanced functional network integration, and modulation of dopaminergic circuits, all of which can be assessed using current neuroimaging techniques. An overview of the proposed pathways linking exercise-related factors (physical activity and aerobic fitness) and cognitive outcomes through key neuroimaging markers of PD is shown in Figure 2. These same brain markers may also be relevant to gait function in PD, particularly under conditions requiring greater cognitive-motor integration, thereby providing a conceptual bridge to the following section.

Figure 2. Neurobiological Links between Fitness-related Factors and Cognition in Parkinson's Disease.

The diagram depicts the pathway through which fitness-related factors (physical activity and aerobic capacity) influence behavioral performance (global cognition, episodic memory, executive function, and visuospatial abilities) through specific brain MRI markers.

Cognitive-Motor Network Alteration Underlying Gait Dysfunction in PD

Gait disturbance is a prominent and functionally disabling feature of PD and is traditionally attributed to dopaminergic deficits within the basal ganglia circuitry [21]. However, accumulating evidence from structural and functional neuroimaging studies indicates that gait impairment in PD reflects alterations not only in subcortical motor circuits, but also in distributed cortical regions implicated in executive control and motor planning [22,23]. This relationship is particularly evident during complex walking conditions such as dual-task walking, which increases cognitive demands and challenges cognitive-motor integration in PD. This has led to a conceptual shift toward understanding gait dysfunction in PD as a consequence of the disruption of cognitive-motor neural networks.

Structural and functional changes in prefrontal regions have also been implicated in the regulation of gait and cognition in PD. Fling et al. (2016) used diffusion MRI to examine the integrity of the genu of the corpus callosum, which mediates connectivity between the bilateral prefrontal cortices, in patients with PD and in patients with frontal gait disorder (FGD), a condition often referred to as lower-body Parkinsonism due to its predominant effects on the lower limbs. They reported that patients with FGD showed reduced cognitive performance, gait speed, and a wider-based gait compared to PD patients. Cognitive and gait impairments were associated with disrupted interhemispheric connectivity mediated by the genu of the corpus callosum. This finding suggests that reduced prefrontal interhemispheric connectivity, along with cognitive dysfunction, may contribute to gait disturbances in PD.

Supporting this interpretation, Albrecht et al. (2022) identified three subtypes of PD (cognitive-dominant, motor-dominant, and motor-cognitive) in patients using random forest clustering based on multimodal data comprising clinical, motor, and cognitive assessments, as well as structural and resting-state MRI. The cognitive-dominant subtype was characterized by prominent cognitive impairment accompanied by structural and functional alterations in the frontoparietal regions. In contrast, the motor-dominant subtype was characterized by motor impairment without any observable brain alterations. Interestingly, the motor-cognitive subtype was characterized by widespread changes in both brain structure and function, as well as impairments in both motor and cognitive abilities. Specifically, the cognitive-dominant group showed the least impairment in motor variables, as reflected in their relatively preserved gait speed during normal walking (127.3 cm/s), compared to the motor-dominant (120.75 cm/s) and motor-cognitive (106.60 cm/s) groups. However, when walking under dual-task conditions compared to normal walking, the cognitive-dominant group showed a comparable decrease in gait speed (-5.2 cm/s) to that of the motor-cognitive group (-6.20 cm/s). By contrast, the motor-dominant group showed minimal changes in gait speed (+1.05 cm/s) during the dual-task condition. This suggests that cognitive impairment may affect gait performance in individuals with PD, even in the context of relatively mild motor symptoms.

Task-based fMRI studies have demonstrated that patients with PD exhibit reduced activation in the globus pallidus and related ganglia during imagined gait tasks, particularly in complex walking scenarios. These findings suggest impairment in the internal generation and sequencing of movement, potentially reflecting dysfunction in the basal ganglia-thalamocortical pathways [26]. In contrast, during complex walking tasks in virtual environments, increased activation was observed in the prefrontal and parietal cortices. These patterns have been interpreted as compensatory engagement of higher-order cognitive control systems in response to impaired automatic motor execution [27]. Figure 3 summarizes brain activity patterns in patients with PD during complex walking tasks.

Figure 3. Brain Activation of PD Patients during Complex Walking Task (fMRI).

PD patients exhibit increased activation (+) in the prefrontal and parietal cortices during complex gait tasks. This recruitment of higher-order cognitive areas is thought to be a compensatory mechanism for the functional decline in the basal ganglia region.

Notably, Shine et al. (2013) examined neural activation patterns in individuals with and without freezing of gait (FoG) during a virtual reality-based walking task under cognitive load. They reported that patients with FoG exhibited significantly reduced activation in the pre-supplementary motor area (pSMA), anterior insula, and ventral striatum, even after controlling for motor severity and affective symptoms. This hypoactivation pattern suggests an impaired ability to efficiently engage the cognitive control network (CCN), especially during dual-task conditions, potentially contributing to freezing episodes. In healthy individuals, successful task engagement typically involves coordinated suppression of the DMN and activation of attention and motor planning regions. However, in FoG patients, persistent DMN activity during walking may disrupt this balance, contributing to freezing episodes and impaired gait automaticity. In support of these findings, Lench et al. (2021) compared the resting-state supplementary motor area (SMA) connectivity between PD patients with and without FoG. They found that individuals with FoG exhibited increased functional connectivity between the SMA and cerebellum, along with reduced connectivity to the orbitofrontal cortex. This pattern suggests a shift toward overreliance on motor-executive pathways and reduced top-down cognitive modulation, which may contribute to impaired gait automaticity and freezing episodes.

Expanding on this viewpoint, recent task-based fMRI studies have provided evidence for the contribution of emotional circuits to gait dysfunction in PD. In particular, Sarasso et al. (2021) examined the neural responses of PD patients with FoG during the observation of FoG episodes. Compared to healthy controls, patients with FoG exhibited increased activation in the anterior medial prefrontal cortex, a region implicated in self-referential emotional processing, and reduced activation in the dorsomedial prefrontal cortex and bilateral hippocampus, which are associated with cognitive empathy and contextual memory. These findings suggest that gait-related motor inhibition in patients with FoG may be modulated by self-focused emotional engagement rather than by generalized limbic hyperactivation. In addition, the heterogeneity of neural network alterations in patients with FoG has become an area of growing interest. A study by Martens et al. (2018) investigated functional network signatures of FoG using data-driven clustering of functional connectivity profiles. They identified distinct subtypes of FoG patients, characterized by differential coupling patterns between cognitive, limbic, and motor networks. Strong anti-coupling between cognitive cortical regions and caudate nucleus was observed across subtypes, independent of clinical severity, suggesting a shared disruption in fronto-striatal integration. These findings emphasize that FoG is not a unitary phenomenon but that it reflects a spectrum of network-level dysfunctions involving the affective, executive, and sensorimotor systems. The key task-related activation and connectivity findings in patients with FoG described above are summarized in Figure 4.

Figure 4. Alterations in Brain Activation and Functional Connectivity Patterns Associated with FoG Patients.

(A) Compared to PD patients without FoG, FoG patients show deactivation (-) in the pre-SMA, insula, and ventral striatum during complex walking task. (B) FoG patients show anti-coupling functional connectivity between the caudate nucleus and cognitive control related cortical regions, including the ACC, DLPFC, and PPC. (C) Compared with PD patients without FoG, FoG patients show reduced functional connectivity between the SMA and the orbitofrontal cortex, and increased connectivity between the SMA and the cerebellum (rs-fMRI).

Structural neuroimaging studies have further identified the anatomic correlates of gait adaptability. In particular, greater cortical thickness in the prefrontal cortex and supplementary motor area (SMA) was found to be associated with pronounced improvements in gait parameters following treadmill training. These findings suggest that preservation of frontal cortical integrity may support neuroplastic adaptation to rehabilitation in PD [32]. In addition to cortical contributions, structural markers of subcortical degeneration have been linked to gait decline in patients with PD. In particular, a longitudinal study by Wilson et al. (2021) demonstrated that smaller baseline volumes of the nucleus basalis of Meynert (NBM), a major source of cortical cholinergic innervation, were associated with progressive impairment of gait parameters over a three-year period, including increased step time variability and reduced swing time. These results suggest that non-dopaminergic mechanisms, particularly those involving cholinergic degeneration within the basal forebrain, may play a significant role in the trajectory of gait dysfunction in PD.

Collectively, the evidence reviewed here suggests that gait dysfunction in PD is driven by distinct, yet interacting, pathophysiological processes involving both subcortical and cortical brain regions. Functional neuroimaging studies indicate that individuals with PD increasingly rely on prefrontal and parietal cortices during complex walking, reflecting a shift from automatic to cognitively mediated gait control. Structural studies have further demonstrated that preserved cortical thickness in frontal motor-related regions and the integrity of cholinergic nuclei, such as the NBM, are associated with more stable gait performance and slower progression of gait impairment. These findings support the reconceptualization of PD-related gait disturbance as a multi-dimensional phenomenon encompassing the motor, cognitive, and neurochemical domains and suggest the need for integrative assessment approaches that can adequately reflect these interrelated components.

DISCUSSION

Across the reviewed observational evidence, higher physical activity and aerobic fitness were found to be associated with favorable structural and functional brain correlates in patients with PD [17-20]. Specifically, higher physical activity was associated with greater functional connectivity across distributed regions, including the parietal cortex, cingulate cortex, hippocampus, and brainstem [20]. This pattern may be consistent with more coherent large-scale network organization in systems supporting executive control, attention, and memory—domains that are closely linked to cognitive and cognitive-motor performance in PD.

Similar physical activity-related functional connectivity patterns have also been reported in healthy older adults, suggesting that some aspects of these associations may reflect shared mechanisms of activity-related brain reserve [34]. At the same time, the reviewed evidence indicates that the specific neuroanatomical correlates of aerobic fitness may vary in patients with PD, potentially reflecting disease-related vulnerability within the subcortical motor-cognitive circuitry. For example, ${\dot{\mathrm{VO}}}_{2\mathrm{peak}}$ has been associated with episodic memory and thalamic volume in PD [17]. In contrast, aerobic exercise interventions in healthy older adults have been linked to increased hippocampal volume [35]. Together, these findings raise the possibility that aerobic fitness-related structural correlates may differ according to the pathological context and study design, suggesting the need for future studies that directly compare patients with PD and healthy controls under harmonized conditions.

Building on these observations, we next consider plausible neurobiological pathways that could link physical activity and aerobic fitness to cognitive and cognitive-motor outcomes in PD, with a particular focus on vulnerability within the cognitive-motor network. Normal walking depends on the interaction between intentional/cognitive processes mediated by cortico-basal ganglia loops and more automatic gait execution supported by subcortical and brainstem mechanisms. Brainstem-related circuits contribute to automated motor programs, including gait rhythm generation and regulation of muscle tone, whereas cortico-basal ganglia loops play a key role in cognitive control and action selection [36]. In patients with PD, dysfunction within these circuits may compromise gait automaticity, thereby increasing reliance on cortical cognitive control to maintain stable walking.

Consistent with this framework, as gait impairment worsens, activity in motor-related regions often tends to decrease [26,28]. However, patients with PD may show relatively greater engagement of cognitive-control cortical regions than healthy older adults, an activation pattern [27] that is commonly interpreted as compensatory recruitment, although it may also reflect reduced neural efficiency under heightened task demands. From this perspective, the anti-coupling between cognitive-control cortical regions and the caudate nucleus observed in patients with FoG [31] could be interpreted as an indication that cognitive-striatal compensation is insufficient or inconsistently engaged when gait-control demands are high, rather than as a network abnormality alone. Accordingly, gait impairment in patients with PD may relate not only to degradation of automatic motor circuitry but also to limitations in the cognitive-motor network’s capacity to compensate for the loss of automaticity.

In this context, physical activity and aerobic fitness may represent potential modulators of vulnerable cognitive-motor networks in PD. Synthesizing existing evidence, higher physical activity and fitness levels are associated with better cognitive function, altered striatal function, and enhanced large-scale network connectivity, which may provide a neurobiological basis for more efficient handling of the increased cognitive demands arising from impaired gait automaticity. Thus, physical activity and aerobic fitness may help support gait function by facilitating the cognitive-motor integration processes required for walking, particularly under conditions that increase cognitive load.

This study has several limitations that warrant consideration. First, most included studies were observational or cross-sectional, limiting causal inference regarding the proposed physical activity-cognition-gait axis and leaving open the possibility of residual confounding and reverse causality. Second, MRI-based neuroimaging constraints—particularly the need to minimize head and body motion—have led many investigations to rely on virtual, simulated, or imagined gait paradigms, which may not fully capture the dynamic neural processes and sensorimotor integration required for real-world, ground-based walking. Third, heterogeneity in physical activity/fitness assessment, cohort characteristics, and analytic approaches may contribute to inconsistent neural correlates across studies. Despite these limitations, physical activity and exercise interventions may warrant consideration as non-pharmacological approaches with potential relevance to compensatory neural processes and cognitive-motor integration related to gait stability. Future studies should prioritize harmonized longitudinal and, where feasible, interventional designs that combine ecologically valid gait assessments (including higher cognitive-load conditions such as dual-task walking) with wearable neuroimaging modalities (e.g., fNIRS) to clarify whether changes in physical activity are associated with corresponding alterations in the cognitive-motor neural markers and improvements in complex gait performance. More specifically, such approaches may help determine whether exercise-related neural changes translate into improvements in cognition and gait and whether changes in physical activity or aerobic fitness are accompanied by corresponding changes in neural activity during real-world walking. Multimodal approaches integrating wearable activity monitoring with neuroimaging methods may be particularly valuable for clarifying these brain-behavior relationships in ecologically valid settings.

Supplementary Material

Supplementary Table 1.

Magnetic resonance imaging studies on cognition and structural/functional brain changes in Parkinson’s disease