Anticipatory reaching motor behavior characterizes patients within the Alzheimer’s disease continuum in a virtual reality environment

Abstract

Background

Alzheimer’s Disease (AD) is characterized by progressive declines in cognitive and motor functions, impairing daily activities. Traditionally, AD diagnosis relies on cognitive assessments, but emerging evidence highlights motor function deficits as early indicators of AD and Mild Cognitive Impairment (MCI). These motor declines, which often precede cognitive symptoms, include slower and less accurate reaching movements. This study explored reaching actions in a Virtual Reality (VR) environment in AD and MCI patients to identify motor deficits and their link to cognitive decline.

Methods

The study involved 61 right-handed participants (19 AD, 21 MCI, and 21 healthy age-matched controls), screened for cognitive health using a Mini-Mental State Examination (MMSE). Participants performed upper-limb motor tasks (sequentially reaching targets) in a Virtual Reality (VR). Kinematic data was recorded and analyzed focusing on task success rate, frequency of anticipatory responses, and direction of anticipatory responses. Statistical analysis was performed using Generalized Linear Mixed Models to differentiate the three groups of participants based on performance metrics, anticipation behavior, and the correlation between anticipation rate and MMSE score.

Results

Both AD and MCI patients showed more anticipatory responses than healthy controls (HC), inversely related to success rates and cognitive function. AD patients exhibited lower success rates and a higher frequency of anticipatory responses, often biased toward previous trial targets, suggesting impaired motor planning or difficulty adapting to new cues. MCI patients showed an intermediate pattern, with more anticipatory responses than HC but comparable success rates. These results highlight the crucial role of anticipatory behavior in motor task performance, with AD patients displaying the most pronounced deficits.

Conclusions

This study highlights significant impairments of reaching movements in AD patients, particularly in terms of anticipatory behavior and success rates. The observed deficits suggest that kinematic metrics could serve as early biomarkers for diagnosis and intervention. These findings emphasize the importance of combining cognitive and sensorimotor assessments for the early detection of AD-related motor dysfunctions. Additionally, they highlight the potential of VR-based motor rehabilitation as a promising approach to address sensorimotor deficits in the AD continuum, improving both motor and cognitive outcomes.

Introduction

Alzheimer’s Disease (AD) is characterized by a progressive decline in cognitive and motor functions, leading to significant impairments in daily living activities. As the prevalence of AD and mild cognitive impairment (MCI) continues to rise globally, there is a pressing need to develop reliable and sensitive methods for early diagnosis and monitoring of the progression of the disease [1]. Traditionally, cognitive assessments have been the cornerstone of diagnosing AD and MCI, focusing on memory, language, and executive functions [2]. However, recent studies have underscored the importance of motor function assessments as complementary tools in identifying and distinguishing these conditions [3].

Poirier et al. [4] and Wu et al. [3] have reviewed the literature on motor function deficits in AD and MCI, highlighting significant deterioration in fine motor control and coordination. This decline is evident in fine motor tasks such as handwriting [5,6,7] or finger tapping [8]. Importantly, these motor deficits often precede cognitive impairments [9], suggesting that motor function assessments could serve as early indicators of disease progression.

Gait analysis further supports the link between motor and cognitive decline. MCI and AD individuals demonstrate slower walking speeds and increased variability in double support time, which correlate with cognitive deterioration [10, 11]. Motor signs such as rigidity and impaired function in the lower extremities have been associated with not only cognitive and functional decline but also increased risk of institutionalization and mortality in AD patients [12,13,14]. Manual dexterity tasks also yield lower scores in individuals at higher risk for neurodegenerative diseases [1, 15], emphasizing the widespread nature of motor impairments in these conditions. Moreover, motor function uncertainty, especially in repetitive elbow flexion tasks, has been strongly linked to cognitive impairment [16].

In light of these findings, motor tasks have proven effective in distinguishing among controls, MCI, and AD subjects, offering a level of sensitivity comparable to cognitive tests of memory and language [17, 18]. Cognitive decline is also associated with generalized movement slowing in peripheral reaching, indicating possible visuomotor deficits in reach planning [19]. Specifically, reaching movements in AD patients are notably slower and less accurate, particularly in the absence of visual feedback [20, 21]. Reaching movements require precise coordination and control, thus providing a valuable window into the motor deficits associated with cognitive decline. These tasks necessitate the integration of sensory information processing and motor planning, which are often compromised in neurodegenerative diseases [22].

Despite the link between brain regions affected in AD and those involved in reaching and grasping actions, detailed studies on the planning of reaching movements in dementia are lacking. Examining fine motor features across stages of reaching could help fill this gap and reveal new therapeutic opportunities.

Previous studies on reaching movements in individuals with dementia were conducted in laboratory settings, which may not fully capture the complexities of real-life motor behavior. In contrast, immersive virtual reality (VR) environments offer a more realistic and dynamic setting that can significantly influence response behavior, particularly in patients with dementia. Hence, immersive VR has recently emerged as a promising tool for assessing and addressing motor deficits. VR systems provide controlled yet flexible environments that can simulate real-world scenarios, which are often challenging to replicate in laboratories. Furthermore, the immersive nature of VR increases patient engagement, which is crucial for successful assessment and rehabilitation [23,24,25,26]. By offering more interactive and motivating experiences, VR holds the potential to improve cognitive and motor rehabilitation outcomes, such as enhancing navigation skills and fostering more effective therapy for individuals with MCI or AD [27,28,29,30].

Therefore, the aim of this study was to examine the planning and execution of a reaching movement to better understand how neurodegeneration impacts them. We assessed reaching in a VR environment performed by AD, MCI, and age-matched healthy control (HC) participants. As we observed that AD and MCI participants struggled to wait for the go signal, we hypothesized that kinematic metrics of the anticipatory phase might characterize their motor impairment that could be used as biomarkers for early diagnosis and monitoring of the disease. Although this aspect is largely unexplored, we believe it could enhance existing knowledge and support new prognostic markers and interventions to improve life quality for affected individuals.

Finally, our approach is innovative as it leverages cutting-edge technologies to explore unique aspects of reaching behavior in individuals with AD and MCI. We believe the findings and results of this study will be of significant interest to both the clinical and motor control research communities.

Methods

Participants

61 individuals (31 females, 30 males; age: 72 ± 6 years) gave their written informed consent to participate in the study, which had been approved by the Ethical Review Board of Fondazione Santa Lucia (Prot. CE/PROG.820) in accordance with the Declaration of Helsinki. Among them, there were 19 patients affected by Alzheimer’s disease (denoted by AD), 21 Mild Cognitive Impairment subjects (denoted by MCI), and 21 healthy age-matched control subjects (denoted by HC). AD and MCI patient recruitment was performed according to current diagnostic criteria [31, 32] from the Memory Clinic of Fondazione Santa Lucia. A further inclusion criterion for the AD group was a Mini-Mental State Examination (MMSE) ranging from 14 to 26 at screening [33, 34]. MCI group corresponded to the early symptomatic phase of AD, characterized by episodic memory loss without general cognitive and functional decline. All patients had to be cognitively intact before the occurrence of cognitive or memory impairment. Major systemic and psychiatric disorders, other neurological conditions, and signs of concomitant cerebrovascular disease on MRI scans were carefully investigated and excluded in all participants. Moreover, HC had to achieve a MMSE score superior to 26.

We employed an inclusion criterion that none of the subjects had played sports involving ball catching at a professional level. Additionally, only right-handed individuals were selected for this study, with handedness determined using a questionnaire based on the Edinburgh Handedness Inventory.

Table 1 displays an overview of the number of participants enrolled in each group, their age, and their MMSE scores.

Table 1 Characteristics of the three groups of participants enrolled in the study

Apparatus

Upper-limb motor tasks were performed in a VR environment, using the HTC Vive Pro 2 (HTC, Taoyuan, TW) VR system. The VR system was composed of four devices: a headset, a controller, a tracker, and a base station. The headset displayed the virtual environment stereoscopically with a refresh rate of 90 Hz and a field of view of 120° and was integrated with earphones. During the task, the controller was held by the participants’ right hand and the tracker was attached to the participants’ right hand through a Velcro strip. Kinematic data were recorded by using the MPU-6500 Six-Axis (Gyro + Accelerometer) MEMS Motion Tracking devices embedded in the controller, in the tracker, and in the headset, in conjunction with the precise spatial tracking provided by the base station.

The virtual environment was created by implementing custom-made C# scripts in Unity (Unity Technologies, San Francisco, CA), a real-time 3D development platform and game engine. The software tool SteamVR (Valve Corporation, Bellevue, WA) was used to interface Unity with the HTC Vive devices. The C# scripts were coded ad-hoc to implement the experimental protocol and to collect and log kinematic data during the execution of the tasks, such as the position of the tracker and controller, and to log the time of specific events.

Experimental protocol

Participants were verbally instructed on how to perform the task. To make them better understand the task, some videos of a laboratory member performing the task were shown to them. Also, the VR scenario was presented to them. Participants were seated during the whole experiment to avoid any risk of falling due to any VR discomfort that may arise.

The virtual environment resembled the room in which the experiment took place to minimize any disorientation when entering the virtual environment. Participants sat on a chair that was positioned at a specific point in the experimental room to match the virtual environment. Figure 1A illustrates an exemplary subject.

Participants saw a cursor (a yellow sphere, diameter 5 cm) attached to the controller they were holding. They were instructed to first move the sphere to a home position, which appeared at the center of a grey circular board (placed in front of the subjects) as a larger blue sphere (diameter 9 cm). After a random time interval (between 1 and 1.5 s), the central sphere disappeared and concomitantly a blue sphere appeared at 1 of 8 possible target positions equally distributed along the edges of the board (at 30 cm distance from the center of the home sphere). To help participants in the reaching task, the targets did not appear in random order but in sequential order: the first one was the one at the top of the board and the successive ones appeared in the counterclockwise direction.

Then, as soon as the central sphere disappeared, participants had to move their cursor from the home position to the target position within 750 ms from the onset of the movement. The onset time was calculated online as the time instant at which the controller’s speed exceeded 5 cm/s or the controller was 5 cm far away from the home position. Initially, a 500 ms threshold was set, but pilot testing revealed that this was too restrictive, as participants rarely managed to reach the target within the given time. Hence, the threshold was adjusted from 500 ms to a more lenient 750 ms to better accommodate participants’ performance. If the participant did not reach the target position within 750 ms, the target sphere disappeared, and the trial was logged as unsuccessful. On the other hand, if the cursor reached the target within 750 ms, participants had to keep the yellow sphere inside the blue sphere, until the latter one disappeared (1s). Then, as soon as the target had disappeared, participants had to return the cursor to the home position.

Figure 1B illustrates the VR view of the participants at two different time instants. On the left, it shows the view when the participant places the cursor within the blue sphere in the home position. On the right, it shows the view at the go signal (i.e. appearance of the target blue sphere).

The protocol involved 10 blocks, after an initial familiarization block. Since there were 8 possible target positions, the total number of trials was 80. Between each pair of blocks, participants took a short break to avoid fatigue. During these breaks, participants were verbally asked about their physical well-being, including whether they experienced any dizziness, nausea, or other discomfort related to VR immersion. If participants reported any issues, the break time was extended to allow for adequate recovery. If the discomfort persisted, the task was terminated for the participants’ safety. This check-in process was also repeated at the end of the tasks to ensure their continued comfort. However, participants included in this study did not experience any of the discomforts. Actually, they were all engaged in the task and reported enjoying the VR environment.

Data analysis

Kinematic data recorded by the controller were analyzed in this study. Data analysis was performed with custom software written in MATLAB (MathWorks, Natick, MA). The position and orientation of the controller, sampled at 90 Hz by the HTC Vive system, were resampled at 100 Hz. Kinematic data were filtered by applying a low-pass zero-lag Butterworth filter, through the MATLAB filtfilt function (order 4 and cut-off frequency of 10 Hz).

Several temporal parameters of interest were extracted. The start time was defined as the time instant at which the trial began. The go time was the time instant at which the central sphere disappeared, and the target appeared. The onset time was the time instant at which the controller’s speed was 10% of its maximum. The impact time was the time instant at which the cursor hit the target (only in successful trials). The end time was the time instant at which the trial ended.

Participants’ performance was quantified by success rate, anticipation rate, and anticipation direction. Success rate was defined as the fraction of successful trials (i.e. the ones in which the cursor arrived at the target position within 750 ms). As participants were instructed to move the cursor only after the go signal, anticipatory responses were identified as the occurrence of a peak in the speed profile before the go signal. The onset of the anticipatory peak was computed in the speed profile as the time at which the speed was 20% of its maximum before the go signal. One of the parameters used to characterize anticipatory responses is the anticipation rate, which was defined as the fraction of trials displaying anticipation. Additionally, two displacement vectors were computed: the vector between the center of the home sphere and the position of the cursor at the time of maximum distance from the center of the home sphere during the time interval of the anticipation response (i.e. before the go signal), and the vector between the center of the home sphere and the center of the target sphere. The angle between the projections of these two vectors on the frontal plane was defined as the anticipation direction, introduced to characterize the aiming direction during an anticipatory response. Figure 1C displays the two vectors used to define the anticipation direction angle.

Fig. 1

(A) Illustration of an individual equipped with all the sensors employed in the study, poised to initiate the task within the experimental setup. The person is wearing the headset, is holding the controller with their right hand, and has the tracker attached to their right wrist. (B) Virtual Reality view of participants in two distinct temporal snapshots: immediately following the start time on the left and at the designated go time on the right. The Euclidean distance between the target position and the home positioning was always equal to 30 cm, regardless of the position of the target ball on the circular board. (C) Computation of the anticipatory direction. The red arrow represents the displacement of the cursor at its maximum distance from the home position during the anticipatory response. The blue arrow represents the direction of the target of the current trial. The angle θ between these two vectors indicates the direction of the anticipatory movement, i.e. the “anticipation direction”

Statistical analysis

Statistical analysis of the kinematic metrics was performed using Generalized Linear Mixed Models (GLMMs) in R (R Core Team, Vienna, Austria). GLMMs incorporate both fixed and random effects, making them particularly useful for analyzing binomial responses taking into account participants’ differences when analyzing single trials.

The statistical analysis was implemented to differentiate the three groups of participants on the different metrics of interest: the success rate, the anticipation rate, the task performance in the presence or absence of anticipatory behavior, and the correlation between anticipation rate and MMSE score.

The general equation used to test the difference between the groups has the form

$$\eqalign{& {Y_{ij}} = {\beta _{0i}} + {\beta _{group}}Grou{p_{ij}} + {\beta _{target}}Targe{t_{ij}} + {\beta _{block}}Bloc{k_{ij}} + {S_i} + {\varepsilon _{ij}} \cr} $$

where Y is the latent variable, β are the fixed-effects coefficients, S are the random-effects coefficients, and ${\varepsilon _{ij}}$are the residual error terms for each subject i and each trial j. Group is a categorical variable with three levels (AD, MCI, and HC). Similarly, Target is a categorical variable with 8 levels. Block is a continuous variable with 10 levels, one for each block.

To assess the effect of the presence of anticipatory behavior on performance we also tested the model in the form

$$\eqalign{& Succes{s_{ij}} = {\beta _{0i}} + {\beta _{ant}}An{t_{ij}} + {\beta _{group}}Grou{p_{ij}} + {S_i} + {\varepsilon _{ij}} \cr} $$

where ${Success_{ij}}$ is the binary variable indicating a successful reaching movement of subject i in trial j, and Ant is the presence of anticipation in that trial.  

Additionally, we tested another GLMM to investigate the dependency of the anticipation rate from the MMSE score and the group, as reported in the following equation:

$$\eqalign{& Anticipatio{n_{ij}} = {\beta _{0i}} + {\beta _{MMSE}}MMS{E_{ij}} + {\beta _{group}}Grou{p_{ij}} + {S_i} + {\varepsilon _{ij}} \cr} $$

To identify the model that best fitted the experimental data, i.e. to test whether to include interactions between factors and random effects, we adopted an iterative procedure [35,36,37]. According to the confirmatory hypothesis testing method [38], we iteratively compared models to assess whether it was necessary to include interaction terms or random-effects slopes.

Given that the direction of the anticipatory movements is a circular variable, standard statistics could not be applied [39]. Therefore, we tested whether the angle distribution for each group was different from a random uniform distribution by means of the Rayleigh Test of Uniformity which applies the necessary corrections for circular statistics.

Results

Qualitative differences between the three groups were investigated by visually comparing the paths and speed profiles of the controller. Figure 2 represents the paths and speed profiles of 3 exemplary subjects, one for each group. Speed profiles for each trial are aligned to the go time, thus the time interval before the vertical dashed line includes anticipatory responses. It can be observed that the trajectories are much less smooth for the AD participant with respect to the other two. Furthermore, the AD participants show the highest number of anticipatory responses, whereas almost no anticipatory behavior is present in the HC participants.

Fig. 2

Controller’s paths and speed profiles of the whole trial of 3 exemplary participants, one for each group. Each line represents one trial. The 8 grey circles represent the 8 target positions. Times in the abscissa of the speed profiles plots are referred to the go time, thus the area before the vertical dashed line (centered at the zero of the x-axis) comprehends the anticipatory responses

Accordingly, to assess whether a heightened propensity for anticipatory responses was a general characteristic of AD individuals, we examined anticipation rates in all three groups (Fig. 3). Consistent with our expectations, the lowest level of anticipatory behavior was observed in HC subjects, with a statistically significant difference compared both to the MCI group (β = 1.06; p < 0.05) and AD group (β = 1.86; p < 0.001). No significant difference between the MCI and AD groups was found.

Fig. 3

Anticipation rate of the three different groups of participants. For each group, the boxplot indicates the distribution of anticipation rates, while each point represents a single subject. There is a statistical difference between the HC and the MCI group (p < 0.05) and between the HC and the AD group (p < 0.001)

In addition, anticipation behavior is more likely to occur during the early blocks (as indicated by the negative slope of the block coefficient in the GLMM: β = -0.05; p < 0.01), and when reaching towards the first targets (as indicated by the negative slope of the target coefficient in the GLMM: β = -0.09; p < 0.001).

These findings align with those obtained from the analysis of the success rate (Fig. 4A). Notably, AD subjects exhibit the lowest success rate, with significantly different performance between HC and AD (β_group = -2.31; p < 0.001). Performance improved with repetitions, i.e. success rate increased with blocks (β_block = 0.18; p < 0.001), but this was true only for the healthy group as shown by the significant interaction between Block and Group (MCI × block: β = -0.15; p < 0.001; AD × block: β = -0.18; p < 0.001). Figure 4B shows the GLMM prediction for each individual in each group. It is worth noting that within a block, participants were more successful with the last targets of the sequence (β_target = 0.04; p < 0.05). These results suggest that participants displaying lower levels of anticipation are more likely to successfully accomplish the task.

Fig. 4

Success rate of the three different groups of participants. (A) Each point represents one single subject. (B) The success rate is shown for each participant in each block. Continuous lines indicate the prediction of the GLMM

Considering the preceding findings, the likelihood of trial success given the presence or absence of anticipatory responses was evaluated. To this aim, we tested a GLMM to investigate the dependency of the success rate from the anticipation rate and the group. Results indicate that there is a significant effect of anticipation, i.e., participants were more likely to reach the target if they did not move before the go signal (β = -1.19; p < 0.001).

Accordingly, Fig. 5 displays the coefficients of the GLMM for the group and the absence/presence of anticipatory responses. This figure highlights the inverse relationship between the probability of trial success and anticipatory behavior: individuals with enhanced levels of anticipation are less likely to achieve trial success. These findings emphasize how anticipatory behavior negatively affects task performance, with varying impacts observed across different participant groups.

Fig. 5

GLMM coefficient for the regression of the probability of a successful trial as a function of participant group and anticipation behavior. Each color represents a specific group of subjects, with lighter/darker shades indicating the absence/presence of anticipatory responses. Each marker denotes the average GLMM coefficient, and each error bar spans minimum to maximum values. A statistical difference in the GLMM regression coefficient was found between the anticipation condition and no-anticipation condition, regardless of the group of subjects (p < 0.001 for all 3 of them)

Figure 6 shows the correlation between the anticipation rate and the MMSE score. Consistent with expectations, elevated MMSE scores, indicative of less impairment, exhibit a corresponding decrease in anticipation rate (β = -0.15, p < 0.01). This observation underscores the broader relationship between cognitive function, as assessed by the MMSE, and anticipatory behavior.

Fig. 6

Correlation between anticipation rate and MMSE score for each of the three groups of subjects. Groups are denoted by colors: blue for HC, pink for MCI, and green for AD. Each point indicates one single subject

The direction of the anticipatory movements provided additional insights. In Fig. 7, polar histograms visually portray the distribution of direction angles of anticipatory movements on the frontal plane across all trials and subjects for each group. In the polar histograms, targets of the current, preceding, and subsequent trials are oriented at 0°, 345°, and 45°, respectively. We tested whether the angle distribution for each population was different from a random uniform distribution (Rayleigh Test of Uniformity). The distribution for HC participants, who exhibited the fewest anticipatory movements, was not significantly different from a uniform distribution (p = 0.73). Similarly, the distribution of MCI subjects, who showed an increase in anticipatory movements, did not differ from a uniform distribution (p = 0.10). In contrast, a more peaked distribution could be appreciated for the AD group, indicating that during an anticipatory behavior AD participants aimed mostly towards the previous target (µ = 308° ± 3°, p < 0.05).

Fig. 7

Polar histograms of the anticipation direction for all trials of all subjects in each group. Current and previous targets are pictured, respectively in the form of light grey and dark grey circles, respectively. Since the current target is positioned at 0° and targets appeared in the counterclockwise direction, the previous target is portrayed at 345°. AD participants tend to move towards the target of the previous trial

Discussion

Our results reveal pronounced differences in the performance and accuracy of reaching movements in MCI and AD patients when tested in a VR environment. AD patients showed a lower success rate and a higher number of anticipatory responses during the preparation of reaching movements. These responses were goal-oriented, likely influenced by the previous trial. MCI patients showed an intermediate pattern, with a higher number of anticipatory responses as compared with the HC group but without a clear difference in terms of success rate. These findings provide novel evidence that goal-directed reaching movements are impaired in AD, supporting the relevance of early assessment of motor impairment in AD patients [3].

AD represents one of the most invalidating age-related neurodegenerative disorders and, therefore, has received increasing attention in recent decades. Although mainly characterized by cognitive and memory symptoms, AD also leads to several motor disabilities. Alterations in gait, balance, and postural control have been commonly reported [40,41,42,43]. In particular, the visuomotor, visuospatial, and attentional deficits that affect AD patients can interfere with fine motor control of reaching movements, which are one of the most frequent actions in daily life. Here, the increased anticipatory responses observed in AD patients suggest a disruption in the motor planning and execution phases for these participants. This might indicate an impaired ability to inhibit anticipatory movements, which aligns with known deficits in executive function and motor planning in AD [4].

The higher anticipation rates observed in earlier blocks suggest that participants might initially struggle to perform the task and then adjust their behavior over time, with cognitively impaired individuals showing less improvement. Although initially this result may seem in contrast with previous findings showing a greater slowness in initiating movements by AD patients [44, 45], it is necessary to consider that participants had to reach the target in 750 ms forcing patients to be as fast as possible. Given their greater slowness in performing the movement, this likely led them to anticipate the go signal more. All of this led to the emergence of an alteration in inhibitory control, which is essential to avoid the anticipatory initiation of actions. This impairment can represent both cognitive and motor symptoms. Indeed, in this type of task, healthy subjects show reduced corticospinal excitability prior to initiating movement [46], an expression of increased inhibition in the motor cortex. This represents precisely the motor aspect of inhibitory control. Recent transcranial magnetic stimulation (TMS) studies in AD patients have observed an hyperexcitability and an altered GABAergic activity of the primary motor cortex [47,48,49]. These findings, parallel to the evidence of altered glutamatergic activity [50] suggest an imbalance between excitatory and inhibitory activity in the cerebral motor areas of this group of patients. This imbalance could represent one of the bases of this inefficient inhibitory control mechanism.

The lower success rates in AD patients strongly highlight the challenges they face in executing motor tasks. The improvement in success rates with repetition in HC participants suggests a learning effect since targets are presented in a systematic (thus predictable) order. However, this effect is strongly diminished in MCI and AD groups. This result highlights not only the motor difficulties present in these groups but, more specifically, the difficulties in motor learning and motor plan updating [17]. Additionally, MCI patients showed a higher number of anticipatory responses than the HC group but a comparable success rate. This result can be considered in line with a population with an early stage of cognitive and motor impairment. This condition could, though not necessarily, evolve into a neurodegenerative disease.

Furthermore, participants with lower levels of anticipation were more likely to successfully complete the reaching task. This inverse relationship highlights the potential impact of anticipatory behavior on task performance, reinforcing the idea that anticipatory movements, likely due to impaired inhibitory control, reduce overall task performance.

Moreover, higher MMSE scores, indicative of more preserved cognitive function, were associated with lower rates of anticipatory responses. This inverse relationship also emphasizes the relation between cognitive decline and motor control deficits. The anticipatory behavior observed in AD patients can thus be seen as a manifestation of underlying cognitive and motor impairments, particularly in executive functions involved in motor planning and control.

Additional insights into the motor learning and planning deficits in neurodegenerative patients are provided by the directional analysis, which reveals that AD participants tend to aim toward the previous target during anticipatory movements. This behavior is absent in MCI and HC participants, suggesting that in AD patients the impairment of the inhibitory control mechanism is paralleled by a motor learning deficit. This leads the AD patients not to discover the sequence in which the targets were presented and to remain anchored to the last motor plan implemented. This pattern reflects impaired cognitive flexibility and motor learning in AD patients, in contrast to more adaptive and context-appropriate motor responses in HC participants (due to the lack of significant directional bias). Indeed, previous studies in healthy subjects have shown that a fixed sequence of keys embedded within a larger random sequence of keys was performed more quickly. This was the case even though the participants were unaware of the existence of a repeating sequence [51, 52]. These results highlight how AD patients rely more heavily on previous motor plans because of their impaired ability to learn a motor sequence and consequently to anticipate and update new motor plan accurately.

The tendency of individuals with Alzheimer’s Disease (AD) to repeatedly aim for the same target may reflect perseverative traits. Perseveration, defined as the inappropriate continuation or repetition of a response or activity [53, 54], has been linked to frontal lobe dysfunction in several studies [53, 55,56,57,58,59,60]. Frontal lobe involvement is a common neuropathological feature of AD [61, 62]. These observations suggest that the repetitive targeting behavior observed in AD patients during anticipatory tasks may be a manifestation of perseveration resulting from frontal lobe dysfunction. Quantifying this behavior could offer clinicians a valuable tool for diagnosing and assessing the stage and progression of AD.

Further insight into the motor deficits of these patients and their neural basis has been gained through the qualitative observation of movement trajectories during the task. As illustrated in Fig. 2, the trajectories of movement exhibited by patients with AD demonstrate a notable degree of dispersion, unlike those performed by the HC and MCI groups. This difficulty in reaching the target, characterized by more disorganized trajectories, is the most distinctive symptom of another clinical population: the patients with optic ataxia. These patients, following posterior parietal cortex (PPC) injury, show severe impairment in their ability to make coordinated reaching movements, particularly for the peripheral target [63, 64]. At the neural level, this sensorimotor integration, which is necessary for the performance of reaching movements, is sustained by an extensive parietal-frontal cortical network that allows the transformation of the visual information into a precise motor plan. In particular, there appears to be a posterior-to-anterior gradient in the PPC, one of the critical areas of the dorsal visual stream [65], with a greater visual dominance for saccades and visual information in the occipital cortex and in the posterior part of the PPC. In the medial intraparietal sulcus (IPS) and superior parietal lobule (SPL), visual and reaching responses became mixed, and finally, in the anterior IPS and anterior SPL reaching responses became predominant. A similar gradient was also observed in frontal premotor areas active during reaching actions. The rostral part of the dorsal premotor cortex (PMd) has shown mixed saccade and reach responses, while the more caudal part of PMd showed reaching-only responses [66]. Hence the PPC is one of the critical nodes of the dorsal visual stream, or where pathway, that supports the localization of objects in space and the visually-guided behaviors [65, 67]. This brain region, which is damaged in patients with optic ataxia, appears to be one of the first to show functional and structural changes in AD. These alterations affecting PPC are likely to be the neural correlate that explains the difficulty that AD patients have in reaching targets and their more dispersed movement trajectories observed here.

Another important factor is the role of subcortical areas in planning and executing reaching actions. Patients with Parkinson’s disease exhibit deficits in initiating and executing reaching movements, similar to those observed in this study, particularly during the movement initiation phase. These impairments have been linked to degeneration of the basal ganglia-thalamus-cortical circuitry [68]. Notably, this circuitry is also disrupted in AD patients [69], suggesting it may contribute to the deficits identified in this study. In light of this, the present study offers new insights into subtle motor symptoms in AD patients, presenting strong evidence that anticipatory behavior and success rate correlate with preserved cognitive and motor functions across participant groups. Our findings suggest that kinematic metrics could be valuable biomarkers for early diagnosis and monitoring of neurodegenerative diseases. However, this study’s short assessment period limits conclusions about long-term motor behavior changes. Future research should address this through longitudinal studies, assessing participants across multiple sessions to understand motor deficit progression as a potential early marker of cognitive decline.

Additionally, participants were required to reach the target within 750 ms rather than at a self-selected speed. While this constraint added task difficulty and revealed planning deficits, future studies could compare relaxed and fast-reaching tasks to examine the influence of speed on motor responses. Although VR provides a controlled environment, it may not fully replicate real-world complexity. Future research should integrate more realistic VR tasks to better mimic daily life, refine kinematic metrics, and further explore the cognitive-motor relationship in AD and MCI.

Conclusions

This is one of the first studies assessing motor behavior in a VR environment specifically in the AD continuum. The use of VR offers novel opportunities for the assessment and rehabilitation of motor functions. VR provides an immersive and controlled environment that can simulate real-life scenarios with high precision. This allows for detailed tracking of motor responses, which might not be feasible in traditional settings. Furthermore, patients reported enjoying the VR experience, which enhances engagement and compliance during assessment and rehabilitation sessions. The immersive nature of VR can also be leveraged to develop novel rehabilitation protocols that are both effective and enjoyable, potentially leading to better outcomes in motor and cognitive recovery. Moreover, motor and cognitive function maintenance therapy using this technology could be carried out by patients at home, opening the possibility of new telerehabilitation programs. Finally, the diagnostic protocols within a VR environment can offer more consistent and repeatable conditions, reducing variability and improving the accuracy of motor function assessments.