A Robotic Assessment of Cognitive and Sensorimotor Functions

A Robotic Assessment of Cognitive and Sensorimotor Functions Valentina Massone ReWing s.r.l. & DIBRIS (University of Genoa) Genoa, Italy [email protected] Jessica Carloni Pragma Engineering s.r.l. Perugia, Italy [email protected] Aurora Freccero ReWing s.r.l. & DIBRIS (University of Genoa) Genoa, Italy [email protected] Riccardo Magni Pragma Engineering s.r.l. Perugia, Italy [email protected] Jacopo Zenzeri ReWing s.r.l. Genoa, Italy [email protected] Giulia A. Albanese ReWing s.r.l. Genoa, Italy [email protected] Maura Casadio DIBRIS (University of Genoa) Genoa, Italy [email protected] Abstract—Traditional assessments often overlook the interaction between cognitive and motor impairments. Robotic systems allow precise measurement of both domains, supporting more comprehensive rehabilitation. This study presents a preliminary upper-limb robotic cognitive-sensorimotor assessment protocol and evaluates its feasibility in 7 unimpaired participants with a robotic wrist device. Kinematic and cognitive outcomes were recorded, demonstrating the potential of integrated assessments for more detailed user profiling. Keywords—robotic rehabilitation, cognitive-sensorimotor assessment, upper limb I. INTRODUCTION Sensorimotor assessments alone do not fully capture the multidimensional impairments experienced by individuals with neurological diseases, where cognitive deficits often coexist and interact with sensorimotor dysfunctions. Key cognitive functions assessed include attention and working memory, often embedded within complex motor tasks [1]. Executive functions and visuospatial attention have been shown to influence motor performance, highlighting how cognitive decline can impair movement precision [2]. Robotic systems enable precise, automated quantification of both cognitive and motor impairments, using upper-limb movements as outputs of cognitive processing, which facilitates more tailored rehabilitation approaches. Traditional cognitive measures can be adapted to these platforms, while kinematic analysis reveals movement patterns during task execution. This approach allows differentiation of performance difficulties due to sensorimotor impairments from those arising from cognitive deficits [3]. The application of these methods has proven effective in different clinical populations. Post-stroke individuals with cognitive deficits exhibit significantly poorer performance on robotic motor tests compared to unimpaired subjects [2], while integrated cognitive-motor robotic assessment demonstrates greater precision than traditional clinical scales in individuals with Multiple Sclerosis [1]. Current rehabilitation approaches often study cognitive and physical exercises separately. There is a clear need to develop multimodal assessment capable of analyzing kinematics while interacting with cognitive tasks. Addressing cognitive-sensorimotor integration represents a step toward developing more comprehensive and personalized user profiling and, consequently, rehabilitation strategies. This study introduces a preliminary upper limb robotic cognitivesensorimotor assessment system aimed at user profiling, along with acceptability tests conducted on unimpaired subjects, aimed at assessing its feasibility and functionality. II. METHODS A. Participants Seven unimpaired participants took part in the acceptability test of the robotic cognitive-sensorimotor assessment system. The group consisted of six females and one male (30.14 ± 3.31 years). B. Experimental Setup The experimental setup includes a robotic wrist device known as EDUSA® PRO-R [4], a three-degree-of-freedom end-effector device designed for wrist flexion/extension (F/E), radial/ulnar deviation (RD/UD), and pronation/supination (P/S), with a human-like range of motion (ROM) along all planes. It is equipped with four brushless motors, high-resolution incremental encoders, and a grip sensor to measure the user’s grasp on the device handle. Additionally, an integrated virtual reality environment provides visual feedback to the user, and an ergonomic wrist and forearm locking system prevents undesired movements. C. Experimental Protocol and Data Analysis The experimental protocol consisted of two sensorimotor tasks and three cognitive-sensorimotor tasks targeting visuospatial attention, working memory, and spatial selective attention, domains particularly relevant for daily life. The experimental protocol included: 1) Active RoM Assessment: the active range of motion was measured as the maximum distance in degrees each participant could reach along different target directions. The minimum RoM across directions was then used to set the target position in the subsequent tasks, ensuring both the feasibility of cognitive-sensorimotor tasks and the comparability of kinematic evaluations between sensorimotor and cognitive-sensorimotor tasks. 2) Active Reaching: participants were instructed to move the robotic device handle to reach a target as quickly as possible. The targets were positioned according to the minimum RoM and the directions assessed in the previous task. 3) Audiovisual Attention task: during this assessment, the participants were required either to perform an action or to remain still, depending on the stimulus presented. The possible stimuli consist of a low sound, a high sound, two red targets or two blue targets, placed at a distance corresponding to the minimum RoM along the flexion and the extension directions, measured during the Active RoM task. The task consisted of four sequences. Each sequence included the four possible stimuli. At the beginning of every sequence, new 2025 I-RIM Conference October 17-19, Rome, Italy ISBN: 9788894580570 10.5281/zenodo.17629884 243 instructions specifying the response required for each stimulus appeared on the screen. The possible responses were flexion, extension, a grip on the device handle, or no movement. 4) Working Memory task: the task was structured into three levels of increasing difficulty, each requiring the memorization of a list of objects followed by a motor selection phase. At Level 1, the task presented the participants with a list of four target objects, chosen among common and easily recognizable items. The list remained visible on the screen for 15 seconds (memorization phase). Subsequently, eight target objects were displayed, positioned along the same directions of the targets of the previously performed Active Reaching task. The participant had 30 seconds to actively reach the objects seen in the memorization phase in the exact order of presentation, by moving the device’s handle through active reaching movements (reaching phase). A maximum of three attempts was allowed to complete the task correctly. If errors occurred, the same list was shown again for 10 seconds before the second attempt and for 5 seconds before the third. If the participant successfully completed the sequence on the first or second attempt, they advanced to Level 2. Otherwise, the task ended. At Level 2, the procedure was repeated with a new list of five target objects. The allowed time for the reaching phase increased to 35 seconds. At Level 3, a list of six objects had to be memorized and selected under the same conditions, with the reaching phase lasting 40 seconds. 5) Spatial Selective Attention task: during this assessment, participants were presented with a matrix of 25 black objects on a white background, visually similar to each other and positioned along concentric rectangles (within the participant’s minimum RoM). Among these, the target object appeared six times. Participants had 30 seconds to identify and actively reach all and only the six target objects. The task had to be completed as quickly as possible, minimizing unnecessary movements and avoiding errors. Data was analyzed offline using custom code in Python. The cognitive and sensorimotor outcome measures considered optimal for user profiling were identified. These measures allow to assess how sensorimotor indicators can be integrated with cognitive metrics to provide a more comprehensive representation of the user’s functional profile. The selected outcomes included: mean and maximum reaching speed, reaction time (defined as the time from target appearance to movement initiation exceeding 2°/s), arrest rate (representing the percentage of time the user remained stationary during the reaching movement). Cognitive measures included errors, omissions, wrong starting movement, wrong combination of movements, incorrect target selections, errors in selection order, and span, representing the highest completed difficulty level. III. RESULTS For the Active Reaching task, participants showed a mean reaching speed of 39.61 ± 7.55 °/s (mean ± std), and a peak reaching speed of 120.69 ± 27.65 °/s. The mean reaction time was 0.14 ± 0.04 s, while the arrest rate was 34.70 ± 7.56 %. Regarding the cognitive-sensorimotor tasks, for the Audiovisual attention task, cognitive measures showed 1.29 ± 1.67 errors, 0.29 ± 0.70 omissions, and 0.29 ± 0.45 wrong starting movements and wrong movement combinations. Sensorimotor measures revealed a mean reaching speed of 81.78 ± 22.87 °/s, higher than in Active Reaching, likely due to computation limited to flexion–extension directions, and a peak speed of 123.32 ± 35.29 °/s, comparable to Active Reaching. Reaction time increased to 0.81 ± 0.14 s, reflecting higher cognitive load. For the Working Memory task, cognitive measures included 0.43 ± 0.73 incorrect target selections, 1.85 ± 3.04 errors in selection order, no omissions, a memory span of 5.14 ± 0.35, and 3.71 ± 0.88 repetitions required to complete the task. Sensorimotor measures indicated a mean reaching speed of 21.53 ± 1.51 °/s and a peak speed of 83.31 ± 5.01 °/s, both lower than in Active Reaching, with a slightly increased reaction time of 0.26 ± 0.06 s. For the Spatial Selective Attention task, cognitive measures showed no errors or omissions. Sensorimotor measures indicated a mean reaching speed of 34.35 ± 5.75 °/s and an arrest rate of 27.13 ± 9.45 %, both comparable to Active Reaching. IV. DISCUSSION This preliminary work aimed to develop and test the feasibility of a robotic assessment of cognitive-sensorimotor integration, a process that constantly underlies activities of daily living where motor execution is modulated by cognitive demands. Results indicate that the robotic platform was able to quantify both motor and cognitive performance in unimpaired participants showing the interaction between the two domains. From a cognitive perspective, errors and omissions remained low overall. However, participants exhibited higher speeds and shorter reaction times in the Active Reaching task than in the Audiovisual Attention and Working Memory tasks, since the introduction of cognitive load led to a reduced motor performance, confirming the interaction between motor and cognitive demands. Interestingly, the Spatial Selective Attention task showed motor outcomes comparable to the Active Reaching task, suggesting that the cognitive challenge did not substantially interfere with execution in this condition. These results confirm the feasibility of adapting traditional cognitive tasks into robotic environments, where kinematic measures enrich the interpretation of performance by separating sensorimotor from cognitive contributions. Although the study was limited by the small number of unimpaired participants, it provided relevant information for the refinement of the protocol, particularly regarding task duration and difficulty levels. Future work will include tests on clinical populations such as individuals post-stroke or with Multiple Sclerosis, who typically present sensorimotor impairments along with cognitive deficits that may influence both the course of sensorimotor rehabilitation and daily life activities. The goal is to understand how cognitive factors affect sensorimotor recovery and functional abilities, in order to enhance both domains. In these populations, the integration of cognitive and motor outcome measures may prove informative, enabling more comprehensive user profiling and the development of personalized rehabilitation strategies. ACKNOWLEDGMENT This work was supported by the European Union - NextGenerationEU project ‘‘VITALITY’’ (ECS00000041). REFERENCES [1] L. E. Simmatis, et al. «The feasibility of assessing cognitive and motor function in multiple sclerosis patients using robotics», Multiple Sclerosis Journal - Experimental, Translational and Clinical, vol. 6, fasc. 4, p. 205521732096494, ott. 2020 [2] K. D. Bui e M. J. Johnson, «Robot-Based Measures of Upper Limb Cognitive-Motor Interference Across the HIV-Stroke Spectrum», IEEE Int Conf Rehabil Robot, vol. 2019, pp. 530–535, giu. 2019 [3] S. H. Scott e S. P. Dukelow, «Potential of robots as next-generation technology for clinical assessment of neurological disorders and upper-limb therapy», JRRD, vol. 48, fasc. 4, p. 335, 2011 [4] I. Pippo et al., «Existing Devices in Wrist Rehabilitation Robotics: Mechanical Design, Control Systems, and User Requirements for Commercialization.» 2024, p. 8. 244