A serious-game-based pipeline for rehabilitation and assessment of upper limb impairment

The progressive ageing of the population has resulted in an increase of neurodegenerative disorders, including Parkinson’s Disease. In order to ensure the preservation and the enhancement of motor functions, continuous rehabilitation programs would be necessary. However, these programmes are often costly and not conceived for home-based use. Exergames, videogames designed to perform therapeutic physical exercises, represent a cost-effective solution suitable for telerehabilitation. They enable continuous patient monitoring and increase engagement throughout rehabilitation process. This study analyzes the data collected by the Palestra exergame, developed by CNR-IEIIT, which employs a single RGB camera combined to Google MediaPipe Pose for human pose estimation. The system is designed for parkinsonian subjects, and focused on upper limb motor recovery, through simple arm-raising exercises (single arm, alternating and simultaneous movements, performed both laterally and frontally). The analysis was performed using a dedicated automatic pipeline, designed to extract angolar trajectories and a set of parameters aimed to evaluate the user’s motor performance. In order to assess the quality of the acquired data and to define limits and potential of the exergame, a validation procedure was carried out. This validation was done against an optoelecronic system and included a total of five healthy subjects. Moreover, the system was tested on parkinsonian patients with different degrees of motor impairments. The results show that the pipeline produces reliable angular trajectories in both healthy and parkinsonian users, thus allowing the extraction of robust parameters in both cases. The Bland-Altman analysis conducted on the entire trajectories demonstrate a good agreement in lateral raises. Temporal parameters, such as the Peak-to-Peak Time (PPT), exhibit a high concordance between the two systems (Bias= 0.00 s, LoA=[−0.09 s, 0.09 s]). Angular parameters, like Range of Motion (ROM), present marginally lower agreement (Bias= 2.03 °, LoA=[−9.16 °, 13.22 °]). In frontal tasks, the Bland-Altman analysis reveals slightly wider discrepancies in temporal parameters, even though they remain comparable with those identified in lateral tasks (PPT: Bias= 0.01 s, LoA=[−0.23 s, 0.24 s]). Larger differences are observed in the angular parameters (ROM: Bias= −3.56 °, LoA=[−23.53 °, 16.42 °]). Nevertherless, these discrepancies, both temporal and angular, are still considered accetable for the intended context of use. In conclusion, Palestra represent a promising solution for telerehabilitation, due to its ease of use, cost-effectiveness and ability to collect data in a continuous and non-invasive manner. The accuracy levels of the system, as demonstrated by the validation analysis, are accetable and suitable for the intendend telerehabilitation applications.