Design and validation of a serious game for rehabilitation and monitoring of individuals with Parkinson’s disease

Abstract

Introduction: Parkinson’s disease (PD) is a neurological disorder that affects body movements and causes symptoms such as bradykinesia and tremor. Bradykinesia, characterised as slowness of movement, is the main motor symptom of PD; while tremor, a rhythmic and involuntary oscillatory movement of a body part, is usually the first and the most troublesome motor symptom of the disease. Currently, Movement Disorder Society–Unified Parkinson’s Disease Rating Scale (MDS–UPDRS) is the most widely used clinical scale for assessing both symptoms. However, this type of assessment has limitations, such as subjectivity and dependence on experienced assessors. These problems can be overcome with the use of inertial sensors for objective measurements of movement, incorporated, for example, in control devices for serious games (SGs), increasingly promising tools for rehabilitation and movement assessment. Objectives: To develop a serious game with high usability and accessibility to individuals with PD, and, by using the signals recorded from the movements made during the game, to objectively assess bradykinesia and tremor. Methods: First, a literature review was carried out to identify existing technologies related to games for individuals with PD. Subsequently, a questionnaire was developed to assess multidimensional aspects of patients and guide the development of the game. Based on this research, a serious game about bees, so-called RehaBEElitation, was designed in Unity 3D. The game presents four phases, each one representing a bee’s task in the real world. The user has to control the movements of a bee in a 3D environment, executing the same movements found in the MDS–UPDRS. Fifteen individuals with PD and fifteen healthy individuals interacted with the game, constituting the experimental group (EG) and the control group (CG), respectively, with matched ages and sexes. Data was collected in the ON and OFF states of the medication for the EG. The usability of the game was assessed using the Nielsen-Shneiderman heuristics and user-centred approaches (evaluation questionnaires and semi-structured interview); and the accessibility, using the performance (scores) of the players in each phase. Bradykinesia was evaluated by detecting the trend of the gyroscope signals using Singular Spectrum Analysis (SSA), a time series decomposition method. The response time (RT) and angular velocity (AV) of the participants while playing RehaBEElitation were then estimated. Tremor was detected by removing the trend from the original signal and assessed by estimating the mean absolute value (MAV), root mean square (RMS) and power spectral density (PSD) of the gyroscope signals. In game usability evaluation, the paired-samples t-test and Wilcoxon test were used to confirm differences for questionnaire responses before and after game improvements. While assessing game accessibility and bradykinesia, one-way analysis of variance (ANOVA), t-test with Bonferroni correction, Kruskal-Wallis test and Wilcoxon test with Bonferroni correction were used to confirm the differences between groups. Results and Discussion: The heuristic evaluation contemplated all the 14 heuristics proposed by Shneiderman, which allowed the implementation of improvements in the game. The questionnaires and the semi-structured interview also contributed to improve the usability of the game. The accessibility evaluation revealed no statistically significant differences (p<0.05) between groups, except for the healthy group and the PD group in both states of medication during Phase 3. In bradykinesia assessment, significant differences were found between groups for all comparisons (p<0.05). In most phases, CG participants obtained the best results for RT and AV, followed by EG in the ON state and EG in the OFF state. Regarding tremor, the probability that the peak frequencies of the PD–OFF signals fall within the frequency range of Parkinson’s tremor (4-7 Hz) was higher than for the PD–ON group, which in turn was higher than for CG. For both MAV and RMS, the chance that the distribution having a bimodal aspect, which represents the presence of PD tremor, was higher for the EG in the ON and OFF states than for the CG. The PD–OFF group had the highest probability of having the highest peak power in the signal than the other groups. In summary, MAV and RMS showed strong correlation with MDS–UPDRS. Conclusion: The RehaBEElitation serious game is an alternative tool for assessing the main motor symptoms of PD in an objective and entertaining way. The game is compatible with the mental models of the target audience, which allowed users to easily interact with the developed system. In general, the EG participants in the OFF state showed increased bradykinesia and tremor compared to the ON state.