Introduction

Parkinson’s Disease (PD) is caused by the progressive degeneration of dopaminergic neurons in the substantia nigra pars compacta, reduced striatal dopamine, and the presence of Lewy Bodies. Its most common form, Idiopathic Parkinson’s Disease, is determined by cardinal symptoms: Rest tremor, asymmetry, bradykinesia and a good response to dopamine confirm the pathology in 99% of cases.

Besides motor symptoms, cognitive dysfunctions occur very often in PD Patients [1, 2]. Once cognitive impairment can be objectified in PD patients, it is called Mild Cognitive Impairment in PD (PD-MCI). Criteria for a diagnosis of PD-MCI following the Petersen criteria includes a gradual decline in cognitive ability in the context of PD, cognitive deficits on a formal neuropsychological test and a cognitive deficit which is not sufficient to interfere with functional independence, but presenting subtle difficulties on complex functional tasks [3]. The prevalence of PD-MCI among PD patients is around 25% [4]. Furthermore, PD-MCI is a risk factor for a further cognitive decline into PD Dementia (PDD), which is characterized by cognitive symptoms in at least two domains (e.g. memory, executive function) which cause dysfunction in activities of daily living (ADL). PDD is an exclusion criterion for some PD treatments [5] such as Deep Brain Stimulation, and there is no strategy to prevent cognitive decline in PD patients or approved pharmacological approach to treat PD-MCI at the moment [5]. However, recent research suggests that cognitive function can be improved, or stabilized, in patients with PD through cognitive training [6,7,8]. Although at this point the ideal type or intervention frequency of cognitive training therapy for PD patients with either PD-MCI or initial PDD is not clear, several clinical studies indicate cognitive training improves executive and memory functions in PD patients. Meta-analyses also stress the need to provide clear inclusion and exclusion criteria in interventions for cognitive training and PD, considering the three stages of cognitive decline in PD [9]. It is also discussed that a combined treatment of cognitive and physical training seems to be a good option [10, 11]. Additional advantages of this strategy are transfer effects between cognitive and motor skills, for example, the positive effects of cognitive training in physical symptoms such as freezing of gait [12].

It is precisely due to the potential transfer effects that exergames [13] show great potential for PD rehabilitation since one of the best methods to combine both cognitive and physical training is the use of video games which require the user to perform physical movements while conducting cognitive exercises. Exergames, a portmanteau of exercise and games, aim to combine the motivational aspects of playing with the physical benefits of exercise. In regard to target groups and activities, such as exergames for rehabilitation, adapting to the particular needs of the users, and to their physical and cognitive capabilities, are usually mentioned as significant advantages over alternative rehabilitation methods [14]. Additionally, the confirmed reliability of sensory feedback from traditional exergaming platforms such as the Wii Balance Board (WBB) or Microsoft Kinect is backed by numerous publications from recent years [15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33].

On the specific topic of exergames for PD rehabilitation, a systematic review published in 2014 by Barry et al. [34] found a total of 1121 abstracts related, to some degree, to the use of exergames in PD. Seven clinical studies out of those publications were selected and analyzed. Their conclusions showed both, the promising future of exergame-based PD therapy, as well as the lack of rigorous clinical studies to evaluate its effectiveness. Research in this area in recent years has been extensive, but to this point, recent meta-analyses do not cover the advance in the specific research area of exergame-based PD rehabilitation. The goal of the present systematic review is thus to study the progress of this area during the years 2014 to 2017, with its focus set on sensors and controllers, specifically comparing sensors based on RCT outcomes where possible, as well as proposing the use of additional sensors where available, thus providing an up-to-date critical view of the work performed so far, and identifying the remaining steps required until exergame-based therapy may become a clinical standard for PD patients.

Methods

In order to update the previous systematic review [34], which covered publications until December 2013, we initially searched for publications using Barry et al’s previous searchline:

Exergam* OR active video gaming OR Microsoft Kinect OR Kinect OR Nintendo Wii OR Wii OR Sony EyeToy OR IREX OR Dance Dance Revolution AND Parkinson*.

However, we found that Barry’s searchline excluded significant research, and we thus updated the query „active video gaming “ to the more comprehensive „video gam* “ instead. We also removed duplicates, with a final result of:

Exergam* OR Video Gam* OR Kinect OR Wii OR Sony EyeToy OR IREX OR Dance Dance Revolution AND Parkinson*.

Inclusion criteria were articles published in the indexed databases Pubmed, Scopus, Sciencedirect, the Institute of Electrical and Electronics Engineers (IEEE) and Cochrane published since January 1, 2014. Publications were not included for duplicates or if the topic did not relate to this review despite a string match, that is, if (A) the target group was not PD patients exclusively, or (B) the employed therapy was not based on exergames. Given the variety of outcomes of publications regarding exergame therapy for PD, and considering most studies provide two or three outcomes, we limited our analysis to the main three outcomes given in case more than three were provided. In case we had to limit the outcomes of a study, standard ones were chosen where possible. The search was performed on February 26, 2018, and again on November 27, 2018. The methods are presented in the flow diagram shown in Fig. 1. This yielded the following results, excluding the aforementioned systematic review and the randomized clinical trials (RCT) identified in it, which were nevertheless also included in our analysis:

  • Pubmed: 87 matches, of which we excluded 36 papers, leaving us with 51 results.

  • Scopus: 163 matches, 8 Additional non-duplicate publications using the same criteria.

  • Sciencedirect: 249 matches, 5 Additional non-duplicate publications using the same criteria.

  • IEEE: 25 matches, 0 further results.

  • Cochrane Library: 1 match, 0 further results.

Fig. 1
figure 1

Flow Diagram of the study selection process

Full size image

From the 109 non-duplicate excluded papers, 77 were excluded due to criterion A and 32 due to criterion B. This reduced our corpus to a total of 64 publications, which we then proceeded to classify into the following groups:

  • Group 1. RCTs of game- or game-technology based methods in PD therapy. These are the primary focus of this paper.

  • Group 2. Pilot studies of game- or game-technology based methods in PD therapy (non-RCTs). These are the secondary focus of this paper.

  • Group 3. Technical papers, with no specific therapeutic focus (for example reliability tests). The goal here is to provide an overview on recent technical advances in the sector. This is summarized in the discussion section.

  • Group 4. Systematic reviews and meta-analyses relating to the topic. The contents and goals of these reviews are summarized in the discussion section.

In order for a study to qualify as an RCT, we used the question and scoring criteria presented in Barry et al’s systematic review [34], in which a study has RCT consideration if satisfactorily answers the following questions:

  • Is inclusion and exclusion criteria stated?

  • Are participant characteristics described in detail? (number, age, sex, disease severity)

  • Is sample size justified?

  • Is group randomization explained?

  • Is the design clear?

  • Were exergaming sessions explained in detail?

  • Were baseline and post-test data collected?

We also designated a publication as an RCT if all of these questions were satisfactorily answered. We also decided to include the previously existing RCT [35], for comparison.

Results

As a first step, we studied the input devices used on each publication in groups 1 to 3. The results are presented in Table 1, while technical data about the sensors mentioned in this publication is presented in Table 7. We noticed there is a strong focus on using the Microsoft Kinect for PD rehabilitation (36 of 49 publications used the Kinect) in comparison with other approaches, such as the WBB, custom sensors or, in two cases, a Wii hand controller (WiiMote). RCTs seem to focus on the Kinect as well (3 out of 8 studies).

Table 1 Input devices used in the analyzed publications. A pilot study [54] used both the Kinect and Wii sensors, and thus represented in parenthesis in both columns. For the total count it is considered as Kinect
Full size table

Randomized Clinical Trials (RCTs)

The quality of studies in the area has drastically improved in the last 4 years, and 8 new RCTs analyzing the use of exergames in PD rehabilitation were identified in this review [36,37,38,39,40,41,42,43]. Therefore, from this point onwards we refer to a total of 9 RCTs including the one identified in the previous review. A comprehensive list of the identified RCTs is presented in Additional file 1.

Given the variety of outcomes of these works, we limited this section to the main three outcomes. Standard outcomes, such as UPDRS Scores, MOCA Scores or standard tests (Berg Balance Scale (BBS), Timed Up and Go Test (TUG) were chosen where possible. Information about the cognitive and motor dysfunctions of the control and intervention groups can be extrapolated from pre-test UPDRS and MMSE scores, which are provided where available. Tables 2 and 3 present a summary of all RCTs, particularly participants, aim, duration and type of exergame, condition trained and methods, settings, outcomes, and conclusions. Table 4 compares RCT outcomes where possible. All studies had active control groups, and present results equal or better than this control groups (which performed traditional rehabilitation therapy or regular exercise), although at this point it is still difficult to determine which therapy works best. We consider the results of Allen et al. [36], and Song et al. [41] to be of particular interest since the outcomes of the intervention and control groups were similar although the evaluation was performed at home in a minimally supervised scenario. Both authors insist on the need to consider task specificity in the future.

Table 2 RCT Experiment Settings. ADL = Activities of Daily Living, BBS = Berg Balance Scale, C = Control Group, H&Y = Hoehn and Yahr Scale, I = Intervention Group, MMSE = Mini Mental Scale Examination, PD = Parkinsons Disease, UPDRS = Universal Parkinsons Disease Rating Scale, WBB = Wii Balance Board
Full size table
Table 3 RCT Experiment Outcomes. MOCA = Montreal Cognitive Assessment, PDQ-39 = Parkinsons Disease Questionnaire, TUG = Time Up and Go Test. Results are presented as mean (standard deviation) unless stated otherwise. Outcomes are adimensional where no units are stated
Full size table
Table 4 Comparison of RCT outcomes. Results are presented as mean (standard deviation) unless stated otherwise. Outcomes are adimensional where no units are stated
Full size table

We still observe several deficiencies in more recent RCTs, namely the lack of standardized outcomes, follow-up protocols and patient assessment methods, which are further described in the discussion of this paper. Regarding the risk of bias, seven of the clinical trials are single-blinded and thus at risk of selection bias [35, 37, 38, 40,41,42,43]. Out of these studies, three specifically mention this risk. [37, 40, 43]. Only [43] mentions outcome effect sizes. None of the identified RCTs report any conflict of interest.

Pilot Studies

A total of 11 Pilot Studies were identified in our search [44,45,46,47,48,49,50,51,52,53,54]. These studies were not included in the RCT group because of the lack of a control group, randomization and/or lack of detail of sessions and data. Nevertheless, these pilot studies present significant results in terms of usability and safety. The same analysis procedure used for RCT was performed in these publications, and results are summarized in Tables 5 and 6. A comprehensive list of the identified pilot studies is presented in Additional file 1.

Table 5 Pilot Study Experiment Settings
Full size table
Table 6 Pilot Study Experiment Outcomes. BES = Balance Evaluation System, FIM = Functional Independence Measure, nMWT = n Minutes Walking Test. S&E = Schwab and England Independence Scale. Results are presented as mean (standard deviation) unless stated otherwise. Outcomes are adimensional where no units are stated
Full size table

Pilot studies are usually focused on proving the feasibility of rehabilitation scenarios, in most cases regarding its safety as well as ease of use in PD patients. Interestingly, most studies reported improvement, and none reported deterioration or safety risk as a consequence of the study. However, due to low sample sizes and insufficient intervention durations, outcome changes were not statistically significant in some cases. As it is the case for the RCTs, the evaluated scenarios were a mixture of custom and commercial games, with both of these providing similar results. Two of the pilot studies [45, 50] mention the possibility of adapting their modules to a minimally supervised home scenario, a level of supervision that could be achieved, for example, by relatives. The question of therapy adherence is mentioned, but, as mentioned in the previous section, challenges regarding safety remain largely unaddressed.

We also discovered a very recent study offering the first comparison between exergaming platforms specifically for PD rehabilitation, specifically WBB and Kinect [54]. Outcomes of this study suggest the WBB may be more adequate for PD rehabilitation, at least in the disease factors that can be evaluated through the 10MWT or TUG tests. This aspect is further mentioned in the discussion.

Discussion

The research activity in the area of exergames for the rehabilitation of PD patients has improved impressively in recent years. The previous systematic review by Barry et al. [34] could only analyze seven studies and just one of these qualified as RCT. By turn, in this systematic review, we included 64 new studies, eight of which qualified as a RCT. Our findings indicate that the putative improvement in motor skills suggested before 2014 has been confirmed by several independent studies including both, RCTs and pilot studies. As a whole, from a total of 19 studies where exergame therapy was evaluated (both RCTs and pilot studies), 17 demonstrated improvements in the measured outcomes (8 out of 9 RCTs and 9 out of 11 pilot studies). Additionally, in 7 out of 9 RCTs the intervention group provided better results than the control group in motor outcomes. Finally, none of the studies showed exergames were in any way worse than traditional rehabilitation therapy. Additionally, exergames seem to improve cognitive skills (MOCA scores, attention) in the aforementioned RCT study populations (PD patients with MMSE scores > 24) as well.

From our analysis on recent RCTs, we can conclude that the focus on using the WBB is switching towards the Microsoft Kinect v2, which apparently provides similar results while being easier to use since patients do not have to stand on an elevated platform. Recent research shows how this sensor can provide relevant and accurate gait parameters which can be potentially used to monitor PD patients remotely during rehabilitation sessions. However, RCTs using this platform remain scarce, while positive WBB results are more numerous. Up to this point, all RCTs include an MMSE score of 24 or higher as an exclusion criterion. Thus, the effect of patients with lower MMSE scores is yet to be determined. Considering recent studies show how exergaming does improve cognitive capabilities in PD patients [35], it seems necessary to extend these studies into PD-MCI as well as initial PDD, and evaluate the feasibility of game-based therapy on these patients as well.

An additional advantage presented by the Kinect over the WBB or Wiimote seems to be that the risk of patient not liking the input device used in the game is lower [55]. How PD patients will react to speech control or an improved version of a WBB is unknown, and the number of RCTs involving the Kinect is up to this point insufficient. Future research should also consider new sensors and methods, such as the Leap Motion finger tracking sensor (see Table 7), which, as with the Kinect sensor, does not require the user to wear it or carry any weight. We believe this tendency will continue towards other sensors and scenarios where the user can perform natural movements, such as fine motor training with the Leap Motion sensor.

Table 7 Technical Information about the sensors mentioned in this publication
Full size table

Regarding outcome comparisons between exergaming platforms. A recent pilot study [54] suggests the WBB may be more optimal for Kinect in PD rehabilitation, based on a comparison performed between these two platforms. However, the meta-analysis of RCT outcomes (Table 4) suggests otherwise: We found higher improvements in the TUG test using WBB (1 WBB-Based RCT vs. 1 Kinect-based RCT) and better results with Kinect in BBS (2 WBB-based RCTs vs 2 Kinect-based RCTs).

Considering technical papers, 19 [15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33] out of the 30 [15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33, 55,56,57,58,59,60,61,62,63,64,65] publications in this group focused towards linking sensor parameters to disease severity indicators. The use of these sensors for screening is suggested [22, 33]. The limitations of the current generation of sensors, particularly in fine motorics, is also discussed [29]. In this case, a more dedicated sensor, such as a data glove [56], or the Leap Motion sensor [65] may be used. In addition, there is also clear interest in developing rehabilitation methods that may be performed in an unsupervised home scenario [56, 59,60,61,62,63]. Patient preferences regarding sensors are rarely discussed [55].

In parallel to these findings, 15 meta-analyses and systematic reviews were found [66,67,68,69,70,71,72,73,74,75,76,77,78,79,80]. These relate to a wide array of topics, however, none of the studies we found related to the specific topic of exergame-based rehabilitation for PD specifically.

This study presents some limitations. Firstly, the lack of standardized outcomes in this research area makes it still difficult to determine which exergaming platform, and method, works best for the specific needs of PD patients. The meta-analysis we could perform in this aspect was limited to the studies presented in Table 4 and, although illustrative, requires significant extension before conclusions may be extracted. It is also early to quantitatively determine the effect of exergame therapy in PD-MCI and initial PDD patients, either in improvement or slowdown of deterioration. Given that the effectiveness of exergames, at least in comparison with traditional rehabilitation, seems to be comparable, it is critical that future RCTs establish the use of a reduced number of standard outcomes such as MDS-UPDRS, BBS or TUG in order to start providing a more significant insight into what scenarios and sensors are actually more effective to train which difficulties of PD patients in particular or motor afflictions in general. If authors adhere to these principles, a repetition of this exercise performed another 5 years from now may very well provide critical insight in the development of game-based PD-specific rehabilitation: Which sensors to use, what game environments are optimal, and which are the ideal difficulty adaptation curves based on UPDRS or MMSE assessments.

Considering the effect of RCTs in cognition, performing a follow-up analysis is also essential to determine long-term patient improvement in all outcomes, especially considering follow-up results seem to be good for cognitive skills, but not for balance.

Although it is still difficult to determine, it seems there is not a big difference between PD-specific games and custom-developed games. For the patients included in the analyzed clinical studies, that is, patients with MMSE scores over 24, and although it is too early to make clear conclusions, it seems that cognitive specific therapy and commercial games have no significant difference in cognitive outcomes, as seen on Table 3, although whether this also applies to PD-MCI patients remains to be determined. An RCT mentions the importance of task specificity in this case [41]. Other studies support the idea that therapy should adapt to the patient’s skills, and this would be possible with custom-developed adaptive exergames, but not with commercial ones.

Regarding sensors, it is clearly essential to further determine what constitutes the right array of sensors to ensure a pleasant experience while the most amount of medically relevant information is captured in the background, adapting to the patients’ skills where possible. Unfortunately, the number of studies in this regard is very limited so far.

Another research field, parallel to the one presented in this work, is the use of Virtual Reality in rehabilitation in general and for PD in particular. A recent systematic review [81] summarizes recent advances in this area in the last years. Similarly to exergame-based rehabilitation, limited evidence shows positive effects, but additional studies are required to confirm the results.

Conclusions

In our opinion, the main conclusion of this study is that exergame PD rehabilitation is indeed safe, feasible and effective, and in some cases [37, 38, 40, 42] even more effective than traditional rehabilitation and regular exercise. Exergames also provide the possibility of a home rehabilitation and remote monitoring scenario, which are limitations of traditional rehabilitation methods. There is contradictory evidence regarding which exergame platform performs better at this point.

Regarding future work, in addition to providing further, high-quality clinical trials in order to confirm the hypotheses so far presented, studies should focus on adapting existing therapies to home scenarios, adding further data collection procedures, such as monitoring hand tremors, perhaps with a finger sensor [82]. Additional variables such as dysphonia [83] and heart rate variability [84] may also be of interest. Evaluating the feasibility of expanding PD Exergame therapy to patients with PD-MCI and PDD, and including new interaction techniques such as fine motoric training and speech control are also points of interest. Finally, programs should be tailored to patient characteristics, both physical and cognitive.