1. Introduction
The application of Virtual Reality (VR) as a diagnostic and therapeutic tool is emerging as a viable alternative for musculoskeletal rehabilitation of the upper limb [
1,
2,
3]. Unlike in orthopedic rehabilitation, VR has been widely used in the field of neurorehabilitation, such as stroke, brain injury, and cerebral palsy [
4,
5,
6,
7,
8,
9]. On the other hand, several studies have been conducted on the effectiveness and application of VR in orthopedics for educational and training purposes in shoulder, elbow, knee, hip, or ankle surgery [
1,
10,
11,
12,
13,
14].
VR technology is an advanced human–computer interface simulating a virtual environment where people can move and interact with unreal objects [
15,
16]. Based on the degree of immersion in the virtual environment, VR systems can be classified into non-immersive, semi-immersive, or immersive [
16,
17]. The most recent immersive VR devices offer a three-dimensional stereoscopic vision through a head-mounted display (HMD) and input devices (VR controllers), in which users are immersed and can interact with the virtual environment [
18,
19]. Among VR devices, Oculus Quest 2 showed good performance and accuracy during controlled translational and rotational displacements, establishing itself as an excellent VR device candidate for use during shoulder rehabilitation exercises [
2]. Given that VR focuses users’ attention on the perceived experience, designing applications tailored to the final population and context of use is crucial. VR applications could be extremely interesting and impactful in shoulder rehabilitation. Indeed, in an increasingly digitized society, VR can be considered an adjunct to standard physiotherapy to increase patients’ motivation, compliance, and engagement during therapy sessions. However, delineating a perfect balance between the efficiency of rehabilitation exercises in the virtual environment intended for patients with rotator cuff (RC) diseases, ease of use, and engagement is challenging.
RC diseases have a high incidence in the working population, entailing pain, reduced range of motion (ROM), and absence from work [
20,
21,
22]. Treatments of RC tears can be conservative or surgical [
23,
24,
25]. The surgical approach, eventually selected in patients unresponsive to conservative treatment, requires a postoperative phase in which the perfect balance between immobilization and mobilization is crucial to preserve the integrity of the repaired tendon and, at the same time, avoid shoulder stiffness [
21]. Shoulder rehabilitation after rotator cuff repair (RCR) aims to restore full painless shoulder ROM, gradually introducing muscle-strengthening exercises under the supervision of expert physiotherapists who monitor and tune exercises’ difficulty based on the patient’s recovery [
26].
In the orthopedic rehabilitation field, few studies have evaluated the effectiveness of kinematic intervention in virtual environments in patients with shoulder stiffness, subacromial impingement syndrome, scapular dyskinesis, and RC diseases [
3,
27,
28,
29,
30]. Chang et al., in a mono-centric randomized controlled trial with patients undergoing RCR, compared a control group following conventional home rehabilitation to the experimental one, which used an augmented reality (AR)-based digital healthcare system [
28]. Overall, the authors support the effectiveness of the developed AR-based post-surgical rehabilitation following arthroscopic RCR [
28]. In a single-blind randomized trial, the efficiency of video-based game rehabilitation exercises and closed kinetic chain was investigated related to conventional care in patients with partial RC tears, showing significant improvement in the outcomes of the experimental groups [
31]. To the best of our knowledge, regarding VR applications designed for the rehabilitation following RCR, only one study presented an approach more similar to ours [
32]. Baldominos et al. tested a custom application reproducing abduction and adduction movements with four professionals of physiotherapy after RCR, using the VR glasses Oculus Rift DK2 in association with the motion-tracking system Inter RealSense [
32]. However, none of the previous studies mention evaluating the performance of developed VR applications against pre-established criteria inherent in purely clinical guidelines and staged ROM goals.
In this study, we propose an immersive VR application (VR app) designed for the rehabilitation of patients after RC arthroscopic repair, following recommendations developed by the American Society of Shoulder and Elbow Therapists [
33]. The VR app includes four propaedeutic levels reproducing increasing levels of ROM similar to what is required during rehabilitation pathways. The objective of this study was to investigate the performance of the custom VR app by comparing the movements performed during the different levels of the VR app with a stereophotogrammetric system considered the gold standard.
3. Results
Results for the first and second rounds of flexion, seen in
Table 2 and
Table 3, respectively, and for abduction (
Table 4), displayed low values for mean absolute error (MAE) and for the number of times conditions were not met overall. Results for external rotation (
Table 5) and internal rotation (
Table 6) displayed total MAE values similar to those seen for abduction and flexion; however, the number of times conditions went unmet was significantly higher. This discrepancy in results is attributed to the inability of volunteers to perform isolated shoulder rotations and thus use trunk movements as compensation.
3.1. Flexion
3.1.1. First Round of Flexion
Results for the first series of flexion are reported in
Table 2. In level 1, volunteers were found to have a total inferior MAE equivalent to 6.4; this value gradually decreased with each level. Participants did not meet inferior conditions 8 times. Total superior MAE was found to be 2.9, and superior conditions were not met 5 times.
In level 2, volunteers had a total inferior MAE equivalent to 5.1 and a total of 15 unmet inferior conditions, representing the highest value of unmet conditions for both the first and second rounds of flexion. No error or unmet conditions were recorded for the superior limit.
Volunteers for level 3 had a total inferior MAE equivalent to 1.3 and a total of 4 unmet conditions. The total superior MAE was found to be 8.6, and conditions were unmet 3 times.
In level 4, volunteers only had one recorded instance of an unmet condition for inferior limit and a total inferior MAE equivalent to 0.8, which is the lowest error value for inferior flexion in both the first and second rounds.
3.1.2. Second Round of Flexion
Results for the second series of flexion are reported in
Table 3. In level 1, volunteers were found to have a total inferior MAE equivalent to 4.6, and participants did not meet inferior conditions 6 times. The total superior MAE was found to be 10.5, the highest for both rounds of flexion. The superior conditions were not met 6 times.
In level 2, volunteers had a total inferior MAE equivalent to 12.3, the highest value of mean error for inferior flexion overall. Participants did not meet inferior conditions 8 times. The total superior MAE was found to be 9.2, and only one volunteer (V4) contributed to this error value. In total, conditions were unmet 7 times.
In level 3, total inferior and superior MAE values were 4.8 and 2.1, respectively, while in both cases, conditions were not met twice.
In level 4, volunteers did not meet conditions 8 times and had a total inferior MAE equivalent to 2.6.
3.2. Abduction
Results for abduction are reported in
Table 4. In level 1, abduction had the highest total superior MAE value and the highest number of times superior conditions were not met; values are 6.1 and 14, respectively.
In level 2, abduction had no error for the inferior condition, and volunteers never failed to meet conditions. For superior condition, the total superior MAE was found to be 4.7, and conditions were not met 4 times.
In level 3, abduction had a total inferior MAE of 8.6, the highest out of all levels, and conditions were not met 15 times. No mean error value was found, and no conditions were unmet for superior condition.
In level 4, abduction had a total inferior MAE of 6.6, and conditions were not met 22 times, which represents the highest value measured for both superior and inferior conditions during abduction.
3.3. External Rotation
Results for external rotation are reported in
Table 5. In level 1, external rotation has a total superior MAE value of 6.5, while conditions were not met 32 times.
In level 2, external rotation has a total inferior MAE equivalent to 2.3, and volunteers did not meet conditions only once. The total superior MAE was found to be 2.6, and conditions were not met 12 times.
In level 3, external rotation has a total inferior MAE equivalent to 6.3, and volunteers did not meet conditions 16 times. The total superior MAE was found to be 2.6, and conditions were not met 5 times.
In level 4, external rotation had a total inferior MAE of 19.6, and conditions were not met 26 times; 19.6 represents the highest total MAE value for all recorded movements.
3.4. Internal Rotation
Results for internal rotation are reported in
Table 6. In level 2, internal rotation has a total superior MAE value of 9.8, while conditions were not met 33 times. This was the highest number of times conditions were not met for all recorded movements.
In level 3, internal rotation has a total inferior MAE equivalent to 2.0, and volunteers did not meet conditions 6 times. For the superior condition, the internal rotation had no error and no conditions went unmet.
In level 4, internal rotation has a total inferior MAE equivalent to 8.0, and volunteers did not meet conditions 30 times.
4. Discussion
Among VR devices, OQ2 is an interesting tool for creating virtual scenarios in which patients can perform rehabilitative exercises according to clinical needs. The device allows the creation of specific games with different objectives in which patients can have an immersive and interactive experience. With OQ2 Touch controllers, patients’ hands and gestures can be directly conveyed into the game so that movements can appear in the VR intuitively and engagingly. The application of games and varied virtual scenarios has the potential to increase patient compliance levels, directly increasing the efficacy of rehabilitation programs [
2,
28,
39]. Furthermore, machine learning algorithms and the modifiable nature of this tool combined can allow professionals to adapt rehabilitation programs to individual patients [
39]. This application has the potential to increase the efficacy of rehabilitation for each individual, and it enables patients to become more engaged with the assigned tasks, given that they have been tailored to their clinical parameters [
28]. Beyond increasing patient compliance and efficacy of treatment plans, these devices are also able to generate digital records of rehabilitation sessions [
39]. Access to digital records aids healthcare professionals, allowing them to dedicate more time to patient-centered activities. Furthermore, having detailed records always available has the potential to aid in the evaluation and possible modification of treatment plans.
Despite the many advantages that VR is projected to bring to the field of shoulder rehabilitation, the validity and effectiveness of kinematic intervention in virtual environments for orthopedic rehabilitation have only been explored by a few studies [
3,
27,
28,
29], none of which mention evaluating the performance of VR applications against pre-established criteria like clinical guidelines and/or staged ROM goals. Hence, the performed investigation compared the movements performed during different levels of the VR app against the gold-standard values using a motion capture system.
The analysis was approached with a methodology similar to that of a study published in 2015 in which authors combined motion tracking and virtual reality, using Oculus Rift DK2, to create a video game [
32]. The game explains certain movements to patients that must then be replicated; patient movements are simultaneously tracked as they are being performed to evaluate whether they are being carried out properly [
32]. However, the study did not subject patients to clinical experimentation; instead, they had experts in the field test the products and provide an evaluation. Thus, although there is a similarity in methodology, results cannot be compared.
Although present results are currently stand alone, given that, to our knowledge, the literature has not yet produced similar findings, the VR app seems to have a good performance when compared to gold-standard values measured via motion tracking. In terms of the first and second series of flexion, the total inferior MAE did not exceed 6.4, excluding one case in which the value was equivalent to 12.3. Regarding the number of times inferior flexion conditions were not met by the five volunteers, only once did the value exceed 8. For the superior flexion condition, volunteers performed with a margin of error only 6 times out of 30, while for the remaining 24 trials, volunteers had no individual MAE value. For inferior and superior conditions for abduction, volunteers performed no errors in levels 2 and 3, respectively. In levels 3 and 4, for inferior conditions, the number of times conditions were not met was slightly higher reaching a value of 22, while the total inferior MAE did not exceed 8.6. Overall results for flexion and abduction seem promising, with patients often reaching the gold-standard goal.
Experimentally internal and external rotation were found to have similar values for total MAE as those for flexion and abduction; however, the number of times patients did not meet conditions was higher compared to the other trials. This discrepancy is attributed to the fact that patients were unable to perform an isolated shoulder rotation. Instead, they aided the internal and external rotation of the shoulder by moving their trunk. Overall, more evidence is necessary to definitively confirm that VR rehabilitation programs can in fact have patients reach the same gold standard of movements as can be carried out with traditional rehabilitation programs. Nevertheless, current findings point to this being a possibility in the near future, and further exploration into the topic is necessary to better evaluate the effectiveness of these applications and their development into very effective therapeutic tools.
The growing impact of VR on shoulder rehabilitation is undeniable, as seen by the promising results of the present study, as well as the several new experimental trials working on the development of these VR applications. One double-blinded pilot study found that manipulated virtual real-time presentation of a subject’s movements during shoulder exercises promoted greater active flexion ROM [
3]. Another study investigated the influence of game features and practice type, task-oriented or imitation-oriented, on shoulder muscle activity [
40]. Findings showed that task-oriented practice elicited more intensive shoulder movements and muscular efforts [
40]. These findings combined continue to support the hypothesis that VR is going to enhance patient adherence to rehabilitation programs, increase the ability of professionals to ‘personalize’ rehabilitation for each patient, and, most importantly, have the potential to improve patient functional outcomes.
A small cohort of healthy volunteers was included in this study, which could have a negative impact on the external validity and generalizability of data, in addition to not providing very strong statistical significance to present findings. Despite the negative effects of using a small cohort of healthy volunteers, given the novel topic and the little literature available, present results can serve as a starting point for further, larger-scale investigations, providing support for future research on the topic. Factors such as age, pathology, level of pain, and emotional state may affect the use of VR applications. Investigating the effect of these factors on the use of VR devices while performing rehabilitation protocols was beyond the scope of the present study. Nevertheless, in future studies, stratification for these factors can be very useful in analyzing how these factors could affect the use of VR devices among patients and further explore the validity of VR apps for shoulder rehabilitation after arthroscopic rotator cuff repair.