This paper is in the following e-collection/theme issue:

mHealth for Rehabilitation (313) mHealth for Wellness, Behavior Change and Prevention (3131) Innovations in Exercise Rehabilitation (246) Portable and Mobile Technologies for Rehabilitation (181)

Published on in Vol 13 (2025)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/63022, first published .
Effects of a Computer Vision–Based Exercise Application for People With Knee Osteoarthritis: Randomized Controlled Trial

Effects of a Computer Vision–Based Exercise Application for People With Knee Osteoarthritis: Randomized Controlled Trial

Effects of a Computer Vision–Based Exercise Application for People With Knee Osteoarthritis: Randomized Controlled Trial

Authors of this article:

Dian Zhu1 Author Orcid Image ;   Jianan Zhao2 Author Orcid Image ;   Tong Wu1 Author Orcid Image ;   Beiyao Zhu3 Author Orcid Image ;   Mingxuan Wang1 Author Orcid Image ;   Ting Han1 Author Orcid Image
Article Authors Cited by Tweetations Metrics

    1School of Design, Shanghai Jiao Tong University, Dong Chuan rd, No 800, Shanghai, China

    2College of Fashion and Design, Donghua University, Donghua University, Shanghai, China

    3Department of Plastic and Reconstructive Surgery, Shanghai Jiao Tong University Ninth People's Hospital, Shanghai, China

    Corresponding Author:

    Ting Han, PhD


    Background: Exercise is a primary recommended treatment for knee osteoarthritis (KOA), as it helps alleviate symptoms and improves joint functionality. Personalized exercise programs, tailored to individual patient needs, have demonstrated promising results in maintaining physical fitness and enhancing overall well-being. In recent years, digital health applications have emerged as innovative tools for supervising and facilitating rehabilitation programs. Leveraging computer vision (CV) technology, these applications offer the potential to provide precise feedback and support personalized exercise interventions for patients with KOA in a scalable and accessible manner.

    Objective: This study aims to evaluate the impact of a CV–graded exercise intervention application over a 6-week period on clinical outcomes in patients with KOA . The outcomes were compared to those achieved through conventional exercise education by videos.

    Methods: A randomized controlled trial was conducted with 60 participants aged 60‐80 years, recruited through community administrators between July 2023 and September 2023. Participants were randomly assigned to one of two groups: the graded exercise application group (n=32) and the exercise education brochure group (n=28). The primary outcomes assessed were short-term changes in pain, physical function, and stiffness as measured by the Western Ontario and McMaster Universities Arthritis Index (WOMAC). Secondary outcomes included assessments of participants’ affective state, self-efficacy, quality of life, and user experience.

    Results: The study recruited 60 participants, including 26 males and 34 females. Analysis revealed statistically significant improvements in physical function (P=.02) and self-efficacy (P=.04) in the graded exercise application group compared to the exercise education brochure group after the intervention. While improvements in pain and stiffness were observed in both groups, these changes were not statistically significant. In addition, participants in the graded exercise application group reported a positive user experience, highlighting the application’s usability and engagement features as beneficial to their rehabilitation process.

    Conclusions: The findings suggest that the CV-based graded exercise intervention application effectively improves physical function and self-efficacy among patients with KOA . This digital tool demonstrates the potential to enhance the quality and personalization of exercise rehabilitation compared to traditional methods. Future studies should explore the application’s long-term efficacy and replicability in larger community-based populations, with a focus on sustained engagement and adherence to rehabilitation programs.

    Trial Registration: ClinicalTrials.gov NCT06220565; https://clinicaltrials.gov/study/NCT06220565

    JMIR Mhealth Uhealth 2025;13:e63022

    doi:10.2196/63022

    Keywords

    knee osteoarthritis; behavior change theory; application; digital health; randomized controlled trial; knee; osteoarthritis; behavior; vision; exercise; app; applications; exercise rehabilitation; rehabilitation; older adults; self-efficacy; physical function 


    Knee osteoarthritis (KOA) is a common and chronic joint disorder that is growing in prevalence as the world’s population ages [1,2]. Exercise has been acknowledged as a nonpharmacological intervention modality for the treatment and prevention of musculoskeletal disorders, including osteoarthritis, osteoporosis, back pain, and rheumatoid arthritis [3]. Specifically, it has been empirically demonstrated that participating in appropriate physical exercise while being monitored by a physiotherapist effectively maintains physical health and athletic ability [4]. In order to optimize joint flexibility, strengthen muscles, and reduce strain, they develop individualized training plans that are prescribed in accordance with the patient’s condition [5]. Nevertheless, this procedure imposes considerable financial burdens on the patients and requires a significant time commitment from the caregivers [6].

    Digital health interventions have the potential to mitigate the time and resource limitations faced by patients with KOA by offering education and self-management through web-based platforms or applications that are scalable, inexpensive, readily available, and high coverage [7-9]. Digital health interventions have been shown to potentially offer benefits in the management of musculoskeletal disorders [10,11]. Nevertheless, there are still several digital intervention tools that demonstrate inconsistent levels of effectiveness. For instance, physical therapy sessions and digital applications were shown to be successful in improving physical function [12], while a program using wearable devices to deliver physical activity counseling was deemed ineffective in promoting physical function [13]. This is attributed to the varying functionalities that are accessible and the absence of support for these functionalities in influencing user behavior.

    Research has indicated that the integration of behavior change theory offers the potential for both positive behavior modification and amelioration of negative emotions [14]. Behavior modification approaches are comprised of three essential components: goals and planning, feedback and monitoring, and repetition and replacement [15]. Goals and planning promote increased general physical activity by setting short-term goals and developing evidence-based progressive individualized exercise plans [16]. Monitoring and feedback oversee progress and acquire immediate feedback on the accomplishment of objectives [17]. The application primarily consists of recurring and substituted generalized and graded tasks, which incorporate behavioral exercises and target behaviors [18,19]. A systematic review revealed that exercise programs through positive feedback, effort reinforcement, motivating techniques, and graded interventions had a higher likelihood of adhering to therapeutic exercise [20].

    Previous research has established that patients can improve their management of illnesses through the implementation of goal-setting and planning strategies (monitoring, documenting, etc) as well as the consistent practice and substitution of behaviors (performing repetitive movements, engaging in daily timed exercise, etc) [21]. Clinical treatment is facilitated through the utilization of sensor technology, which provides objective monitoring data for noninvasive evaluation of knee function [22,23]. Despite the availability of various feedback programs formulated to assist patients in adhering to exercise regimens, a subset of patients diagnosed with KOA continue to encounter suboptimal exercise outcomes [9,24,25]. Previous reviews have highlighted that one of the primary objectives of CV-based applications is to calculate joint angles, enabling the identification and counting of correct postures during rehabilitation training. In addition, accurately assessing whether exercises are performed correctly is another critical goal, ensuring that the movement sequence aligns with rehabilitation standards [26,27]. Given the requirement for enhanced and immediate visual input to enable the execution of rehabilitation exercises, the application of computer vision (CV)–algorithms for body position tracking appears to hold considerable promise [28,29]. Nevertheless, clinical investigations evaluating most of these CV-based rehabilitation programs have not yet been conducted.

    Given the tremendous potential of these applications for rehabilitation exercises, a program based on CV was developed. The application facilitates the development of a graded exercise rehabilitation program for patients and aids them in self-monitoring the program’s implementation. The main objective was to assess the effects of using a CV-graded exercise intervention application (after 6 weeks) on clinical outcomes (pain and physical function) among patients diagnosed with KOA. The secondary outcome was to investigate the effects of application implementation on the affective state and self-efficacy of patients with KOA.


    Study Design

    This parallel, 2-arm, unbalanced randomized, single-blinded controlled trial was conducted following the CONSORT (Consolidated Standards of Reporting Trials) Checklist 1 [30] and CONSORT-EHEALTH (Electronic and Mobile Health Applications and Online Telemedicine Reporting Trials) [31] guidelines. The trial was carried out from January 2024 to May 2024 in community activity centers in Shanghai. Participants in the intervention group used CV-based exercise assessment and intervention system designed specifically for patients with KOA, while those in the control group used an exercise rehabilitation education video.

    However, during the pilot study of 4 participants, it was observed that some of them lacked adequate digital health literacy and were unable to independently configure and operate the experimental devices. To address this issue, the research team modified the experimental protocol, organizing in-person sessions at community centers 3 times per week, where trained personnel facilitated the intervention tasks for participants.

    Participants Recruitment

    Participants underwent a 6-week intervention comprising CV-based graded exercise sessions 3 times per week, designed to deliver personalized rehabilitation exercise plans and instructional videos. Recruitment was conducted from July 2023 to January 2024[d1] in Shanghai through invitations extended by community managers and posters distributed in nursing homes.

    Inclusion criteria were as follows: (1) age ≥50 years, (2) radiologically confirmed diagnosis (KL grade ≥2, pain in affected joints) in accordance with American College of Rheumatology clinical criteria, (3) mean overall pain severity ≥4-point numeric rating scale (NRS), (4) good communication and comprehension skills without significant cognitive deficits, and (5) the ability to operate an electronic device with some proficiency.

    Exclusion criteria were as follows: (1) participants with severe knee pain and discomfort, (2) participants with severe organic lesions of vital organs such as heart, brain, and kidney, and (3) participants who had been undergoing knee arthroplasty.

    Only participants who finished the final 6-week assessment of the experiment and did not withdraw were included in this study.

    Interventions

    Intervention Group (Applications)

    This study developed an exercise evaluation and intervention system for individuals suffering from KOA (Multimedia Appendix 1). The objective was to assist participants in preserving and enhancing their knee joint motor function by implementing a tailored progressive rehabilitation regimen and providing immediate feedback. The intervention consists of three components: (1) establishment of a comprehensive knee function assessment system; (2) delivery of secure and precise exercise interventions for participants and offer immediate feedback on the impact of exercise intervention by continuously collecting knee function metrics in real-time; (3) issuing timed tasks to encourage participants to exercise accordingly.

    Status Assessment

    A comprehensive knee function assessment system was developed using CV technology, an electronic Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) questionnaire, joint mobility detection, and mobile gait data monitoring. The tool offers a scientific and all-encompassing evaluation instrument for individuals afflicted with knee joint disorders (Figure 1).

    First, a methodical evaluation of the participants’ joint pain, rigidity, and dyskinesia was conducted with the WOMAC scale. Furthermore, CV technology was used to identify and document the frequency and quantity of movements executed by the participants throughout the physical function assessments. A physical function test involves the participants repeatedly performing a designated movement, and physical function is evaluated by tally-marking the number of movements completed within a specified time interval. CV is also used to dynamically gather the utmost angle of knee flexion. The remote gait data tracking feature relies on Apple’s HealthKit, which enables the system to collect gait data over a period of time. This includes measurements such as step length, step frequency, walking speed, bipedal symmetry, and bipedal support duration. However, the user must provide authorization for the system to access this data. Continuous and extended monitoring of this data allows for evaluation of the advancement of the condition and offers participants insights into their walking patterns and behavioral habits.

    ‎
    Figure 1. Knee function evaluation system.
    Action Detection

    Standardizing movements during exercise interventions has an impact on the body’s dynamics during exercise and physical function testing. Hence, aside from the completion status of the action, they also prioritize factors such as the precision and efficacy of the movement (Figure 2).

    Through the application of CV technology, the system is able to capture the participants’ movement posture. The system collects data from posture capture to create a real-time analysis of movement completion as the participants conduct the scheduled exercise. This enables immediate feedback to be given to the participants during the training regimen. The screen displays a physical function test demonstration, which the older adults observe and subsequently replicate. The software generates a tally of the older adults’ completed motions within a specified time frame. Throughout the entire procedure, the program is equipped with a function that can assess hazardous scenarios. If a fall is imminently possible, the physical function exercise program will promptly halt and emit an audible alert. Upon completion of the test, the older adults could access and review the results displayed on the screen. By conducting regular physical function assessments, individuals can acquire an impartial comprehension of the condition of their knee joint and obtain an accurate assessment of their physical mobility that is supported by the knee joint.

    ‎
    Figure 2. Application of computer vision technology.
    Exercise Rehabilitation

    Following the evaluation process, participants are granted the ability to implement a customized exercise regimen that aligns with their rehabilitation strategy and integrates digital evaluation techniques to obtain immediate feedback and suggestions for modifications. The goal is to prioritize the rehabilitation requirements of the participants through the implementation of a customized rehabilitation program and consistent remote care.

    The primary purpose of the exercise intervention is to implement an interactive exercise program that captures and assesses the participants’ movements by CV during exercise. The user performs the prescribed movements in a sequential manner in order to achieve the daily workout objectives. The subsequent interface will present the exhibited actions, and older individuals can imitate each movement by following the demonstration video (Table 1). A brief interval is observed between each set of movements, and the exercise routine concludes once the task of the designated movements has been completed. The degree of standardization of the movement is denoted by the range of motion, and feedback pertaining to this degree of standardization is additionally acquired following each movement completion.

    Table 1. Description of rehabilitation exercise movements.
    ActionAction instructions
    Calf raisesWith this calf raise exercise, calf muscles can be strengthened so that patients can walk and climb stairs with ease.
    Lateral raise (leg)This side leg raise will strengthen the hip abductors and make everyday tasks like getting in and out of the car easier.
    Knee lift exerciseStrengthen the muscles around the buttocks with this knee lift exercise.
    Seated knee liftIt can be used to improve the ability to get up from a chair, get in and out of bed or a car. Patients can start with the easiest and then slowly progress to the hardest.
    Seated riseIncrease the flexibility and strength of the hip extensors with this standing hip extension exercise.
    Hip stretchesMaintain range of motion in the legs with heel and toe-tapping exercises.
    Heel and toe tappingImproved balance and coordination through toe-tapping exercises.
    Remote Follow-Up

    The primary functions also encompass the formation of a rehabilitation community and conducting remote follow-up. In the context of rehabilitation, incentives like as competition and leader-boards are implemented to motivate participants to actively engage in remote therapy and foster mutual contact within the rehabilitation community. Another purpose of the remote follow-up feature is to establish remote communication channels with outpatient doctors in order to minimize the time and effort participants spend on medical treatment.

    Control Group (Exercise Rehabilitation Education Video)

    Participants in the control group received a video with instructions for daily exercise movements, which were similar to those in the application (Table 1).

    Outcomes

    Participants received validated digital questionnaires at baseline, after 6 weeks of intervention, or face-to-face questionnaire completion at offline follow-up. Participants did not receive (financial) incentives or other compensation for completing the questionnaire or the study. Demographic data were collected at baseline and described in detail in the outcome measures section.

    Main Outcomes

    Pre and postintervention status was assessed by the WOMAC scale [32]. Respondents reported the severity of their usual arthritis pain from 0 (none) to 20 (severe), functional impairment due to osteoarthritis from 0 (none) to 68 (severe), and knee stiffness from 0 (none) to 8 (severe).

    Secondary Outcomes

    The Arthritis Self-Efficacy Scale

    The Arthritis Self-Efficacy Scale (ASES-8) assesses participants’ confidence in their ability to manage arthritis pain and its impact on function [33,34]. Scores range from 1 (very uncertain) to 8 (very certain). Responses are averaged, with higher scores indicating higher self-efficacy.

    Geriatric Depression Scale (GDS)

    Negative emotions were measured by the Geriatric Depression Scale (GDS), whose 30 entries represent the symptoms of depression in old age and contain low mood, decreased activity, irritability, withdrawal from painful thoughts, and negative appraisal of the past, present, and future [35]. Scores range from 0 (none) to 30 (extremely). Higher scores indicate greater positive or negative affect.

    Quality of Life

    Quality of Life (AQoL-6D [Assessment of Quality of Life-6D] version) ranges from -0.04 to 1.00, with higher scores indicating better quality of life [34].

    Range of motion

    Range of motion [36], consisting of both extension (ie, the ability to straighten the leg) and flexion (ie, the ability to bend the leg), with scores ranging from 0 to 135 (higher scores indicate better joint mobility).

    User Experience Questionnaire

    The User Experience Questionnaire (UEQ), which is used to quickly assess the user experience of an interactive product, includes the traditional metrics of ease-of-use aspects: efficiency, comprehensibility, trustworthiness, attractiveness, motivation, and freshness [37].

    Sample Size

    The sample size for this study was determined using a “Sample Size Calculation for Two-Group Mean Comparison.” Considering the characteristics of the intervention and control groups (digital application group and education group, respectively), the calculation was informed by the parameters reported in previous studies [38,39].

    The sample size calculation was conducted assuming a difference between the intervention and control groups (μ₁-μ₂) of 8.8 (SD 9) Using a 2-sided t test with a significance level (α) of .05 and a statistical power (1-β) of 80%, the minimum required sample size was estimated to be 36 participants, with 18 participants in each group. This ensured adequate power to detect meaningful differences between the groups.

    Based on previous experience from our research team’s extensive work in community-based digital rehabilitation, we have observed a high dropout rate among older adults in such settings. To address this, we implemented over-recruitment, increasing the sample size by at least 50% beyond the calculated minimum to ensure sufficient statistical power for final analyses.

    Randomization, Allocation Concealment, and Blinding

    The randomization process will involve stratifying the 60 participants into 15 blocks, each containing 4 participants. Within each block, participants will be randomly allocated to either the intervention group or the control group (received the exercise rehabilitation education video). Experienced statisticians used SAS (SAS Institute) for randomization grouping to ascertain that the random assignment of researchers was concealed for treatment assignment. Due to the nature of the study, participants were not blinded as they knew whether they received the program during the intervention period, while the statisticians were blinded to the group assignment.

    Following the assignment, individuals in the intervention group were sent an email by the researcher (JZ) containing instructions on how to access the computerized visual grading exercise application and details regarding the field trial arrangements. Participants in the control group were notified via email about their assignment and received an exercise rehabilitation education video. Throughout the trial, participants were provided with the option to contact the researchers by phone or email if they had any inquiries regarding the application or the study.

    Statistical Methods

    The analyses were performed by statisticians (MW and TW) on SPSS 26.0 (IBM) software and following the intention-to-treat principle. Baseline characteristics of participants who provided and did not provide the primary outcome were compared using t tests or Χ2 tests. The Mann-Whitney U test was employed to examine between-group and within-group data pre and post experiment. For this work, a significance level (α) of .05 was established for all statistical analyses. Differences were deemed statistically significant for the purposes of this study if the P value was less than .05.

    Ethical Considerations

    The clinical trial was registered (NCT06220565) and approved by the Ethics Committee of Shanghai Jiao Tong University (H20240039I). All participants were required to complete an informed consent form and were informed that their data would remain anonymous. After the completion of the experiment, each participant will receive a US $50 financial compensation.


    Baseline

    Between June and September 2023, a randomization process involving a total of 60 people was conducted who fulfilled the recruiting criteria (32 in the application group and 28 in the educational video group) (Figure 3). A total of 41 (68.33%) participants successfully fulfilled the major outcome measure after 6 weeks, including 24/32 (75%) participants from the application group and 17/28 (60.7%) participants from the instructional video group. The average (SD) age of the sample was 68.8 (SD 6.4), and most of the sample was female (37/60, 61.67%). There were no significant differences between the groups at baseline (Table 2).

    Initially, with the exception of 2 individuals, all participants maintained the belief that the rehabilitation application would aid them in the more precise rehabilitation of their motions. In addition, 87% (135/155) of participants believed that the application would aid them in recovering physical function more effectively. A considerable percentage indicated that they were not presently engaged in active therapy, despite experiencing knee discomfort. There were very few negative incidents reported, with only one instance reporting a negative incident.

    ‎
    Figure 3. The flowchart of the experiment. AQol: Assessment of Quality of Life; ASES: Arthritis Self‐Efficacy Scale; GDS: Geriatric Depression Scale; ROM: range of motion; UEQ: User Experience Questionnaire; WOMAC: Western Ontario and McMaster Universities Arthritis Index.
    Table 2. Baseline characteristics of participants.
    OutcomeIntervention (n=32）Control (n=28)All (N=60）P value
    Age (years), mean (SD)68.59 (7.16)67.64 (7.85)68.15 (7.28).82
    Female, n (%）19 (60)15 (54)35 (57).79
    Medication, n (%）
    Yes11 (34)10 (36)21 (35).43
    No21 (66)18 (64)39 (65)
    Symptom duration (years), n (%）
    <27 (22)6 (23)13 (22).98
    2‐1021 (66)18 (63)39 (65)
    >104 (12)4 (14)8 (13)
    Primary outcome (WOMAC)a, mean (SD)
    Physical function24.13 (2.7)25.39 (2.99)24.72 (2.9).12
    Pain8.87 (1.59)8.64 (1.52)8.76 (1.57).98
    Symptoms3.47 (1.19)3.39 (1)3.43 (1.1).9
    Secondary outcome, mean (SD)
    ASES-8b49.97 (7.49)48.32 (7.25)49.97 (7.37).41
    GDSc6.28 (1.84)6.32 (0.08)6.30 (1.8).9
    AQoL-6Dd0.65 (0.1)0.58 (0.08)0.62 (0.09).06
    ROMe104.75 (7.16)103.93 (7.93)104.37 (7.48).57

    aWOMAC: Western Ontario and McMaster Universities Arthritis Index.

    bASES-8: Arthritis Self-Efficacy Scale–8.

    cGDS: Geriatric Depression Scale.

    dAQoL-6D: 6-Dimensional Assessment of Quality of Life scale.

    eROM: range of motion.

    Primary Outcomes

    Table 3 and Figure 4 describe the application program in the intervention and control groups. At week 6, the application group demonstrated improvement in physical activity impairment (P=.02) compared to the control group, and 15/24 participants reported subsequent adherence to home exercise or community fitness. Although there was a significant within-group difference in the intervention group before and after the intervention (P<.001), there was no evidence of a significant improvement in pain (P=.96) between the intervention group and the control group. In addition, there were no within-group (P=.21) or between-group (P=.49) differences in stiffness before or after the intervention group experiment.

    Table 3. Primary and secondary intervention outcomes.
    Intervention group (n=24）Within groupControl group (n=17）Within groupBetween group
    Baseline6-wkDifference (95% CI)P valueBaseline6-wkDifference 95% CIP valueP value
    Primary outcomes (WOMAC)a
    Physical function24.17 (3.24)17.08 (2.7)−7.08 (−9.98 to 4.18).00b24.76 (3.27)18.71 (2.99)−6.06 (−7.97 to −4.15).42c.02c
    Pain8.83 (1.79)7.21 (1.58)−1.63 (−3.38 to 0.06).00b8.94 (1.82)7.76 (1.52)−1.18 (−3.3 to 0.94).00b.96
    Symptoms3.5 (1.25)3.25 (1.19)−0.25 (−1.19 to 0.69).213.41 (1)3.35 (0.99)−0.06 (−0.81 to 0.69).74.49
    Secondary outcomes, mean (SD)
    ASES-8d49.79 (6.96)51.75 (6.54)1.96 (-2.3 to 6.22).003c48.32 (7.25)49.47 (7.37)−0.47 (−4.44 to 3.5).79.04c
    GDSe6.33 (1.86)5.58 (1.83)−0.75 (−3.06 to 1.56).136.24 (1.3)5.65 (1.8)−0.59 (−2.9 to 1.72).42.98
    AQoLf0.45 (0.18)0.48 (0.19)0.05 (−0.09 to 0.19).080.55 (0.19)0.52 (0.21)−0.03 (−0.23 to 0.17).56.09
    ROMg104.63 (7.72)104.96 (7.79)0.33 (−8.58 to 9.24).76103.29 (7.82)104.23 (9.52)0.94 (−8.62 to 8.62).59.58

    aWOMAC: Western Ontario and McMaster Universities Arthritis Index.

    bP<.001.

    cP<0.05.

    dASES-8: Arthritis Self-Efficacy Scale–8.

    eGDS: Geriatric Depression Scale.

    fAQoL: Assessment of Quality of Life.

    gROM: range of motion.

    ‎
    Figure 4. Box plots illustrating pre and postintervention comparisons of the outcomes.

    Secondary Outcomes

    The secondary outcome data showed a significant difference between the 2 groups in terms of the change in self-efficacy after 6 weeks (P=.04). Nevertheless, the study found no evidence to suggest that the application program intervention led to a noteworthy enhancement in geriatric depression (P=.93) and range of motion (P=.58) as compared to the control group. The intervention group showed a slight improvement in quality of life. However, this improvement was not statistically significant overall (P=.09).

    User Experience

    The scoring results of the 6 dimensions of the UEQ are shown in Table 4. According to the scoring results, it can be visualized that “Attractiveness,” “Clarity” and “Novelty” received high scores (Figure 5). The exercise intervention and functional assessment interface used in this study received the highest score for “attractiveness,” indicating that it really appealed to the patients in the experimental group. The “Clarity” metric, representing the second highest score, suggests that the product presentation is straightforward and easily comprehensible for the user. The elevated score of “Novelty” suggests that the product possesses a commendable level of innovation and captivation, hence enticing customers to engage with it. However, Conversely, “Reliability” receives the lowest rating.

    Table 4. User Experience Questionnaire score.
    DimensionMean (SD)Rating
    Attractive2.05 (0.33)Excellent
    Clarity1.89 (0.25)Good
    Efficiency1.56 (0.17)Good
    Reliability1.48 (0.48)Above average
    Promotion1.72 (0.2)Excellent
    Novelty1.76 (0.35)Excellent
    ‎
    Figure 5. Quality rating score chart.

    Principal Findings

    The extensive incorporation of digital applications into the rehabilitation process for patients with KOA has been driven by the overarching goal of minimizing the duration of time for them to access specialized rehabilitation experts [5]. More precisely, it is developed based on the thorough enhancement of the widely used framework called “Behavior Change Technique,” as it has a greater propensity to support patients in actively participating in the day-to-day management of their chronic illness [16,25,39]. In addition, the upward trend in the proportion of older individuals who access the Internet demonstrates the potential for internet-based interventions designed to assist this demographic [40]. In addition, studies have yielded inconsistent results concerning the effectiveness of remote rehabilitation for individuals with KOA, primarily attributable to the lack of oversight and guidance [41]. Consequently, a suite of CV-based applications was created to aid patients in the execution of rehabilitation maneuvers with accuracy and to track their advancements through the utilization of noninvasive technology. The results of the study indicated that individuals with KOA who made use of the application reported significantly improved physical function and self-efficacy when compared to the control group.

    Research has shown that CV technology–based applications can enhance participants’ physical functioning and alleviate discomfort. This is due to the fact that the capability of integrating behavioral change into digital intervention technologies has been designed to facilitate and encourage adjustments to the patients’ decision-making framework, thereby promoting behavioral change among participants with KOA [42]. The educational aspect of the digital tool encompasses knowledge about the pathology and etiology of osteoarthritis, treatment based on established standards, exercise for osteoarthritis, and methods to alleviate pain and symptoms through adopting healthy behaviors [43,44]. This readily accessible, semisupervised intervention has the potential to be an effective means of alleviating pain and physical dysfunction [39]. Furthermore, our study revealed that the application group experienced a significantly greater improvement in physical functioning compared to the control group. The finding is reminiscent of a study conducted by Tore et al [41], which showed that the quality of physical therapy obtained by telerehabilitation was notably superior to self-management. The difference in quality could perhaps be attributed to the kind and length of the training activities undertaken [45,46]. By means of CV, our application tracks and evaluates the motion of users, thereby facilitating the integration of a customized exercise regimen and aerobic regimen. Contrary to previous research, the investigation did not identify any correlation between the application program and an enhancement in rigidity. Variations in the duration of the studies and the quality and scope of functional support provided may account for this [39]. The reduction in stiffness can be attributed to the fortification of the leg and abdominal muscles; however, a short-term intervention fails to yield substantial outcomes [47,48].

    Furthermore, our research revealed that the utilization of a CV-based application resulted in a favorable impact on self-efficacy. Consistent with findings from previous research [11,49,50], digital self-monitoring programs have been associated with relative increases in self-efficacy activation. The promotion of self-management behaviors is facilitated by the provision of information and individualized instruction pertaining to arthritis care, which encourages and assists in the resolution of individual obstacles [51]. Since the factors that influence exercise behavior are motivational and volitional in nature [52,53]. it can be concluded that the application’s monitoring features provided users with the capability to observe their progress and receive prompt feedback on the degree to which their goals were achieved. Conversely, the intervention did not yield any discernible statistically significant effects on negative affect or quality of life. This discovery is consistent with previous inquiries [54], which have shown that digital health interventions have minimal to no effect on adverse emotions and quality of life. Future work should underlie mechanisms responsible for these feelings in patients.

    In conclusion, the researchers determined, based on the responses to the user experience questionnaire, that the design principles’ objectives for our application had been effectively met. A majority of the respondents indicated that their encounter with the e-exercise intervention was positive. They believed that the interactive design and real-time feedback enhanced the exercisers’ enjoyment and promoted rehabilitation initiatives. Furthermore, apart from guaranteeing a heightened level of individualized attention, the participants perceived the e-enabled exercise intervention to be more tailored to their rehabilitation needs, thus exceeding apprehensions regarding comfort. The participants articulated that the real-time monitoring effectively captured the motion of the joints, thereby augmenting the efficacy of the measurements. Nevertheless, further investigation is necessary to determine the dependability and accuracy of the sensor-based measurement medium in comparison to the traditional measurement method.

    Limitations

    This study has several limitations that warrant discussion. First, the small sample size and the recruitment of participants from communities in close proximity may have led to a concentration of the population distribution, thereby limiting the generalizability of the findings. Second, the pre-experiment revealed that older individuals with insufficient digital health literacy were unable to independently configure and operate the experimental devices. As a result, the experimental protocol was adjusted to involve experimenters facilitating tri-weekly centralized sessions in collaboration with multiple community centers.

    While the experimenters’ involvement was limited to device debugging and their interaction with participants was minimal, their presence may have inadvertently influenced the experimental group’s motivation and engagement with the intervention.

    The broad age range of 60-80 years is another limitation, as physical capabilities vary significantly within this demographic, potentially introducing variability in the intervention’s outcomes. Future studies should consider stratifying participants into narrower age ranges to enhance sample homogeneity and provide more precise insights into the intervention’s effects across age subgroups.

    In addition, the intervention duration was relatively short at 6 weeks, which may not have been sufficient to capture the long-term impact of the application. Future research should extend the study duration, further refine the application design to better accommodate older adults’ needs, and explore tailored strategies to improve digital health literacy. These adjustments would enhance the application’s utility and scalability in diverse community settings, enabling broader adoption and more robust evaluations of its effectiveness.

    To address these limitations and enhance future research outcomes, it is recommended to extend the experiment duration, refine the application design to better accommodate older adults’ needs, and explore strategies to improve digital health literacy among the target population. This would enable a more comprehensive evaluation of the application’s capabilities and its potential impact on long-term rehabilitation outcomes.

    Conclusions

    The findings of this study demonstrated that the application based on CV technology effectively improved the physical functioning and self-efficacy of participants compared to conventional interventions. This suggests that the application holds promise for replication and implementation within community environments for patients with KOA.

    A key novelty of this research is its validation, through a randomized controlled trial, of an application designed according to behavioral change theory principles. The study confirmed that such an application can partially substitute the guidance of a rehabilitation therapist, thereby enhancing exercise rehabilitation outcomes for participants.

    In addition, the potential scalability and adaptability of the application in diverse settings warrant further exploration. Its implementation could significantly support older adults in community health management, particularly in addressing the varying levels of digital health literacy among the population. Future research should focus on extending the application’s reach to broader demographics, optimizing its design to better accommodate older adults, and tailoring strategies to improve engagement and usability. This will maximize the application’s practical impact and facilitate its adoption in real-world scenarios.

    Acknowledgments

    We thank the members of our steering committee for their oversight and expert guidance. We also thank the stakeholders who contributed to our project management group and people with lived experience of receiving or providing exercise therapy for KOA patients who took part in the stakeholder workshops and online survey for their useful feedback. The study was funded by National Key R&D Program of China (Grant No. 2022YFB3303303) Fundamental Research Funds for the Central Universities (Grant No. YG2023ZD10) and Lushan Lab Research Funding. We would also like to extend our special thanks to Tian Xia, Bochen Cao, and Wenhui Zhang for their contributions to the product design.

    Data Availability

    The datasets generated or analyzed during this study are available from the corresponding author on reasonable request.

    Authors' Contributions

    DZ contributed to writing–original draft, conceptualization, data curation, and methodology. JZ was responsible was

    conceptualization, data curation, software, and methodology. TH handled writing–review and editing, project administration, and funding acquisition. TW conducted formal analysis and visualization. MW was involved in software and validation. BZ performed investigation and data curation.

    Conflicts of Interest

    None declared.

    Multimedia Appendix 1

    Description of the Motion Tracking and Evaluation Algorithm in the CV Application.

    DOCX File, 1653 KB
    Checklist 1

    CONSORT checklist.

    PDF File, 1235 KB
    1. Uivaraseanu B, Vesa CM, Tit DM, et al. Therapeutic approaches in the management of knee osteoarthritis (Review). Exp Ther Med. May 2022;23(5):328. [CrossRef] [Medline]
    2. Jormand H, Mohammadi N, Khani Jeihooni A, et al. Self-care behaviors in older adults suffering from knee osteoarthritis: Application of theory of planned behavior. Front Public Health. 2022;10:958614. [CrossRef] [Medline]
    3. Arden NK, Perry TA, Bannuru RR, et al. Non-surgical management of knee osteoarthritis: comparison of ESCEO and OARSI 2019 guidelines. Nat Rev Rheumatol. Jan 2021;17(1):59-66. [CrossRef] [Medline]
    4. Sharma L. Osteoarthritis of the knee. N Engl J Med. Jan 7, 2021;384(1):51-59. [CrossRef] [Medline]
    5. Kraus VB, Sprow K, Powell KE, et al. Effects of physical activity in knee and hip osteoarthritis: a systematic umbrella review. Med Sci Sports Exerc. Jun 2019;51(6):1324-1339. [CrossRef] [Medline]
    6. Kampmeijer R, Pavlova M, Tambor M, et al. The use of e-health and m-health tools in health promotion and primary prevention among older adults: a systematic literature review. BMC Health Serv Res. Sep 5, 2016;16 Suppl 5(Suppl 5):290. [CrossRef] [Medline]
    7. Wilson R, Chua J, Briggs AM, et al. The cost-effectiveness of recommended adjunctive interventions for knee osteoarthritis: results from a computer simulation model. Osteoarthr Cartil Open. Dec 2020;2(4):100123. [CrossRef] [Medline]
    8. Kostic AM, Leifer VP, Gong Y, et al. Cost-effectiveness of surgical weight-loss interventions for patients with knee osteoarthritis and class III obesity. Arthritis Care Res (Hoboken). Mar 2023;75(3):491-500. [CrossRef] [Medline]
    9. Bennell KL, Nelligan R, Dobson F, et al. Effectiveness of an internet-delivered exercise and pain-coping skills training intervention for persons with chronic knee pain: a randomized trial. Ann Intern Med. Apr 4, 2017;166(7):453-462. [CrossRef] [Medline]
    10. Hewitt S, Sephton R, Yeowell G. The effectiveness of digital health interventions in the management of musculoskeletal conditions: systematic literature review. J Med Internet Res. Jun 5, 2020;22(6):e15617. [CrossRef] [Medline]
    11. Shah N, Costello K, Mehta A, et al. Applications of digital health technologies in knee osteoarthritis: narrative review. JMIR Rehabil Assist Technol. Jun 8, 2022;9(2):e33489. [CrossRef] [Medline]
    12. Kloek CJJ, Bossen D, Spreeuwenberg PM, et al. Effectiveness of a blended physical therapist intervention in people with hip osteoarthritis, knee osteoarthritis, or both: a cluster-randomized controlled trial. Phys Ther. Jul 1, 2018;98(7):560-570. [CrossRef] [Medline]
    13. Li LC, Feehan LM, Xie H, et al. Effects of a 12-week multifaceted wearable-based program for people with knee osteoarthritis: randomized controlled trial. JMIR Mhealth Uhealth. Jul 3, 2020;8(7):e19116. [CrossRef] [Medline]
    14. Dobson F, Hinman RS, French S, et al. Internet-mediated physiotherapy and pain coping skills training for people with persistent knee pain (IMPACT - knee pain): a randomised controlled trial protocol. BMC Musculoskelet Disord. Aug 13, 2014;15(1):279. [CrossRef] [Medline]
    15. Bouma AJ, van Wilgen P, Dijkstra A. The barrier-belief approach in the counseling of physical activity. Patient Educ Couns. Feb 2015;98(2):129-136. [CrossRef] [Medline]
    16. Bennell KL, Campbell PK, Egerton T, et al. Telephone coaching to enhance a home-based physical activity program for knee osteoarthritis: a randomized clinical trial. Arthritis Care Res (Hoboken). Jan 2017;69(1):84-94. [CrossRef] [Medline]
    17. Bailey JF, Agarwal V, Zheng P, et al. Digital care for chronic musculoskeletal pain: 10,000 participant longitudinal cohort study. J Med Internet Res. May 11, 2020;22(5):e18250. [CrossRef] [Medline]
    18. Michie S, Richardson M, Johnston M, et al. The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: building an international consensus for the reporting of behavior change interventions. Ann Behav Med. Aug 2013;46(1):81-95. [CrossRef] [Medline]
    19. Zhu D, Zhao J, Wang M, et al. Rehabilitation applications based on behavioral therapy for people with knee osteoarthritis: systematic review. JMIR Mhealth Uhealth. May 2, 2024;12:e53798. [CrossRef] [Medline]
    20. Bowden JL, Hunter DJ, Deveza LA, et al. Core and adjunctive interventions for osteoarthritis: efficacy and models for implementation. Nat Rev Rheumatol. Aug 2020;16(8):434-447. [CrossRef] [Medline]
    21. Bossen D, Veenhof C, Dekker J, et al. The usability and preliminary effectiveness of a web-based physical activity intervention in patients with knee and/or hip osteoarthritis. BMC Med Inform Decis Mak. May 28, 2013;13:61. [CrossRef] [Medline]
    22. Cui X, Zhao Z, Ma C, et al. A gait character analyzing system for osteoarthritis pre-diagnosis using RGB-D camera and supervised classifier. In: World Congress on Medical Physics and Biomedical Engineering 2018. Springer; 2018:297-301. URL: https://link.springer.com/chapter/10.1007/978-981-10-9035-6_53#citeas [Accessed 2025-04-16] [CrossRef]
    23. Bolink S, van Laarhoven SN, Lipperts M, et al. Inertial sensor motion analysis of gait, sit-stand transfers and step-up transfers: differentiating knee patients from healthy controls. Physiol Meas. Nov 2012;33(11):1947-1958. [CrossRef] [Medline]
    24. Fitzgibbon ML, Tussing-Humphreys L, Schiffer L, et al. Fit and Strong! Plus: Twelve and eighteen month follow-up results for a comparative effectiveness trial among overweight/obese older adults with osteoarthritis. Prev Med. Dec 2020;141:106267. [CrossRef] [Medline]
    25. Bennell K, Nelligan RK, Schwartz S, et al. Behavior change text messages for home exercise adherence in knee osteoarthritis: randomized trial. J Med Internet Res. Sep 28, 2020;22(9):e21749. [CrossRef] [Medline]
    26. Debnath B, O’Brien M, Yamaguchi M, et al. A review of computer vision-based approaches for physical rehabilitation and assessment. Multimed Syst. Feb 2022;28(1):209-239. [CrossRef]
    27. Eswaran U, Khang A. Artificial intelligence (AI)-aided computer vision (CV) in healthcare system. In: Computer Vision and AI-Integrated IoT Technologies in the Medical Ecosystem. CRC Press; 2024:125-137. [CrossRef]
    28. Sanford S, Liu M, Selvaggi T, et al. Effects of visual feedback complexity on the performance of a movement task for rehabilitation. J Mot Behav. 2021;53(2):243-257. [CrossRef] [Medline]
    29. Bing F, Wang Y, Chen SF, et al. Effects of cycling rehabilitation training on patients with knee osteoarthritis: a systematic review and meta‑analysis. Research Square. Preprint posted online on Jan 6, 2023. [CrossRef]
    30. Boutron I, Moher D, Altman DG, et al. Extending the CONSORT statement to randomized trials of nonpharmacologic treatment: explanation and elaboration. Ann Intern Med. Feb 19, 2008;148(4):295-309. [CrossRef] [Medline]
    31. Eysenbach G, CONSORT-EHEALTH Group. CONSORT-EHEALTH: improving and standardizing evaluation reports of Web-based and mobile health interventions. J Med Internet Res. Dec 31, 2011;13(4):e126. [CrossRef] [Medline]
    32. Bellamy N, Buchanan WW, Goldsmith CH, et al. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol. Dec 1988;15(12):1833-1840. [Medline]
    33. Brady TJ. Measures of self‐efficacy: arthritis self‐efficacy scale (ASES), arthritis self‐efficacy scale‐8 item (ASES‐8), children’s arthritis self‐efficacy scale (CASE), chronic disease self‐efficacy scale (CDSES), parent’s arthritis self‐efficacy scale (PASE), and rheumatoid arthritis self‐efficacy scale (RASE). Arthritis Care Res (Hoboken). Nov 2011;63(S11):S473-S485. [CrossRef]
    34. Allen J, Inder KJ, Lewin TJ, et al. Construct validity of the assessment of quality of life - 6D (AQoL-6D) in community samples. Health Qual Life Outcomes. 2013;11:1. [CrossRef] [Medline]
    35. Kurlowicz L, Greenberg SA, Kurlowicz L, Greenberg SA. The geriatric depression scale (GDS). Am J Nurs. 2007;107(10):67-68. [CrossRef]
    36. Li T, Ma J, Zhao T, et al. Application and efficacy of extracorporeal shockwave treatment for knee osteoarthritis: a systematic review and meta-analysis. Exp Ther Med. Oct 2019;18(4):2843-2850. [CrossRef] [Medline]
    37. Schrepp M, Hinderks A, Thomaschewski J. Construction of a benchmark for the user experience questionnaire (UEQ). IJIMAI. 2017;4(4):40. [CrossRef]
    38. Fernandes L, Storheim K, Sandvik L, et al. Efficacy of patient education and supervised exercise vs patient education alone in patients with hip osteoarthritis: a single blind randomized clinical trial. Osteoarthr Cartil. Oct 2010;18(10):1237-1243. [CrossRef]
    39. Nelligan RK, Hinman RS, Kasza J, et al. Effects of a self-directed web-based strengthening exercise and physical activity program supported by automated text messages for people with knee osteoarthritis: a randomized clinical trial. JAMA Intern Med. Jun 1, 2021;181(6):776-785. [CrossRef] [Medline]
    40. Seiferth C, Vogel L, Aas B, et al. How to e-mental health: a guideline for researchers and practitioners using digital technology in the context of mental health. Nat Mental Health. 2023;1(8):542-554. [CrossRef]
    41. Tore NG, Oskay D, Haznedaroglu S. The quality of physiotherapy and rehabilitation program and the effect of telerehabilitation on patients with knee osteoarthritis. Clin Rheumatol. Mar 2023;42(3):903-915. [CrossRef] [Medline]
    42. Tack C. A model of integrated remote monitoring and behaviour change for osteoarthritis. BMC Musculoskelet Disord. Aug 9, 2021;22(1):669. [CrossRef] [Medline]
    43. Jönsson T, Dell’Isola A, Lohmander LS, et al. Comparison of face-to-face vs digital delivery of an osteoarthritis treatment program for hip or knee osteoarthritis. JAMA Netw Open. Nov 1, 2022;5(11):e2240126. [CrossRef] [Medline]
    44. Cenamor J. Use of health self-management platform features: the case of a specialist ehealth app. Technol Forecast Soc Change. Dec 2022;185:122066. [CrossRef]
    45. Batrakoulis A, Jamurtas AZ, Metsios GS, et al. Comparative efficacy of 5 exercise types on cardiometabolic health in overweight and obese adults: a systematic review and network meta-analysis of 81 randomized controlled trials. Circ Cardiovasc Qual Outcomes. Jun 2022;15(6):e008243. [CrossRef] [Medline]
    46. Solis-Navarro L, Gismero A, Fernández-Jané C, et al. Effectiveness of home-based exercise delivered by digital health in older adults: a systematic review and meta-analysis. Age Ageing. Nov 2, 2022;51(11):afac243. [CrossRef] [Medline]
    47. Assar S, Gandomi F, Mozafari M, et al. The effect of Total resistance exercise vs. aquatic training on self-reported knee instability, pain, and stiffness in women with knee osteoarthritis: a randomized controlled trial. BMC Sports Sci Med Rehabil. 2020;12(1):27. [CrossRef] [Medline]
    48. Chen H, Zheng X, Huang H, et al. The effects of a home-based exercise intervention on elderly patients with knee osteoarthritis: a quasi-experimental study. BMC Musculoskelet Disord. Apr 9, 2019;20(1):160. [CrossRef] [Medline]
    49. Seppen BF, den Boer P, Wiegel J, et al. Asynchronous mHealth interventions in rheumatoid arthritis: systematic scoping review. JMIR Mhealth Uhealth. Nov 5, 2020;8(11):e19260. [CrossRef] [Medline]
    50. Baumeister H, et al. Persuasive E-health design for behavior change. In: Montag C, Baumeister H, editors. Digital Phenotyping and Mobile Sensing, New Developments in Psychoinformatics. Springer; 2023:347-364. [CrossRef]
    51. Olsson CB, Ekelund J, Degerstedt Å, et al. Change in self-efficacy after participation in a supported self-management program for osteoarthritis - an observational study of 11 906 patients. Disabil Rehabil. Jul 2020;42(15):2133-2140. [CrossRef] [Medline]
    52. Durst J, Roesel I, Sudeck G, et al. Effectiveness of human versus computer-based instructions for exercise on physical activity-related health competence in patients with hip osteoarthritis: randomized noninferiority crossover trial. J Med Internet Res. Sep 28, 2020;22(9):e18233. [CrossRef] [Medline]
    53. Xie SH, Wang Q, Wang LQ, et al. Effect of internet-based rehabilitation programs on improvement of pain and physical function in patients with knee osteoarthritis: systematic review and meta-analysis of randomized controlled trials. J Med Internet Res. Jan 5, 2021;23(1):e21542. [CrossRef] [Medline]
    54. Valentijn PP, Tymchenko L, Jacobson T, et al. Digital health interventions for musculoskeletal pain conditions: systematic review and meta-analysis of randomized controlled trials. J Med Internet Res. Sep 6, 2022;24(9):e37869. [CrossRef] [Medline]

    ‎
    ASES: Arthritis Self-Efficacy Scale
    CONSORT: Consolidated Standards of Reporting Trials
    CV: computer vision
    GDS: Geriatric Depression Scale
    KOA: knee osteoarthritis
    NRS: numeric rating scale
    ROM: Range of motion
    UEQ: User Experience Questionnaire
    WOMAC: Western Ontario and McMaster Universities Arthritis Index

    Edited by Lorraine Buis; submitted 08.06.24; peer-reviewed by Mohammad Abu Shaphe, Reza Mansoorizade; final revised version received 19.12.24; accepted 23.02.25; published 12.05.25.

    Copyright

    © Dian Zhu, Jianan Zhao, Tong Wu, Beiyao Zhu, Mingxuan Wang, Ting Han. Originally published in JMIR mHealth and uHealth (https://mhealth.jmir.org), 12.5.2025.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on https://mhealth.jmir.org/, as well as this copyright and license information must be included.