With the rapid global aging of the population, degenerative cervical spine disease has become a major public health challenge affecting older adults. Approximately 85% of individuals above 60 years old show radiographic evidence of cervical degeneration, and the number of people living with neck pain reached 203 million—an increase of 77.3% since 1990—placing neck pain among the leading contributors to musculoskeletal disability [1]. Also, the global age-standardized prevalence of neck pain is 27 per 1000 in 2019, causing substantial health care expenditure and productivity loss [2].

For degenerative cervical spine disease accompanied by spinal cord or nerve root compression, anterior cervical discectomy and fusion (ACDF) remains the surgical gold standard [3]. However, shorter inpatient stays indicate that most functional recovery occurs after discharge, and traditional outpatient physiotherapy often lacks continuous monitoring, real-time feedback, and psychological support [4]. It is further revealed that postoperative patients and their caregivers widely seek support to close these postdischarge guidance gaps [4].

Digital rehabilitation has the potential to address these shortcomings and provide remote, real-time, and individualized exercise guidance. In previous studies involving chronic low back pain, AI-assisted tele-rehabilitation significantly reduced pain and improved function within 4‐8 weeks [5,6], while systematic reviews confirmed that digital health interventions lessen neck and back pain and enhance quality of life [7,8]. Other studies further indicate that spine-surgery patients using digital platforms experience lower rates of postoperative emergency visits and readmissions, highlighting their potential in postoperative care [9].

However, evidence for digital rehabilitation specifically designed for ACDF postoperative patients remains limited [9]. Thus, this study evaluates the effectiveness and safety of an AI-driven digital rehabilitation program in promoting functional recovery and improving compliance following ACDF, thereby providing evidence-based postoperative digital rehabilitation strategies.

This study aims to evaluate the clinical efficacy of a digital rehabilitation program that combines computer-vision–guided exercise coaching, real-time inertial-sensor monitoring, and cognitive behavioral interventions for patients recovering from ACDF. By systematically comparing neck-pain intensity, quality of life, and functional recovery before and after the intervention, we will assess the program’s effectiveness in relieving symptoms, enhancing quality of life, and promoting functional recovery.

This single-centered, prospective cohort study was carried out at Peking Union Medical College Hospital to compare a 12-week digital rehabilitation program with standard in-person physiotherapy after ACDF [10-15]. A total of 336 eligible patients were enrolled.

Postoperatively, participants selected their rehabilitation modality based on personal preference and logistical convenience, forming the digital rehabilitation group and the in-person rehabilitation group (IRG). Within the digital rehabilitation group, compliance criteria further divided patients into a digital rehabilitation completion group (DCG) and a digital rehabilitation noncompletion group (DNG).

All groups followed an identical evidence-based rehabilitation protocol. Digital groups (DCG and DNG) used the healbone rehabilitation system (HRS), a remote rehabilitation platform developed by Jiakangzhongzhi Technology Co, paired with inertial measurement unit (IMU) wearables, completing three 30-minute sessions per week. IRG attended one 45-minute, in-person session with a physical therapist each week and performed the remaining exercises independently at home using a printed manual. Pain, functional status, health-related quality of life, and cervical range of motion were assessed at baseline (Week 0), Week 12, and Week 24, whereas patient satisfaction was measured only at Week 24.

All participants were recruited at Peking Union Medical College Hospital. Two licensed physicians enrolled postoperative ACDF candidates according to the inclusion and exclusion criteria. Each participant received detailed information about the study’s objectives, procedures, and potential risks. Enrollment proceeded only after written informed consent was obtained. The inclusion and exclusion criteria were as follows:

Because of the distinct modes of intervention, conventional double‐blinding was not feasible. To minimize bias, we implemented several safeguards. Follow-up assessments were conducted by clinicians blinded to group allocation. Group identifiers were replaced with coded labels during data analysis, and data analysis was completed without knowledge of group allocation.

The digital and in-person groups followed an identical rehabilitation protocol, designed by licensed physicians and physical therapists with reference to clinical guidelines. The program included cervical range-of-motion exercises, cervical-muscle strengthening exercises, scapular-stability exercises, and functional exercises. Full details are provided in Multimedia Appendix 1. All patients underwent follow-up assessments at baseline, Week 12, and Week 24.

Before discharge from the hospital, participants installed the training app on their smartphones and logged in with their registered accounts. Patients then followed the software app, which provided tutorials on how to complete each exercise. The cell phone’s front-facing camera captured cervical angles in real time, and the software provided verbal and auditory feedback to patients when they were completing their exercises in real time [19-22]. Each session lasted about 30 minutes, 3 times per week, for a total of 36 sessions over 12 weeks according to clinical guidelines [23-27].

The digital rehabilitation app delivers eight micro-lessons designed to reduce postoperative chronic pain and psychological stress. Each lesson lasts 15‐20 minutes and is completed once per week during the first 8 weeks of training. Licensed therapists created the curriculum with reference to clinical protocols. The lessons progress through the following topics:

After completing each lesson, patients complete an interactive homework assignment and write an electronic diary. Licensed therapists review submissions online each week and provide feedback within 24 hours, thereby fostering patients’ self-management skills.

Patients in the digital rehabilitation group were instructed to wear a 9-axis IMU on the forehead for real-time tracking of cervical flexion and rotation. The device had to be worn for more than 4 hours per day during waking activities. If forward flexion exceeded or axial rotation surpassed a preset range, the app sent an alert. Each day, the software generated a posture-monitoring report and uploaded it to the cloud for therapist review. If no data were received for 48 hours, the system issued an automatic reminder to the patient, followed by a therapist phone check-in to improve engagement [28-31].

Within the digital rehabilitation arm, patients were classified as either the DCG or the DNG according to the following criteria:

Exercise compliance: all 36 scheduled sessions had to be completed. Each session required every prescribed exercise to be completed.

CBT compliance: all 8 CBT micro-lessons had to be viewed, with the associated homework submitted.

Posture-monitoring compliance: the IMU had to be worn for ≥4 hours per day, on ≥70% of the total days for the 12-week digital program [32-34].

Every DCG patient completed the full set of 36 sessions and met the CBT and posture-monitoring requirements. By contrast, most DNG participants discontinued the program early and did not meet completion criteria.

All statistical analyses were performed using SPSS software (version 29.0; IBM Corp). Descriptive statistics for baseline characteristics were reported as mean (SD) for continuous variables or number (%) for categorical variables. Between-group comparisons for continuous variables were conducted using independent-sample 2-tailed t tests under the assumption of normality. Categorical variables were compared using χ² tests. Primary outcomes (VAS and NDI) and secondary outcomes (SF-36 MCS, cervical ROM, muscle endurance) were assessed at baseline (0 week), 12 weeks, and 24 weeks postoperatively. Results were presented as means, SDs, and 95% CIs. Pairwise group comparisons between the DCG, DNG, and IRG were conducted at each time point using independent-sample 2-tailed t tests. A P value less than .05 was considered statistically significant.

This study was supported by CAMS Innovation Fund for Medical Sciences (CIFMS; Grant No. 2023-I2M-C&T-B-002) and the National High-Level Hospital Clinical Research Funding (Grant No. 2022-PUMCH-A-123). The protocol adhered to the principles of the Declaration of Helsinki and was reviewed and approved by the Ethics Committee of Peking Union Medical College Hospital, Chinese Academy of Medical Sciences (approval No. I-25PJ0581). Written informed consent was obtained from all participants prior to enrollment. All personal data were coded and deidentified to ensure participant privacy and data security. No financial or other compensation was offered to participants.

Between January and December 2023, we screened 368 candidates for the study. Fifteen were excluded because of severe spinal-canal stenosis, 9 because of comorbid conditions that limited exercise, and 8 declined participation. The remaining 336 patients met all criteria and completed the baseline assessment (Figure 3). All were followed longitudinally and allocated to one of 3 cohorts according to actual rehabilitation uptake: 192 in the DCG, 78 in the DNG, and 66 in the IRG.

Baseline demographic and neurological profiles were well matched. Mean age was 54.3 (10.1) years in the DCG, 54.1 (9.7) years in the DNG, and 53.4 (9.5) years in the IRG. Women accounted for 50%, 52%, and 51% of each group, respectively. On preoperative neurological examination, the prevalence of unilateral radicular pain was similar (56%, 59%, and 57%, respectively), as was bilateral radicular pain (13%, 14%, and 15%). Rates of muscle weakness (57%, 55%, and 56%), sensory deficit (43%, 45%, and 44%), and a positive upper-limb traction test (80%, 77%, and 78%) also showed no meaningful differences. These findings confirm baseline comparability among the 3 groups (Table 1).

At Week 24, patient-reported satisfaction was high and comparable between the DCG and the IRG: DCG, 95.7 (6.1) (95% CI 94.8‐96.6) versus IRG, 95.9 (6.5) (95% CI 94.3‐97.5); P=.83. The DNG averaged 84.3 (7.9) (95% CI 82.5‐86.1). Satisfaction in the DNG was lower than in the DCG (P<.001) and IRG (P<.001), reaching statistical significance (Table 4).

We enrolled 336 patients who underwent ACDF and compared 3 rehabilitation strategies—a complete digital program, conventional in-person physiotherapy, and an incomplete digital program—at 12 and 24 weeks. Patients who finished the digital program (DCG) achieved pain (VAS) and neck-specific disability (NDI) reductions comparable to those in the in-person group (IRG) and superior to those in the digital group who did not complete the program (DNG). Changes in 3-plane cervical range of motion and in flexor/extensor muscle endurance were greater in both DCG and IRG than in the DNG. The DCG also reported the largest improvement in mental health (SF-36 MCS). Patient satisfaction remained high and similar between the DCG and IRG groups, whereas inadequate compliance in DNG substantially weakened treatment benefits.

This study validated the effectiveness of a digital rehabilitation program in facilitating functional recovery following ACDF surgery. At 24 weeks postsurgery, the DCG demonstrated superior improvements in cervical spine mobility across sagittal, coronal, and horizontal planes compared to the DNG, with a range of motion improved reaching 23.7 degrees, 19.4 degrees, and 19.7 degrees for DCG versus 11.9 degrees, 9.8 degrees, and 9.2 degrees for DNG, respectively. Moreover, significant improvements in neck muscle endurance were observed in the DCG, for anterior and posterior muscle groups at 11.8 seconds and 14.8 seconds, respectively, compared to 6 seconds and 3.6 seconds in the DNG. A notable reduction in the NDI of 13.3 for DCG and 6.4 for DNG was also documented, further showing the digital rehabilitation program’s effect in promoting cervical spine functional recovery.

This study highlighted the effectiveness of combining computer vision-guided exercises with sensor-based monitoring to deliver precise rehabilitation guidance postsurgery. The computer vision-guided approach allowed the physician to monitor whether patients completed the exercises correctly and detected when there was difficulty performing an exercise. By providing precise and real-time feedback via computer vision, the digital program allows the physician to monitor the patient remotely and to adjust the intensity of the exercise, ensuring that patients are engaged in suitable exercises, therefore improving the effectiveness of the rehabilitation process. The wearable sensor further monitored patient behavior to correct postures, which could facilitate recovery. These findings are consistent with previous studies that showed digital rehabilitation is comparable to traditional rehabilitation methods in terms of effectiveness [39]. This is further supported by DCG and IRG demonstrating similar outcomes, both superior to DNG, showcasing the potential of technology-enhanced programs to offer more customized and flexible patient management solutions [40].

Additionally, this study evaluated the impact of integrating CBT into a digital rehabilitation program to improve pain management and psychological health. The usage of a digital platform for CBT delivery allowed the patients to have continuous access to pain management content and strategies, irrespective of physical location or mobility restrictions. The digital program taught patients how to manage postsurgical pain and educated patients to focus on breathing exercises and guided exercises to decrease pain. The results showed that a notable reduction in the VAS pain score of 4.1 for DCG by the 24th week, compared to 1.3 for the DNG. This finding is consistent with previous studies that demonstrated CBT techniques, such as cognitive restructuring and coping strategy enhancement, significantly reduce pain intensity and distress associated with pain conditions [36]. This is consistent with the gate control theory of pain, which proposes that when non-nociceptive input predominates, such as tactile stimulation or other sensory modalities, the perception of pain is reduced [40]. Moreover, the emphasis on behavior modification strategies, encouraging active engagement in exercise and social activities, resonates with previous work that found lifestyle modifications play a vital role in mitigating the influence of pain on daily functioning and overall quality of life [41]. These findings demonstrate the importance of the cognitive-behavioral component of pain management.

This study’s findings showed that patients in the DCG reported high satisfaction scores comparable to those in the IRG, with scores of 95.7 and 95.9, respectively. In contrast, the DNG reported a significantly lower satisfaction score of 84.3 points. The higher patient satisfaction observed in both the DCG and IRG for cervical spine rehabilitation post-ACDF was potentially caused by several key factors. The digital rehabilitation program tailored the rehabilitation plans to individual needs and introduced the convenience of undergoing therapy at home, the ease of access, and personalization potentially improved patient satisfaction. This is supported by previous studies that highlighted digital health programs in promoting patient engagement and satisfaction by leveraging the convenience and flexibility of technology-based solutions [42].

Moreover, the interactive software used in the digital rehabilitation program made rehabilitation exercises more engaging and interactive, similar to gameplay. The incorporation of interactive experiences within the digital rehabilitation program introduces patient satisfaction in the rehabilitation process, a factor that not only aids in maintaining high levels of patient compliance but also enriches the psychological well-being of patients undergoing treatment [43,44]. This finding is similar to previous research, which asserts that the interactive components of digital health programs contribute to positive patient experiences and outcomes [45]. These elements together contribute to the overall increase in patient satisfaction.

The SF-36 MCS results highlight the impact of the rehabilitation types on mental health post-ACDF cervical spine rehabilitation. Over the 24 weeks, the DCG and IRG saw significant mental health improvements, with scores rising of 21.4 and 12.9 points, respectively. The DNG showed a moderate improvement of 7.8 points. This is consistent with previous studies that showed digital rehabilitation platforms, particularly those incorporating cognitive-behavioral strategies, have a significant positive impact on mental health [46]. The improved mental health score could be due to the comprehensive approach of these systems. Particularly in digital rehabilitation, the inclusion of CBT provides essential psychological support, directly contributing to mental health improvements. Also, software increases the patient education frequency, which could have a more profound impact on the patient’s mental health improvement. This is in contrast with the standard of care patient education, which solely focused on postoperative wound care and joint function training, which did not address the mental well-being of the patient.

Our study shows that patients who completed the 12-week digital rehabilitation program experienced pain relief, functional gains, and improvements in three-plane cervical range of motion and muscle endurance that matched those achieved through conventional, clinic-based therapy at both 12 and 24 weeks after ACDF. In addition, the digital group reported better mental health scores and higher satisfaction, whereas participants with poor adherence realized only modest improvements. These findings reinforce earlier telerehabilitation research indicating that smartphone- and wearable-based interventions can equal in-person physiotherapy in clinical effectiveness while offering greater accessibility and adherence [5,47].

The platform we evaluated combines computer-vision motion analysis, inertial-sensor posture monitoring, and brief CBT modules within a single mobile app. This training-and-feedback loop provides patients with real-time corrective cues and allows licensed clinicians to remotely review performance data and provide feedback to the patients. The CBT component addresses postoperative mood and motivation, thereby supporting sustained engagement [48]. High satisfaction and completion rates in the DRG further attest to the platform’s ability to keep users actively involved [6,49].

From a health services perspective, digital rehabilitation offers several advantages. First, clinicians can monitor recovery more frequently without requiring patients to make repeated clinic visits, reducing travel time and associated costs. Second, continuous data captured by the software enables providers to identify early signs of nonadherence or postoperative psychological distress and intervene promptly. Finally, because its underlying technologies are disease-agnostic, a similar framework can be applied to other musculoskeletal diseases, providing a scalable template for modernizing postoperative rehabilitation care.

This investigation was conducted at a single, high-volume tertiary medical center, which may limit geographic and institutional generalizability; however, the center’s standardized surgical pathways and comprehensive follow-up enhance internal validity and provide a benchmark for future multicenter studies. Second, rehabilitation modality was determined by participant preference rather than random allocation, introducing the possibility of selection bias. Third, although patients who underwent multilevel fusion were excluded to ensure surgical homogeneity, the resulting cohort offers a clear view of single-level ACDF recovery and establishes a foundation for subsequent work in more complex cases. These considerations suggest that in the future, randomized and multicenter clinical trials can extend the present findings to broader patient populations and surgical scenarios.

This investigation demonstrates that a 12-week digital rehabilitation program delivers postoperative pain relief and functional recovery comparable to conventional, clinic-based therapy following ACDF. Beyond matching these core outcomes, the digital pathway confers additional advantages in psychological well-being and maintains high levels of patient-reported satisfaction. By combining computer-vision–guided exercise, wearable sensor monitoring, and embedded cognitive-behavioral support, the platform enables effective remote rehabilitation for ACDF patients. High compliance translates directly into greater clinical benefit, underscoring engagement as a pivotal determinant of success in remote care models. These findings support digital rehabilitation as a clinically effective, resource-efficient, and patient-centered option for postoperative rehabilitation, with potential for broader applications in musculoskeletal care.