This was a single-center, prospective, open-label, randomized controlled trial. We used an offline approach to screen adult patients who underwent radical lung cancer surgery in the Department of Thoracic Surgery at Zhujiang Hospital, Southern Medical University, between June 2024 and January 2025, and the study protocol adhered to the CONSORT-EHEALTH (V 1.6; Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and Online Telehealth) checklist (Multimedia Appendix 1). The study was prospectively registered in the Chinese Clinical Trial Registry (ChiCTR2500113139).

The sample size was determined based on the primary outcome—6-minute walk distance (6MWD). Previous studies have demonstrated that telerehabilitation can significantly improve 6MWD in patients following lung cancer surgery, and the minimal clinically important difference for 6MWD in this population has been reported to be 50 meters [16]. Accordingly, we hypothesized that the motion capture telerehabilitation intervention would yield a clinically meaningful improvement of 50 meters in 6MWD compared with the control group. Assuming a 2-sided α value of .05 and a power of 80%, the required sample size was calculated to be 23 participants per group. To account for an anticipated dropout rate of 20%, the sample size was inflated to 29 participants per group. Therefore, a total of 58 participants were required for this study.

The clinical study was conducted in accordance with the Declaration of Helsinki and China regulations on clinical research and has been approved by the medical ethics committee. All participants signed informed consent forms. Prior to the formal initiation of this study, it was approved by the Ethics Committee of Zhujiang Hospital, Southern Medical University (approval 2024-KY-327-01, Medical Research Registration and Filing Information System Registration MR-44-24-052709). The study has been registered in the Chinese Clinical Trial Registry (ChiCTR2500113139). The study was conducted by trained investigators, who were responsible for patient recruitment, data collection, and test administration. The data collected included patients’ basic information, clinical data, exercise tolerance, pulmonary function, quality of life scales, mobility, treatment adherence, and adverse event incidence. For patients capable of completing the paper-based questionnaire independently, the investigators provided standardized questionnaires and guided them through the completion process. For patients unable to complete the questionnaire independently due to physical conditions or other reasons, the investigators conducted oral inquiries and recorded responses accordingly. During the recording process, the investigators verified information with patients to ensure data accuracy and completeness, and completed the questionnaire after confirming no errors. Throughout the research process, the investigators strictly adhered to ethical guidelines, safeguarding patient privacy and data security. All collected data were processed anonymously and used solely for research purposes, with no disclosure to third parties without patient consent. Participants received complimentary health education resources and were granted access to the AI-driven telerehabilitation platform throughout the intervention phase, while no monetary incentives were offered to participants for taking part in the study.

The study selected patients with NSCLC who underwent thoracic surgery and were hospitalized at Zhujiang Hospital of Southern Medical University from June 2024 to January 2025 as the research participants. A total of 58 eligible patients were enrolled. The research team provided comprehensive and detailed explanations to all enrolled patients and their families, covering the purpose, specific content, and significance of the study.

Inclusion criteria included (1) histopathologic diagnosis of NSCLC per National Comprehensive Cancer Network Guidelines 2024.V4 [17], (2) age 18-75 years, (3) cognitive competence adequate for assessment and exercise adherence, and (4) completed radical resection (lobectomy, segmentectomy, or wedge) with stable postoperative vital signs. Exclusion criteria included (1) clinically significant impairment of cardiac, hepatic, or renal function precluding exercise; (2) uncontrolled postoperative complications such as pneumonia, atelectasis, pleural effusion, or respiratory failure; (3) concurrent malignancy likely to influence recovery; (4) enrollment in another interventional study within the preceding 90 days; (5) enrollment in another interventional study within the preceding 90 days; and (6) inability to use a smartphone or absence of reliable internet connectivity required for motion capture telerehabilitation.

Randomization was conducted using a computer-generated random number sequence in a 1:1 ratio. Simple randomization without blocking or stratification was applied. The allocation sequence was generated by an independent researcher (QZ) and concealed using sequentially numbered, opaque, and sealed envelopes to ensure allocation concealment. All researchers involved in patient recruitment or data collection will not be involved in the generation of the random sequence and the preparation and production of the random allocation concealment envelopes. The primary and all secondary outcome measures were collected offline through face-to-face interviews.

Table 1 presents the structured components of the postoperative home rehabilitation program implemented in the experimental group. The intervention was organized according to specific time points, including discharge day, 4 weeks after discharge, and the follow-up assessment. It outlines the intervention items, detailed content, implementation methods, and prescribed frequency and duration.

On the day of discharge, participants underwent an in-hospital discharge assessment to evaluate exercise tolerance, pulmonary function, activities of daily living, and quality of life. During the 4-week exercise after discharge, the intervention consisted of structured exercise training delivered through a digital platform, incorporating aerobic, strength, and flexibility exercises.

Remote supervision and follow-up management were conducted by a multidisciplinary team via online communication tools such as the WeChat app and telephone contact to enhance adherence and provide timely feedback. At 4 weeks after discharge, outcome assessments were performed in the outpatient setting using the same evaluation domains as baseline.

All analyses were performed in SPSS (version 26.0; IBM Corp). Normally distributed continuous data are presented as mean (SD); group differences and pre-to-post changes were examined with independent-samples and paired t tests, respectively. Non–normally distributed variables are summarized by medians; between-group contrasts were evaluated using the Mann-Whitney U test, and within-group shifts via Wilcoxon signed-rank tests. Categorical variables are described with counts and percentages, and baseline comparability was verified by chi-square analysis. The last available primary end point value was carried forward for withdrawals. Missing data mechanisms guided handling: missing completely at random observations were listwise-deleted, missing at random values were multiply imputed, and missing not at random scenarios were explored with sensitivity analyses using pattern mixture or selection models.

A total of 80 applicants were initially recruited and underwent preliminary screening. Following the eligibility assessment, 22 individuals were excluded, and the remaining participants were included in the trial. Fifty-eight patients were randomized (29 per group). During follow-up, 4 participants in each group withdrew from the study. In the intervention group, reasons for dropout included seeking treatment at another hospital, residing too far from the hospital, and discontinuation of exercise. In the control group, dropouts were due to relocation far from the hospital or discontinuation of exercise. Ultimately, 25 participants in each group completed the follow-up assessment. Consequently, 50 participants (25 experimental and 25 control) provided complete data for the final analysis (Figure 3).

A total of 50 patients with lung cancer were enrolled and randomly allocated to the experimental group (n=25) and the control group (n=25). The mean age was 54.56 (SD 12.12) years in the experimental group and 56.96 (SD 11.02) years in the control group, with no statistically significant difference between groups (P=.62). In terms of sex distribution, 40% (10/25) of the participants in the experimental group and 48% (12/25) in the control group were male, and the difference was not statistically significant (P=.39). No significant differences were observed between the 2 groups with respect to height, body weight, or BMI (all P>.05). Similarly, baseline characteristics including educational attainment, occupational status, and exercise habits were evenly distributed between groups, with no significant between-group differences (all P>.05), suggesting comparability at baseline (Table 2).

There were no significant differences between the 2 groups in smoking history, medical insurance type, history of chemoradiotherapy, pathological type, surgical method, length of hospital stay distribution, or postoperative hospital stay duration (all P≥.05). The majority of participants in both groups had stage I lung cancer, and all patients were pathologically confirmed at stage I, indicating homogeneity in disease severity. Adenocarcinoma was the predominant pathological type, accounting for 92% in the experimental group and 100% in the control group. Regarding surgical procedures, most patients underwent wedge resection or segmentectomy, with comparable distributions between groups. Postoperative hospital stay was similar between the 2 groups, with a mean duration of 6.08 (SD 0.60) days in the experimental group and 4.84 (SD 0.48) days in the control group (P=.10). Overall, baseline clinical and perioperative characteristics were balanced between groups (Table 3).

At baseline, there were no statistically significant differences between the experimental and control groups across all FACT-L subscales and total score (all P>.05), indicating good comparability before intervention. For PWB, the experimental group improved from 14.68 (SD 2.27) to 21.68 (SD 2.27; T=–4.55; P<.001), while the control group improved from 13.72 (SD 2.23) to 17.68 (SD 2.27; T=–4.52; P<.001). The between-group comparison revealed a statistically significant difference favoring the motion capture cohort (mean difference 4.00 points, 95% CI 2.66-5.34; T=6.24; P<.001). For SWB, the experimental group improved from 16.04 (SD 1.24) to 20.08 (SD 1.32; T=–4.33; P<.001), and the control group improved from 16.80 (SD 1.53) to 19.80 (SD 1.38; T=–4.40; P<.001). However, the between-group difference was not statistically significant (mean difference 0.28 points, 95% CI –0.76 to 1.32; T=–0.53; P=.60). For EWB, the experimental group improved from 13.48 (SD 1.16) to 15.80 (SD 1.19; T=–4.48; P<.001), and the control group improved from 13.72 (SD 1.49) to 15.92 (SD 1.38; T=–4.44; P<.001). The between-group difference did not reach statistical significance (mean difference –0.12 points, 95% CI –0.98 to 0.74; T=–0.27; P=.79). For FWB, the experimental group improved from 12.28 (SD 2.23) to 18.68 (SD 2.27; T=–4.55; P<.001), while the control group improved from 11.04 (SD 2.25) to 14.00 (SD 1.94; T=–4.44; P<.001). The between-group comparison revealed a statistically significant difference favoring the motion capture cohort (mean difference 4.68 points, 95% CI 3.44-5.92; T=7.85; P<.001). For the Additional Concerns–Lung (ACL) subscale, the experimental group improved from 9.72 (SD 1.93) to 17.24 (SD 2.11; T=–4.41; P<.001), and the control group improved from 9.64 (SD 1.91) to 12.96 (SD 1.77; T=–4.41; P<.001). The between-group difference was statistically significant (mean difference 4.28 points, 95% CI 3.14-5.42; T=7.78; P<.001). For the total FACT-L score, the experimental group improved from 66.20 (SD 6.83) to 93.36 (SD 6.72; T=–51.12; P<.001), while the control group improved from 64.92 (SD 6.85) to 80.36 (SD 6.04; T=–33.94; P<.001). The between-group comparison revealed a statistically significant difference favoring the experimental group (mean difference 13.00 points, 95% CI 9.30-16.70; T=7.19; P<.001). Complete data are provided in Table 7.

Home-based program adherence was markedly greater in the motion capture group (18/25, 72% vs 10/25, 40%; P=.02), and no exercise-related adverse events were recorded in either group (Table 8).

This study demonstrated that motion recognition–based telerehabilitation was associated with greater improvements in postoperative recovery among patients with NSCLC compared with conventional home-based video-guided training. Specifically, participants in the intervention group showed superior gains in exercise tolerance, pulmonary function, selected aspects of functional mobility, and the physical and functional domains of quality of life. Rehabilitation adherence was also higher in the motion-guided group, and no intervention-related adverse events were observed in either group.

Taken together, these findings suggest that integrating real-time kinematic feedback into home-based pulmonary rehabilitation may enhance movement quality, functional recovery, and patient engagement beyond standard remote instruction. Importantly, these results are consistent with emerging evidence from recent systematic reviews and network meta-analyses [27] indicating that telerehabilitation can achieve improvements in physical function and quality of life comparable with, or in some contexts exceeding, conventional face-to-face pulmonary rehabilitation programs. This study extends this body of evidence by highlighting the potential added value of motion recognition–based feedback within a postoperative population with lung cancer.

Exercise tolerance constitutes a primary indicator of postoperative functional recovery in patients undergoing NSCLC resection [28]. Baseline 6MWD did not differ between groups, confirming equivalent preintervention capacity. Although both cohorts achieved after-intervention gains—reflecting the efficacy of any structured training—the experimental group demonstrated a substantially larger increment, underscoring the supplementary value of motion capture–guided telerehabilitation. These observations align with prior evidence.

The superior improvement in the motion capture group can be attributed to several interconnected mechanisms. First, real-time kinematic feedback facilitates motor learning through augmented feedback loops. According to the guidance hypothesis (“Comparative Effects of Sensory Motor and Virtual Reality Interventions to Improve Gait, Balance, and Quality of Life MS Patients”), concurrent feedback enhances skill acquisition by providing immediate error information, allowing patients to correct aberrant movement patterns (eg, knee valgus and excessive trunk sway) before they become ingrained. This process engages the cerebellar-cortical circuit, promoting neuroplastic changes that optimize movement efficiency [29].

Second, real-time, AI-guided motion correction may augment walking economy by refining neuromuscular synchrony and dynamic stability. Earlier work documents that cancer survivors exhibit quadriceps weakness approximately one-third to two-fifths greater than healthy peers [30]. Instantaneous joint-angle feedback during tasks such as squats promptly redirects aberrant mechanics—knee valgus or excessive trunk sway—reducing metabolic cost per stride.

Third, patients in the experimental group covered more ground during the 6MWT without parallel rises in Borg ratings; superior gait efficiency and balance translated into greater distance with lower perceived exertion. Moreover, quantitative visual cues (eg, “posture accuracy: 92%”) reinforced engagement and intensity, yielding a compliance rate of 72% (18/25) versus 40% (10/25) in controls. Sustained participation drove cumulative cardiopulmonary and musculoskeletal adaptations—the principal determinants of improved exercise capacity [31].

The Borg scale gauges perceived exertion and complements objective measures of exercise tolerance [32]. In our trial, Borg ratings did not differ between the motion capture and control groups. Nevertheless, instantaneous kinematic feedback appeared to bolster self-efficacy and refine movement efficiency, allowing the experimental cohort to increase walking distance without proportionate fatigue. Conversely, controls—hampered by suboptimal mechanics—covered shorter distances and yet reported greater perceived exertion per meter.

Enhancing pulmonary function represents a key objective in postoperative care for patients with lung cancer [33]. The superior improvements in FVC, FEV₁, and PEF observed in the motion capture group likely stem from the system’s capacity to ensure optimal respiratory muscle synergy through real-time corrective feedback. While participants in both cohorts completed identical training protocols, only those in the intervention group received real-time monitoring of biomechanical parameters with prompt adjustments for aberrant movement patterns [34].

While participants in both cohorts completed identical exercise protocols, only those in the intervention group received continuous monitoring of trunk alignment during core exercises. This distinction is critical because respiratory function is intimately linked to core stability. The diaphragm, the primary muscle of inspiration, shares fascial connections with the transversus abdominis and pelvic floor muscles, forming the thoracoabdominal pump [35]. Proper activation of this integrated system requires precise coordination—a skill that benefits from augmented feedback.

Consider the plank—an evidence-based core stabilization maneuver engaging key musculature including the transversus abdominis, rectus abdominis, and erector spinae. Regular plank training (3 times per week) enhances both respiratory function and general fitness in aging populations [36]. Yet, its benefits are substantially attenuated by technical errors such as hyperlordosis [37], which specifically impairs transversus abdominis recruitment [38]. Such deviations not only undermine therapeutic effects but also elevate spinal loading and injury potential.

The motion recognition system provided instantaneous proprioceptive feedback that helped patients maintain neutral spine alignment, thereby optimizing the feedforward control mechanism of core muscles. This feedforward activation—where trunk muscles contract in anticipation of limb movement—is essential for maintaining spinal stability during dynamic activities [39]. By reinforcing this mechanism during training, the intervention likely enhanced the neural drive to respiratory muscles, translating into improved spirometric outcomes. Furthermore, the observed improvements in PEF suggest enhanced expiratory muscle function, possibly through better recruitment of abdominal muscles during forced expiration.

Functional mobility assessment serves as a critical component in evaluating rehabilitation success. Standardized tests such as the TUG [40], single-leg balance assessment [41], 30-second CST [42], FTSST [43], and 2-minute stepping task [44] collectively offer a comprehensive evaluation of movement capacity and functional performance. These validated measures enable clinicians to objectively determine patients’ functional capacity for activities of daily living and community participation, while effectively tracking recovery milestones.

Among these assessments, the TUG represents a composite functional task requiring coordinated execution of multiple motor components, including sit-to-stand transition, gait initiation, turning, and resitting. Performance on this test reflects not only lower limb strength but also dynamic balance, motor sequencing, and intertask coordination [45]. The observed improvement in TUG performance in the motion recognition group may be related to enhanced movement precision and temporal coordination facilitated by real-time kinematic feedback. The motion recognition system provides continuous monitoring of joint angles and movement trajectories, potentially supporting more accurate sensorimotor integration during exercise execution—the process by which sensory feedback refines motor output [46]. Similar improvements in postural control and motor stability have been reported in neurological populations receiving telemedicine-based rehabilitation, including individuals with Parkinson disease, suggesting that remote feedback mechanisms may enhance movement regulation across different clinical conditions [47]. Real-time visual feedback of joint angles and movement trajectories may have facilitated implicit motor learning, whereby patients internalize correct movement patterns without extensive conscious processing. This contrasts with explicit learning approaches that rely primarily on verbal instruction and are more cognitively demanding and potentially less durable under stress [48]. The improved TUG performance in the experimental group likely reflects enhanced procedural memory for the sequence of movements involved in the test.

The lack of significant between-group differences in single-leg stance, 30-second CST, FTSST, and 2-minute step tests does not contradict this interpretation. These tests predominantly assess isolated components of physical function—static balance, lower extremity strength, and repetitive endurance—rather than the integrated motor control required for complex tasks. Both groups received identical exercise prescriptions targeting lower limb musculature, likely producing comparable gains in gross strength (as reflected in the similar improvements on these tests). The unique value of the motion-aware system lies in optimizing movement quality and interjoint coordination—domains captured by composite measures such as the TUG but not by simpler functional tests. This dissociation between strength gains and coordination improvements is consistent with the principle that motor skill acquisition and muscular adaptation are mediated by distinct neural and physiological mechanisms [49].

Quality of life represents a critical multidimensional end point in postsurgical NSCLC rehabilitation, encompassing physical, psychological, and social well-being as primary markers of recovery success [50]. Our intervention group showed statistically superior improvements in PWB, FWB, ACL, and the total FACT-L score relative to controls. This pattern implies that motion-aware telerehabilitation offers particular advantages for restoring physical and functional aspects of quality of life, potentially attributable to its capacity for optimizing functional recovery through millimeter-precise movement analysis and instantaneous form correction. Additionally, the system’s ability to provide symptom-specific movement modifications and breathing technique guidance may explain this therapeutic advantage, empowering patients to better control their condition-related challenges.

The analysis revealed no statistically significant intergroup differences in either SWB or EWB domain. This observation may be explained by several key considerations. First, meaningful enhancement of social support systems and psychological state typically necessitates prolonged intervention durations and often benefits from dedicated mental health support, posing challenges for demonstrating immediate effects within a limited study period [51]. Second, the psychological status of postsurgical patients with NSCLC is frequently influenced by complex psychosocial factors—particularly disease recurrence anxiety and treatment-related financial stress [52]—that extend beyond the scope of physical rehabilitation and may require targeted psychotherapeutic approaches combined with robust social services.

Program adherence represents a critical determinant of rehabilitation efficacy [53]. Our data revealed substantially better compliance rates in the intervention group than in controls, indicating that motion-aware telerehabilitation systems may optimize treatment engagement for postsurgical patients with NSCLC. This effect appears mediated by the platform’s multimodal feedback mechanisms—including real-time performance metrics, achievement scoring, and graphical progress tracking—which collectively enhance treatment motivation through positive reinforcement. Contemporary research confirms that instantaneous performance feedback bolsters rehabilitation adherence by reinforcing patient self-confidence and perceived competence [54]. Furthermore, the system’s interactive design elements convert conventional exercises into dynamic, goal-oriented tasks, offering superior psychological engagement compared with passive video demonstrations.

Regarding safety outcomes, neither study group reported intervention-related adverse events, confirming the acceptable risk profile of home-based motion capture rehabilitation. The technology’s ability to continuously monitor and correct biomechanical execution helps ensure proper movement patterns, consequently minimizing injury risks associated with faulty technique [55].

Despite the promising findings, several constraints warrant consideration. First, assessments were restricted to the 4-week end point, precluding insight into durability; subsequent studies should extend follow-up to 3, 6, and 12 months to chart longer-term recovery. Second, the modest sample size may limit both external validity and statistical power. Although significant between-group differences were observed for several primary outcomes, the study may have been underpowered to detect smaller yet clinically meaningful effects in some secondary measures. Accordingly, the absence of statistically significant differences for certain functional indicators should be interpreted cautiously, as the possibility of type II error cannot be excluded. Larger, adequately powered multicenter trials are required to validate and refine these findings. Third, the lack of preoperative functional baselines hampers delineation of the full perioperative trajectory; future protocols should incorporate presurgical assessments to construct comprehensive longitudinal recovery profiles.

To our knowledge, this is the first randomized controlled trial to integrate motion capture analytics into telerehabilitation following NSCLC surgery. Over 4 weeks, the approach proved both safe and superior to standard video-guided exercise, yielding larger gains in exercise tolerance, spirometry, functional mobility, and quality of life. These data establish a clinically viable strategy that can readily be scaled across telemedicine and postoperative care pathways.