跳到正文
原文
Nature 健康护理主题·· 8 天前AI 评分31

基于视频的ReMAP运动质量指标用于乳腺癌幸存者康复监测:临床可解释性、响应性与可靠性评估

Clinically interpretable video-derived movement quality metrics for rehabilitation monitoring in breast cancer survivors

AI 导读

一项来自韩国七家大学医院康复中心的探索性研究分析了ReMAP(Rehabilitation Movement Assessment Protocol)框架的视频驱动运动质量指标,用于评估乳腺癌治疗后的功能恢复。

正文

Introduction

Breast cancer survivors frequently experience persistent impairments in physical function following cancer treatment, including deficits in balance, mobility, coordination, and overall movement confidence1,2. Surgical treatment, chemotherapy, radiation therapy, and endocrine therapy may contribute to neuromuscular dysfunction, sensorimotor impairment, and reduced physical capacity, resulting in limitations in daily activities and increased risk of falls. Consequently, accurate assessment of functional recovery has become a central component of cancer rehabilitation and survivorship care.

Conventional rehabilitation assessment relies primarily on patient-reported outcomes and performance-based measures. Patient-reported instruments provide important information regarding disability, symptom burden, and perceived physical function, whereas performance-based measures such as the Timed Up and Go (TUG) test, 6-Minute Walk Test (6MWT), grip strength, and balance tests quantify functional capacity under standardized conditions. These measures are widely used in clinical practice and remain essential for rehabilitation evaluation. However, they primarily describe what patients can do rather than how movement is performed during task execution. This distinction may be clinically important. Individuals with similar performance outcomes may use substantially different movement strategies, levels of postural control, coordinative patterns, or compensatory behaviors. Likewise, improvements in movement organization may occur before meaningful changes are observed in disability, walking capacity, strength, or balance outcomes. As a result, clinically relevant aspects of recovery may remain undetected when assessment is limited to conventional rehabilitation outcomes alone.

Recent advances in computer vision and markerless motion analysis have created new opportunities for video-based rehabilitation assessment3. Standard RGB video recordings can now be transformed into quantitative movement descriptors without specialized motion-capture laboratories, wearable sensors, or marker-based biomechanical systems4. Such approaches offer several advantages for rehabilitation applications, including low cost, scalability, remote assessment capability, and feasibility for repeated longitudinal monitoring5. Emerging markerless assessment platforms have demonstrated that clinically relevant movement information can be extracted from routine video recordings and incorporated into rehabilitation assessment and monitoring processes6.

For clinical use, video-derived movement metrics should demonstrate acceptable measurement properties, including reliability, construct validity, responsiveness, and clinical interpretability7,8. In addition, quantitative movement measures should remain linked to recognizable aspects of motor behavior that are meaningful to rehabilitation professionals and patients.

Movement quality represents a potentially important dimension of rehabilitation assessment. Characteristics such as postural stability9, movement consistency, adaptive coordination10, and movement efficiency11,12 may provide information regarding how movement is generated, organized, and adjusted during task performance13. Although movement-quality assessment is increasingly recognized as a potentially valuable component of rehabilitation evaluation, its clinical meaning remains insufficiently understood. In particular, it remains unclear whether video-derived movement-quality metrics simply reflect existing rehabilitation outcomes such as disability, mobility, strength, endurance, and fatigue, or whether they provide complementary information regarding recovery processes that are not fully represented by conventional measures.

To address this gap, we evaluated the Rehabilitation Movement Assessment Protocol (ReMAP), a standardized markerless video-based framework that quantifies postural stability, motor control, coordination, and movement efficiency from standardized movement-task videos. The aim of this study was to examine the clinical interpretability, responsiveness, and reliability of ReMAP-derived movement-quality metrics in breast cancer survivors undergoing rehabilitation. We hypothesized that selected movement-quality metrics would show group-level responsiveness to rehabilitation-related change and examined the extent to which they were associated with conventional rehabilitation outcomes.

Methods

Study design

This study was a secondary measurement-property analysis of data collected within a prospective, multicenter randomized wait-list controlled trial in breast cancer survivors. The primary objective was to evaluate the reliability, measurement error, construct validity, and group-level responsiveness of ReMAP-derived video movement metrics. The present analysis does not establish individual-level longitudinal monitoring capability or clinically meaningful change thresholds.

Study size

Because this study was designed as an exploratory evaluation of the clinical interpretability and measurement properties of ReMAP-derived movement-quality metrics, all eligible participants recruited during the study period were included. No formal a priori sample-size calculation was performed. This approach is consistent with exploratory studies evaluating the measurement properties of novel rehabilitation outcomes.

Participants

Women aged 18–65 years with a history of breast cancer were recruited from rehabilitation centers affiliated with seven university hospitals in Korea between January 2023 and June 2024. Participants were eligible if they had completed primary cancer treatment at least three months before enrollment, were able to ambulate independently, and could safely perform standardized movement tasks. Individuals with acute musculoskeletal injury, non-cancer-related neurological disorders, severe cardiopulmonary limitations, or other conditions that could interfere with task performance were excluded. This study adhered to the Declaration of Helsinki and was approved by the Institutional Review Board of all participating institutions, including Daerim St. Mary’s Hospital (IRB No. DMC-22-047), Pusan National University Hospital (2305-025-126), CHA University Bundang Medical Center (CHAMC 2023-04-062-002), National Health Insurance Service Ilsan Hospital (NHIMC 2023-03-056-001), Jeju National University Hospital (JEJUNUH 2023-03-006), Chungnam National University Hospital (CNUH 2022-12-075-001), and Jeonbuk National University Hospital (CUH 2023-04-003-002). All participants provided written informed consent after receiving detailed information on study procedures, risks, and benefits.

Video-based movement assessment

Movement quality was evaluated using the Rehabilitation Movement Assessment Protocol (ReMAP), a standardized video-based assessment consisting of six movement subtasks organized into three paired task sequences. Tasks were designed to challenge complementary dimensions of movement behavior, including postural stability, motor control, coordination, and movement efficiency. Representative movement tasks included in the ReMAP assessment are described in Supplementary Table S1, and the overall processing workflow is illustrated in Supplementary Figure S1. An overview of the video-based movement assessment procedure is shown in Supplementary Figure S1.

All assessments were recorded using standardized RGB video cameras positioned to capture anterior-posterior and lateral views. Videos were recorded at 30 frames per second and processed using MediaPipe Pose (Google LLC, Mountain View, CA, USA), which generated 33 anatomical landmarks for each video frame13. For the Generation 1 analysis used in this study, a two-dimensional body-center trajectory was calculated as the arithmetic mean of bilateral shoulder and hip landmarks (MediaPipe landmarks 11, 12, 23, and 24) after conversion of normalized image coordinates to pixel coordinates; the z-coordinate was not used. Frames without a detected pose were excluded. Trajectories were smoothed using a 9-frame, second-order Savitzky-Golay filter. Metric-specific computational definitions and view/task aggregation procedures are provided in Supplementary Table S2. Standardized task instructions, camera positioning, and processing procedures were used across participating centers. For the four manuscript metrics, trajectory variability was summarized across both AP and lateral views, whereas energy ratio was derived from the lateral view and the sway metric from the lateral-view horizontal trajectory. For lateral correction rate, the descriptive and within-session task-segment consistency analyses used the lateral-view metric, whereas the responsiveness analysis used the AP/lateral combined metric from the multi-view pipeline.

Video-derived movement quality metrics

Four predefined movement-quality metrics were evaluated to represent postural stability9, motor control, coordination10,14 and movement efficiency11,12. Detailed computational definitions and extraction procedures are provided in Supplementary Table S2.

Reference outcome measures

To evaluate construct validity, relationships between ReMAP-derived movement metrics and conventional rehabilitation outcomes were examined. Functional mobility was assessed using the Timed Up and Go (TUG) test. Walking capacity was assessed using the 6-Minute Walk Test (6MWT). Disability was evaluated using the World Health Organization Disability Assessment Schedule (WHODAS 2.0). Patient-reported physical function and fatigue were assessed using the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ-C30). Handgrip strength was assessed using a digital dynamometer (T.K.K. 5401; Takei Scientific Instruments Co., Ltd., Niigata, Japan). Two maximal grip trials were performed with each hand, and the highest value was used for analysis. One-leg standing balance was assessed with the eyes open. Two trials were performed on each side, with a maximum duration of 30 s per trial, and the best performance was used for analysis.

These outcomes were selected because they represent commonly used rehabilitation endpoints reflecting mobility, physical capacity, disability, and symptom burden. Associations with ReMAP-derived metrics were interpreted as construct-validity evidence. Weak associations were not considered evidence of construct independence, nonredundancy, or complementary clinical utility.

Responsiveness

Group-level responsiveness was evaluated by comparing participants assigned to the 8-week ReMAP rehabilitation program with wait-list controls. The intervention was delivered twice weekly for approximately 40–50 min per supervised session, with progression from static alignment to bilateral/reciprocal coordination and dynamic multi-joint balance tasks at an approximate OMNI intensity of 5–6/10. Controls maintained usual activity and received advice regarding walking, range-of-motion exercise, and light stretching. Baseline-adjusted between-group differences were estimated using ANCOVA, and standardized responsiveness magnitude was summarized using Cohen’s d. These analyses were interpreted as group-level responsiveness and not as evidence of individual clinically meaningful change.

Reliability

Within-session task-segment consistency was evaluated using a two-way mixed-effects, single-measure, consistency intraclass correlation coefficient [ICC(3,1)] with 95% confidence intervals. The analysis characterizes consistency of derived metric values across sequential ReMAP task segments within the same assessment, rather than inter-rater reliability or longitudinal test-retest reproducibility. Measurement error was summarized using the standard error of measurement (SEM) and minimal detectable change at the 95% confidence level (MDC95), calculated as SEM = SD × sqrt(1 - ICC) and MDC95 = 1.96 x sqrt(2) x SEM. Because these estimates were derived from within-session task-segment consistency, they were not interpreted as longitudinal thresholds for individual clinical change.

Construct validity

Construct validity was examined using Spearman correlation coefficients between ReMAP-derived movement metrics and conventional rehabilitation outcomes. Baseline and benefit-aligned change-score associations were treated as exploratory construct-validity analyses. Weak correlations were interpreted as limited statistical overlap in this sample and not as proof of independent or complementary constructs.

Statistical analysis

Continuous variables are presented as mean ± standard deviation and categorical variables as frequencies and percentages. Baseline balance in the analytic cohort was characterized descriptively using absolute standardized differences calculated from available observations; no imputation was performed for missing baseline demographic values. Construct validity was examined using Spearman correlation coefficients between ReMAP-derived movement metrics and conventional rehabilitation outcomes, including TUG, 6MWT, WHODAS 2.0, EORTC physical function and fatigue, handgrip strength, and one-leg standing. Within-session task-segment consistency was evaluated using ICC(3,1), with SEM and MDC95 calculated as described above. Group-level responsiveness was evaluated using analysis of covariance (ANCOVA) models with the 8-week outcome as the dependent variable, study group as the exposure, and the corresponding baseline value as a covariate. Cohen’s d was calculated from between-group differences in change scores standardized by the pooled standard deviation of the change scores; absolute values are presented as responsiveness magnitudes, while direction is conveyed by the adjusted between-group difference. Analyses used available-case data; no imputation was performed. Outcome-specific baseline, 8-week, and paired availability is summarized in Supplementary Table S4. These secondary analyses were exploratory and no multiplicity adjustment was applied. Statistical significance was defined as a two-sided p-value < 0.05. All analyses were performed using Stata version 19 (StataCorp LLC, College Station, TX, USA).

Results

Participant characteristics

A total of 71 breast cancer survivors were included in the analysis, comprising 30 participants in the control group and 41 participants in the rehabilitation group. Participant flow is shown in Supplementary Figure S2. Baseline demographic, clinical, and functional characteristics are summarized in Table 1. Education data were available for 27 control and 38 intervention participants. The largest absolute standardized differences in the analytic cohort were observed for age (0.50), WHODAS 2.0 (0.45), and receipt of radiation therapy (0.40); baseline characteristics are therefore presented descriptively rather than interpreted as demonstrating complete balance. Overall, the cohort represented ambulatory breast cancer survivors with relatively preserved baseline functional status.

Table 1 Baseline characteristics of study participants.

Full size table

Responsiveness of ReMAP-derived movement quality metrics

Responsiveness results are presented in Table 2; Fig. 1. Among conventional rehabilitation outcomes, TUG demonstrated the largest responsiveness to rehabilitation-related change (Cohen’s d = 0.86, p < 0.001). Among ReMAP-derived movement-quality metrics, trajectory variability demonstrated the strongest responsiveness (adjusted difference = -20.92, 95% CI -30.34 to -11.50, p < 0.001; d = 0.64), followed by mediolateral sway (d = 0.63) and energy ratio (adjusted difference = -12.80, 95% CI -21.27 to -4.33, p = 0.003; d = 0.50). Lateral correction rate demonstrated a smaller responsiveness magnitude (d = 0.38) and did not reach statistical significance. In contrast, several conventional rehabilitation outcomes demonstrated weaker responsiveness. WHODAS disability showed a small effect size (d = 0.29), grip strength demonstrated limited responsiveness (d = 0.24), and the 6MWT showed minimal responsiveness (d = 0.03).

Table 2 Responsiveness of ReMAP-derived movement-quality metrics and conventional rehabilitation outcomes.

Full size table

Fig. 1

Comparative responsiveness of ReMAP-derived movement-quality metrics and conventional rehabilitation outcomes. Bars represent benefit-aligned Cohen’s d effect sizes, with larger values indicating greater responsiveness to rehabilitation-related change. Solid black bars indicate ReMAP-derived movement-quality metrics, whereas white bars indicate conventional rehabilitation outcomes.

Reliability of ReMAP-derived movement-quality metrics

Within-session task-segment consistency results are presented in Table 3. Consistency varied substantially across metrics. Lateral correction rate showed the highest ICC (0.94, 95% CI 0.92–0.96), trajectory variability showed an ICC of 0.61 (95% CI 0.52–0.70), energy ratio 0.47 (95% CI 0.35–0.59), and mediolateral sway 0.17 (95% CI 0.04–0.30). These estimates characterize consistency across sequential task segments within the same assessment and should not be interpreted as longitudinal test–retest reproducibility.

Table 3 Within-session task-segment consistency and associated measurement error of ReMAP-derived movement-quality metrics.

Full size table

Construct validity and associations with conventional rehabilitation outcomes

Baseline associations between ReMAP-derived movement metrics and conventional rehabilitation outcomes are presented in Fig. 2; Table 4. Associations were generally weak. Lateral correction rate showed the strongest relationship with TUG performance (ρ = 0.353, p = 0.003), whereas mediolateral sway, trajectory variability, and energy ratio showed negligible associations. These findings indicate limited cross-domain overlap in this sample; they do not establish construct independence, nonredundancy, or complementary clinical utility.

Fig. 2

Baseline associations between ReMAP-derived movement-quality metrics and conventional rehabilitation outcomes. Heatmap of Spearman correlation coefficients (ρ) among patient-reported outcomes, objective performance measures, and ReMAP-derived movement-quality metrics at baseline assessment. Darker shading indicates stronger absolute correlations. ReMAP metrics demonstrated generally weak associations with conventional rehabilitation outcomes, whereas stronger correlations were observed among metrics within the same measurement domain.

Table 4 Baseline associations between ReMAP-derived movement-quality metrics and conventional rehabilitation outcomes.

Full size table

Similarly, relationships between movement-quality metrics and patient-reported outcomes, including WHODAS disability, EORTC physical functioning, fatigue, and global health status, were generally small. Collectively, these findings suggest that movement-quality metrics are related only modestly to conventional rehabilitation outcomes and may capture aspects of function not fully represented by disability, mobility, endurance, or symptom measures.

To further examine cross-domain associations, benefit-aligned change-score correlations were evaluated (Supplementary Table S3). Correlations between changes in ReMAP-derived metrics and changes in conventional outcomes were generally weak. These exploratory findings were interpreted as limited statistical overlap and not as evidence that changes occurred independently or that the measures are clinically complementary.

Exploratory associations with mobility performance

Exploratory regression analyses examining associations between movement-quality metrics and TUG performance are presented in Supplementary Figure S3. At baseline, coordination demonstrated a significant association with TUG performance (β = 0.69, p = 0.034), whereas other movement-quality metrics were not significantly associated. Longitudinally, improvement in energy ratio was associated with improvement in TUG performance (β = 0.39, p = 0.043), whereas changes in the remaining movement-quality metrics were not significantly associated with changes in TUG performance.

Discussion

This exploratory secondary analysis evaluated reliability, measurement error, construct validity, and group-level responsiveness of video-derived movement metrics in breast cancer survivors. Measurement properties varied substantially across metrics. Lateral correction rate showed high within-session task-segment consistency, whereas mediolateral sway and energy ratio showed lower consistency. Trajectory variability and energy ratio demonstrated between-group differences in change over 8 weeks. Associations with conventional rehabilitation outcomes were generally weak, indicating limited cross-domain overlap in this sample but not establishing construct independence or complementary clinical utility.

The generally weak correlations observed between the video-derived metrics and TUG indicate limited statistical overlap in this sample. However, weak correlations alone do not establish construct independence or complementary value and may also reflect measurement error, restricted functional range, ceiling effects, sample heterogeneity, or incomplete representation of the intended constructs.

Among the evaluated descriptors, trajectory variability and energy ratio showed the clearest between-group differences i change. These findings support further investigation of their group-level responsiveness, but they should not be interpreted as evidence of specific sensorimotor mechanisms15,16,17 or as validated thresholds for individual patient monitoring.

This observation is consistent with measurement-theory principles demonstrating that reliability and responsiveness represent distinct psychometric properties and should be interpreted independently when evaluating rehabilitation outcome measures. Together, these findings emphasize the importance of evaluating multiple measurement properties when considering candidate movement-quality outcomes and support the use of trajectory variability and movement efficiency as potentially sensitive indicators of rehabilitation-related adaptation18,19.

Construct-validity analyses demonstrated generally weak associations between the video-derived metrics and conventional rehabilitation outcomes. These findings are compatible with limited overlap between task-completion performance and selected descriptors of movement execution20,21,22 but they do not establish that the measures are independent or clinically complementary.

The present findings support continued evaluation of video-derived movement metrics within rehabilitation measurement frameworks. The current results provide preliminary evidence regarding heterogeneous measurement properties in this cancer rehabilitation cohort, while substantial uncertainty remains regarding longitudinal test–retest reliability, individual-level interpretation, external validity, and clinical utility.

From a clinical perspective, the present metrics should be regarded as candidate video-derived movement measures rather than established clinical monitoring tools. Their potential value for repeated or remote assessment23,24 requires validation using independently repeated acquisitions, external clinical anchors, and reference biomechanical measurements.

Several limitations should be acknowledged. First, this was an exploratory secondary measurement-property analysis with no separate a priori precision-based sample-size calculation. Second, the cohort consisted of independently ambulatory, relatively high-functioning Korean women treated for breast cancer; restricted functional range and ceiling effects may have attenuated cross-domain correlations and limit generalizability. Third, the reliability analysis reflected within-session task-segment consistency rather than longitudinal test–retest reproducibility, so the corresponding SEM and MDC95 values should not be interpreted as validated thresholds for individual longitudinal change. Fourth, the present design cannot partition technical measurement variability from task-dependent biological variability. Fifth, construct-validity analyses were exploratory and weak correlations do not establish independence or complementary clinical utility. Recruitment was distributed across seven centers, several of which contributed only 2–3 participants; therefore, stable center-specific random-effects or cluster-robust sensitivity models were not considered supportable, and residual center effects cannot be excluded. Finally, external validation in more diverse populations and against reference biomechanical measures and clinical anchors is required.

Future research should focus on external validation across diverse rehabilitation populations, evaluation of home-based and community-based assessment settings, and investigation of the ability of movement-quality metrics to predict functional recovery trajectories. Additional studies should also examine relationships with patient-reported outcomes, participation measures, and other clinically relevant indicators of rehabilitation success.

In this exploratory analysis, the measurement properties of the video-derived movement metrics varied substantially across domains. Lateral correction rate showed high within-session task-segment consistency, whereas mediolateral sway and energy ratio showed lower consistency. Trajectory variability and energy ratio showed between-group differences in change over the 8-week intervention period. Because the ICC, SEM, and MDC95 estimates were derived from within-session task-segment consistency rather than longitudinal test–retest assessments, they should not be interpreted as thresholds for individual longitudinal change. Associations with conventional rehabilitation outcomes were generally weak, indicating limited cross-domain overlap in this sample but not establishing construct independence or complementary clinical utility. These findings warrant further evaluation of video-derived movement metrics as exploratory group-level measures. Longitudinal test–retest studies, external validation in more diverse populations and settings, and clinically anchored thresholds are needed before these metrics can be considered for individual-level clinical application.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to participant privacy and confidentiality considerations but are available from the corresponding author on reasonable request, subject to applicable institutional and ethical requirements.

References

  1. Campbell, K. L. et al. Exercise Guidelines for Cancer Survivors: Consensus Statement from International Multidisciplinary Roundtable. Med. Sci. Sports Exerc. 51, 2375–2390. https://doi.org/10.1249/MSS.0000000000002116 (2019).

    Article  PubMed  PubMed Central  Google Scholar 

  2. Stout, N. L., Mina, S., Lyons, D., Robb, K. D., Silver, J. & K. & K. A systematic review of rehabilitation and exercise recommendations in oncology guidelines. CA Cancer J. Clin. 71, 149–175. https://doi.org/10.3322/caac.21639 (2021).

    Article  PubMed  Google Scholar 

  3. Uhlrich, S. D. et al. Human movement dynamics from smartphone videos. PLoS Comput. Biol. 19, e1011462. https://doi.org/10.1371/journal.pcbi.1011462 (2023). OpenCap.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Kanko, R. M., Laende, E. K., Davis, E. M., Selbie, W. S. & Deluzio, K. J. Concurrent assessment of gait kinematics using marker-based and markerless motion capture. J. Biomech. 127, 110665. https://doi.org/10.1016/j.jbiomech.2021.110665 (2021).

    Article  PubMed  Google Scholar 

  5. Needham, L. et al. The accuracy of several pose estimation methods for 3D joint centre localisation. Sci. Rep.11, 20673. https://doi.org/10.1038/s41598-021-00212-x (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  6. Lam, W. W. T., Tang, Y. M. & Fong, K. N. K. A systematic review of the applications of markerless motion capture (MMC) technology for clinical measurement in rehabilitation. J. Neuroeng. Rehabil. 20, 57. https://doi.org/10.1186/s12984-023-01186-9 (2023).

    Article  PubMed  PubMed Central  Google Scholar 

  7. Natraj, S. et al. 3D pose estimation for scalable remote gait kinematics assessment. NPJ Digit. Med. 9https://doi.org/10.1038/s41746-025-02211-y (2025).

  8. Lin, J. et al. Clinical reliability and validity of a video-based markerless gait evaluation method. Front. Pediatr. 11, 1331176. https://doi.org/10.3389/fped.2023.1331176 (2023).

    Article  PubMed  PubMed Central  Google Scholar 

  9. Mancini, M. et al. Digital gait biomarkers in Parkinson’s disease: susceptibility/risk, progression, response to exercise, and prognosis. NPJ Parkinsons Dis. 11https://doi.org/10.1038/s41531-025-00897-1 (2025).

  10. Roemmich, R. T. & Bastian, A. J. Closing the Loop: From Motor Neuroscience to Neurorehabilitation. Annu. Rev. Neurosci. 41, 415–429. https://doi.org/10.1146/annurev-neuro-080317-062245 (2018).

    Article  CAS  PubMed  Google Scholar 

  11. Mico-Amigo, M. E. et al. Assessing real-world gait with digital technology? Validation, insights and recommendations from the Mobilise-D consortium. J. Neuroeng. Rehabil. 20, 78. https://doi.org/10.1186/s12984-023-01198-5 (2023).

    Article  PubMed  PubMed Central  Google Scholar 

  12. Brown, G. L., Seethapathi, N. & Srinivasan, M. A unified energy-optimality criterion predicts human navigation paths and speeds. Proc. Natl. Acad. Sci. U S A. 118https://doi.org/10.1073/pnas.2020327118 (2021).

  13. Edriss, S. et al. Pose estimation for pickleball players’ kinematic analysis through MediaPipe-based deep learning: A pilot study. J. Sports Sci. 43, 1860–1870. https://doi.org/10.1080/02640414.2025.2524283 (2025).

    Article  PubMed  Google Scholar 

  14. Latash, M. L. Understanding and Synergy: A Single Concept at Different Levels of Analysis? Front. Syst. Neurosci. 15, 735406. https://doi.org/10.3389/fnsys.2021.735406 (2021).

    Article  PubMed  PubMed Central  Google Scholar 

  15. Harbourne, R. T. & Stergiou, N. Movement variability and the use of nonlinear tools: principles to guide physical therapist practice. Phys. Ther.89, 267–282. https://doi.org/10.2522/ptj.20080130 (2009).

    Article  PubMed  PubMed Central  Google Scholar 

  16. Hausdorff, J. M. Gait dynamics, fractals and falls: finding meaning in the stride-to-stride fluctuations of human walking. Hum. Mov. Sci. 26, 555–589. https://doi.org/10.1016/j.humov.2007.05.003 (2007).

    Article  PubMed  PubMed Central  Google Scholar 

  17. Levin, M. F., Kleim, J. A. & Wolf, S. L. What do motor recovery and compensation mean in patients following stroke? Neurorehabil Neural Repair. 23, 313–319. https://doi.org/10.1177/1545968308328727 (2009).

    Article  CAS  PubMed  Google Scholar 

  18. Mokkink, L. B. et al. The COSMIN study reached international consensus on taxonomy, terminology, and definitions of measurement properties for health-related patient-reported outcomes. J. Clin. Epidemiol. 63, 737–745. https://doi.org/10.1016/j.jclinepi.2010.02.006 (2010).

    Article  PubMed  Google Scholar 

  19. Husted, J. A., Cook, R. J., Farewell, V. T. & Gladman, D. D. Methods for assessing responsiveness: a critical review and recommendations. J. Clin. Epidemiol. 53, 459–468. https://doi.org/10.1016/s0895-4356(99)00206-1 (2000).

    Article  CAS  PubMed  Google Scholar 

  20. Stergiou, N., Harbourne, R. & Cavanaugh, J. Optimal movement variability: a new theoretical perspective for neurologic physical therapy. J. Neurol. Phys. Ther. 30, 120–129. https://doi.org/10.1097/01.npt.0000281949.48193.d9 (2006).

    Article  PubMed  Google Scholar 

  21. Latash, M. L. The bliss (not the problem) of motor abundance (not redundancy). Exp. Brain Res. 217, 1–5. https://doi.org/10.1007/s00221-012-3000-4 (2012).

    Article  PubMed  PubMed Central  Google Scholar 

  22. Dingwell, J. B. & Cusumano, J. P. Re-interpreting detrended fluctuation analyses of stride-to-stride variability in human walking. Gait Posture. 32, 348–353. https://doi.org/10.1016/j.gaitpost.2010.06.004 (2010).

    Article  PubMed  PubMed Central  Google Scholar 

  23. Cornman, H. L., Stenum, J. & Roemmich, R. T. Video-based quantification of human movement frequency using pose estimation: A pilot study. PLoS One. 16, e0261450. https://doi.org/10.1371/journal.pone.0261450 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. Boswell, M. A. et al. Smartphone videos of the sit-to-stand test predict osteoarthritis and health outcomes in a nationwide study. NPJ Digit. Med. 6, 32. https://doi.org/10.1038/s41746-023-00775-1 (2023).

    Article  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

The authors thank all participants for their cooperation in this study.

Funding

This work was supported by the National R&D Program for Cancer Control, Ministry of Health and Welfare, Republic of Korea [HA21C0216].

Author information

Authors and Affiliations

  1. Department of Physical Medicine and Rehabilitation, National Cancer Center, Goyang, Republic of Korea

    Seung Hyun Chung

  2. Department of Rehabilitation Medicine, Seoul National University Bundang Hospital, Seongnam, Republic of Korea

    Eun Joo Yang

Authors

  1. Seung Hyun Chung
  2. Eun Joo Yang

Contributions

E.J.Y. conceptualized and designed the study, collected and analyzed the data, and drafted the manuscript. S.H.C. contributed to study design and data interpretation. E.J.Y.and S.H.C. assisted with data collection and critically reviewed the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Eun Joo Yang.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Chung, S.H., Yang, E.J. Clinically interpretable video-derived movement quality metrics for rehabilitation monitoring in breast cancer survivors. Sci Rep 16, 37086 (2026). https://doi.org/10.1038/s41598-026-76353-8

Download citation

  • Received: 20 June 2026

  • Accepted: 13 October 2026

  • Published: 12 November 2026

  • Version of record: 12 November 2026

  • DOI: https://doi.org/10.1038/s41598-026-76353-8

Keywords

来源:Nature 健康护理主题 · nature.com