Video-Based Finger Kinematics for Degenerative Cervical Myelopathy: A Smartphone-Based Computer Vision Approach for Screening

Video-Based Finger Kinematics for Degenerative Cervical Myelopathy: A Smartphone-Based Computer Vision Approach for Screening | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Video-Based Finger Kinematics for Degenerative Cervical Myelopathy: A Smartphone-Based Computer Vision Approach for Screening Viprav B. Raju, Ramesh M. Arnest, Huy Truong, Anjishnu Banerjee, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7454858/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 02 Feb, 2026 Read the published version in European Spine Journal → Version 1 posted 7 You are reading this latest preprint version Abstract Purpose: Degenerative cervical myelopathy (DCM) is a progressive spinal cord disorder that is frequently underdiagnosed, with diagnostic delays averaging 1 to 4 years. A key limitation in current clinical practice is the lack of objective and accessible screening tools. The 10-second grip-and-release test is commonly used to assess hand dysfunction in DCM, but its diagnostic performance is limited, particularly in older individuals with comorbid hand conditions such as osteoarthritis or peripheral neuropathy. To address this limitation, we developed and evaluated a smartphone-based computer vision tool that quantifies finger kinematics during the grip-and-release test. Our primary objective was to determine whether video-derived finger kinematics can provide superior diagnostic performance compared to grip count alone. A secondary objective was to assess how these video features correlate with cervical spinal cord compression on Magnetic Resonance Imaging (MRI). Methods: We collected smartphone videos of 58 participants with DCM and 65 age-matched controls (including healthy individuals and those with non-DCM hand dysfunction) performing the 10-second grip-and-release test. Finger landmarks were extracted using MediaPipe, and 250 kinematic features per finger were computed and combined across both hands. Feature selection was performed using ANOVA (p 0.01). A CatBoost classifier was trained on selected features using an 80/20 train-test split and five-fold cross-validation. A logistic regression model was trained using grip count alone. Model performance was evaluated using AUC, F1-score, sensitivity, and specificity. For the secondary analysis, we used linear regression models to evaluate associations between video-derived kinematics and cervical spinal cord compression, quantified on MRI, in 56 DCM participants. Results: Mean grip count was significantly lower in the DCM group (7.92 ± 3.27) compared to controls (10.26 ± 3.78; p < 0.001). The CatBoost model trained on 66 selected kinematic features achieved an AUC of 0.90, F1-score of 0.83, sensitivity of 83.3%, and specificity of 84.7%. The grip count-only model achieved lower performance (AUC 0.69, F1-score 0.67, sensitivity 75.0%, specificity 46.2%). Video-derived features were associated with MRI-derived measures of spinal cord compression including transverse diameter (R² = 0.43, p = 0.002), sagittal diameter (R² = 0.45, p = 0.001), compression ratio (R² = 0.42, p = 0.003), and maximum spinal cord compression ratio (R² = 0.36, p = 0.018). Conclusion: We demonstrated that a smartphone-based computer vision tool can accurately detect hand motor impairment specific to DCM. Finger kinematic analysis demonstrated significantly higher diagnostic accuracy than grip count alone and was associated with spinal cord compression on MRI. This approach offers a promising tool for early and scalable screening for DCM in both clinical and community settings. Degenerative Cervical Myelopathy Computer Vision Finger Kinematics Grip-and-Release Test Spinal Cord Compression Objective Motor Screening Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Introduction Degenerative cervical myelopathy (DCM) is the most common cause of non-traumatic spinal cord injury worldwide [ 1 , 2 ] and is characterized by compression of the spinal cord in the cervical spine, leading to sensorimotor deficits in the upper extremities (impaired hand dexterity) [ 3 ]. The diagnosis of DCM is made based on clinical history and examination, and magnetic resonance imaging (MRI) is used to confirm cervical spinal cord compression. These findings are used to guide the decision on surgical intervention, which is the primary treatment for DCM. However, many of the early symptoms of DCM are subtle and difficult to confirm on clinical evaluation. Over 20% of patients with DCM undergoing surgery may not exhibit physical signs of myelopathy on clinical examination [ 4 ]. Additionally, the degree of spinal cord compression on conventional MRI correlates poorly with neurological dysfunction in myelopathy [ 4 , 2 ] Together, this leads to a delay in diagnosis (1–2 years on average) and surgical intervention, which contributes to poorer recovery of neurological function. There is an unmet clinical need for objective clinical tools to detect subtle neurological deficits associated with DCM to facilitate early diagnosis and treatment. Existing symptom-based scales to assess the severity of DCM are subjective, have low sensitivity, and use arbitrary categories for a wide range of clinical severity [ 5 – 9 ]. Acknowledging these limitations, screening tests for hand dexterity and balance have been attempted including the 10 second grip and release test [ 10 ]. However, these screening tests are subjective, exhibit high interobserver variability [ 11 , 12 ] and lack sensitivity for detecting early-stage DCM. Additionally, none of these tools have been evaluated against a control group with non-DCM related hand dysfunction (due to osteoarthritis, prior hand injury or surgery), which is common among older adults. Artificial intelligence based analytical tools demonstrate high levels of validity, reliability, and responsiveness for clinical screening. These tools have been tested for neurological conditions such as amyotrophic lateral sclerosis [ 13 ] and Parkinson’s to replace self-reported and clinician assessed screening tools [ 14 ]. In DCM, computer vision-based video analysis has been used to detect subtle hand dysfunction during the 10 second grip and release test [ 11 , 15 , 16 ]. By recording and analyzing hand motion patterns, computer vision can identify subtle motor deficits associated with spinal cord pathology [ 17 ]. However, existing methods focus on hand dexterity assessment and require complex setup and instrumentation. With further refinement and validation, computer vision approaches for clinical screening in DCM offer the advantage of non-invasive and real-time assessment, potentially enhancing diagnostic accuracy and streamlining clinical workflow. To address these challenges, we developed a smartphone-based framework that leverages video-based tracking of hand and movements, coupled with machine learning-based classification, to enable automated screening of DCM. Our approach captures distinct video signatures associated with the DCM phenotype, enabling objective, scalable, and rapid identification of upper extremity motor dysfunctions. The use of a smartphone platform ensures portability, ease of use, and widespread applicability in both clinical and community environments. Methods This study was a prospective, cross-sectional observational study designed to evaluate the diagnostic performance of a smartphone-based computer vision tool for assessing hand motor dysfunction in individuals with DCM. The primary objective was to compare the diagnostic accuracy of video-derived finger kinematics with traditional grip count during the 10-second grip-and-release test. A secondary objective was to examine the relationship between video-derived features and spinal cord compression metrics from cervical spine MRI. The study was approved by the Institutional Review Board at the blinded for review (Approval Number: blinded for review ). All participants provided written informed consent prior to participation. All methods were performed in accordance with relevant guidelines and regulations, including the Declaration of Helsinki. We followed the STARD 2015 guidelines (Standards for Reporting Diagnostic Accuracy Studies) for reporting the diagnostic evaluation. Participants We recruited a total of 123 participants, comprising 58 pre-surgical patients with DCM and 65 age- and sex-matched controls, from the Spine and Hand Surgery clinics at a single academic medical center. The control group was stratified into three subgroups of approximately 20 individuals each: (1) individuals with impaired hand dexterity due to various other conditions such as arthritis, peripheral neuropathy, or prior hand injury or surgery; (2) individuals with symptomatic lumbar stenosis; and (3) healthy individuals without any known hand or gait dysfunction. DCM participants were diagnosed by the spine surgical team based on clinical history, neurological examination, and supporting MRI findings. Eligible participants were adults aged 18 to 85 years who were capable of providing informed consent. Eligibility criteria for the four groups included: patients diagnosed with DCM or cervical myeloradiculopathy exhibiting at least one clinical sign and symptom of myelopathy, MRI evidence of cervical spinal cord compression, and scheduled for decompression surgery; patients with impaired hand dexterity due to non-DCM causes such as arthritis, injury, or peripheral neuropathy; patients with impaired gait from non-DCM causes including lumbar stenosis, radiculopathy, or neurogenic claudication; and healthy controls with no history of neck or arm pain requiring medical or surgical treatment and no clinical signs or symptoms of cervical myelopathy or radiculopathy. Participants were excluded if they were unable to independently perform hand and gait tasks or unable to provide informed consent. Additional exclusion criteria included cognitive impairment, limb amputation, need for assistance with standing, or complete paralysis of the hands or legs. While a formal power calculation was not feasible due to the exploratory and machine learning-based nature of the study, a simplified power analysis based on previously reported area under the curve (AUC) values from similar classification tasks indicated that a minimum of 20 participants per group would provide at least 80% statistical power at the 0.05 significance level. Of the 60 participants initially enrolled in the DCM group, two were excluded from the final analysis: one due to inadequate lighting that rendered video analysis unreliable, and another due to severe hand dysfunction that prevented completion of the grip-and-release task. Thus, the final dataset included 58 DCM participants and all 65 controls (18 hand dysfunction, 22 lumbar stenosis, 25 healthy controls), all of whom successfully completed the video-recorded 10-second grip-and-release test and were included in the diagnostic accuracy analysis. Among the DCM participants, 56 had clinical cervical spine MRI scans available at the time of the study and were included in the secondary analysis assessing the relationship between video-derived features and the degree of spinal cord compression. Figure 1. Experimental setup for the 10-second grip-and-release (G&R) test. A smartphone is placed on a tabletop with the front camera facing upwards while a ring light illuminates the scene from below. Participants perform the G&R task with palms facing downward over the setup, allowing the camera to record from below. Protocol All participants underwent video capture of the 10-second grip-and-release task using the front-facing camera of a Galaxy Z Flip5 smartphone (Samsung, Inc.). Videos were recorded at 60 frames per second with a resolution of 1080p (Full HD) under indoor lighting conditions of at least 500 lux, measured using a digital luxmeter. The Galaxy Z Flip5 features a 10-megapixel front-facing camera (f/2.2 aperture, 1.22µm pixel size) capable of recording smooth high-resolution video suitable for motion analysis. The device was positioned horizontally to provide a clear, unobstructed view of the participant’s hand during the task (Fig. 1). Participants were asked to perform the grip-and-release task either while seated or standing, depending on comfort and balance ability. A standardized countdown was given before the start of each trial to ensure readiness, and video recording was initiated using a Bluetooth remote control to minimize movement artifacts. During the 10-second trial, participants were instructed to close (grip) and open (release) their hand as quickly and fully as possible while keeping their hand centered within the camera’s field of view. Video capture was supervised by a study team member to ensure proper positioning and lighting. In addition to the video recording, all participants completed the modified Japanese Orthopaedic Association (mJOA) questionnaire to assess upper extremity function. Video Analysis Video recordings from the 10-second grip-and-release task were analyzed using MediaPipe [ 18 ] Hands, an open-source, markerless pose estimation framework developed by Google. MediaPipe extracted 3D coordinates (x, y, z) for 21 anatomical landmarks of each hand, generating frame-wise joint positions at a sampling rate of 60 frames per second (Fig. 2 ). For each 10-second video, this resulted in approximately 75,600 raw data points per participant (2 hands × 21 landmarks × 3 coordinates × 60 frames/second × 10 seconds). The framework enabled detailed capture of hand joint trajectories across all five fingers, including metacarpophalangeal (MCP), proximal interphalangeal (PIP; Interphalangeal (IP) joint for the thumb), distal interphalangeal (DIP), and fingertip joints. From the MediaPipe raw outputs, directional vectors between adjacent joint landmarks were computed to represent joint segment orientations. These were used to derive Euler angles (yaw, pitch, and roll) for each segment, providing a frame-wise measure of finger joint rotations. The resulting time series of angular motion formed the basis for further feature extraction. In the second stage, pitch angles from the MCP–PIP joints were used for all five fingers to compute higher-order motion features (Fig. 2 ). The time series were normalized by subtracting the mean of the first five frames (assumed rest position), filtered using a fourth-order low-pass Butterworth filter with a cutoff frequency of 10 Hz, and then smoothed using a moving average with a five-frame window. Peaks and troughs were detected from the processed signal using a minimum distance of 20 frames and a prominent threshold of 10 degrees to ensure physiologically meaningful oscillations. From these events, the signal was segmented into opening (peak to trough) and closing (trough to peak) phases of hand motion. Kinematic features (see supplementary table 1 and Fig. 3) were derived from these angular trajectories, including temporal metrics (e.g., grip count, frequency, interval), angular measures (e.g., amplitude, total rotation), and derivative-based dynamics (e.g., velocity, acceleration, and jerk). Opening and closing phases of finger movement were identified using peak and trough analysis on the smoothed angular signal. These features were then used in downstream statistical comparisons and machine learning models for diagnostic classification of DCM. All analyses were conducted using custom Python scripts developed in-house. Finally, participants with at least 570 valid frames (9.5 seconds of data, < 5% data loss) were retained for analysis to ensure temporal consistency across participants. To ensure uniformity across all samples, we interpolated missing frames using values from the preceding and succeeding frames to obtain a consistent length of 600 frames. Feature extraction was completed separately for the left and right hands. The final dataset included 250 kinematic features (125 per hand) per participant. These handcrafted features were used for subsequent statistical comparisons and to train machine learning models evaluating diagnostic performance. Figure 3. Pitch angle of the index finger's segment connecting the Metacarpophalangeal (MCP) and Proximal Interphalangeal (PIP) joints. Data represented is from a sample participant during a grip-and-release task. Filtered signal (yellow lines) is overlaid on the raw signal (blue lines). Red circles indicate detected peaks (maximum flexion during grip), and green crosses denote troughs (maximum extension during extension). Data Analyses DCM Classification To assess diagnostic performance, a CatBoost [ 19 ] machine learning classifier was trained using an 80/20 train-test split, with all feature selection, hyperparameter tuning, and five-fold cross-validation performed exclusively on the training set. Feature selection was conducted on the training set only using analysis of variance (ANOVA) with a significance threshold of p 0.01 to retain the most informative features. Hyperparameters were optimized using Optuna [ 20 ] over 100 trials using the training set. A baseline logistic regression model was trained using grip count as the sole input feature to evaluate its standalone diagnostic performance. To identify the most influential features contributing to the CatBoost classifier’s predictions, SHapley Additive exPlanations (SHAP) [ 21 ] values were calculated, and the top 20 features were retained for further analysis. MRI Prediction via Video Features To further assess the clinical relevance of the extracted kinematic features, linear regression with recursive feature elimination was used to identify predictors of spinal cord morphology in 56 DCM patients who underwent both video-based testing and spinal MRI. MRI-based anatomical parameters included: transverse diameter: The lateral width of the spinal cord at the point of maximal compression; maximum spinal cord compression (MSCC) [ 22 ]: Defined as the ratio of the narrowest sagittal cord diameter to the average of the nearest normal diameters above and below the lesion; sagittal diameter: The anterior-posterior diameter of the cord at the site of compression; compression ratio: Calculated as the ratio of sagittal to transverse diameters, reflecting cord flattening severity. To identify kinematic features predictive of each MRI parameter, a multistep process was used: features with low variance and high intercorrelation were excluded, missing values were imputed, and recursive feature elimination (RFE) with linear regression was performed to automatically select up to ten of the most predictive features for each MRI outcome. Selected features were standardized before model fitting to ensure comparability of regression coefficients. Results A total of 58 participants with Degenerative Cervical Myelopathy (DCM) were included in the final analysis, consisting of 29 males (50%) and 29 females (50%), with a mean (SD) age of 64.2 (11.3) years (Table 1 ). The mean mJOA total score was 13.5, with a mean mJOA upper extremity of 4.1. Based on established thresholds, 22 participants were classified as having mild DCM, 19 as moderate, and 17 as severe. The median duration of symptoms among DCM participants was 24 months. For comparison, 65 control participants were also analyzed, with a combined mean (SD) age of 62.6 (12.8) years. The control group included 25 males and 40 females. The STARD flow diagram [ 23 ] is shown in Fig. 4 (see also supplementary materials for STARD-2015-Checklist ). DCM Classification Group differences in grip count were evaluated using t-tests. The mean grip count in DCM Table 1 Cohort Characteristics: Age, Gender Distribution, and DCM Severity Variable Summary Participants by Group DCM 58 Controls 65 Healthy Controls 25 Hand Dysfunction 18 Lumbar Stenosis 22 Age DCM 64.24 ± 11.25 yrs Controls 62.63 ± 12.78 yrs Healthy Controls 62.44 ± 12.58 yrs Hand Dysfunction 62.28 ± 14.29 yrs Lumbar Stenosis 63.14 ± 12.30 yrs Sex DCM 29 F (50.0%) 29 M (50.0%) Controls 25 F (38.5%) 40 M (61.5%) mJOA scores (DCM only) Mean upper extremity score 4.12 (SD- 0.94) Mean mJOA Total 13.53 (SD -2.96) mJOA Severity (DCM only) Mild (15–18) 22 (37.9%) Moderate (12–14) 19 (32.8%) Severe (≤ 11) 17 (29.3%) Duration of Symptoms (DCM) Median (25th -75th ) 24 (10,60) months participants (7.92 ± 3.27) was significantly lower (p < 0.001) than that of the control group (10.26 ± 3.78). Among control subgroups, the hand dysfunction group averaged 7.09 grips, the lumbar stenosis group averaged 10.65, and healthy controls achieved 12.13 grips. The CatBoost model trained on detailed kinematic features outperformed the grip-count-only model, achieving an area under the ROC curve (AUC) of 0.90 (95% CI: 0.75–1.00), area under the precision-recall curve (AP) of 0.91 (95% CI: 0.78–1.00), F1-score of 0.88 (95% CI: 0.74–1.00), sensitivity of 91.56% (95% CI: 75–100%), and specificity of 84.17% (95% CI: 61.54–100%), as estimated via 2000-iteration bootstrapping on the test set (Fig. 5 ). In contrast, the logistic regression model yielded an AUC of 0.69, F1-score of 0.67, sensitivity of 75.0%, and specificity of 46.2%. A bootstrap test with 2,000 replicates confirmed that the CatBoost model had significantly higher diagnostic accuracy (D = -2.707, p = 0.006). ROC analysis was performed (Fig. 5 ), highlighting the robustness of the top 20 features (Fig. 6 ) contributing to model performance. The thumb contributed the most (~ 34%) followed by the Index and the Middle to classification decisions, while the ring finger showed no contribution (Fig. 7 ). MRI Prediction via Video Features Regression models predicting each MRI parameter from video-derived kinematic features demonstrated significant associations across all four spinal cord compression metrics (Fig. 8 ). Model fits were strongest for sagittal diameter (R² = 0.45, p = 0.001) and transverse diameter (R² = 0.43, p = 0.002), followed by compression ratio (R² = 0.42, p = 0.003) and MSCC (R² = 0.36, p = 0.018). For MSCC, significant predictors included reductions in maximum velocity during closing movements (β = -0.057, p = 0.026), minimum acceleration during opening (β = -0.050, p = 0.023), and minimum velocity during opening (β = -0.089, p = 0.047). The Sagittal Diameter model, significant predictors included variability in range of motion and maximum jerk during closing phases (all p < 0.05). For Transverse Diameter, significant predictors involved average minimum velocities, energy balance measures, and velocity variability during opening and closing movements (p-values ranging from 0.004 to 0.041), reflecting preserved motor function associated with reduced spinal cord compression. Lastly, the Compression Ratio (CR) regression identified significant negative associations with velocity, acceleration, and velocity variability metrics (all p < 0.05), suggesting that increased compression ratio corresponds to reduced finger movement dynamics. (For further details on feature definitions and full regression summaries, please see Supplementary Materials— Table 1 and Regression Summaries) . Discussion Our study demonstrates the potential of a smartphone-based computer vision tool analyzing finger kinematics for DCM diagnosis. Our tool showed significantly improves the diagnostic accuracy compared to traditional grip count, when evaluated on our laboratory data. By capturing subtle movement features during the 10-second grip-and-release test, our model was able to detect motor deficits that are difficult to identify during a standard clinical evaluation, particularly among individuals with comorbid hand conditions. The classifier outperformed grip count in every metric—AUC, sensitivity, specificity, and F1-score—underscoring the added diagnostic value of detailed motion analysis. The high accuracy of this tool, combined with the scalability of smartphone video capture, supports its potential use as a non-invasive, accessible screening approach in both clinical and remote settings. Furthermore, the dominance of range of motion and dynamic features (e.g., velocity, acceleration, jerk) in driving classification decisions highlights disease-specific finger kinematics in DCM-related hand dysfunction. Recent advancements in computer vision and motion capture technologies have sought to address limitations in standard clinical tests. Ibara et al. [ 11 ] developed a smartphone-based machine learning model to analyze hand motion during the grip and release test, reporting high sensitivity (90.9%) and specificity (88.2%). However, the study’s small sample size and use of fixed 20-frame segments may have limited its ability to capture individualized motion patterns, reducing subject-specific insights. Ye et al. [ 24 ] improved the grip and release test using deep learning, showing that the test duration could be reduced to 6 seconds without loss of diagnostic accuracy (AUC 0.83; sensitivity 81.8%, specificity 70.3%). Yet, their control group did not include participants with other conditions affecting hand movement, and it remains unclear whether the examination can effectively distinguish DCM from other disorders. Li et al. [ 25 ] combined kinematic data from motion sensors with clinical variables in a multimodal deep learning framework. Their model achieved 79.97% accuracy and an F1 score of 0.79 using only kinematics, improving to 83.06% accuracy and an F1 score of 0.82 when clinical data were included. However, their reliance on specialized body sensors limits scalability and ease of use in routine clinical settings. While these studies highlight the potential of video-based analysis for DCM, key limitations were noted, including small sample sizes, non-representative control groups, and reliance on specialized hardware. Our study addresses these gaps by using a substantially larger cohort than prior smartphone-based work, including a clinically relevant control group with other causes of hand dysfunction, and eliminating need for specialized sensors. Thumb kinematics emerged as the strongest predictors of DCM-related hand dysfunction, contributing ~ 34% of the top 20 features in ROC analysis (Figs. 5 – 6 ), highlighting the importance of thumb kinematics in distinguishing DCM from controls in our cohort. Our results align with recent work by Ibara et al. [ 11 ], who showed that a classification model based on computer vision relied heavily on thumb interphalangeal joint movements to classify DCM and controls. Similarly, Koyama et al. [ 26 ] used a Leap Motion sensor system and reported that thumb-derived features provided the strongest discriminative power for DCM detection (AUC > 0.85), outperforming models based on ulnar fingers. Together, these studies provide converging evidence that the thumb plays a disproportionately important role in the manifestation and detection of DCM-related hand dysfunction. Neuroanatomically, the large cortical representation of the thumb and corticospinal projections devoted to thumb control at the cervical level may contribute to these findings. Indeed, precision grip between the thumb and index finger has been shown to depend on both corticospinal and subcortical pathways, which are particularly vulnerable to cervical spinal cord injury [ 27 ]. Another study [ 28 ] reported that people with DCM exhibit specific changes in pinching movements with the thumb and index fingers suggesting that radial finger movements are significantly altered. This mechanistic basis may explain why thumb kinematics consistently emerge as the most sensitive predictors of DCM across independent studies and modalities. We also found that video-derived kinematic features significantly correlate with MRI-based measures of spinal cord compression, supporting their relevance as functional biomarkers. These relationships suggest that impaired finger dynamics captured through video are associated with underlying spinal pathology. Importantly, these findings position smartphone video-based kinematics as not just diagnostic tools, but also as non-invasive assessment tools for estimating radiographic disease severity for screening in DCM. Feature selection identified distinct kinematic predictors for each MRI measure. Video features that were the strongest predictors of radiographic spinal cord compression were distinct from video features that distinguished DCM participants from controls suggesting that unique finger kinematics are associated with radiographic spinal cord compression. Our use of MediaPipe for video-based motion capture is supported by prior validation studies demonstrating its accuracy and reliability in human movement analysis. The use of computer vision technologies for motion tracking has garnered increasing attention in recent years, particularly in the context of neurological and musculoskeletal assessments. Several studies have evaluated the accuracy of MediaPipe, a markerless, real-time pose estimation framework, for human motion analysis. Maggioni et al. [ 29 ] compared MediaPipe and the Leap Motion Controller against a gold-standard marker-based motion capture system across five hand movements in 15 healthy participants. Their findings demonstrated lower root-mean-square error (RMSE) for MediaPipe (10.9°) compared to the Leap Motion Controller (14.7°), indicating superior accuracy. Similarly, Wagh et al. [ 30 ] validated MediaPipe during a touchscreen shape-tracing task with 10 participants, comparing the predicted trajectories to ground-truth touchscreen data. They reported an average RMSE of 0.28 normalized pixels and noted that shape similarity improved with geometric transformation techniques, though minor distortions persisted. Latreche et al. [ 31 ] assessed the utility of MediaPipe in measuring shoulder range-of-motion in 50 healthy volunteers by comparing it to traditional tools such as a universal goniometer and a digital angle ruler. Their results demonstrated high reliability and validity for telerehabilitation use, with mean differences close to zero and 95% limits of agreement ranging from − 9.5° to + 11.2°. This study has several limitations that warrant consideration. First, although our sample included a diverse control group, the cross-sectional design limits longitudinal insight into disease progression or post-surgical recovery. Second, while MediaPipe enabled robust hand tracking, the system may be less accurate in individuals with severe deformities or tremor, potentially introducing measurement variability. Third, although the tool shows promise for clinical use, real-world deployment would require user-friendly interfaces, standardized protocols, and external validation across different camera types and lighting conditions. Lastly, the absence of formal power calculations and the exploratory nature of machine learning limit the generalizability of statistical findings. Future studies should address these gaps through larger, longitudinal trials and integration with clinical decision-making frameworks. Our current framework could be extended to other functional assessments or body regions affected by neurological conditions, including at-home monitoring for DCM and other motor disorders. Conclusion In this study, we presented a novel, smartphone-based computer vision tool that analyzes finger kinematics to improve the diagnosis of DCM. Our results demonstrate that video-derived motion features, extracted during a simple 10-second grip-and-release task, significantly outperform traditional grip count in identifying DCM. Moreover, regression analyses revealed strong associations between selected kinematic features and MRI-based metrics of spinal cord compression validating the physiological relevance of these motion signatures. With further longitudinal validation, this approach could offer an objective method for early detection, monitoring, and stratification of patients with DCM in both clinical and community settings. Declarations Author Contribution VR and AV contributed to the conceptualization and methodology. VR was responsible for data collection. VR and HT were responsible for MRI data acquisition. VR, RA, AV, YL and AB conducted the data analysis and statistics. VR, YL and AV contributed to supervising the study. VR wrote the manuscript, and all authors reviewed and approved the final manuscript. Acknowledgement We thank Angela Jolivette, Tracy Gallenberger, Caroline Treis, and Stephanie Flaherty for their help and support. Data Availability Summary results and data supporting the findings of this study are available from the corresponding author upon reasonable request, subject to ethical and institutional approvals. References New PW, Cripps RA, Bonne Lee B (2014) Global maps of non-traumatic spinal cord injury epidemiology: towards a living data repository. Spinal Cord 52:97–109 Nagata K, Yoshimura N, Muraki S, Hashizume H, Ishimoto Y, Yamada H et al (2012) Prevalence of cervical cord compression and its association with physical performance in a population-based cohort in Japan: the Wakayama Spine Study. Spine 37:1892–1898 Choi SH, Kang C-N (2020) Degenerative Cervical Myelopathy: Pathophysiology and Current Treatment Strategies. Asian Spine J 14:710–720 Rhee JM, Heflin JA, Hamasaki T, Freedman B (2009) Prevalence of physical signs in cervical myelopathy: a prospective, controlled study. Spine 34:890–895 Harsh GR, Sypert GW, Weinstein PR, Ross DA, Wilson CB (1987) Cervical spine stenosis secondary to ossification of the posterior longitudinal ligament. J Neurosurg 67:349–357 Vitzthum H-E, Dalitz K (2007) Analysis of five specific scores for cervical spondylogenic myelopathy. Eur Spine J Off Publ Eur Spine Soc Eur Spinal Deform Soc Eur Sect Cerv Spine Res Soc 16:2096–2103 Ranawat CS, O’Leary P, Pellicci P, Tsairis P, Marchisello P, Dorr L (1979) Cervical spine fusion in rheumatoid arthritis. J Bone Joint Surg Am 61:1003–1010 Nurick S (1972) The pathogenesis of the spinal cord disorder associated with cervical spondylosis. Brain J Neurol 95:87–100 Cooper PR, Epstein F (1985) Radical resection of intramedullary spinal cord tumors in adults: Recent experience in 29 patients. J Neurosurg 63:492–499 Ono K, Ebara S, Fuji T, Yonenobu K, Fujiwara K, Yamashita K (1987) Myelopathy hand. New clinical signs of cervical cord damage. J Bone Joint Surg Br 69:215–219 Ibara T, Matsui R, Koyama T, Yamada E, Yamamoto A, Tsukamoto K et al (2023) Screening for degenerative cervical myelopathy with the 10-second grip-and-release test using a smartphone and machine learning: A pilot study. Digit Health 9:20552076231179030 Hilton B, Gardner EL, Jiang Z, Tetreault L, Wilson JRF, Zipser CM et al (2022) Establishing Diagnostic Criteria for Degenerative Cervical Myelopathy [AO Spine RECODE-DCM Research Priority Number 3]. Glob Spine J 12:55S–63S Bandini A, Green JR, Taati B, Orlandi S, Zinman L, Yunusova Y Automatic Detection of Amyotrophic Lateral Sclerosis (ALS) from Video-Based Analysis of Facial Movements: Speech and Non-Speech Tasks. 2018 13th IEEE Int Conf Autom Face Gesture Recognit FG 2018 [Internet]. Xi’an: IEEE; 2018 [cited 2024 Mar 11]. pp. 150–7. Available from: https://ieeexplore.ieee.org/document/8373824/ Choy WJ, Chen L, De Quel C, Verhagen AP, Damodaran O, Anderson DB (2022) Gait assessment tools for degenerative cervical myelopathy: a systematic review. J Spine Surg Hong Kong 8:149–162 Matsui R, Koyama T, Fujita K, Saito H, Sugiura Y et al (2021) Video-Based Hand Tracking for Screening Cervical Myelopathy. In: Bebis G, Athitsos V, Yan T, Lau M, Li F, Shi C, editors. Adv Vis Comput [Internet]. Cham: Springer International Publishing; [cited 2023 May 31]. pp. 3–14. Available from: https://link.springer.com/ 10.1007/978-3-030-90436-4_1 Koyama T, Fujita K, Watanabe M, Kato K, Sasaki T, Yoshii T et al (2022) Cervical Myelopathy Screening with Machine Learning Algorithm Focusing on Finger Motion Using Noncontact Sensor. Spine 47:163–171 Su X-J, Hou C-L, Shen B-D, Zhang W-Z, Wu D-S, Li Q et al (2020) Clinical Application of a New Assessment Tool for Myelopathy Hand Using Virtual Reality. Spine 45:E1645–E1652 Lugaresi C, Tang J, Nash H, McClanahan C, Uboweja E, Hays M et al (2019) MediaPipe: A Framework for Building Perception Pipelines [Internet]. arXiv; [cited 2024 Oct 15]. Available from: https://arxiv.org/abs/1906.08172 Prokhorenkova L, Gusev G, Vorobev A, Dorogush AV, Gulin A (2019) CatBoost: unbiased boosting with categorical features [Internet]. arXiv; [cited 2025 Apr 21]. Available from: http://arxiv.org/abs/1706.09516 Akiba T, Sano S, Yanase T, Ohta T, Koyama M, Optuna A Next-generation Hyperparameter Optimization Framework. Proc 25th ACM SIGKDD Int Conf Knowl Discov Data Min [Internet]. Anchorage AK USA: ACM; 2019 [cited 2025 Apr 30]. pp. 2623–31. Available from: https://dl.acm.org/doi/10.1145/3292500.3330701 Lundberg S, Lee S-I A Unified Approach to Interpreting Model Predictions [Internet]. arXiv; 2017 [cited 2025 Aug 5]. Available from: http://arxiv.org/abs/1705.07874 Nouri A, Martin AR, Mikulis D, Fehlings MG (2016) Magnetic resonance imaging assessment of degenerative cervical myelopathy: a review of structural changes and measurement techniques. Neurosurg Focus 40:E5 Moher D, Hopewell S, Schulz KF, Montori V, Gotzsche PC, Devereaux PJ et al (2010) CONSORT 2010 Explanation and Elaboration: updated guidelines for reporting parallel group randomised trials. BMJ 340:c869–c869 Ye Y, Chang Y, Wu W, Liao T, Yu T, Chen C et al (2024) Deep Learning-Enhanced Hand Grip and Release Test for Degenerative Cervical Myelopathy: Shortening Assessment Duration to 6 Seconds. Neurospine 21:46–56 Li X, Fei N, Wan K, Pui Yin Cheung J, Hu Y (2025) A deep learning-based hand motion classification for hand dysfunction assessment in cervical spondylotic myelopathy. Biomed Signal Process Control 99:106884 Koyama T, Matsui R, Yamamoto A, Yamada E, Norose M, Ibara T et al (2022) High-Dimensional Analysis of Finger Motion and Screening of Cervical Myelopathy With a Noncontact Sensor: Diagnostic Case-Control Study. JMIR Biomed Eng 7:e41327 Bunday KL, Tazoe T, Rothwell JC, Perez MA (2014) Subcortical control of precision grip after human spinal cord injury. J Neurosci Off J Soc Neurosci 34:7341–7350 Sakai N (2005) Finger Motion Analysis of the Patients With Cervical Myelopathy: Spine. ;30:2777–82 Maggioni V, Azevedo-Coste C, Durand S, Bailly F (2025) Optimisation and Comparison of Markerless and Marker-Based Motion Capture Methods for Hand and Finger Movement Analysis. Sensors 25:1079 Wagh V, Scott MW, Kraeutner SN (2024) Quantifying Similarities Between MediaPipe and a Known Standard to Address Issues in Tracking 2D Upper Limb Trajectories: Proof of Concept Study. JMIR Form Res 8:e56682–e56682 Latreche A, Kelaiaia R, Chemori A, Kerboua A (2023) Reliability and validity analysis of MediaPipe-based measurement system for some human rehabilitation motions. Measurement 214:112826 Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterialRegressionSummaries.docx SupplementaryMaterialTable1.pdf STARD2015Checklist.docx Cite Share Download PDF Status: Published Journal Publication published 02 Feb, 2026 Read the published version in European Spine Journal → Version 1 posted Editorial decision: Revision requested 11 Nov, 2025 Reviews received at journal 10 Nov, 2025 Reviewers agreed at journal 03 Nov, 2025 Reviewers invited by journal 03 Nov, 2025 Editor assigned by journal 28 Aug, 2025 Submission checks completed at journal 27 Aug, 2025 First submitted to journal 25 Aug, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7454858","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":543093293,"identity":"8e707795-4307-4931-b865-3aefe4699640","order_by":0,"name":"Viprav B. Raju","email":"","orcid":"","institution":"Medical College of Wisconsin","correspondingAuthor":false,"prefix":"","firstName":"Viprav","middleName":"B.","lastName":"Raju","suffix":""},{"id":543093294,"identity":"d1745998-953e-4ca6-b244-a00a1925ae7b","order_by":1,"name":"Ramesh M. Arnest","email":"","orcid":"","institution":"Medical College of Wisconsin","correspondingAuthor":false,"prefix":"","firstName":"Ramesh","middleName":"M.","lastName":"Arnest","suffix":""},{"id":543093295,"identity":"fc5b87cb-2389-4364-a10e-7f28f758eae8","order_by":2,"name":"Huy Truong","email":"","orcid":"","institution":"Medical College of Wisconsin","correspondingAuthor":false,"prefix":"","firstName":"Huy","middleName":"","lastName":"Truong","suffix":""},{"id":543093302,"identity":"76ad9061-ca2d-4295-8026-2e4d1fef3541","order_by":3,"name":"Anjishnu Banerjee","email":"","orcid":"","institution":"Medical College of Wisconsin","correspondingAuthor":false,"prefix":"","firstName":"Anjishnu","middleName":"","lastName":"Banerjee","suffix":""},{"id":543093303,"identity":"d3e076d5-42a9-4b72-836a-7b417dcad9ef","order_by":4,"name":"Yin Li","email":"","orcid":"","institution":"University of Wisconsin–Madison","correspondingAuthor":false,"prefix":"","firstName":"Yin","middleName":"","lastName":"Li","suffix":""},{"id":543093307,"identity":"9cb95744-2cee-479e-a2ea-248bc9bebf79","order_by":5,"name":"Aditya Vedantam","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwklEQVRIiWNgGAWjYJACCTDJ3gCmGBuI1GLAwMBzgGQtEglEauGfkXzwxs8df+R1Z75O/MzDYCO74QAhG26kJVv2njEw3HY7d7M0D0OaMUEtDGfOmEnwthkwArVsY85hOJxIUIv8mfPfJP+2Gdhvu3kWpOU/YS0Gx3vYpIG2JG67wQvScoCwFsPjbcbWsm3GydvOAP3yxyDZeCYhLXKHmR/efNsmZ7vt+NmNH2dU2Mn2EdKC7k7SlI+CUTAKRsEowAEA0NhGkct3GfEAAAAASUVORK5CYII=","orcid":"","institution":"Medical College of Wisconsin","correspondingAuthor":true,"prefix":"","firstName":"Aditya","middleName":"","lastName":"Vedantam","suffix":""}],"badges":[],"createdAt":"2025-08-25 14:38:27","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7454858/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7454858/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00586-026-09764-w","type":"published","date":"2026-02-02T15:57:27+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":96240801,"identity":"0e1ea6da-4397-41c2-b11a-8b9567ae2018","added_by":"auto","created_at":"2025-11-19 07:09:35","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5347563,"visible":true,"origin":"","legend":"","description":"","filename":"DCMGripandReleaseFinal.docx","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/8728b0d840a2df1e15e86f70.docx"},{"id":95844887,"identity":"b3fd2cc3-9819-4a66-ab65-af044e477357","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9187,"visible":true,"origin":"","legend":"","description":"","filename":"0d59a6067834488c9deaed5388ed7a66.json","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/152c76787394c15fd61f6b3b.json"},{"id":95844892,"identity":"c3e56152-53fd-4816-bd01-6dbe2f2a45f8","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35158,"visible":true,"origin":"","legend":"","description":"","filename":"STARD2015Checklist.docx","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/24e1b22198d27d528669043e.docx"},{"id":96240455,"identity":"fe5d4bc5-a498-4899-9117-bc4122505be9","added_by":"auto","created_at":"2025-11-19 07:08:55","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":31261,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterialRegressionSummaries.docx","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/e8bf95d0380b8aba273ce60b.docx"},{"id":95844903,"identity":"fcc15912-dba6-49cb-8cfb-8cb4f94b54c7","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":135737,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterialTable1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/e00018e98aea3a18ce78b25c.pdf"},{"id":96240225,"identity":"1db2ab5f-1b82-4e7c-8ddb-529777a34ec2","added_by":"auto","created_at":"2025-11-19 07:08:37","extension":"xml","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":94050,"visible":true,"origin":"","legend":"","description":"","filename":"0d59a6067834488c9deaed5388ed7a661enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/7681c0b4be7e2b1a3feec19d.xml"},{"id":95844909,"identity":"c7bbf40f-c0ce-40e0-becb-d36d47df7926","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3197771,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/b4091611c59a97b5f6919bc3.png"},{"id":96240081,"identity":"ba12ded5-ec5f-4771-b941-8c59ee5c2c56","added_by":"auto","created_at":"2025-11-19 07:08:21","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":915437,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/3ec86f07a168377525860c8f.png"},{"id":95844914,"identity":"cd698aa6-aaa9-4e11-b31f-e9a2df078c0d","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":340468,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/8ec825907f584c8bbff995a2.png"},{"id":96240530,"identity":"a2eb2640-556e-4dee-bc25-b16c5640b2d9","added_by":"auto","created_at":"2025-11-19 07:09:02","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":22746,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/71249882a4d45c9ca93b194b.png"},{"id":95844900,"identity":"05c50336-70b8-45c6-8b6f-2620e5e50808","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":48676,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/2ecfcf0c3da60baaeb4465ea.png"},{"id":96241311,"identity":"501adcdd-49e5-4136-8c9a-829a9ff9a22a","added_by":"auto","created_at":"2025-11-19 07:10:34","extension":"jpeg","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":263729,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/c0df4815311b84691d1be5e1.jpeg"},{"id":96240404,"identity":"573e2d9b-2658-45a6-9ae0-241ae9b94c1b","added_by":"auto","created_at":"2025-11-19 07:08:53","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":114781,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/ae20a6915d3d0a222f3e8ba3.png"},{"id":96240441,"identity":"f991986d-ca91-4a1c-9f00-2d2d7fb12b0d","added_by":"auto","created_at":"2025-11-19 07:08:55","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":204711,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/bb6861b46b6f6a16ff83b18f.png"},{"id":96241371,"identity":"1e737a0f-7416-491f-9d5e-90d7a9edaed0","added_by":"auto","created_at":"2025-11-19 07:10:39","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":342335,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/c8982ff8895b863fb716325a.png"},{"id":95844912,"identity":"373cc708-4fdf-4eb8-a957-4bae8a09ccbd","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":85720,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/46ffbe183e34d028adf704d2.png"},{"id":95844916,"identity":"5398bbf2-8c85-4874-8dd3-5f9a9bcd034c","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":38721,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/58a602ca47e508155a5fc743.png"},{"id":95844911,"identity":"5ac5064b-e5b0-42b6-8cc0-e11746c31904","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13559,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/1064dbc011049ac60e235dbc.png"},{"id":95844913,"identity":"7735d829-873e-42a7-898a-a77332dec653","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15303,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/447ef088357d23bc1fe143b0.png"},{"id":96240313,"identity":"8c095d41-0750-4425-90b9-866718c5cef5","added_by":"auto","created_at":"2025-11-19 07:08:47","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":66485,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/ff2265bac452fab84ab07657.png"},{"id":96241155,"identity":"668045c2-e13d-419b-b303-d58c74c5cafb","added_by":"auto","created_at":"2025-11-19 07:10:20","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":25012,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/b7db3d0440701f33313f5036.png"},{"id":95844915,"identity":"49488593-a890-4393-99f9-79b0e59e60b6","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":41796,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/87dfce26c318bcb24018e713.png"},{"id":95844919,"identity":"e69c8eed-cf4a-40b4-96df-2749a323c8a1","added_by":"auto","created_at":"2025-11-13 14:46:58","extension":"xml","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":88627,"visible":true,"origin":"","legend":"","description":"","filename":"0d59a6067834488c9deaed5388ed7a661structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/953d3cd4c820e277d8704d8c.xml"},{"id":96240271,"identity":"92ec7289-24a1-4fbb-bc4f-b6d6cbc02048","added_by":"auto","created_at":"2025-11-19 07:08:43","extension":"html","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":99318,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/0b62ee383d5099ea9a4a6008.html"},{"id":95844886,"identity":"5b7332b2-c700-4183-aada-5c09d46c6a8b","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":367346,"visible":true,"origin":"","legend":"\u003cp\u003eExperimental setup for the 10-second grip-and-release (G\u0026amp;R) test. A smartphone is placed on a tabletop with the front camera facing upwards while a ring light illuminates the scene from below. Participants perform the G\u0026amp;R task with palms facing downward over the setup, allowing the camera to record from below.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/60d2ea25b6d91ba39e70a739.png"},{"id":96240359,"identity":"a2404c1e-098a-4349-b7ff-a619c30291b1","added_by":"auto","created_at":"2025-11-19 07:08:51","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":292004,"visible":true,"origin":"","legend":"\u003cp\u003eHand tracking results from the 10-second grip-and-release test. Euler pitch angles were estimated for finger segments of interest using tracked hand landmarks connecting the Metacarpophalangeal (MCP) and Proximal Interphalangeal (PIP) joints (Interphalangeal (IP) joint for the thumb).\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/8ad7afb2bf4902c93ef2fd25.png"},{"id":96240348,"identity":"adff6130-6e68-4eca-b25e-f616a9d543ae","added_by":"auto","created_at":"2025-11-19 07:08:50","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":118739,"visible":true,"origin":"","legend":"\u003cp\u003ePitch angle of the index finger's segment connecting the Metacarpophalangeal (MCP) and Proximal Interphalangeal (PIP) joints. Data represented is from a sample participant during a grip-and-release task. Filtered signal (yellow lines) is overlaid on the raw signal (blue lines). Red circles indicate detected peaks (maximum flexion during grip), and green crosses denote troughs (maximum extension during extension).\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/46960fd2a997fa74be18c24b.png"},{"id":96239905,"identity":"11cbb815-504f-4562-a614-4173be8791fb","added_by":"auto","created_at":"2025-11-19 07:07:56","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":65800,"visible":true,"origin":"","legend":"\u003cp\u003eStandards for the Reporting of Diagnostic Accuracy (STARD-2015) flowchart illustrating participant inclusion, exclusion, and analysis for the validation of the video-based grip and release task.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/f977d09c9a3b900f2aa98ed8.png"},{"id":96240832,"identity":"143c8f3e-6754-48da-8c2c-2d87f9729f00","added_by":"auto","created_at":"2025-11-19 07:09:37","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":29908,"visible":true,"origin":"","legend":"\u003cp\u003eModel performance comparison for classifying individuals with Degenerative Cervical Myelopathy (DCM) versus controls. Receiver Operating Characteristic (ROC) curves for the CatBoost model using detailed kinematic features (blue) and a logistic regression model using grip count only (orange). The CatBoost model shows superior discrimination, as indicated by a higher area under the curve (AUC). The CatBoost model demonstrates better calibration across the probability range, indicating improved reliability in its predictions.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/8835adeb4f62db379e98b5f7.png"},{"id":96240413,"identity":"77611cf4-aaa8-48da-be5d-035c7ee2b37c","added_by":"auto","created_at":"2025-11-19 07:08:54","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":82959,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP-Based Feature Importance for DCM vs. Control Classification. Histograms of SHAP values for each feature, illustrating their influence and direction in the classification of DCM vs. controls. (See \u003cem\u003esupplementary materials – Table 1\u003c/em\u003e for feature definitions)\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/8944f806d9ebbd6736dec861.png"},{"id":95844898,"identity":"483c343b-3225-4884-a96b-17cc280689e0","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":54046,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP-Based Feature Importance for DCM vs. Control Classification. Left: Pie chart showing grouped importance of the top 20 kinematic features by finger. The thumb contributes most (~34%), while the ring finger has no contribution. Right: Pie chart showing grouped importance of the top 20 kinematic features by feature type. Features associated with Range of Motion contribute most (~61.1%). (see \u003cem\u003esupplementary materials – Table 1\u003c/em\u003efor feature definitions)\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/7efdf4e0d6d3f000c845cb41.png"},{"id":96240396,"identity":"f2b811e7-1ab7-4c00-91e3-9a768477f2f4","added_by":"auto","created_at":"2025-11-19 07:08:53","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":106803,"visible":true,"origin":"","legend":"\u003cp\u003eActual vs. Predicted values for four MRI parameters—Transverse Diameter, MSCC, Sagittal Diameter, and Compression Ratio—in individuals with Degenerative Cervical Myelopathy (DCM). Each plot shows the perfect prediction line (red dashed) and shaded ±10% (green) and ±20% (yellow) error bands. Most predictions fell within ±10% for Transverse Diameter (45/56), MSCC (36/56), and Sagittal Diameter (36/56), while Compression Ratio showed greater variability, with only 23/56 within ±10%. These results demonstrate the utility of video-derived features for estimating MRI-based markers of spinal cord compression.\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/e5ad8c223d83429d7a5272c8.png"},{"id":102236218,"identity":"ebf57b1e-fcf8-4a43-8064-f990ebde3a15","added_by":"auto","created_at":"2026-02-09 16:19:42","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1771866,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/ad1a717d-1b33-4a2e-936f-25b2b03d3907.pdf"},{"id":96240309,"identity":"f6222676-5acd-4bfa-903d-3071e7b0a3d3","added_by":"auto","created_at":"2025-11-19 07:08:47","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":31261,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterialRegressionSummaries.docx","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/129223e582dad28c53887866.docx"},{"id":95844895,"identity":"e82ab2f3-65ce-4d04-b764-48c24254e290","added_by":"auto","created_at":"2025-11-13 14:46:57","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":135737,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterialTable1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/7c2e943737a2b91f1632ca55.pdf"},{"id":96240520,"identity":"80eb3e55-3fab-4680-ab22-440c0ff00b3c","added_by":"auto","created_at":"2025-11-19 07:09:02","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":35158,"visible":true,"origin":"","legend":"","description":"","filename":"STARD2015Checklist.docx","url":"https://assets-eu.researchsquare.com/files/rs-7454858/v1/9b5a876228c8bc9814b88005.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Video-Based Finger Kinematics for Degenerative Cervical Myelopathy: A Smartphone-Based Computer Vision Approach for Screening","fulltext":[{"header":"Introduction","content":"\u003cp\u003eDegenerative cervical myelopathy (DCM) is the most common cause of non-traumatic spinal cord injury worldwide [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e] and is characterized by compression of the spinal cord in the cervical spine, leading to sensorimotor deficits in the upper extremities (impaired hand dexterity) [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. The diagnosis of DCM is made based on clinical history and examination, and magnetic resonance imaging (MRI) is used to confirm cervical spinal cord compression. These findings are used to guide the decision on surgical intervention, which is the primary treatment for DCM. However, many of the early symptoms of DCM are subtle and difficult to confirm on clinical evaluation. Over 20% of patients with DCM undergoing surgery may not exhibit physical signs of myelopathy on clinical examination [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Additionally, the degree of spinal cord compression on conventional MRI correlates poorly with neurological dysfunction in myelopathy [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e] Together, this leads to a delay in diagnosis (1\u0026ndash;2 years on average) and surgical intervention, which contributes to poorer recovery of neurological function. There is an unmet clinical need for objective clinical tools to detect subtle neurological deficits associated with DCM to facilitate early diagnosis and treatment.\u003c/p\u003e\u003cp\u003eExisting symptom-based scales to assess the severity of DCM are subjective, have low sensitivity, and use arbitrary categories for a wide range of clinical severity [\u003cspan additionalcitationids=\"CR6 CR7 CR8\" citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Acknowledging these limitations, screening tests for hand dexterity and balance have been attempted including the 10 second grip and release test [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. However, these screening tests are subjective, exhibit high interobserver variability [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] and lack sensitivity for detecting early-stage DCM. Additionally, none of these tools have been evaluated against a control group with non-DCM related hand dysfunction (due to osteoarthritis, prior hand injury or surgery), which is common among older adults.\u003c/p\u003e\u003cp\u003eArtificial intelligence based analytical tools demonstrate high levels of validity, reliability, and responsiveness for clinical screening. These tools have been tested for neurological conditions such as amyotrophic lateral sclerosis [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e] and Parkinson\u0026rsquo;s to replace self-reported and clinician assessed screening tools [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. In DCM, computer vision-based video analysis has been used to detect subtle hand dysfunction during the 10 second grip and release test [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. By recording and analyzing hand motion patterns, computer vision can identify subtle motor deficits associated with spinal cord pathology [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. However, existing methods focus on hand dexterity assessment and require complex setup and instrumentation. With further refinement and validation, computer vision approaches for clinical screening in DCM offer the advantage of non-invasive and real-time assessment, potentially enhancing diagnostic accuracy and streamlining clinical workflow.\u003c/p\u003e\u003cp\u003eTo address these challenges, we developed a smartphone-based framework that leverages video-based tracking of hand and movements, coupled with machine learning-based classification, to enable automated screening of DCM. Our approach captures distinct video signatures associated with the DCM phenotype, enabling objective, scalable, and rapid identification of upper extremity motor dysfunctions. The use of a smartphone platform ensures portability, ease of use, and widespread applicability in both clinical and community environments.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eThis study was a prospective, cross-sectional observational study designed to evaluate the diagnostic performance of a smartphone-based computer vision tool for assessing hand motor dysfunction in individuals with DCM. The primary objective was to compare the diagnostic accuracy of video-derived finger kinematics with traditional grip count during the 10-second grip-and-release test. A secondary objective was to examine the relationship between video-derived features and spinal cord compression metrics from cervical spine MRI. The study was approved by the Institutional Review Board at the \u003cem\u003eblinded for review\u003c/em\u003e (Approval Number: \u003cem\u003eblinded for review\u003c/em\u003e). All participants provided written informed consent prior to participation. All methods were performed in accordance with relevant guidelines and regulations, including the Declaration of Helsinki. We followed the STARD 2015 guidelines (Standards for Reporting Diagnostic Accuracy Studies) for reporting the diagnostic evaluation.\u003c/p\u003e\n\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003eParticipants\u003c/h2\u003e\n \u003cp\u003eWe recruited a total of 123 participants, comprising 58 pre-surgical patients with DCM and 65 age- and sex-matched controls, from the Spine and Hand Surgery clinics at a single academic medical center. The control group was stratified into three subgroups of approximately 20 individuals each: (1) individuals with impaired hand dexterity due to various other conditions such as arthritis, peripheral neuropathy, or prior hand injury or surgery; (2) individuals with symptomatic lumbar stenosis; and (3) healthy individuals without any known hand or gait dysfunction. DCM participants were diagnosed by the spine surgical team based on clinical history, neurological examination, and supporting MRI findings.\u003c/p\u003e\n \u003cp\u003eEligible participants were adults aged 18 to 85 years who were capable of providing informed consent. Eligibility criteria for the four groups included: patients diagnosed with DCM or cervical myeloradiculopathy exhibiting at least one clinical sign and symptom of myelopathy, MRI evidence of cervical spinal cord compression, and scheduled for decompression surgery; patients with impaired hand dexterity due to non-DCM causes such as arthritis, injury, or peripheral neuropathy; patients with impaired gait from non-DCM causes including lumbar stenosis, radiculopathy, or neurogenic claudication; and healthy controls with no history of neck or arm pain requiring medical or surgical treatment and no clinical signs or symptoms of cervical myelopathy or radiculopathy.\u003c/p\u003e\n \u003cp\u003eParticipants were excluded if they were unable to independently perform hand and gait tasks or unable to provide informed consent. Additional exclusion criteria included cognitive impairment, limb amputation, need for assistance with standing, or complete paralysis of the hands or legs.\u003c/p\u003e\n \u003cp\u003eWhile a formal power calculation was not feasible due to the exploratory and machine learning-based nature of the study, a simplified power analysis based on previously reported area under the curve (AUC) values from similar classification tasks indicated that a minimum of 20 participants per group would provide at least 80% statistical power at the 0.05 significance level.\u003c/p\u003e\n \u003cp\u003eOf the 60 participants initially enrolled in the DCM group, two were excluded from the final analysis: one due to inadequate lighting that rendered video analysis unreliable, and another due to severe hand dysfunction that prevented completion of the grip-and-release task. Thus, the final dataset included 58 DCM participants and all 65 controls (18 hand dysfunction, 22 lumbar stenosis, 25 healthy controls), all of whom successfully completed the video-recorded 10-second grip-and-release test and were included in the diagnostic accuracy analysis. Among the DCM participants, 56 had clinical cervical spine MRI scans available at the time of the study and were included in the secondary analysis assessing the relationship between video-derived features and the degree of spinal cord compression.\u003c/p\u003e\n \u003cp\u003eFigure 1. Experimental setup for the 10-second grip-and-release (G\u0026amp;R) test. A smartphone is placed on a tabletop with the front camera facing upwards while a ring light illuminates the scene from below. Participants perform the G\u0026amp;R task with palms facing downward over the setup, allowing the camera to record from below.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eProtocol\u003c/h3\u003e\n\u003cp\u003eAll participants underwent video capture of the 10-second grip-and-release task using the front-facing camera of a Galaxy Z Flip5 smartphone (Samsung, Inc.). Videos were recorded at 60 frames per second with a resolution of 1080p (Full HD) under indoor lighting conditions of at least 500 lux, measured using a digital luxmeter. The Galaxy Z Flip5 features a 10-megapixel front-facing camera (f/2.2 aperture, 1.22\u0026micro;m pixel size) capable of recording smooth high-resolution video suitable for motion analysis. The device was positioned horizontally to provide a clear, unobstructed view of the participant\u0026rsquo;s hand during the task (Fig.\u0026nbsp;1). Participants were asked to perform the grip-and-release task either while seated or standing, depending on comfort and balance ability. A standardized countdown was given before the start of each trial to ensure readiness, and video recording was initiated using a Bluetooth remote control to minimize movement artifacts. During the 10-second trial, participants were instructed to close (grip) and open (release) their hand as quickly and fully as possible while keeping their hand centered within the camera\u0026rsquo;s field of view. Video capture was supervised by a study team member to ensure proper positioning and lighting. In addition to the video recording, all participants completed the modified Japanese Orthopaedic Association (mJOA) questionnaire to assess upper extremity function.\u003c/p\u003e\n\u003ch3\u003eVideo Analysis\u003c/h3\u003e\n\u003cp\u003eVideo recordings from the 10-second grip-and-release task were analyzed using MediaPipe [\u003cspan class=\"CitationRef\"\u003e18\u003c/span\u003e] Hands, an open-source, markerless pose estimation framework developed by Google. MediaPipe extracted 3D coordinates (x, y, z) for 21 anatomical landmarks of each hand, generating frame-wise joint positions at a sampling rate of 60 frames per second (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). For each 10-second video, this resulted in approximately 75,600 raw data points per participant (2 hands \u0026times; 21 landmarks \u0026times; 3 coordinates \u0026times; 60 frames/second \u0026times; 10 seconds). The framework enabled detailed capture of hand joint trajectories across all five fingers, including metacarpophalangeal (MCP), proximal interphalangeal (PIP; Interphalangeal (IP) joint for the thumb), distal interphalangeal (DIP), and fingertip joints.\u003c/p\u003e\n\u003cp\u003eFrom the MediaPipe raw outputs, directional vectors between adjacent joint landmarks were computed to represent joint segment orientations. These were used to derive Euler angles (yaw, pitch, and roll) for each segment, providing a frame-wise measure of finger joint rotations. The resulting time series of angular motion formed the basis for further feature extraction. In the second stage, pitch angles from the MCP\u0026ndash;PIP joints were used for all five fingers to compute higher-order motion features (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). The time series were normalized by subtracting the mean of the first five frames (assumed rest position), filtered using a fourth-order low-pass Butterworth filter with a cutoff frequency of 10 Hz, and then smoothed using a moving average with a five-frame window. Peaks and troughs were detected from the processed signal using a minimum distance of 20 frames and a prominent threshold of 10 degrees to ensure physiologically meaningful oscillations. From these events, the signal was segmented into opening (peak to trough) and closing (trough to peak) phases of hand motion.\u003c/p\u003e\n\u003cp\u003eKinematic features (see \u003cem\u003esupplementary table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e\u003c/em\u003e and Fig. 3) were derived from these angular trajectories, including temporal metrics (e.g., grip count, frequency, interval), angular measures (e.g., amplitude, total rotation), and derivative-based dynamics (e.g., velocity, acceleration, and jerk). Opening and closing phases of finger movement were identified using peak and trough analysis on the smoothed angular signal. These features were then used in downstream statistical comparisons and machine learning models for diagnostic classification of DCM. All analyses were conducted using custom Python scripts developed in-house.\u003c/p\u003e\n\u003cp\u003eFinally, participants with at least 570 valid frames (9.5 seconds of data, \u0026lt;\u0026thinsp;5% data loss) were retained for analysis to ensure temporal consistency across participants. To ensure uniformity across all samples, we interpolated missing frames using values from the preceding and succeeding frames to obtain a consistent length of 600 frames. Feature extraction was completed separately for the left and right hands. The final dataset included 250 kinematic features (125 per hand) per participant. These handcrafted features were used for subsequent statistical comparisons and to train machine learning models evaluating diagnostic performance.\u003c/p\u003e\n\u003cp\u003eFigure 3. Pitch angle of the index finger\u0026apos;s segment connecting the Metacarpophalangeal (MCP) and Proximal Interphalangeal (PIP) joints. Data represented is from a sample participant during a grip-and-release task. Filtered signal (yellow lines) is overlaid on the raw signal (blue lines). Red circles indicate detected peaks (maximum flexion during grip), and green crosses denote troughs (maximum extension during extension).\u003c/p\u003e\n\u003ch3\u003eData Analyses\u003c/h3\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n \u003ch2\u003eDCM Classification\u003c/h2\u003e\n \u003cp\u003eTo assess diagnostic performance, a CatBoost [\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e] machine learning classifier was trained using an 80/20 train-test split, with all feature selection, hyperparameter tuning, and five-fold cross-validation performed exclusively on the training set. Feature selection was conducted on the training set only using analysis of variance (ANOVA) with a significance threshold of p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, and mutual information scores\u0026thinsp;\u0026gt;\u0026thinsp;0.01 to retain the most informative features. Hyperparameters were optimized using \u003cem\u003eOptuna\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e] over 100 trials using the training set. A baseline logistic regression model was trained using grip count as the sole input feature to evaluate its standalone diagnostic performance. To identify the most influential features contributing to the CatBoost classifier\u0026rsquo;s predictions, \u003cem\u003eSHapley Additive exPlanations\u003c/em\u003e (SHAP) [\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e] values were calculated, and the top 20 features were retained for further analysis.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003eMRI Prediction via Video Features\u003c/h2\u003e\n \u003cp\u003eTo further assess the clinical relevance of the extracted kinematic features, linear regression with recursive feature elimination was used to identify predictors of spinal cord morphology in 56 DCM patients who underwent both video-based testing and spinal MRI. MRI-based anatomical parameters included: transverse diameter: The lateral width of the spinal cord at the point of maximal compression; maximum spinal cord compression (MSCC) [\u003cspan class=\"CitationRef\"\u003e22\u003c/span\u003e]: Defined as the ratio of the narrowest sagittal cord diameter to the average of the nearest normal diameters above and below the lesion; sagittal diameter: The anterior-posterior diameter of the cord at the site of compression; compression ratio: Calculated as the ratio of sagittal to transverse diameters, reflecting cord flattening severity. To identify kinematic features predictive of each MRI parameter, a multistep process was used: features with low variance and high intercorrelation were excluded, missing values were imputed, and recursive feature elimination (RFE) with linear regression was performed to automatically select up to ten of the most predictive features for each MRI outcome. Selected features were standardized before model fitting to ensure comparability of regression coefficients.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eA total of 58 participants with Degenerative Cervical Myelopathy (DCM) were included in the final analysis, consisting of 29 males (50%) and 29 females (50%), with a mean (SD) age of 64.2 (11.3) years (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The mean mJOA total score was 13.5, with a mean mJOA upper extremity of 4.1. Based on established thresholds, 22 participants were classified as having mild DCM, 19 as moderate, and 17 as severe. The median duration of symptoms among DCM participants was 24 months. For comparison, 65 control participants were also analyzed, with a combined mean (SD) age of 62.6 (12.8) years. The control group included 25 males and 40 females. The STARD flow diagram [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e] is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e4\u003c/span\u003e \u003cem\u003e(see also supplementary materials for STARD-2015-Checklist\u003c/em\u003e).\u003c/p\u003e\n\u003ch3\u003eDCM Classification\u003c/h3\u003e\n\u003cp\u003eGroup differences in grip count were evaluated using t-tests. The mean grip count in DCM\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eCohort Characteristics: Age, Gender Distribution, and DCM Severity\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVariable\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSummary\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eParticipants by Group\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDCM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e58\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eControls\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e65\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHealthy Controls\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e25\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHand Dysfunction\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e18\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLumbar Stenosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e22\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDCM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e64.24\u0026thinsp;\u0026plusmn;\u0026thinsp;11.25 yrs\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eControls\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e62.63\u0026thinsp;\u0026plusmn;\u0026thinsp;12.78 yrs\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHealthy Controls\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e62.44\u0026thinsp;\u0026plusmn;\u0026thinsp;12.58 yrs\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHand Dysfunction\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e62.28\u0026thinsp;\u0026plusmn;\u0026thinsp;14.29 yrs\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLumbar Stenosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e63.14\u0026thinsp;\u0026plusmn;\u0026thinsp;12.30 yrs\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDCM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e29 F (50.0%)\u003c/p\u003e\u003cp\u003e29 M (50.0%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eControls\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e25 F (38.5%)\u003c/p\u003e\u003cp\u003e40 M (61.5%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003emJOA scores (DCM only)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMean upper extremity score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4.12 (SD- 0.94)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMean mJOA Total\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e13.53 (SD -2.96)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003emJOA Severity (DCM only)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMild (15\u0026ndash;18)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e22 (37.9%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModerate (12\u0026ndash;14)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e19 (32.8%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSevere (\u0026le;\u0026thinsp;11)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e17 (29.3%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDuration of Symptoms (DCM)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedian (25th -75th )\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e24 (10,60) months\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eparticipants (7.92\u0026thinsp;\u0026plusmn;\u0026thinsp;3.27) was significantly lower (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) than that of the control group (10.26\u0026thinsp;\u0026plusmn;\u0026thinsp;3.78). Among control subgroups, the hand dysfunction group averaged 7.09 grips, the lumbar stenosis group averaged 10.65, and healthy controls achieved 12.13 grips.\u003c/p\u003e\u003cp\u003eThe CatBoost model trained on detailed kinematic features outperformed the grip-count-only model, achieving an area under the ROC curve (AUC) of 0.90 (95% CI: 0.75\u0026ndash;1.00), area under the precision-recall curve (AP) of 0.91 (95% CI: 0.78\u0026ndash;1.00), F1-score of 0.88 (95% CI: 0.74\u0026ndash;1.00), sensitivity of 91.56% (95% CI: 75\u0026ndash;100%), and specificity of 84.17% (95% CI: 61.54\u0026ndash;100%), as estimated via 2000-iteration bootstrapping on the test set (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003e). In contrast, the logistic regression model yielded an AUC of 0.69, F1-score of 0.67, sensitivity of 75.0%, and specificity of 46.2%. A bootstrap test with 2,000 replicates confirmed that the CatBoost model had significantly higher diagnostic accuracy (D = -2.707, p\u0026thinsp;=\u0026thinsp;0.006). ROC analysis was performed (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003e), highlighting the robustness of the top 20 features (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e6\u003c/span\u003e) contributing to model performance. The thumb contributed the most (~\u0026thinsp;34%) followed by the Index and the Middle to classification decisions, while the ring finger showed no contribution (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eMRI Prediction via Video Features\u003c/h2\u003e\u003cp\u003eRegression models predicting each MRI parameter from video-derived kinematic features demonstrated significant associations across all four spinal cord compression metrics (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e8\u003c/span\u003e). Model fits were strongest for sagittal diameter (R\u0026sup2; = 0.45, p\u0026thinsp;=\u0026thinsp;0.001) and transverse diameter (R\u0026sup2; = 0.43, p\u0026thinsp;=\u0026thinsp;0.002), followed by compression ratio (R\u0026sup2; = 0.42, p\u0026thinsp;=\u0026thinsp;0.003) and MSCC (R\u0026sup2; = 0.36, p\u0026thinsp;=\u0026thinsp;0.018).\u003c/p\u003e\u003cp\u003eFor MSCC, significant predictors included reductions in maximum velocity during closing movements (β = -0.057, p\u0026thinsp;=\u0026thinsp;0.026), minimum acceleration during opening (β = -0.050, p\u0026thinsp;=\u0026thinsp;0.023), and minimum velocity during opening (β = -0.089, p\u0026thinsp;=\u0026thinsp;0.047). The Sagittal Diameter model, significant predictors included variability in range of motion and maximum jerk during closing phases (all p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). For Transverse Diameter, significant predictors involved average minimum velocities, energy balance measures, and velocity variability during opening and closing movements (p-values ranging from 0.004 to 0.041), reflecting preserved motor function associated with reduced spinal cord compression. Lastly, the Compression Ratio (CR) regression identified significant negative associations with velocity, acceleration, and velocity variability metrics (all p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), suggesting that increased compression ratio corresponds to reduced finger movement dynamics. (For further details on feature definitions and full regression summaries, please see \u003cem\u003eSupplementary Materials\u0026mdash;\u003c/em\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e \u003cem\u003eand Regression Summaries)\u003c/em\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur study demonstrates the potential of a smartphone-based computer vision tool analyzing finger kinematics for DCM diagnosis. Our tool showed significantly improves the diagnostic accuracy compared to traditional grip count, when evaluated on our laboratory data. By capturing subtle movement features during the 10-second grip-and-release test, our model was able to detect motor deficits that are difficult to identify during a standard clinical evaluation, particularly among individuals with comorbid hand conditions. The classifier outperformed grip count in every metric\u0026mdash;AUC, sensitivity, specificity, and F1-score\u0026mdash;underscoring the added diagnostic value of detailed motion analysis. The high accuracy of this tool, combined with the scalability of smartphone video capture, supports its potential use as a non-invasive, accessible screening approach in both clinical and remote settings. Furthermore, the dominance of range of motion and dynamic features (e.g., velocity, acceleration, jerk) in driving classification decisions highlights disease-specific finger kinematics in DCM-related hand dysfunction.\u003c/p\u003e\u003cp\u003eRecent advancements in computer vision and motion capture technologies have sought to address limitations in standard clinical tests. Ibara et al. [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] developed a smartphone-based machine learning model to analyze hand motion during the grip and release test, reporting high sensitivity (90.9%) and specificity (88.2%). However, the study\u0026rsquo;s small sample size and use of fixed 20-frame segments may have limited its ability to capture individualized motion patterns, reducing subject-specific insights. Ye et al. [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e] improved the grip and release test using deep learning, showing that the test duration could be reduced to 6 seconds without loss of diagnostic accuracy (AUC 0.83; sensitivity 81.8%, specificity 70.3%). Yet, their control group did not include participants with other conditions affecting hand movement, and it remains unclear whether the examination can effectively distinguish DCM from other disorders. Li et al. [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e] combined kinematic data from motion sensors with clinical variables in a multimodal deep learning framework. Their model achieved 79.97% accuracy and an F1 score of 0.79 using only kinematics, improving to 83.06% accuracy and an F1 score of 0.82 when clinical data were included. However, their reliance on specialized body sensors limits scalability and ease of use in routine clinical settings. While these studies highlight the potential of video-based analysis for DCM, key limitations were noted, including small sample sizes, non-representative control groups, and reliance on specialized hardware. Our study addresses these gaps by using a substantially larger cohort than prior smartphone-based work, including a clinically relevant control group with other causes of hand dysfunction, and eliminating need for specialized sensors.\u003c/p\u003e\u003cp\u003eThumb kinematics emerged as the strongest predictors of DCM-related hand dysfunction, contributing\u0026thinsp;~\u0026thinsp;34% of the top 20 features in ROC analysis (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e6\u003c/span\u003e), highlighting the importance of thumb kinematics in distinguishing DCM from controls in our cohort. Our results align with recent work by Ibara et al. [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e], who showed that a classification model based on computer vision relied heavily on thumb interphalangeal joint movements to classify DCM and controls. Similarly, Koyama et al. [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e] used a Leap Motion sensor system and reported that thumb-derived features provided the strongest discriminative power for DCM detection (AUC\u0026thinsp;\u0026gt;\u0026thinsp;0.85), outperforming models based on ulnar fingers. Together, these studies provide converging evidence that the thumb plays a disproportionately important role in the manifestation and detection of DCM-related hand dysfunction. Neuroanatomically, the large cortical representation of the thumb and corticospinal projections devoted to thumb control at the cervical level may contribute to these findings. Indeed, precision grip between the thumb and index finger has been shown to depend on both corticospinal and subcortical pathways, which are particularly vulnerable to cervical spinal cord injury [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Another study [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e] reported that people with DCM exhibit specific changes in pinching movements with the thumb and index fingers suggesting that radial finger movements are significantly altered. This mechanistic basis may explain why thumb kinematics consistently emerge as the most sensitive predictors of DCM across independent studies and modalities.\u003c/p\u003e\u003cp\u003eWe also found that video-derived kinematic features significantly correlate with MRI-based measures of spinal cord compression, supporting their relevance as functional biomarkers. These relationships suggest that impaired finger dynamics captured through video are associated with underlying spinal pathology. Importantly, these findings position smartphone video-based kinematics as not just diagnostic tools, but also as non-invasive assessment tools for estimating radiographic disease severity for screening in DCM. Feature selection identified distinct kinematic predictors for each MRI measure. Video features that were the strongest predictors of radiographic spinal cord compression were distinct from video features that distinguished DCM participants from controls suggesting that unique finger kinematics are associated with radiographic spinal cord compression.\u003c/p\u003e\u003cp\u003eOur use of MediaPipe for video-based motion capture is supported by prior validation studies demonstrating its accuracy and reliability in human movement analysis. The use of computer vision technologies for motion tracking has garnered increasing attention in recent years, particularly in the context of neurological and musculoskeletal assessments. Several studies have evaluated the accuracy of MediaPipe, a markerless, real-time pose estimation framework, for human motion analysis. Maggioni et al. [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e] compared MediaPipe and the Leap Motion Controller against a gold-standard marker-based motion capture system across five hand movements in 15 healthy participants. Their findings demonstrated lower root-mean-square error (RMSE) for MediaPipe (10.9\u0026deg;) compared to the Leap Motion Controller (14.7\u0026deg;), indicating superior accuracy. Similarly, Wagh et al. [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e] validated MediaPipe during a touchscreen shape-tracing task with 10 participants, comparing the predicted trajectories to ground-truth touchscreen data. They reported an average RMSE of 0.28 normalized pixels and noted that shape similarity improved with geometric transformation techniques, though minor distortions persisted. Latreche et al. [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e] assessed the utility of MediaPipe in measuring shoulder range-of-motion in 50 healthy volunteers by comparing it to traditional tools such as a universal goniometer and a digital angle ruler. Their results demonstrated high reliability and validity for telerehabilitation use, with mean differences close to zero and 95% limits of agreement ranging from \u0026minus;\u0026thinsp;9.5\u0026deg; to +\u0026thinsp;11.2\u0026deg;.\u003c/p\u003e\u003cp\u003eThis study has several limitations that warrant consideration. First, although our sample included a diverse control group, the cross-sectional design limits longitudinal insight into disease progression or post-surgical recovery. Second, while MediaPipe enabled robust hand tracking, the system may be less accurate in individuals with severe deformities or tremor, potentially introducing measurement variability. Third, although the tool shows promise for clinical use, real-world deployment would require user-friendly interfaces, standardized protocols, and external validation across different camera types and lighting conditions. Lastly, the absence of formal power calculations and the exploratory nature of machine learning limit the generalizability of statistical findings. Future studies should address these gaps through larger, longitudinal trials and integration with clinical decision-making frameworks. Our current framework could be extended to other functional assessments or body regions affected by neurological conditions, including at-home monitoring for DCM and other motor disorders.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn this study, we presented a novel, smartphone-based computer vision tool that analyzes finger kinematics to improve the diagnosis of DCM. Our results demonstrate that video-derived motion features, extracted during a simple 10-second grip-and-release task, significantly outperform traditional grip count in identifying DCM. Moreover, regression analyses revealed strong associations between selected kinematic features and MRI-based metrics of spinal cord compression validating the physiological relevance of these motion signatures. With further longitudinal validation, this approach could offer an objective method for early detection, monitoring, and stratification of patients with DCM in both clinical and community settings.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eVR and AV contributed to the conceptualization and methodology. VR was responsible for data collection. VR and HT were responsible for MRI data acquisition. VR, RA, AV, YL and AB conducted the data analysis and statistics. VR, YL and AV contributed to supervising the study. VR wrote the manuscript, and all authors reviewed and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eWe thank Angela Jolivette, Tracy Gallenberger, Caroline Treis, and Stephanie Flaherty for their help and support.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eSummary results and data supporting the findings of this study are available from the corresponding author upon reasonable request, subject to ethical and institutional approvals.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eNew PW, Cripps RA, Bonne Lee B (2014) Global maps of non-traumatic spinal cord injury epidemiology: towards a living data repository. Spinal Cord 52:97\u0026ndash;109\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNagata K, Yoshimura N, Muraki S, Hashizume H, Ishimoto Y, Yamada H et al (2012) Prevalence of cervical cord compression and its association with physical performance in a population-based cohort in Japan: the Wakayama Spine Study. Spine 37:1892\u0026ndash;1898\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChoi SH, Kang C-N (2020) Degenerative Cervical Myelopathy: Pathophysiology and Current Treatment Strategies. Asian Spine J 14:710\u0026ndash;720\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRhee JM, Heflin JA, Hamasaki T, Freedman B (2009) Prevalence of physical signs in cervical myelopathy: a prospective, controlled study. Spine 34:890\u0026ndash;895\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHarsh GR, Sypert GW, Weinstein PR, Ross DA, Wilson CB (1987) Cervical spine stenosis secondary to ossification of the posterior longitudinal ligament. J Neurosurg 67:349\u0026ndash;357\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVitzthum H-E, Dalitz K (2007) Analysis of five specific scores for cervical spondylogenic myelopathy. Eur Spine J Off Publ Eur Spine Soc Eur Spinal Deform Soc Eur Sect Cerv Spine Res Soc 16:2096\u0026ndash;2103\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRanawat CS, O\u0026rsquo;Leary P, Pellicci P, Tsairis P, Marchisello P, Dorr L (1979) Cervical spine fusion in rheumatoid arthritis. J Bone Joint Surg Am 61:1003\u0026ndash;1010\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNurick S (1972) The pathogenesis of the spinal cord disorder associated with cervical spondylosis. Brain J Neurol 95:87\u0026ndash;100\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCooper PR, Epstein F (1985) Radical resection of intramedullary spinal cord tumors in adults: Recent experience in 29 patients. J Neurosurg 63:492\u0026ndash;499\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOno K, Ebara S, Fuji T, Yonenobu K, Fujiwara K, Yamashita K (1987) Myelopathy hand. New clinical signs of cervical cord damage. J Bone Joint Surg Br 69:215\u0026ndash;219\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIbara T, Matsui R, Koyama T, Yamada E, Yamamoto A, Tsukamoto K et al (2023) Screening for degenerative cervical myelopathy with the 10-second grip-and-release test using a smartphone and machine learning: A pilot study. Digit Health 9:20552076231179030\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHilton B, Gardner EL, Jiang Z, Tetreault L, Wilson JRF, Zipser CM et al (2022) Establishing Diagnostic Criteria for Degenerative Cervical Myelopathy [AO Spine RECODE-DCM Research Priority Number 3]. Glob Spine J 12:55S\u0026ndash;63S\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBandini A, Green JR, Taati B, Orlandi S, Zinman L, Yunusova Y Automatic Detection of Amyotrophic Lateral Sclerosis (ALS) from Video-Based Analysis of Facial Movements: Speech and Non-Speech Tasks. 2018 13th IEEE Int Conf Autom Face Gesture Recognit FG 2018 [Internet]. Xi\u0026rsquo;an: IEEE; 2018 [cited 2024 Mar 11]. pp. 150\u0026ndash;7. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ieeexplore.ieee.org/document/8373824/\u003c/span\u003e\u003cspan address=\"https://ieeexplore.ieee.org/document/8373824/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChoy WJ, Chen L, De Quel C, Verhagen AP, Damodaran O, Anderson DB (2022) Gait assessment tools for degenerative cervical myelopathy: a systematic review. J Spine Surg Hong Kong 8:149\u0026ndash;162\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMatsui R, Koyama T, Fujita K, Saito H, Sugiura Y et al (2021) Video-Based Hand Tracking for Screening Cervical Myelopathy. In: Bebis G, Athitsos V, Yan T, Lau M, Li F, Shi C, editors. Adv Vis Comput [Internet]. Cham: Springer International Publishing; [cited 2023 May 31]. pp. 3\u0026ndash;14. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://link.springer.com/\u003c/span\u003e\u003cspan address=\"https://link.springer.com/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/978-3-030-90436-4_1\u003c/span\u003e\u003cspan address=\"10.1007/978-3-030-90436-4_1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKoyama T, Fujita K, Watanabe M, Kato K, Sasaki T, Yoshii T et al (2022) Cervical Myelopathy Screening with Machine Learning Algorithm Focusing on Finger Motion Using Noncontact Sensor. Spine 47:163\u0026ndash;171\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSu X-J, Hou C-L, Shen B-D, Zhang W-Z, Wu D-S, Li Q et al (2020) Clinical Application of a New Assessment Tool for Myelopathy Hand Using Virtual Reality. Spine 45:E1645\u0026ndash;E1652\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLugaresi C, Tang J, Nash H, McClanahan C, Uboweja E, Hays M et al (2019) MediaPipe: A Framework for Building Perception Pipelines [Internet]. arXiv; [cited 2024 Oct 15]. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://arxiv.org/abs/1906.08172\u003c/span\u003e\u003cspan address=\"https://arxiv.org/abs/1906.08172\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eProkhorenkova L, Gusev G, Vorobev A, Dorogush AV, Gulin A (2019) CatBoost: unbiased boosting with categorical features [Internet]. arXiv; [cited 2025 Apr 21]. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://arxiv.org/abs/1706.09516\u003c/span\u003e\u003cspan address=\"http://arxiv.org/abs/1706.09516\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAkiba T, Sano S, Yanase T, Ohta T, Koyama M, Optuna A Next-generation Hyperparameter Optimization Framework. Proc 25th ACM SIGKDD Int Conf Knowl Discov Data Min [Internet]. Anchorage AK USA: ACM; 2019 [cited 2025 Apr 30]. pp. 2623\u0026ndash;31. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://dl.acm.org/doi/10.1145/3292500.3330701\u003c/span\u003e\u003cspan address=\"https://dl.acm.doi/10.1145/3292500.3330701\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLundberg S, Lee S-I A Unified Approach to Interpreting Model Predictions [Internet]. arXiv; 2017 [cited 2025 Aug 5]. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://arxiv.org/abs/1705.07874\u003c/span\u003e\u003cspan address=\"http://arxiv.org/abs/1705.07874\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNouri A, Martin AR, Mikulis D, Fehlings MG (2016) Magnetic resonance imaging assessment of degenerative cervical myelopathy: a review of structural changes and measurement techniques. Neurosurg Focus 40:E5\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMoher D, Hopewell S, Schulz KF, Montori V, Gotzsche PC, Devereaux PJ et al (2010) CONSORT 2010 Explanation and Elaboration: updated guidelines for reporting parallel group randomised trials. BMJ 340:c869\u0026ndash;c869\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYe Y, Chang Y, Wu W, Liao T, Yu T, Chen C et al (2024) Deep Learning-Enhanced Hand Grip and Release Test for Degenerative Cervical Myelopathy: Shortening Assessment Duration to 6 Seconds. Neurospine 21:46\u0026ndash;56\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi X, Fei N, Wan K, Pui Yin Cheung J, Hu Y (2025) A deep learning-based hand motion classification for hand dysfunction assessment in cervical spondylotic myelopathy. Biomed Signal Process Control 99:106884\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKoyama T, Matsui R, Yamamoto A, Yamada E, Norose M, Ibara T et al (2022) High-Dimensional Analysis of Finger Motion and Screening of Cervical Myelopathy With a Noncontact Sensor: Diagnostic Case-Control Study. JMIR Biomed Eng 7:e41327\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBunday KL, Tazoe T, Rothwell JC, Perez MA (2014) Subcortical control of precision grip after human spinal cord injury. J Neurosci Off J Soc Neurosci 34:7341\u0026ndash;7350\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSakai N (2005) Finger Motion Analysis of the Patients With Cervical Myelopathy: Spine. ;30:2777\u0026ndash;82\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMaggioni V, Azevedo-Coste C, Durand S, Bailly F (2025) Optimisation and Comparison of Markerless and Marker-Based Motion Capture Methods for Hand and Finger Movement Analysis. Sensors 25:1079\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWagh V, Scott MW, Kraeutner SN (2024) Quantifying Similarities Between MediaPipe and a Known Standard to Address Issues in Tracking 2D Upper Limb Trajectories: Proof of Concept Study. JMIR Form Res 8:e56682\u0026ndash;e56682\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLatreche A, Kelaiaia R, Chemori A, Kerboua A (2023) Reliability and validity analysis of MediaPipe-based measurement system for some human rehabilitation motions. Measurement 214:112826\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"european-spine-journal","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"esjo","sideBox":"Learn more about [European Spine Journal](http://link.springer.com/journal/586)","snPcode":"586","submissionUrl":"https://submission.springernature.com/new-submission/586/3","title":"European Spine Journal","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Degenerative Cervical Myelopathy, Computer Vision, Finger Kinematics, Grip-and-Release Test, Spinal Cord Compression, Objective Motor Screening","lastPublishedDoi":"10.21203/rs.3.rs-7454858/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7454858/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePurpose:\u003c/strong\u003e\u003cbr\u003e\n Degenerative cervical myelopathy (DCM) is a progressive spinal cord disorder that is frequently underdiagnosed, with diagnostic delays averaging 1 to 4 years. A key limitation in current clinical practice is the lack of objective and accessible screening tools. The 10-second grip-and-release test is commonly used to assess hand dysfunction in DCM, but its diagnostic performance is limited, particularly in older individuals with comorbid hand conditions such as osteoarthritis or peripheral neuropathy. To address this limitation, we developed and evaluated a smartphone-based computer vision tool that quantifies finger kinematics during the grip-and-release test. Our primary objective was to determine whether video-derived finger kinematics can provide superior diagnostic performance compared to grip count alone. A secondary objective was to assess how these video features correlate with cervical spinal cord compression on Magnetic Resonance Imaging (MRI).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods:\u003c/strong\u003e\u003cbr\u003e\nWe collected smartphone videos of 58 participants with DCM and 65 age-matched controls (including healthy individuals and those with non-DCM hand dysfunction) performing the 10-second grip-and-release test. Finger landmarks were extracted using MediaPipe, and 250 kinematic features per finger were computed and combined across both hands. Feature selection was performed using ANOVA (p \u0026lt; 0.05) and mutual information scores (\u0026gt; 0.01). A CatBoost classifier was trained on selected features using an 80/20 train-test split and five-fold cross-validation. A logistic regression model was trained using grip count alone. Model performance was evaluated using AUC, F1-score, sensitivity, and specificity.\u003c/p\u003e\n\u003cp\u003eFor the secondary analysis, we used linear regression models to evaluate associations between video-derived kinematics and cervical spinal cord compression, quantified on MRI, in 56 DCM participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003e\u003cbr\u003e\nMean grip count was significantly lower in the DCM group (7.92 ± 3.27) compared to controls (10.26 ± 3.78; p \u0026lt; 0.001). The CatBoost model trained on 66 selected kinematic features achieved an AUC of 0.90, F1-score of 0.83, sensitivity of 83.3%, and specificity of 84.7%. The grip count-only model achieved lower performance (AUC 0.69, F1-score 0.67, sensitivity 75.0%, specificity 46.2%). Video-derived features were associated with MRI-derived measures of spinal cord compression including transverse diameter (R² = 0.43, \u003cem\u003ep\u003c/em\u003e= 0.002), sagittal diameter (R² = 0.45, \u003cem\u003ep\u003c/em\u003e = 0.001), compression ratio (R² = 0.42, \u003cem\u003ep\u003c/em\u003e = 0.003), and maximum spinal cord compression ratio (R² = 0.36, \u003cem\u003ep\u003c/em\u003e = 0.018).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion:\u003c/strong\u003e\u003cbr\u003e\nWe demonstrated that a smartphone-based computer vision tool can accurately detect hand motor impairment specific to DCM. Finger kinematic analysis demonstrated significantly higher diagnostic accuracy than grip count alone and was associated with spinal cord compression on MRI. This approach offers a promising tool for early and scalable screening for DCM in both clinical and community settings.\u003c/p\u003e","manuscriptTitle":"Video-Based Finger Kinematics for Degenerative Cervical Myelopathy: A Smartphone-Based Computer Vision Approach for Screening","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-13 14:46:52","doi":"10.21203/rs.3.rs-7454858/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-11-11T06:10:03+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-10T12:58:23+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"992702432993510088888554608996437453","date":"2025-11-03T11:42:10+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-11-03T07:36:28+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-28T06:39:49+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-27T07:19:04+00:00","index":"","fulltext":""},{"type":"submitted","content":"European Spine Journal","date":"2025-08-25T14:36:25+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"european-spine-journal","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"esjo","sideBox":"Learn more about [European Spine Journal](http://link.springer.com/journal/586)","snPcode":"586","submissionUrl":"https://submission.springernature.com/new-submission/586/3","title":"European Spine Journal","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"7fbca793-173f-4968-b72a-90afbd4a8b9a","owner":[],"postedDate":"November 13th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-02-09T16:19:37+00:00","versionOfRecord":{"articleIdentity":"rs-7454858","link":"https://doi.org/10.1007/s00586-026-09764-w","journal":{"identity":"european-spine-journal","isVorOnly":false,"title":"European Spine Journal"},"publishedOn":"2026-02-02 15:57:27","publishedOnDateReadable":"February 2nd, 2026"},"versionCreatedAt":"2025-11-13 14:46:52","video":"","vorDoi":"10.1007/s00586-026-09764-w","vorDoiUrl":"https://doi.org/10.1007/s00586-026-09764-w","workflowStages":[]},"version":"v1","identity":"rs-7454858","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7454858","identity":"rs-7454858","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}