pith. sign in

arxiv: 2605.11314 · v2 · pith:UBS25CSFnew · submitted 2026-05-11 · 💻 cs.CV · cs.AI

Quantifying Rodda and Graham Gait Classification from 3D Makerless Kinematics derived from a Single-view Video in a Heterogeneous Pediatric Clinical Cohort

Lauhitya Reddy , Seth Donahue , Jeremy Bauer , Susan Sienko , Anita Bagley , Joseph Krzak , Maura Eveld , Karen Kruger
show 3 more authors
Ross Chafetz Vedant Kulkarni Hyeokhyen Kwon
This is my paper

Pith reviewed 2026-05-22 10:14 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords markerless gait analysiscerebral palsyRodda and Graham classificationsingle-view videoz-scorespediatric gaitkinematicscomputer vision
0
0 comments X

The pith

Single-view video analysis recovers knee and ankle z-scores for Rodda and Graham gait classification in children.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper develops a markerless pipeline that estimates quantitative z-scores for knee and ankle deviations directly from ordinary clinical gait videos. In a cohort of 152 children with mixed diagnoses, the method reaches R-squared of 0.80 for knees and 0.57 for ankles when checked against 3D instrumented gait analysis. It further supports binary screening for excess knee flexion and partial classification into standard gait patterns. Such video-derived numbers could replace subjective observation and reduce the need for costly lab visits while enabling repeated measurements over time.

Core claim

Across 1,058 bilateral limb samples from 529 trials of 152 children with 60 distinct primary diagnoses, the sagittal-view model achieved R² = 0.80 ± 0.02 and CCC = 0.89 ± 0.02 for knee z-scores and R² = 0.57 ± 0.02 and CCC = 0.72 ± 0.02 for ankle z-scores against 3D-IGA. Binary screening for excess knee flexion achieves AUROC = 0.88, correctly identifying 83% of affected children, and applying Rodda and Graham rules yields 43 ± 1% 7-class accuracy with macro-AUROC = 0.78 ± 0.01, with ankle prediction error remaining the primary bottleneck.

What carries the argument

Markerless gait analysis pipeline that derives 3D kinematics from single-view video and maps them to Rodda and Graham knee and ankle z-scores.

If this is right

  • Continuous z-scores enable longitudinal trajectory tracking across multiple patient visits.
  • Quantitative measures provide a substrate for monitoring disease progression and treatment response.
  • Binary screening identifies 83% of children with excess knee flexion.
  • The approach supports scalable objective gait assessment in low-resource clinical settings without 3D equipment.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Adding a second camera angle could reduce the ankle z-score error that currently limits overall classification accuracy.
  • Routine clinic videos could be processed automatically to create standardized gait records for any child who can walk.
  • The same pipeline might be tested on adults with similar gait deviations to check whether the learned mapping holds outside pediatrics.
  • Error patterns in ankle predictions could guide targeted improvements in how the model extracts foot and shank motion.

Load-bearing premise

A single sagittal-view video contains enough information to recover the specific sagittal-plane deviations that define the Rodda and Graham z-scores, and the learned mapping generalizes to new patients and recording conditions.

What would settle it

Retesting the pipeline on an independent set of 100 children from a different clinic yields R² below 0.6 for knee z-scores or AUROC below 0.75 for excess-flexion screening.

Figures

Figures reproduced from arXiv: 2605.11314 by Anita Bagley, Hyeokhyen Kwon, Jeremy Bauer, Joseph Krzak, Karen Kruger, Lauhitya Reddy, Maura Eveld, Ross Chafetz, Seth Donahue, Susan Sienko, Vedant Kulkarni.

Figure 1
Figure 1. Figure 1: Rodda and Graham z-score space. The horizontal axis represents the ankle z-score (negative: excess plantarflexion; positive: excess dorsiflexion) and the vertical axis represents the knee z-score (negative: hyperextension; positive: excess flexion). The gray region represents z-scores between −1 and +1, considered within the normal range. Silhouettes depict each gait class. CP using 2D pose data captured f… view at source ↗
Figure 2
Figure 2. Figure 2: (a) The sagittal video stream. (b) The frontal video stream. (c) Experimental setup mockup showing the relative positioning of the multi-view recording array used during trials. (d) Monocular 3D pose estimation results across all eight camera viewpoints for a representative child. kinematic deviation measure rather than a diagnosis-specific label. A gait cycle is the sequence of motions that occurs from th… view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the experimental pipeline. Monocular 3D pose estimation extracts keypoints from clinical gait video, which are cleaned and represented as raw coordinates or derived joint angles. Both representations are windowed and processed by deep learning models (DCL, ST-GCN, AGCN, and AGCN+ViT) under participant-wise 5-fold cross-validation for ankle and knee z-score regression. May 13, 2026 8/29 [PITH_F… view at source ↗
Figure 4
Figure 4. Figure 4: Trial-level z-score prediction on AGCN+ViT. Top row, predicted vs. true z-scores for (a) knee and (b) ankle. The black line is the regression fit and the gray line is the identity. Red dashed lines mark ±1 classification boundaries. Bottom row, Bland-Altman plots for (c) knee (bias = −0.16, LoA = [−4.06, 3.74]) and (d) ankle (bias = 0.23, LoA = [−4.21, 4.67]). Rodda and Graham 7-Class Gait Classification W… view at source ↗
Figure 5
Figure 5. Figure 5: Confusion matrix for Rodda and Graham 7-class classification from predicted z-scores (AGCN). reliably identified classifications. The confusion matrix ( [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: ROC curve for binary knee flexion classification (AGCN+ViT). AUROC = 0.88. rising in prevalence with age [47]. The clinical decision threshold for this pattern is a single cut-point on the knee z-score (z > 1), making binary detection of z > 1 the most natural first test of whether our continuous z-score predictions can support a clinically useful triage decision. Thresholding predicted knee z-scores at +1… view at source ↗
Figure 7
Figure 7. Figure 7: Per-bin analysis of trial-level AGCN+ViT z-score predictions. The true z-score range is partitioned into contiguous bins of width 0.5, and mean absolute error (MAE) and 3-class accuracy are computed within each bin. (a) Knee per-bin MAE with label-distribution-smoothed inverse sample density overlay (gray). (b) Knee per-bin 3-class accuracy. (c) Ankle per-bin MAE with inverse sample density overlay. (d) An… view at source ↗
Figure 8
Figure 8. Figure 8: Decile-binned calibration plots for knee (left) and ankle (right). The model line (colored) shows mean predicted vs. mean true z-score per decile, where a perfectly calibrated model would follow the identity (dotted). Discussion Direct Biomechanical Calculation From Monocular Video Fails The Rodda and Graham z-score is, by construction, a deterministic function of sagittal-plane joint angles, and we calcul… view at source ↗
read the original abstract

Cerebral Palsy (CP) is a neurological disorder of movement and the most common cause of lifelong physical disability in childhood. Approximately 75% of children with CP are ambulatory, and accurate gait assessment is central to preserving walking function, which deteriorates by mid-adulthood in a quarter to half of adults with CP. The Rodda and Graham classification system quantifies sagittal-plane gait deviations using ankle and knee z-scores derived from 3D Instrumented Gait Analysis (3D-IGA), but 3D-IGA is expensive and limited to specialized centers, while observational assessment shows only moderate inter-rater agreement. We developed a markerless gait analysis pipeline that quantifies Rodda and Graham knee and ankle z-scores directly from single-view clinical gait videos. Across 1,058 bilateral limb samples from 529 trials of 152 children (88 male, 63 female; age 12.1 $\pm$ 4.0 years; 60 distinct primary diagnoses, cerebral palsy the most common at $n=54$), the sagittal-view model achieved $R^2 = 0.80 \pm 0.02$ and CCC $= 0.89 \pm 0.02$ for knee z-scores and $R^2 = 0.57 \pm 0.02$ and CCC $= 0.72 \pm 0.02$ for ankle z-scores against 3D-IGA. Binary screening for excess knee flexion achieves AUROC $= 0.88$, correctly identifying 83% of affected children, and applying Rodda and Graham rules yields $43 \pm 1$% 7-class accuracy with macro-AUROC $= 0.78 \pm 0.01$, ankle prediction error remaining the primary bottleneck. Beyond cross-sectional screening, continuous z-scores support longitudinal trajectory tracking across visits, providing a quantitative substrate for monitoring disease progression and treatment response unavailable from observational scales. These results demonstrate the feasibility of video-based z-score estimation, excess-flexion screening, and longitudinal trajectory tracking as a path toward scalable, objective gait assessment in low-resource clinical settings.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The manuscript describes the development of a markerless gait analysis pipeline that estimates Rodda and Graham knee and ankle z-scores from single-view clinical gait videos in a heterogeneous cohort of 152 children with 60 distinct primary diagnoses. Using data from 1,058 bilateral limb samples across 529 trials, the sagittal-view model reports R² = 0.80 ± 0.02 and CCC = 0.89 ± 0.02 for knee z-scores and R² = 0.57 ± 0.02 and CCC = 0.72 ± 0.02 for ankle z-scores when compared to 3D-IGA. It also evaluates binary screening for excess knee flexion (AUROC = 0.88) and 7-class classification accuracy (43 ± 1%).

Significance. If the reported performance generalizes, the work could enable more accessible quantitative gait assessment in low-resource clinical settings without 3D-IGA. The heterogeneous cohort spanning 60 diagnoses and the large sample size (1,058 limbs) are strengths that support broader applicability. Reporting continuous z-scores for potential longitudinal tracking, along with multiple metrics (R², CCC, AUROC), provides a useful quantitative substrate beyond observational scales.

major comments (1)
  1. [Abstract] The abstract and results reporting do not specify the cross-validation partitioning strategy (e.g., subject-wise vs. trial-wise splits). With 529 trials drawn from only 152 children, this detail is load-bearing for the central generalization claim: without subject-stratified splits, the model could exploit child-specific gait signatures or session artifacts rather than recovering sagittal-plane deviations from video, undermining the reported R² = 0.80 ± 0.02 and CCC = 0.89 ± 0.02 for knee z-scores.
minor comments (1)
  1. [Abstract] The abstract states performance numbers but omits any description of model architecture, training procedure, or video quality handling; adding these would improve reproducibility assessment without altering the core claims.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive assessment of our work's potential impact and for the constructive major comment. We address it directly below and will revise the manuscript to improve clarity on this point.

read point-by-point responses

  1. Referee: [Abstract] The abstract and results reporting do not specify the cross-validation partitioning strategy (e.g., subject-wise vs. trial-wise splits). With 529 trials drawn from only 152 children, this detail is load-bearing for the central generalization claim: without subject-stratified splits, the model could exploit child-specific gait signatures or session artifacts rather than recovering sagittal-plane deviations from video, undermining the reported R² = 0.80 ± 0.02 and CCC = 0.89 ± 0.02 for knee z-scores.

    Authors: We agree that explicit reporting of the partitioning strategy is essential given the multi-trial structure (529 trials from 152 children). Our experiments used subject-wise partitioning via leave-one-subject-out cross-validation: all trials from any given child were assigned entirely to training or held-out test folds. This design was chosen precisely to prevent leakage of child-specific gait signatures or session artifacts and to support the generalization claims. We will revise the abstract to state: 'using subject-wise cross-validation across 152 children' and will add a corresponding sentence in the Methods and Results sections describing the strategy and its rationale. This change directly addresses the concern without altering the reported metrics. revision: yes

Circularity Check

0 steps flagged

No circularity: predictions evaluated against independent external ground truth

full rationale

The paper trains a markerless pipeline to regress Rodda-Graham knee and ankle z-scores from single-view video and reports R², CCC, and AUROC against 3D-IGA measurements obtained from the same subjects. No equations, fitted parameters, or self-citations are presented that would make the reported z-score outputs definitionally identical to any internal model quantity or training target. The performance numbers are therefore empirical comparisons to an external reference standard rather than tautological self-consistency checks. Because the derivation chain contains no self-definitional, fitted-input-renamed-as-prediction, or load-bearing self-citation steps, the result is self-contained.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Abstract-only review provides limited visibility into modeling choices; the central claim rests on standard supervised learning assumptions plus the domain premise that 2D video kinematics suffice for sagittal z-score recovery. No explicit invented entities or ad-hoc constants are described.

free parameters (1)
  • deep learning model weights and hyperparameters
    Fitted during supervised training on paired video and 3D-IGA data to minimize prediction error on z-scores.
axioms (1)
  • domain assumption Single-view clinical gait videos contain sufficient sagittal-plane information to recover Rodda and Graham z-scores
    Implicit in the claim that the markerless pipeline quantifies the z-scores directly from video without 3D reconstruction.

pith-pipeline@v0.9.0 · 5988 in / 1561 out tokens · 51340 ms · 2026-05-22T10:14:50.923666+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

53 extracted references · 53 canonical work pages

  1. [1]

    A report: the definition and classification of cerebral palsy April 2006

    Rosenbaum P, Paneth N, Leviton A, Goldstein M, Bax M, Damiano D, et al. A report: the definition and classification of cerebral palsy April 2006. Developmental Medicine and Child Neurology Supplement. 2007;109:8–14

    work page 2006

  2. [2]

    An update on the prevalence of cerebral palsy: a systematic review and meta-analysis

    Oskoui M, Coutinho F, Dykeman J, Jett´ e N, Pringsheim T. An update on the prevalence of cerebral palsy: a systematic review and meta-analysis. Developmental Medicine and Child Neurology. 2013;55(6):509–519. doi:10.1111/dmcn.12080

    work page doi:10.1111/dmcn.12080 2013

  3. [3]

    Gait analysis in children with cerebral palsy

    Armand S, Decoulon G, Bonnefoy-Mazure A. Gait analysis in children with cerebral palsy. EFORT Open Reviews. 2016;1(12):448–460. doi:10.1302/2058-5241.1.000052

    work page doi:10.1302/2058-5241.1.000052 2016

  4. [4]

    Walking Ability, Participation, and Quality of Life in Children with Spastic Diplegic Cerebral Palsy: A Path Analysis Study

    Vameghi R, Hoseini SA, Heydarian S, Azadeh H, Gharib M. Walking Ability, Participation, and Quality of Life in Children with Spastic Diplegic Cerebral Palsy: A Path Analysis Study. Iranian Journal of Child Neurology. 2023;17(2):75–91. doi:10.22037/ijcn.v17i1.34924

    work page doi:10.22037/ijcn.v17i1.34924 2023

  5. [5]

    Musculoskeletal Pathology in Cerebral Palsy: A Classification System and Reliability Study

    Graham HK, Thomason P, Willoughby K, Hastings-Ison T, Stralen RV, Dala-Ali B, et al. Musculoskeletal Pathology in Cerebral Palsy: A Classification System and Reliability Study. Children. 2021;8(3):252. doi:10.3390/children8030252

    work page doi:10.3390/children8030252 2021

  6. [6]

    Prognostic predictors for ambulation in children with cerebral palsy: a systematic review and meta-analysis of observational studies

    Keeratisiroj O, Thawinchai N, Siritaratiwat W, Buntragulpoontawee M, Pratoomsoot C. Prognostic predictors for ambulation in children with cerebral palsy: a systematic review and meta-analysis of observational studies. Disability and Rehabilitation. 2018;40(2):135–143. doi:10.1080/09638288.2016.1250119

    work page doi:10.1080/09638288.2016.1250119 2018

  7. [7]

    The Effect of Preoperative Gait Analysis on Orthopaedic Decision Making

    Kay RM, Dennis S, Rethlefsen S, Reynolds RAK, Skaggs DL, Tolo VT. The Effect of Preoperative Gait Analysis on Orthopaedic Decision Making. Clinical Orthopaedics and Related Research. 2000;372:217–222. doi:10.1097/00003086-200003000-00023

    work page doi:10.1097/00003086-200003000-00023 2000

  8. [8]

    Clinical gait analysis 1973–2023: Evaluating progress to guide the future

    Stebbins J, Harrington M, Stewart C. Clinical gait analysis 1973–2023: Evaluating progress to guide the future. Journal of Biomechanics. 2023;160:111827. doi:10.1016/j.jbiomech.2023.111827

    work page doi:10.1016/j.jbiomech.2023.111827 1973

  9. [9]

    Efficacy of clinical gait analysis: A systematic review

    Wren TAL, Gorton GE, ˜Ounpuu S, Tucker CA. Efficacy of clinical gait analysis: A systematic review. Gait & Posture. 2011;34(2):149–153. doi:10.1016/j.gaitpost.2011.03.027

    work page doi:10.1016/j.gaitpost.2011.03.027 2011

  10. [10]

    Classification of gait patterns in spastic hemiplegia and spastic diplegia: a basis for a management algorithm

    Rodda J, Graham HK. Classification of gait patterns in spastic hemiplegia and spastic diplegia: a basis for a management algorithm. European Journal of Neurology. 2001;8 Suppl 5:98–108. doi:10.1046/j.1468-1331.2001.00042.x. May 22, 2026 25/30

    work page doi:10.1046/j.1468-1331.2001.00042.x 2001

  11. [11]

    Sagittal gait patterns in cerebral palsy: The plantarflexor–knee extension couple index

    Sangeux M, Rodda J, Graham HK. Sagittal gait patterns in cerebral palsy: The plantarflexor–knee extension couple index. Gait & Posture. 2015;41(2):586–591. doi:10.1016/j.gaitpost.2014.12.019

    work page doi:10.1016/j.gaitpost.2014.12.019 2015

  12. [12]

    Sagittal gait patterns in spastic diplegia

    Rodda JM, Graham HK, Carson L, Galea MP, Wolfe R. Sagittal gait patterns in spastic diplegia. The Journal of Bone and Joint Surgery British Volume. 2004;86-B(2):251–258. doi:10.1302/0301-620x.86b2.13878

    work page doi:10.1302/0301-620x.86b2.13878 2004

  13. [13]

    The Shriners Children’s Gait Model (SCGM)

    Kruger KM, Fischer P, Augsburger S, Feng J, Girouard JF, Gregory DL, et al. The Shriners Children’s Gait Model (SCGM). Gait & Posture. 2024;110:84–109. doi:10.1016/j.gaitpost.2024.03.006

    work page doi:10.1016/j.gaitpost.2024.03.006 2024

  14. [14]

    Reliability of visual classification of sagittal gait patterns in patients with bilateral spastic cerebral palsy

    Kim DJ, Park ES, Sim EG, Kim KJ, Kim YU, Rha DW. Reliability of visual classification of sagittal gait patterns in patients with bilateral spastic cerebral palsy. Annals of Rehabilitation Medicine. 2011;35(3):354–360. doi:10.5535/arm.2011.35.3.354

    work page doi:10.5535/arm.2011.35.3.354 2011

  15. [15]

    A review of observational gait assessment in clinical practice

    Toro B, Nester C, Farren P. A review of observational gait assessment in clinical practice. Physiotherapy Theory and Practice. 2003;19(3):137–149. doi:10.1080/09593980307964

    work page doi:10.1080/09593980307964 2003

  16. [16]

    Interrater reliability of videotaped observational gait-analysis assessments

    Eastlack ME, Arvidson J, Snyder-Mackler L, Danoff JV, McGarvey CL. Interrater reliability of videotaped observational gait-analysis assessments. Physical Therapy. 1991;71(6):465–472. doi:10.1093/ptj/71.6.465

    work page doi:10.1093/ptj/71.6.465 1991

  17. [17]

    Application of Deep Learning Models in Classification of Sagittal Gait Patterns Based on Rodda’s Classification System in Patients with Cerebral Palsy

    Mazidi AA, Banihashemi M, Jamshidian A, Shojaeefard M, Taheri A, Farahmand F. Application of Deep Learning Models in Classification of Sagittal Gait Patterns Based on Rodda’s Classification System in Patients with Cerebral Palsy. In: 2024 31st National and 9th International Iranian Conference on Biomedical Engineering (ICBME); 2024. p. 289–295

    work page 2024

  18. [18]

    Kim YK, Visscher RMS, Viehweger E, Singh NB, Taylor WR, Vogl F. A deep-learning approach for automatically detecting gait-events based on foot-marker kinematics in children with cerebral palsy—Which markers work best for which gait patterns? PLOS ONE. 2022;17(10):e0275878. doi:10.1371/journal.pone.0275878

    work page doi:10.1371/journal.pone.0275878 2022

  19. [19]

    Deep neural networks enable quantitative movement analysis using single-camera videos

    Kidzi´ nski L, Yang B, Hicks JL, Rajagopal A, Delp SL, Schwartz MH. Deep neural networks enable quantitative movement analysis using single-camera videos. Nature Communications. 2020;11(1):4054. doi:10.1038/s41467-020-17807-z

    work page doi:10.1038/s41467-020-17807-z 2020

  20. [20]

    OpenPose: Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields

    Cao Z, Hidalgo Martinez G, Simon T, Wei SE, Sheikh YA. OpenPose: Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2021;43(1):172–186. doi:10.1109/TPAMI.2019.2929257

    work page doi:10.1109/tpami.2019.2929257 2021

  21. [21]

    Pantzar-Castilla E, Balta D, Croce UD, Cereatti A, Riad J. Feasibility and usefulness of video-based markerless two-dimensional automated gait analysis, in providing objective quantification of gait and complementing the evaluation of gait in children with cerebral palsy. BMC Musculoskeletal Disorders. 2024;25(1):747. doi:10.1186/s12891-024-07853-9

    work page doi:10.1186/s12891-024-07853-9 2024

  22. [22]

    Motor Function Assessment of Children with Cerebral Palsy using Monocular Video

    Zhao P, Alencastre-Miranda M, Shen Z, O’Neill C, Whiteman D, Gervas-Arruga J. Motor Function Assessment of Children with Cerebral Palsy using Monocular Video. In: 2023 IEEE 19th International Conference on Body Sensor Networks (BSN); 2023. p. 1–4. May 22, 2026 26/30

    work page 2023

  23. [23]

    Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition

    Yan S, Xiong Y, Lin D. Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 32; 2018. p. 7444–7452

    work page 2018

  24. [24]

    Algorithm based on one monocular video delivers highly valid and reliable gait parameters

    Azhand A, Rabe S, M¨ uller S, Sattler I, Heimann-Steinert A. Algorithm based on one monocular video delivers highly valid and reliable gait parameters. Scientific Reports. 2021;11(1):14065. doi:10.1038/s41598-021-93530-z

    work page doi:10.1038/s41598-021-93530-z 2021

  25. [25]

    Development and reliability of a system to classify gross motor function in children with cerebral palsy

    Palisano R, Rosenbaum P, Walter S, Russell D, Wood E, Galuppi B. Development and reliability of a system to classify gross motor function in children with cerebral palsy. Developmental Medicine & Child Neurology. 1997;39(4):214–223. doi:10.1111/j.1469-8749.1997.tb07414.x

    work page doi:10.1111/j.1469-8749.1997.tb07414.x 1997

  26. [26]

    Variation in kinematic and spatiotemporal gait parameters by Gross Motor Function Classification System level in children and adolescents with cerebral palsy

    ˜Ounpuu S, Gorton G, Bagley A, Sison-Williamson M, Hassani S, Johnson B, et al. Variation in kinematic and spatiotemporal gait parameters by Gross Motor Function Classification System level in children and adolescents with cerebral palsy. Developmental Medicine & Child Neurology. 2015;57(10):955–962. doi:10.1111/dmcn.12766

    work page doi:10.1111/dmcn.12766 2015

  27. [27]

    Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition

    Ord´ o˜ nez FJ, Roggen D. Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition. Sensors. 2016;16(1):115. doi:10.3390/s16010115

    work page doi:10.3390/s16010115 2016

  28. [28]

    Two-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition

    Shi L, Zhang Y, Cheng J, Lu H. Two-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    work page

  29. [29]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

    Dosovitskiy A, Beyer L, Kolesnikov A, Weissenborn D, Zhai X, Unterthiner T, et al. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. In: International Conference on Learning Representations (ICLR); 2021

    work page 2021

  30. [30]

    MeTRAbs: Metric-Scale Truncation-Robust Heatmaps for Absolute 3D Human Pose Estimation

    S´ ar´ andi I, Linder T, Arras KO, Leibe B. MeTRAbs: Metric-Scale Truncation-Robust Heatmaps for Absolute 3D Human Pose Estimation. IEEE Transactions on Biometrics, Behavior, and Identity Science. 2021;3(1):16–30. doi:10.1109/TBIOM.2020.3037257

    work page doi:10.1109/tbiom.2020.3037257 2021

  31. [31]

    Learning 3D Human Pose Estimation from Dozens of Datasets using a Geometry-Aware Autoencoder to Bridge Between Skeleton Formats

    S´ ar´ andi I, Hermans A, Leibe B. Learning 3D Human Pose Estimation from Dozens of Datasets using a Geometry-Aware Autoencoder to Bridge Between Skeleton Formats. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV); 2023. p. 2955–2965

    work page 2023

  32. [32]

    MoVi: A large multi-purpose human motion and video dataset

    Ghorbani S, Mahdaviani K, Thaler A, Kording K, Cook DJ, Blohm G, et al. MoVi: A large multi-purpose human motion and video dataset. PLOS ONE. 2021;16(6):e0253157. doi:10.1371/journal.pone.0253157

    work page doi:10.1371/journal.pone.0253157 2021

  33. [33]

    The frequency content of gait

    Antonsson EK, Mann RW. The frequency content of gait. Journal of Biomechanics. 1985;18(1):39–47. doi:10.1016/0021-9290(85)90043-0

    work page doi:10.1016/0021-9290(85)90043-0 1985

  34. [34]

    ISB recommendation on definitions of joint coordinate system of various joints for the reporting of human joint motion—part I: ankle, hip, and spine

    Wu G, Siegler S, Allard P, Kirtley C, Leardini A, Rosenbaum D, et al. ISB recommendation on definitions of joint coordinate system of various joints for the reporting of human joint motion—part I: ankle, hip, and spine. Journal of Biomechanics. 2002;35(4):543–548. doi:10.1016/S0021-9290(01)00222-6

    work page doi:10.1016/s0021-9290(01)00222-6 2002

  35. [35]

    Gait Analysis: Normal and Pathological Function

    Perry J, Burnfield JM. Gait Analysis: Normal and Pathological Function. 2nd ed. Thorofare, NJ: SLACK Incorporated; 2010. May 22, 2026 27/30

    work page 2010

  36. [36]

    Gait analysis methods in rehabilitation

    Baker R. Gait analysis methods in rehabilitation. Journal of NeuroEngineering and Rehabilitation. 2006;3:4. doi:10.1186/1743-0003-3-4

    work page doi:10.1186/1743-0003-3-4 2006

  37. [37]

    A gait analysis data collection and reduction technique

    Davis RB III, ˜Ounpuu S, Tyburski D, Gage JR. A gait analysis data collection and reduction technique. Human Movement Science. 1991;10(5):575–587. doi:10.1016/0167-9457(91)90046-Z

    work page doi:10.1016/0167-9457(91)90046-z 1991

  38. [38]

    Measurement of lower extremity kinematics during level walking

    Kadaba MP, Ramakrishnan HK, Wootten ME. Measurement of lower extremity kinematics during level walking. Journal of Orthopaedic Research. 1990;8(3):383–392. doi:10.1002/jor.1100080310

    work page doi:10.1002/jor.1100080310 1990

  39. [39]

    An Explainable Spatial-Temporal Graphical Convolutional Network to Score Freezing of Gait in Parkinsonian Patients

    Kwon H, Clifford GD, Genias I, Bernhard D, Esper CD, Factor SA, et al. An Explainable Spatial-Temporal Graphical Convolutional Network to Score Freezing of Gait in Parkinsonian Patients. Sensors. 2023;23(4):1766. doi:10.3390/s23041766

    work page doi:10.3390/s23041766 2023

  40. [40]

    Semi-Supervised Classification with Graph Convolutional Networks

    Kipf TN, Welling M. Semi-Supervised Classification with Graph Convolutional Networks. In: International Conference on Learning Representations (ICLR); 2017

    work page 2017

  41. [41]

    Crouch Gait in Cerebral Palsy: Current Concepts Review

    Pandey RA, Johari AN, Shetty T. Crouch Gait in Cerebral Palsy: Current Concepts Review. Indian Journal of Orthopaedics. 2023;57(12):1913–1926. doi:10.1007/s43465-023-01002-5

    work page doi:10.1007/s43465-023-01002-5 2023

  42. [42]

    Are Clinical Impairments Related to Kinematic Gait Variability in Children and Young Adults With Cerebral Palsy? Frontiers in Human Neuroscience

    Tabard-Foug` ere A, Rutz D, Pouliot-Laforte A, De Coulon G, Newman CJ, Armand S, et al. Are Clinical Impairments Related to Kinematic Gait Variability in Children and Young Adults With Cerebral Palsy? Frontiers in Human Neuroscience. 2022;16:816088. doi:10.3389/fnhum.2022.816088

    work page doi:10.3389/fnhum.2022.816088 2022

  43. [43]

    Long short -term memory,

    Hochreiter S, Schmidhuber J. Long Short-Term Memory. Neural Computation. 1997;9(8):1735–1780. doi:10.1162/neco.1997.9.8.1735

    work page doi:10.1162/neco.1997.9.8.1735 1997

  44. [44]

    Deep, Convolutional, and Recurrent Models for Human Activity Recognition Using Wearables

    Hammerla NY, Halloran S, Pl¨ otz T. Deep, Convolutional, and Recurrent Models for Human Activity Recognition Using Wearables. In: Proceedings of the 25th International Joint Conference on Artificial Intelligence (IJCAI); 2016. p. 1533–1540

    work page 2016

  45. [45]

    Optuna: A Next-generation Hyperparameter Optimization Framework

    Akiba T, Sano S, Yanase T, Ohta T, Koyama M. Optuna: A Next-generation Hyperparameter Optimization Framework. In: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD); 2019. p. 2623–2631

    work page 2019

  46. [46]

    Adam: A Method for Stochastic Optimization

    Kingma DP, Ba J. Adam: A Method for Stochastic Optimization. In: International Conference on Learning Representations (ICLR); 2015

    work page 2015

  47. [47]

    Prevalence of specific gait abnormalities in children with cerebral palsy revisited: influence of age, prior surgery, and Gross Motor Function Classification System level

    Rethlefsen SA, Blumstein G, Kay RM, Dorey F, Wren TAL. Prevalence of specific gait abnormalities in children with cerebral palsy revisited: influence of age, prior surgery, and Gross Motor Function Classification System level. Developmental Medicine & Child Neurology. 2017;59(1):79–88. doi:10.1111/dmcn.13205

    work page doi:10.1111/dmcn.13205 2017

  48. [48]

    Delving into Deep Imbalanced Regression

    Yang Y, Zha K, Chen Y, Wang H, Katabi D. Delving into Deep Imbalanced Regression. In: Proceedings of the 38th International Conference on Machine Learning (ICML). vol. 139 of Proceedings of Machine Learning Research. PMLR

    work page

  49. [49]

    11842–11851

    p. 11842–11851. May 22, 2026 28/30

    work page 2026

  50. [50]

    Assessing the Performance of Prediction Models: A Framework for Traditional and Novel Measures

    Steyerberg EW, Vickers AJ, Cook NR, Gerds T, Gonen M, Obuchowski N, et al. Assessing the Performance of Prediction Models: A Framework for Traditional and Novel Measures. Epidemiology. 2010;21(1):128–138. doi:10.1097/EDE.0b013e3181c30fb7

    work page doi:10.1097/ede.0b013e3181c30fb7 2010

  51. [51]

    Concurrent assessment of gait kinematics using marker-based and markerless motion capture

    Kanko RM, Laende EK, Davis EM, Selbie WS, Deluzio KJ. Concurrent assessment of gait kinematics using marker-based and markerless motion capture. Journal of Biomechanics. 2021;127:110665. doi:10.1016/j.jbiomech.2021.110665

    work page doi:10.1016/j.jbiomech.2021.110665 2021

  52. [52]

    Using deep neural networks for kinematic analysis: Challenges and opportunities

    Cronin NJ. Using deep neural networks for kinematic analysis: Challenges and opportunities. Journal of Biomechanics. 2021;123:110460. doi:10.1016/j.jbiomech.2021.110460

    work page doi:10.1016/j.jbiomech.2021.110460 2021

  53. [53]

    CHEF-VL: Detecting Cognitive Sequencing Errors in Cooking with Vision-Language Models

    Wang R, Gao P, Lynch P, Liu T, Lee Y, Baum CM, et al. CHEF-VL: Detecting Cognitive Sequencing Errors in Cooking with Vision-Language Models. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies. 2025;doi:10.1145/3770714. May 22, 2026 29/30 Supplementary Supplementary T able 1. F ull primary-diagnosis breakdown of the 152-chi...

    work page doi:10.1145/3770714 2025