{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T15:19:22Z","timestamp":1784819962334,"version":"3.55.0"},"reference-count":38,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2023,3,2]],"date-time":"2023-03-02T00:00:00Z","timestamp":1677715200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Robot. AI"],"abstract":"<jats:p><jats:bold>Introduction:<\/jats:bold> Video-based clinical rating plays an important role in assessing dystonia and monitoring the effect of treatment in dyskinetic cerebral palsy (CP). However, evaluation by clinicians is time-consuming, and the quality of rating is dependent on experience. The aim of the current study is to provide a proof-of-concept for a machine learning approach to automatically assess scoring of dystonia using 2D stick figures extracted from videos. Model performance was compared to human performance.<\/jats:p><jats:p><jats:bold>Methods:<\/jats:bold> A total of 187 video sequences of 34 individuals with dyskinetic CP (8\u201323\u00a0years, all non-ambulatory) were filmed at rest during lying and supported sitting. Videos were scored by three raters according to the Dyskinesia Impairment Scale (DIS) for arm and leg dystonia (normalized scores ranging from 0\u20131). Coordinates in pixels of the left and right wrist, elbow, shoulder, hip, knee and ankle were extracted using DeepLabCut, an open source toolbox that builds on a pose estimation algorithm. Within a subset, tracking accuracy was assessed for a pretrained human model and for models trained with an increasing number of manually labeled frames. The mean absolute error (MAE) between DeepLabCut\u2019s prediction of the position of body points and manual labels was calculated. Subsequently, movement and position features were calculated from extracted body point coordinates. These features were fed into a Random Forest Regressor to train a model to predict the clinical scores. The model performance trained with data from one rater evaluated by MAEs (model-rater) was compared to inter-rater accuracy.<\/jats:p><jats:p><jats:bold>Results:<\/jats:bold> A tracking accuracy of 4.5 pixels (approximately 1.5\u00a0cm) could be achieved by adding 15\u201320 manually labeled frames per video. The MAEs for the trained models ranged from 0.21 \u00b1 0.15 for arm dystonia to 0.14 \u00b1 0.10 for leg dystonia (normalized DIS scores). The inter-rater MAEs were 0.21 \u00b1 0.22 and 0.16 \u00b1 0.20, respectively.<\/jats:p><jats:p><jats:bold>Conclusion:<\/jats:bold> This proof-of-concept study shows the potential of using stick figures extracted from common videos in a machine learning approach to automatically assess dystonia. Sufficient tracking accuracy can be reached by manually adding labels within 15\u201320 frames per video. With a relatively small data set, it is possible to train a model that can automatically assess dystonia with a performance comparable to human scoring.<\/jats:p>","DOI":"10.3389\/frobt.2023.1108114","type":"journal-article","created":{"date-parts":[[2023,3,2]],"date-time":"2023-03-02T06:20:19Z","timestamp":1677738019000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":24,"title":["Towards automated video-based assessment of dystonia in dyskinetic cerebral palsy: A novel approach using markerless motion tracking and machine learning"],"prefix":"10.3389","volume":"10","author":[{"given":"Helga","family":"Haberfehlner","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shankara S.","family":"van de Ven","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sven A.","family":"van der Burg","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Florian","family":"Huber","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sonja","family":"Georgievska","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ignazio","family":"Aleo","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jaap","family":"Harlaar","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Laura A.","family":"Bonouvri\u00e9","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Marjolein M.","family":"van der Krogt","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Annemieke I.","family":"Buizer","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2023,3,2]]},"reference":[{"key":"B1","doi-asserted-by":"crossref","DOI":"10.1109\/CVPR.2014.471","article-title":"2D human pose estimation: New benchmark and state of the art analysis","author":"Andriluka","year":"2014"},{"key":"B2","volume-title":"MPII human pose dataset","author":"Andriluka","year":"2014"},{"key":"B3","doi-asserted-by":"publisher","first-page":"1612","DOI":"10.1212\/01.wnl.0000036274.36543.c5","article-title":"Arthur van Gehuchten takes neurology to the movies","volume":"59","author":"Aubert","year":"2002","journal-title":"Neurology"},{"key":"B4","doi-asserted-by":"publisher","first-page":"2082","DOI":"10.1212\/01.wnl.0000127488.13208.1c","article-title":"The origins of scientific cinematography and early medical applications","volume":"62","author":"Barboi","year":"2004","journal-title":"Neurology"},{"key":"B5","doi-asserted-by":"publisher","first-page":"118","DOI":"10.1016\/j.ejpn.2020.08.002","article-title":"Dystonia and choreoathetosis presence and severity in relation to powered wheelchair mobility performance in children and youth with dyskinetic cerebral palsy","volume":"29","author":"Bekteshi","year":"2020","journal-title":"Eur. J. Paediatr. Neurol."},{"key":"B6","doi-asserted-by":"publisher","first-page":"37","DOI":"10.5334\/tohm.651","article-title":"The clinical value of patient home videos in movement disorders","volume":"11","author":"Billnitzer","year":"2021","journal-title":"Tremor Other Hyperkinet Mov. (N Y)."},{"key":"B7","doi-asserted-by":"publisher","first-page":"79","DOI":"10.1002\/ana.25498","article-title":"The effect of intrathecal baclofen in dyskinetic cerebral palsy: The IDYS trial","volume":"86","author":"Bonouvrie","year":"2019","journal-title":"Ann. Neurol."},{"key":"B8","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1080\/09638288.2022.2057600","article-title":"Attainment of personal goals in the first year of intrathecal baclofen treatment in dyskinetic cerebral palsy: A prospective cohort study","author":"Bonouvrie","year":"2022","journal-title":"Disabil. Rehabil."},{"key":"B34","first-page":"1302","article-title":"Realtime multi-person 2D pose estimation using part affinity fields","author":"Cao","year":"2017"},{"key":"B9","doi-asserted-by":"crossref","DOI":"10.1109\/CVPRW53098.2021.00265","article-title":"MobileHumanPose: Toward real-time 3D human pose estimation in mobile devices","volume-title":"Proceedings of the IEEE\/CVF conference on computer vision and pattern recognition","author":"Choi","year":"2021"},{"key":"B10","article-title":"Vox populi: Collecting high-quality labels from a crowd","volume-title":"Annual conference computational learning theory 2009","author":"Dekel","year":"2009"},{"key":"B35","doi-asserted-by":"publisher","first-page":"103042","DOI":"10.1016\/j.humov.2022.103042","article-title":"Perspective on \u201cin the wild\u201d movement analysis using machine learning","volume":"87","author":"Dorschky","year":"2023","journal-title":"Hum. Mov. Sci."},{"key":"B11","doi-asserted-by":"publisher","first-page":"549","DOI":"10.1111\/j.1469-8749.2006.tb01313.x","article-title":"The manual ability classification System (MACS) for children with cerebral palsy: Scale development and evidence of validity and reliability","volume":"48","author":"Eliasson","year":"2006","journal-title":"Dev. Med. Child. Neurol."},{"key":"B12","doi-asserted-by":"publisher","first-page":"2398","DOI":"10.1002\/mds.22341","article-title":"The unified dyskinesia rating scale: Presentation and clinimetric profile","volume":"23","author":"Goetz","year":"2008","journal-title":"Mov. Disord."},{"key":"B13","doi-asserted-by":"publisher","first-page":"e2221325","DOI":"10.1001\/jamanetworkopen.2022.21325","article-title":"Development and validation of a deep learning method to predict cerebral palsy from spontaneous movements in infants at high risk","volume":"5","author":"Groos","year":"2022","journal-title":"JAMA Netw. Open"},{"key":"B14","volume-title":"MODYS-Video: 2D human pose estimation data and Dyskinesia impairment scale scores from children and young adults with dyskinetic cerebral palsy","author":"Haberfehlner","year":"2021"},{"key":"B15","doi-asserted-by":"publisher","first-page":"39","DOI":"10.1186\/s12984-020-00658-6","article-title":"Instrumented assessment of motor function in dyskinetic cerebral palsy: A systematic review","volume":"17","author":"Haberfehlner","year":"2020","journal-title":"J. Neuroeng Rehabil."},{"key":"B36","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1016\/j.jbiomech.2018.09.009","article-title":"Machine learning in human movement biomechanics: Best practices, common pitfalls, and new opportunities","volume":"81","author":"Halilaj","year":"2018","journal-title":"J. Biomech."},{"key":"B16","doi-asserted-by":"publisher","first-page":"921","DOI":"10.1136\/adc.2008.144014","article-title":"Dyskinetic cerebral palsy in Europe: Trends in prevalence and severity","volume":"94","author":"Himmelmann","year":"2009","journal-title":"Arch. Dis. Child."},{"key":"B17","first-page":"34","article-title":"DeeperCut: A deeper, stronger, and faster multi-person pose estimation model","volume-title":"European conference on computer vision","author":"Insafutdinov","year":"2016"},{"key":"B37","doi-asserted-by":"publisher","first-page":"441","DOI":"10.1080\/03091902.2020.1822940","article-title":"Role of machine learning in gait analysis: a review","volume":"44","author":"Khera","year":"2020","journal-title":"J. Med. Eng. Technol."},{"key":"B18","doi-asserted-by":"publisher","first-page":"4054","DOI":"10.1038\/s41467-020-17807-z","article-title":"Deep neural networks enable quantitative movement analysis using single-camera videos","volume":"11","author":"Kidzinski","year":"2020","journal-title":"Nat. Commun."},{"key":"B19","doi-asserted-by":"publisher","first-page":"118","DOI":"10.1016\/j.ejpn.2016.05.015","article-title":"Deep brain stimulation in cerebral palsy: Challenges and opportunities","volume":"21","author":"Koy","year":"2017","journal-title":"Eur. J. Paediatr. Neurol."},{"key":"B20","doi-asserted-by":"publisher","first-page":"5314","DOI":"10.3390\/s21165314","article-title":"Student behavior recognition System for the classroom environment based on skeleton pose estimation and person detection","volume":"21","author":"Lin","year":"2021","journal-title":"Sensors (Basel)."},{"key":"B21","doi-asserted-by":"publisher","first-page":"1281","DOI":"10.1038\/s41593-018-0209-y","article-title":"DeepLabCut: Markerless pose estimation of user-defined body parts with deep learning","volume":"21","author":"Mathis","year":"2018","journal-title":"Nat. Neurosci."},{"key":"B22","doi-asserted-by":"publisher","first-page":"327","DOI":"10.1016\/j.ejpn.2016.09.003","article-title":"The relationship of dystonia and choreoathetosis with activity, participation and quality of life in children and youth with dyskinetic cerebral palsy","volume":"21","author":"Monbaliu","year":"2017","journal-title":"Eur. J. Paediatr. Neurology"},{"key":"B23","doi-asserted-by":"publisher","first-page":"138","DOI":"10.1111\/dmcn.12846","article-title":"Clinical patterns of dystonia and choreoathetosis in participants with dyskinetic cerebral palsy","volume":"58","author":"Monbaliu","year":"2016","journal-title":"Dev. Med. Child. Neurol."},{"key":"B24","doi-asserted-by":"publisher","first-page":"741","DOI":"10.1016\/s1474-4422(17)30252-1","article-title":"Clinical presentation and management of dyskinetic cerebral palsy","volume":"16","author":"Monbaliu","year":"2017","journal-title":"Lancet Neurol."},{"key":"B25","doi-asserted-by":"publisher","first-page":"278","DOI":"10.1111\/j.1469-8749.2011.04209.x","article-title":"The dyskinesia impairment scale: A new instrument to measure dystonia and choreoathetosis in dyskinetic cerebral palsy","volume":"54","author":"Monbaliu","year":"2012","journal-title":"Dev. Med. Child. Neurol."},{"key":"B26","doi-asserted-by":"publisher","first-page":"2152","DOI":"10.1038\/s41596-019-0176-0","article-title":"Using DeepLabCut for 3D markerless pose estimation across species and behaviors","volume":"14","author":"Nath","year":"2019","journal-title":"Nat. Protoc."},{"key":"B38","article-title":"PoseNet 2.0. Real-time human pose estimation in the browser with TensorFlow.js \u2014 The TensorFlow blog","author":"Oved","year":"2018"},{"key":"B27","doi-asserted-by":"publisher","first-page":"744","DOI":"10.1111\/j.1469-8749.2008.03089.x","article-title":"Content validity of the expanded and revised Gross motor function classification System","volume":"50","author":"Palisano","year":"2008","journal-title":"Dev. Med. Child. Neurol."},{"key":"B28","first-page":"2825","article-title":"Scikit-learn: Machine learning in Python","volume":"12","author":"Pedregosa","year":"2011","journal-title":"J. Mach. Learn Res."},{"key":"B29","doi-asserted-by":"publisher","first-page":"2453","DOI":"10.1002\/mds.23536","article-title":"A film of patients with movement disorders made in queen square, London in the mid-1920s by samuel alexander kinnier wilson","volume":"26","author":"Reynolds","year":"2011","journal-title":"Mov. Disord."},{"key":"B30","doi-asserted-by":"publisher","first-page":"333","DOI":"10.1007\/s10072-018-3639-0","article-title":"Observing movement disorders: Best practice proposal in the use of video recording in clinical practice","volume":"40","author":"Sambati","year":"2019","journal-title":"Neurol. Sci."},{"key":"B31","doi-asserted-by":"publisher","first-page":"551","DOI":"10.1177\/08830738060210070201","article-title":"Arm trajectories in dyskinetic cerebral palsy have increased random variability","volume":"21","author":"Sanger","year":"2006","journal-title":"J. Child. Neurol."},{"key":"B32","doi-asserted-by":"publisher","first-page":"205","DOI":"10.3390\/s21010205","article-title":"Skeleton-based emotion recognition based on two-stream self-attention enhanced spatial-temporal graph convolutional network","volume":"21","author":"Shi","year":"2020","journal-title":"Sensors (Basel)"},{"key":"B33","article-title":"Automated video-based assessment of dyskinetic cerebral palsy (v0.1.1)","volume-title":"Zenodo","author":"van de Ven","year":"2021"}],"container-title":["Frontiers in Robotics and AI"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2023.1108114\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,3,2]],"date-time":"2023-03-02T06:20:25Z","timestamp":1677738025000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2023.1108114\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,3,2]]},"references-count":38,"alternative-id":["10.3389\/frobt.2023.1108114"],"URL":"https:\/\/doi.org\/10.3389\/frobt.2023.1108114","relation":{},"ISSN":["2296-9144"],"issn-type":[{"value":"2296-9144","type":"electronic"}],"subject":[],"published":{"date-parts":[[2023,3,2]]},"article-number":"1108114"}}