[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-127413-en":3,"doc-seo-127413-105":29,"detail-sidebar-cat-0-en-105":94},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":11,"language":21,"language_code":22,"site_id":23,"html_lang":22,"table_of_contents":24,"faqs":25,"seo_title":26,"seo_description":14,"update_tm":27,"read_time":28},127413,962085564807,"Aurelia","https://ap-avatar.wpscdn.com/davatar_6f874abed73319feea01a86fa6f0fab8",8,"Research & Report","Machine learning is better than surgeons at assessing unicompartmental knee replacement radiographs","Poor results can occur after unicompartmental knee replacement (UKR), and interpreting post-operative radiographs to explain those outcomes remains challenging even for experienced surgeons. This study compares surgeons and a machine-learning model in predicting whether patients will have poor or excellent one-year outcomes from radiographs. A transfer-learning ResNet50v2 is trained on 924 images and then tested on new radiographs to evaluate classification performance and derive radiographic cues linked to outcome.","Machine learning is better than surgeons at assessing unicompartmental knee replacement radiographs  \nS Jack Tu a, ⇑, Sara Kendrick b, Karthik Saravanan a, Christopher Dodd c, David W Murray a,c, Stephen J Mellon a  \na Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, Windmill Road, Oxford OX3 7LD, United Kingdom b Indiana University School of Medicine, 340 W. 10th Street Fairbanks Hall., Indianapolis, IN 46202, United States  \nc Oxford University Hospitals NHS Foundation Trust, Nuffield Orthopaedic Centre, Old Road, Oxford OX3 7HE, United Kingdom  \na r t i c l e i n f o  \nArticle history:  \nReceived 6 August 2024 Revised 4 November 2024 Accepted 8 November 2024  \nKeywords:  \nConvolutional Neural Network (CNN)  \nTransfer learning Radiograph Clinical outcomes  \na b s t r a c t  \nBackground:: Poor results occasionally occur after unicompartmental knee replacement (UKR). It is often difficult, even for experienced surgeons, to determine why patients have poor outcomes from radiographs. The aim was to compare the ability of experienced surgeons and machine learning to predict whether patients had poor or excellent outcomes from radiographs.  \nMethods:: 924 one-year anterior-posterior radiographs post-UKR were used to train a machine learning model (ResNet50v2) with a transfer learning approach based on their one-year Oxford Knee Score categories. Two experienced surgeons and the model assessed and categorised 70 radiographs (14 Poor scores; 56 Excellent scores) not used for training according to their expected outcome.  \nResults:: The ResNet50v2 model correctly identified 71%(n = 10) of the patients with a poor score and 46 (82%) of those with an excellent score. In contrast, one surgeon could not identify patients with Poor scores (0%) and the other identified one (7%). Both misidentified 3 of those with Excellent scores. The model visualisation method suggested that estimated classifications were made from image features around the implants.  \nConclusion:: The results suggest that there are radiographical features that relate to poor outcomes, which the surgeons are unaware of. Those the model did not identify may have an extra-articular cause for their poor outcome. Further analysis to identify the features associated with poor outcomes could potentially suggest ways that indications or techniques could be improved so as to decrease the incidence of poor results.  \n© 2024 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ([http://creativecommons.org/licenses/by-nc-nd/4.0/](http://creativecommons.org/licenses/by-nc-nd/4.0/)).  \n⇑ Corresponding author.  \nE-mail addresses: [jack.tu@ndorms.ox.ac.uk](jack.tu@ndorms.ox.ac.uk) (S Jack Tu), [stephen.mellon@ndorms.ox.ac.uk](stephen.mellon@ndorms.ox.ac.uk) (S.J Mellon).  \n[https://doi.org/10.1016/j.knee.2024.11.007](https://doi.org/10.1016/j.knee.2024.11.007)  \n0968-0160/© 2024 The Author(s). Published by Elsevier B.V.  \nThis is an open access article under the CC BY-NC-ND license ([http://creativecommons.org/licenses/by-nc-nd/4.0/](http://creativecommons.org/licenses/by-nc-nd/4.0/)).  \nSo what does this mean for the knee surgeon?  \nIt is difficult for surgeons to predict whether a patient will have a poor outcome from radiographs taken after knee replacements or if they have a poor outcome to determine why it occurred. We found that a trained machine learning (ML) model was much better than surgeons at predicting which patients will have a poor outcome following Unicompartmental knee replacement and that it based its decision on image features near the implants and medial tibial spine. Further analysis should allow us to identify these features and modify the operation to decrease the risk of a poor outcome. This approach is different from how ML is usually used. Instead of replicating tasks we can already do, it is identifying unknown image features. If rolled out on a national scale, based","cbCaigD7mnkxFNOx","https://ap.wps.com/l/cbCaigD7mnkxFNOx","pdf",1216826,1,"English","en",105,"# Background and aim\n## Study methods and model training\n## Assessment procedure and test setup\n## Results and performance comparison\n## Interpretation and visualisation findings\n## Conclusion and clinical implications\n# Introduction\n## Clinical challenge after UKR and revisions\n## Rationale for deep convolutional neural networks","[{\"question\":\"What problem does the study address after unicompartmental knee replacement?\",\"answer\":\"Poor outcomes sometimes occur after UKR, and it can be difficult for surgeons to use radiographs to determine why those poor outcomes happened, limiting treatment decisions.\"},{\"question\":\"How is the machine learning model trained and evaluated?\",\"answer\":\"The model uses 924 one-year anterior-posterior radiographs to train a ResNet50v2 via transfer learning based on one-year Oxford Knee Score categories, then assesses 70 new radiographs not used for training.\"},{\"question\":\"How do the surgeons’ predictions compare with the model’s predictions?\",\"answer\":\"The ResNet50v2 identifies poor-score patients better than both surgeons, while surgeons struggle particularly with the poor group; the model also misclassifies some excellent cases.\"},{\"question\":\"What does the model’s visualisation suggest about decision factors?\",\"answer\":\"The visualisation indicates that the model’s estimated classifications are driven by image features around the implants, rather than cues the surgeons were using or aware of.\"}]","Machine learning is better than surgeons at assessing unicompartmental knee replacement radiographs | PDF",1785938748,20,{"code":4,"msg":30,"data":31},"ok",{"site_id":23,"language":22,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":89,"head_meta":91,"extra_data":93,"updated_unix":27},"machine-learning-is-better-than-surgeons-at-assessing-unicompartmental-knee-replacement-radiographs","",{"@graph":35,"@context":88},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/machine-learning-is-better-than-surgeons-at-assessing-unicompartmental-knee-replacement-radiographs/127413/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":22,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-08-05",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80,84],{"name":71,"@type":72,"acceptedAnswer":73},"What problem does the study address after unicompartmental knee replacement?","Question",{"text":74,"@type":75},"Poor outcomes sometimes occur after UKR, and it can be difficult for surgeons to use radiographs to determine why those poor outcomes happened, limiting treatment decisions.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How is the machine learning model trained and evaluated?",{"text":79,"@type":75},"The model uses 924 one-year anterior-posterior radiographs to train a ResNet50v2 via transfer learning based on one-year Oxford Knee Score categories, then assesses 70 new radiographs not used for training.",{"name":81,"@type":72,"acceptedAnswer":82},"How do the surgeons’ predictions compare with the model’s predictions?",{"text":83,"@type":75},"The ResNet50v2 identifies poor-score patients better than both surgeons, while surgeons struggle particularly with the poor group; the model also misclassifies some excellent cases.",{"name":85,"@type":72,"acceptedAnswer":86},"What does the model’s visualisation suggest about decision factors?",{"text":87,"@type":75},"The visualisation indicates that the model’s estimated classifications are driven by image features around the implants, rather than cues the surgeons were using or aware of.","https://schema.org",{"og:url":51,"og:type":90,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":92,"canonical":51},"index,follow",{"doc_id":7,"site_id":23},{"code":4,"msg":5,"data":95},[96,100,104,108,113,118,123,126,130,133,137],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":105,"show_sort_weight":106,"slug":107},"Exam",70,"exam",{"id":109,"doc_module":4,"doc_module_name":45,"category_name":110,"show_sort_weight":111,"slug":112},5,"Comic",60,"comic",{"id":114,"doc_module":4,"doc_module_name":45,"category_name":115,"show_sort_weight":116,"slug":117},6,"Technology",50,"technology",{"id":119,"doc_module":4,"doc_module_name":45,"category_name":120,"show_sort_weight":121,"slug":122},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":124,"slug":125},30,"research-report",{"id":127,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":28,"slug":129},9,"Religion & Spirituality","religion-spirituality",{"id":28,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":28,"slug":132},"World Cup","world-cup",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":134,"slug":136},10,"Lifestyle","lifestyle",{"id":138,"doc_module":4,"doc_module_name":45,"category_name":139,"show_sort_weight":109,"slug":140},19,"General","general"]