{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,15]],"date-time":"2026-09-15T10:10:20Z","timestamp":1789467020511,"version":"build-2803163510"},"reference-count":20,"publisher":"BMJ","issue":"1","license":[{"start":{"date-parts":[[2021,4,6]],"date-time":"2021-04-06T00:00:00Z","timestamp":1617667200000},"content-version":"unspecified","delay-in-days":5,"URL":"http:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"}],"content-domain":{"domain":["bmj.com"],"crossmark-restriction":true},"short-container-title":["Trauma Surg Acute Care Open"],"published-print":{"date-parts":[[2021,4]]},"abstract":"<jats:sec>\n                  <jats:title>Background<\/jats:title>\n                  <jats:p>Pelvic X-ray (PXR) is a ubiquitous modality to diagnose hip fractures. However, not all healthcare settings employ round-the-clock radiologists and PXR sensitivity for diagnosing hip fracture may vary depending on digital display. We aimed to validate a computer vision algorithm to detect hip fractures across two institutions\u2019 heterogeneous patient populations. We hypothesized a convolutional neural network algorithm can accurately diagnose hip fractures on PXR and a web application can facilitate its bedside adoption.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Methods<\/jats:title>\n                  <jats:p>The development cohort comprised 4235 PXRs from Chang Gung Memorial Hospital (CGMH). The validation cohort comprised 500 randomly sampled PXRs from CGMH and Stanford\u2019s level I trauma centers. Xception was our convolutional neural network structure. We randomly applied image augmentation methods during training to account for image variations and used gradient-weighted class activation mapping to overlay heatmaps highlighting suspected fracture locations.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Results<\/jats:title>\n                  <jats:p>Our hip fracture detection algorithm\u2019s area under the receiver operating characteristic curves were 0.98 and 0.97 for CGMH and Stanford\u2019s validation cohorts, respectively. Besides negative predictive value (0.88 Stanford cohort), all performance metrics\u2014sensitivity, specificity, predictive values, accuracy, and F1 score\u2014were above 0.90 for both validation cohorts. Our web application allows users to upload PXR in multiple formats from desktops or mobile phones and displays probability of the image containing a hip fracture with heatmap localization of the suspected fracture location.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Discussion<\/jats:title>\n                  <jats:p>We refined and validated a high-performing computer vision algorithm to detect hip fractures on PXR. A web application facilitates algorithm use at the bedside, but the benefit of using our algorithm to supplement decision-making is likely institution dependent. Further study is required to confirm clinical validity and assess clinical utility of our algorithm.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Level of evidence<\/jats:title>\n                  <jats:p>III, Diagnostic tests or criteria.<\/jats:p>\n               <\/jats:sec>","DOI":"10.1136\/tsaco-2021-000705","type":"journal-article","created":{"date-parts":[[2021,4,7]],"date-time":"2021-04-07T12:54:48Z","timestamp":1617800088000},"page":"e000705","update-policy":"https:\/\/doi.org\/10.1136\/crossmarkpolicy","source":"Crossref","is-referenced-by-count":17,"title":["Practical computer vision application to detect hip fractures on pelvic X-rays: a bi-institutional study"],"prefix":"10.1136","volume":"6","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-1639-8781","authenticated-orcid":false,"given":"Jeff","family":"Choi","sequence":"first","affiliation":[{"name":"General Surgery, Stanford University, Stanford, California, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"James Z","family":"Hui","sequence":"additional","affiliation":[{"name":"Radiology, Stanford University, Stanford, California, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"David","family":"Spain","sequence":"additional","affiliation":[{"name":"Surgery, Stanford University, Stanford, California, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yi-Siang","family":"Su","sequence":"additional","affiliation":[{"name":"Trauma and Emergency Surgery, Chang Gung Memorial Hospital, Taoyuan, Taiwan"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2697-4642","authenticated-orcid":false,"given":"Chi-Tung","family":"Cheng","sequence":"additional","affiliation":[{"name":"Trauma and Emergency Surgery, Chang Gung Memorial Hospital, Taoyuan, Taiwan"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Chien-Hung","family":"Liao","sequence":"additional","affiliation":[{"name":"Trauma and Emergency Surgery, Chang Gung Memorial Hospital, Taoyuan, Taiwan"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"239","published-online":{"date-parts":[[2021,4,7]]},"reference":[{"key":"2025081710564615000_6.1.e000705.1","doi-asserted-by":"publisher","DOI":"10.1007\/s00198-012-1964-3"},{"key":"2025081710564615000_6.1.e000705.2","doi-asserted-by":"publisher","DOI":"10.1007\/s00198-006-0172-4"},{"key":"2025081710564615000_6.1.e000705.3","doi-asserted-by":"publisher","DOI":"10.1016\/j.injury.2018.04.015"},{"key":"2025081710564615000_6.1.e000705.4","doi-asserted-by":"publisher","DOI":"10.3111\/13696998.2010.488969"},{"key":"2025081710564615000_6.1.e000705.5","doi-asserted-by":"publisher","DOI":"10.1056\/NEJMcp1611090"},{"key":"2025081710564615000_6.1.e000705.6","doi-asserted-by":"crossref","first-page":"674","DOI":"10.1016\/j.injury.2015.11.001","article-title":"Missed subtle fractures on the trauma-meeting digital projector","volume":"47","author":"Chellam","year":"2016","journal-title":"Injury"},{"key":"2025081710564615000_6.1.e000705.7","doi-asserted-by":"crossref","first-page":"97","DOI":"10.3390\/jcm9010097","article-title":"Prevalence of fractures and diagnostic accuracy of emergency X-ray in older adults sustaining a low-energy fall: a retrospective study","volume":"9","author":"Lampart","year":"2019","journal-title":"J Clin Med"},{"key":"2025081710564615000_6.1.e000705.8","doi-asserted-by":"publisher","DOI":"10.1007\/s00330-019-06167-y"},{"key":"2025081710564615000_6.1.e000705.9","doi-asserted-by":"crossref","unstructured":"Chollet F , 2017. Xception: deep learning with depthwise separable convolutions. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).","DOI":"10.1109\/CVPR.2017.195"},{"key":"2025081710564615000_6.1.e000705.10","unstructured":"PelviXNet . Hip fracture detection system. http:\/\/medcare.ym.edu.tw:8080\/#\/login (14 Jan 2021)."},{"key":"2025081710564615000_6.1.e000705.11","doi-asserted-by":"crossref","first-page":"2307","DOI":"10.1002\/jbmr.4146","article-title":"Clinical utility of computer-aided diagnosis of vertebral fractures from computed tomography images","volume":"35","author":"Kolanu","year":"2020","journal-title":"J Bone Miner Res"},{"key":"2025081710564615000_6.1.e000705.12","doi-asserted-by":"crossref","first-page":"357","DOI":"10.1016\/j.crad.2019.02.006","article-title":"Artificial intelligence in breast imaging","volume":"74","author":"Le","year":"2019","journal-title":"Clin Radiol"},{"key":"2025081710564615000_6.1.e000705.13","doi-asserted-by":"crossref","first-page":"709","DOI":"10.1016\/j.jtho.2016.01.021","article-title":"Computer vision tool and technician as first reader of lung cancer screening CT scans","volume":"11","author":"Ritchie","year":"2016","journal-title":"J Thorac Oncol"},{"key":"2025081710564615000_6.1.e000705.14","doi-asserted-by":"publisher","DOI":"10.1097\/AOG.0b013e318169ce03"},{"key":"2025081710564615000_6.1.e000705.15","doi-asserted-by":"crossref","first-page":"633","DOI":"10.1002\/ar.23552","article-title":"The human pelvis: variation in structure and function during gait","volume":"300","author":"Lewis","year":"2017","journal-title":"Anat Rec"},{"key":"2025081710564615000_6.1.e000705.16","doi-asserted-by":"crossref","first-page":"2769","DOI":"10.1016\/j.injury.2016.10.019","article-title":"The sensitivity of CT scans in diagnosing occult femoral neck fractures","volume":"47","author":"Sadozai","year":"2016","journal-title":"Injury"},{"key":"2025081710564615000_6.1.e000705.17","doi-asserted-by":"crossref","first-page":"305","DOI":"10.1038\/s42256-020-0186-1","article-title":"Secure, privacy-preserving and federated machine learning in medical imaging","volume":"2","author":"Kaissis","year":"2020","journal-title":"Nat Mach Intell"},{"key":"2025081710564615000_6.1.e000705.18","unstructured":"Rueckert D , Schnabel JA . Model-based and data-driven strategies in medical image computing. ArXiv190910391 Cs. 2019. http:\/\/arxiv.org\/abs\/1909.10391 (30 Jan 2021)."},{"key":"2025081710564615000_6.1.e000705.19","unstructured":"Oakden-Rayner L , Dunnmon J , Carneiro G , R\u00e9 C . Hidden stratification causes clinically meaningful failures in machine learning for medical imaging. ArXiv190912475 Cs Stat. 2019. http:\/\/arxiv.org\/abs\/1909.12475 (30 Jan 2021)."},{"key":"2025081710564615000_6.1.e000705.20","doi-asserted-by":"publisher","DOI":"10.1258\/jrsm.2011.110180"}],"container-title":["Trauma Surgery &amp; Acute Care Open"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.1136\/tsaco-2021-000705","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,8,17]],"date-time":"2025-08-17T17:56:56Z","timestamp":1755453416000},"score":1,"resource":{"primary":{"URL":"https:\/\/tsaco.bmj.com\/lookup\/doi\/10.1136\/tsaco-2021-000705"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2021,4]]},"references-count":20,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2021,4,7]]},"published-print":{"date-parts":[[2021,4]]}},"alternative-id":["10.1136\/tsaco-2021-000705"],"URL":"https:\/\/doi.org\/10.1136\/tsaco-2021-000705","relation":{},"ISSN":["2397-5776"],"issn-type":[{"value":"2397-5776","type":"electronic"}],"subject":[],"published":{"date-parts":[[2021,4]]}}}