Detecting upper extremity native joint dislocations using deep learning: A multicenter study

Published:September 24, 2022DOI:


      • Deep learning models trained to identify native shoulder dislocation achieved high performance on internal and external test sets.
      • Deep learning models trained to identify native elbow dislocations achieved high performance on internal and external test sets.
      • Heatmaps showed emphasis of relevant joints for decision-making.



      Joint dislocations are orthopedic emergencies that require prompt intervention. Automatic identification of these injuries could help improve timely patient care because diagnostic delays increase the difficulty of reduction. In this study, we developed convolutional neural networks (CNNs) to detect elbow and shoulder dislocations, and tested their generalizability on external datasets.


      We collected 106 elbow radiographs (53 with dislocation [50 %]) and 140 shoulder radiographs (70 with dislocation [50 %]) from a level-1 trauma center. After performing 24× data augmentation on training/validation data, we trained multiple CNNs to detect elbow and shoulder dislocations, and also evaluated the best-performing models using external datasets from an external hospital and online radiology repositories. To examine CNN decision-making, we generated class activation maps (CAMs) to visualize areas of images that contributed the most to model decisions.


      On all internal test sets, CNNs achieved AUCs >0.99, and on all external test sets, CNNs achieved AUCs >0.97. CAMs demonstrated that the CNNs were focused on relevant joints in decision-making regardless of whether or not dislocations were present.


      Joint dislocations in both shoulders and elbows were readily identified with high accuracy by CNNs with excellent generalizability to external test sets. These findings suggest that CNNs could expedite access to intervention by assisting in diagnosing dislocations.


      To read this article in full you will need to make a payment

      Purchase one-time access:

      Academic & Personal: 24 hour online accessCorporate R&D Professionals: 24 hour online access
      One-time access price info
      • For academic or personal research use, select 'Academic and Personal'
      • For corporate R&D use, select 'Corporate R&D Professionals'


      Subscribe to Clinical Imaging
      Already a print subscriber? Claim online access
      Already an online subscriber? Sign in
      Institutional Access: Sign in to ScienceDirect


        • Gottlieb M.
        Shoulder dislocations in the emergency department: a comprehensive review of reduction techniques.
        J Emerg Med. 2020; 58: 647-666
        • Modi C.S.
        • Wasserstein D.
        • Mayne I.P.
        • et al.
        The frequency and risk factors for subsequent surgery after a simple elbow dislocation.
        Injury. 2015; 46: 1156-1160
        • Kanji A.
        • Atkinson P.
        • Fraser J.
        • et al.
        Delays to initial reduction attempt are associated with higher failure rates in anterior shoulder dislocation: a retrospective analysis of factors affecting reduction failure.
        Emerg Med J. 2016; 33: 130-133
        • Ding B.T.K.
        • Decruz J.
        • Kunnasegaran R.
        Time-sensitive ambulatory orthopaedic soft-tissue surgery paradigms during the COVID-19 pandemic.
        Int Orthop. 2020; 44: 1531-1538
        • Graber M.
        • Marino D.V.
        • Johnson D.E.
        Anterior hip dislocation. [Updated 2022 Apr 30].
        (Jan-. Available from)in: StatPearls [Internet]. StatPearls Publishing, Treasure Island (FL)2022
        • Suzuki T.
        • Maki S.
        • Yamazaki T.
        • et al.
        Detecting distal radial fractures from wrist radiographs using a deep convolutional neural network with an accuracy comparable to hand orthopedic surgeons.
        J Digit Imaging. 2022; 35: 39-46
        • Krogue J.D.
        • Cheng K.V.
        • Hwang K.M.
        • et al.
        Automatic hip fracture identification and functional subclassification with deep learning.
        Radiol Artif Intell. 2020; 2e190023
        • Images Google
        Date accessed: November 1, 2022
      1., the Wiki-basedCollaborative Radiology Resource.
        Date accessed: November 1, 2022
        • Russakovsky O.
        • Deng J.
        • Su H.
        • et al.
        ImageNet large scale visual recognition challenge.
        Int J Comput Vis. 2015; 115: 211-252
      2. K He X Zhang S Ren J Sun n.d. Deep Residual Learning for Image Recognition. arXiv:1512.03385.

      3. C Szegedy V Vanhoucke S Ioffe et al Rethinking the Inception Architecture for Computer Vision. arXiv:1512.00567.n.d.

      4. G Huang Z Liu L van der Maaten KQ Weinberger n.d. Densely Connected Convolutional Networks. arXiv:1608.06993.

      5. S Xie R Girshick P Dollár et al Aggregated Residual Transformations for Deep Neural Networks. arXiv:1611.05431.n.d.

      6. S Zagoruyko N. Komodakis Wide Residual Networks. arXiv:1605.07146.n.d.

      7. J Xu Y Pan X Pan et alRegNet: Self-regulated Network for Image Classification. arXiv:2101.00590.n.d.

        • Deng J.
        • Dong W.
        • Socher R.
        • et al.
        ImageNet: a large-scale hierarchical image database.
        IEEE Conf Comput Vis Pattern Recognit. 2009; 2009: 248-255
        • DeLong E.R.
        • DeLong D.M.
        • Clarke-Pearson D.L.
        Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach.
        Biometrics. 1988; 44: 837-845
        • Zech J.R.
        • Badgeley M.A.
        • Liu M.
        • et al.
        Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: a cross-sectional study.
        PLoS Med. 2018; 15e1002683
        • Yi P.H.
        • Malone P.
        • Lin C.T.
        • Filice R.W.
        Deep learning algorithms for interpretation of upper extremity radiographs: laterality and technologist initial labels as confounding factors.
        Am J Roentgenol. 2022; 218: 714-715