Incremental inputs improve the automated detection of implant loosening using machine-learning algorithms.

AIMS The aim of this study was to evaluate the ability of a machine-learning algorithm to diagnose prosthetic loosening from preoperative radiographs and to investigate the inputs that might improve its performance. METHODS A group of 697 patients underwent a first-time revision of a total hip (THA) or total knee arthroplasty (TKA) at our institution between 2012 and 2018. Preoperative anteroposterior (AP) and lateral radiographs, and historical and comorbidity information were collected from their electronic records. Each patient was defined as having loose or fixed components based on the operation notes. We trained a series of convolutional neural network (CNN) models to predict a diagnosis of loosening at the time of surgery from the preoperative radiographs. We then added historical data about the patients to the best performing model to create a final model and tested it on an independent dataset. RESULTS The convolutional neural network we built performed well when detecting loosening from radiographs alone. The first model built de novo with only the radiological image as input had an accuracy of 70%. The final model, which was built by fine-tuning a publicly available model named DenseNet, combining the AP and lateral radiographs, and incorporating information from the patient's history, had an accuracy, sensitivity, and specificity of 88.3%, 70.2%, and 95.6% on the independent test dataset. It performed better for cases of revision THA with an accuracy of 90.1%, than for cases of revision TKA with an accuracy of 85.8%. CONCLUSION This study showed that machine learning can detect prosthetic loosening from radiographs. Its accuracy is enhanced when using highly trained public algorithms, and when adding clinical data to the algorithm. While this algorithm may not be sufficient in its present state of development as a standalone metric of loosening, it is currently a useful augment for clinical decision making. Cite this article: Bone Joint J 2020;102-B(6 Supple A):101-106.

[1]  C. Zoccali,et al.  The role of FDG-PET in distinguishing between septic and aseptic loosening in hip prosthesis: a review of literature , 2009, International Orthopaedics.

[2]  John J Callaghan,et al.  Artificial Intelligence and Machine Learning in Lower Extremity Arthroplasty: A Review. , 2019, The Journal of arthroplasty.

[3]  Chris Gibbons,et al.  Supervised Machine Learning Algorithms Can Classify Open-Text Feedback of Doctor Performance With Human-Level Accuracy , 2017, Journal of medical Internet research.

[4]  Olivier Gheysens,et al.  Consensus document for the diagnosis of prosthetic joint infections: a joint paper by the EANM, EBJIS, and ESR (with ESCMID endorsement) , 2019, European Journal of Nuclear Medicine and Molecular Imaging.

[5]  Javad Parvizi,et al.  Diagnosis of Infected Total Knee: Findings of a Multicenter Database , 2008, Clinical orthopaedics and related research.

[6]  Michael Blum,et al.  High-Throughput Classification of Radiographs Using Deep Convolutional Neural Networks , 2016, Journal of Digital Imaging.

[7]  J. Parvizi,et al.  Why are total knee arthroplasties failing today--has anything changed after 10 years? , 2014, The Journal of arthroplasty.

[8]  Andrew Zisserman,et al.  ISSLS PRIZE IN BIOENGINEERING SCIENCE 2017: Automation of reading of radiological features from magnetic resonance images (MRIs) of the lumbar spine without human intervention is comparable with an expert radiologist , 2017, European Spine Journal.

[9]  Anurag Gupta,et al.  Deep neural network improves fracture detection by clinicians , 2018, Proceedings of the National Academy of Sciences.

[10]  Thomas F Hany,et al.  FDG PET for differentiation of infection and aseptic loosening in total hip replacements: comparison with conventional radiography and three-phase bone scintigraphy. , 2004, Radiology.

[11]  Edmund Lau,et al.  Impact of the economic downturn on total joint replacement demand in the United States: updated projections to 2021. , 2014, The Journal of bone and joint surgery. American volume.

[12]  J R van Horn,et al.  Nuclear arthrography: combined scintigraphic and radiographic procedure for diagnosis of total hip prosthesis loosening. , 1996, Journal of nuclear medicine : official publication, Society of Nuclear Medicine.

[13]  Marcus A. Badgeley,et al.  Deep learning predicts hip fracture using confounding patient and healthcare variables , 2018, npj Digital Medicine.

[14]  K. Borgwardt,et al.  Machine Learning in Medicine , 2015, Mach. Learn. under Resour. Constraints Vol. 3.

[15]  Steven M. Kurtz,et al.  SYMPOSIUM: ABJS CARL T. BRIGHTON WORKSHOP ON HEALTH POLICY ISSUES IN ORTHOPAEDIC SURGERY Future Young Patient Demand for Primary and Revision Joint Replacement , 2009 .

[16]  Chien-Hung Liao,et al.  Application of a deep learning algorithm for detection and visualization of hip fractures on plain pelvic radiographs , 2019, European Radiology.

[17]  Nicholas Russell,et al.  The Diagnostic Accuracy of Radionuclide Arthrography for Prosthetic Loosening in Hip and Knee Arthroplasty , 2013, BioMed research international.

[18]  Yiting Xie,et al.  Pre-training on Grayscale ImageNet Improves Medical Image Classification , 2018, ECCV Workshops.

[19]  Yixin Chen,et al.  Image Categorization by Learning and Reasoning with Regions , 2004, J. Mach. Learn. Res..

[20]  Muazzam Maqsood,et al.  Transfer Learning Assisted Classification and Detection of Alzheimer’s Disease Stages Using 3D MRI Scans , 2019, Sensors.

[21]  W. Zimmerli,et al.  Diagnosis and treatment of implant-associated septic arthritis and osteomyelitis , 2008, Current infectious disease reports.

[22]  J. Dudley,et al.  Machine-Learning Algorithms to Automate Morphological and Functional Assessments in 2D Echocardiography. , 2016, Journal of the American College of Cardiology.

[23]  Andrew D. Norden,et al.  Cognitive computing in oncology: A qualitative assessment of IBM Watson for Oncology in Mexico. , 2017 .

[24]  F. Pfeiffer,et al.  Efficient Deep Network Architectures for Fast Chest X-Ray Tuberculosis Screening and Visualization , 2019, Scientific Reports.

[25]  Brian R Wolf,et al.  Total knee arthroplasty volume, utilization, and outcomes among Medicare beneficiaries, 1991-2010. , 2012, JAMA.

[26]  Douglas A. Dennis,et al.  What Is the Value of Component Loosening Assessment of a Preoperatively Obtained Bone Scan Prior to Revision Total Knee Arthroplasty? , 2019, The Journal of arthroplasty.

[27]  Federico Cabitza,et al.  Machine Learning in Orthopedics: A Literature Review , 2018, Front. Bioeng. Biotechnol..

[28]  C. Mackenzie,et al.  A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. , 1987, Journal of chronic diseases.