1. Generalizability of Deep Learning Classification of Spinal Osteoporotic Compression Fractures on Radiographs Using an Adaptation of the Modified-2 Algorithm-Based Qualitative Criteria.
- Author
-
Dong Q, Luo G, Lane NE, Lui LY, Marshall LM, Johnston SK, Dabbous H, O'Reilly M, Linnau KF, Perry J, Chang BC, Renslo J, Haynor D, Jarvik JG, and Cross NM
- Subjects
- Male, Humans, Retrospective Studies, Bone Density, Lumbar Vertebrae diagnostic imaging, Algorithms, Fractures, Compression diagnostic imaging, Deep Learning, Spinal Fractures diagnostic imaging, Osteoporosis complications, Osteoporosis diagnostic imaging
- Abstract
Rationale and Objectives: Spinal osteoporotic compression fractures (OCFs) can be an early biomarker for osteoporosis but are often subtle, incidental, and underreported. To ensure early diagnosis and treatment of osteoporosis, we aimed to build a deep learning vertebral body classifier for OCFs as a critical component of our future automated opportunistic screening tool., Materials and Methods: We retrospectively assembled a local dataset, including 1790 subjects and 15,050 vertebral bodies (thoracic and lumbar). Each vertebral body was annotated using an adaption of the modified-2 algorithm-based qualitative criteria. The Osteoporotic Fractures in Men (MrOS) Study dataset provided thoracic and lumbar spine radiographs of 5994 men from six clinical centers. Using both datasets, five deep learning algorithms were trained to classify each individual vertebral body of the spine radiographs. Classification performance was compared for these models using multiple metrics, including the area under the receiver operating characteristic curve (AUC-ROC), sensitivity, specificity, and positive predictive value (PPV)., Results: Our best model, built with ensemble averaging, achieved an AUC-ROC of 0.948 and 0.936 on the local dataset's test set and the MrOS dataset's test set, respectively. After setting the cutoff threshold to prioritize PPV, this model achieved a sensitivity of 54.5% and 47.8%, a specificity of 99.7% and 99.6%, and a PPV of 89.8% and 94.8%., Conclusion: Our model achieved an AUC-ROC>0.90 on both datasets. This testing shows some generalizability to real-world clinical datasets and a suitable performance for a future opportunistic osteoporosis screening tool., Competing Interests: Declaration of Competing Interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Nathan Cross, Qifei Dong, Sandra Johnston, Jessica Perry, and David Haynor report financial support was provided by the National Institute of Arthritis and Musculoskeletal and Skin Diseases. Gang Luo reports financial support was provided by the National Institute of Arthritis and Musculoskeletal and Skin Diseases and National Heart Lung and Blood Institute. Jeffrey Jarvik reports financial support was provided by the National Institute of Arthritis and Musculoskeletal and Skin Diseases and reports a relationship with the General Electric-Association of University Radiologists Radiology Research Academic Fellowship (GERRAF) that includes travel reimbursement. Nathan Cross reports financial support was provided by the General Electric-Association of University Radiologists Radiology Research Academic Fellowship (GERRAF). Nathan M. Cross, Jeffrey Jarvik, David R. Haynor, Gang Luo, Sandra Johnston, Qifei Dong, Jonathan Renslo, Brian Chang, and Jessica Perry have patent #63/463823 pending to the University of Washington., (Copyright © 2023 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.)
- Published
- 2023
- Full Text
- View/download PDF