ACR Meeting Abstracts

ACR Meeting Abstracts

Abstract Number: 0106

A Machine Learning-derived Radiomics Nomogram for Diagnosis of Osteoporosis and Osteopenia

Qianrong Xie1, Yue Chen2, Yimei Hu3, Fanwei Zeng4, Pingxi Wang4, Lin Xu5, Jianhong Wu6, Jie Li1, Jing Zhu7, Ming Xiang8 and Fanxin Zeng9, 1Department of Clinical Research Center, Dazhou Central Hospital, Dazhou, China (People's Republic), 2Department of Clinical Medicine, Chengdu University of Traditional Chinese Medicine, Chengdu, China (People's Republic), 3Department of Orthopedics, Hospital of Chengdu University of Traditional Chinese Medicine, Chengdu, China (People's Republic), 4Department of bone disease, Dazhou Central Hospital, Dazhou, China (People's Republic), 5Department of medical imaging, Dazhou Central Hospital, Dazhou, China (People's Republic), 6Department of Rheumatology, Dazhou Central Hospital, Dazhou, China (People's Republic), 7Department of Rheumatology and Immunology, Sichuan Academy of Medical Sciences & Sichuan Provincial People’s Hospital, Chengdu, China (People's Republic), 8Department of Orthopedics, Sichuan Provincial Orthopedic Hospital, Chengdu, China (People's Republic), 9Department of Clinical Research Center, Dazhou Central Hospital, Dazhou, Sichuan, China (People's Republic)

Meeting: ACR Convergence 2020

Keywords: , , , ,

Session Information

Date: Friday, November 6, 2020

Title: Osteoporosis & Metabolic Bone Disease Poster

Session Type: Poster Session A

Session Time: 9:00AM-11:00AM

Background/Purpose: To discriminate osteoporosis and osteopenia using a quantitative computed tomography (QCT) radiomics signatures and clinical variables.

Methods: This retrospective study enrolled 635 patients with QCT images and clinical characteristics from November 2016 to November 2019. 851 radiomics features extracted from the QCT images of the third lumbar vertebra of each patients. Minimum redundancy and maximum relevance (mRMR) and least absolute shrinkage and selection operator (LASSO) were used for features selection. A radiomic nomogram was constructed based on the radiomics signatures and clinical characteristics for diagnosing osteoporosis and osteopenia. We systematically evaluated the diagnostic performance of the combined radiomics model, clinical model and radiomics model. Moreover, we investigated the value of the radiomics score in normal group and abnormal group.

Results: A total of 6 optimal radiomics features were selected to construct quantitative computed tomography (QCT) based signatures. The individualized nomogram included these features and 3 clinical characteristics (age, alkaline phosphatase, homocysteine) achieved good discrimination performance in both the training cohort (N=414; area under the curve (AUC) 0.96, 95% CI 0.95–0.98) and the validation cohort (N=176; AUC 0.96, 95% CI 0.92–1.00). The alone radiomics score also demonstrated significant differences in osteoporosis and osteopenia (P < 0.001).

Conclusion: This study presents a radiomics nomogram that incorporates the radiomics score and clinical risk factors, which can serve as a reliable and powerful tool for discriminating osteoporosis and osteopenia.

Disclosure: Q. Xie, None; Y. Chen, None; Y. Hu, None; F. Zeng, None; P. Wang, None; L. Xu, None; J. Wu, None; J. Li, None; J. Zhu, None; M. Xiang, None; F. Zeng, None.

To cite this abstract in AMA style:

Xie Q, Chen Y, Hu Y, Zeng F, Wang P, Xu L, Wu J, Li J, Zhu J, Xiang M, Zeng F. A Machine Learning-derived Radiomics Nomogram for Diagnosis of Osteoporosis and Osteopenia [abstract]. Arthritis Rheumatol. 2020; 72 (suppl 10). https://acrabstracts.org/abstract/a-machine-learning-derived-radiomics-nomogram-for-diagnosis-of-osteoporosis-and-osteopenia/. Accessed .

« Back to ACR Convergence 2020

ACR Meeting Abstracts - https://acrabstracts.org/abstract/a-machine-learning-derived-radiomics-nomogram-for-diagnosis-of-osteoporosis-and-osteopenia/