A Hierarchical Opportunistic Screening Model for Osteoporosis Using Machine Learning Applied to Clinical Data and CT Images

Abstract

Background: Bone mineral density (BMD) measured by dual-energy X-ray absorptiometry (DXA) is the gold standard for osteoporosis detection. However, only a limited percentage of people with osteoporosis risks actually undergo DXA test. This study aimed to detect osteoporosis using clinical data (demographic characteristics and routine laboratory tests data) and CT images rather than DXA data via machine learning. Methods: 2210 individuals over age 40 were collected retrospectively, among which 246 individuals’ clinical data and CT images are both available. The cohort was divided into osteoporotic and non-osteoporotic groups according to DXA, without considering osteopenia individuals. Lumbar vertebral bodies (L1-L4) were segmented manually as regions of interest (ROIs). A hierarchical model with three layers was proposed to identify osteoporosis. The first layer utilized demographic characteristics only, with clinical data for the second layer, and clinical data together with texture features extracted from ROIs for the third layer. Logistic regression (LR), random forests (RF), and eXtreme Gradient Boosting (XGBoost) were employed as classifiers. The dataset was randomly split into training and test set repeated five times and model performances were evaluated by area under the receiver operating characteristic curve (AUC). Findings: The hierarchical model based on LR showed better performances with an AUC of 0.818±0.039, 0.838±0.050, and 0.962±0.012 for three layers, respectively in distinguishing individuals with osteoporosis and normal BMD. Interpretation: The proposed model showed great potential in opportunistic screening for osteoporosis without additional expense. Funding: National Natural Science Foundation of China, Research and Development Program of Jinan. Funding Information: This work was supported by the National Natural Science Foundation of China [grant numbers 61673246, 81301294]; and the Research and Development Program of Jinan [grant number 201907064]. Declaration of Interests: We declare that we have no competing interests. Ethics Approval Statement: We conducted a retrospective study, which complied with the World Medical Association Declaration of Helsinki. The study obtained local ethics committee approval (KYLL-2020(KS)-743) and informed consent was not required owing to the retrospective nature of the study.