Department of Radiology, Affiliated Xiaoshan Hospital, Hangzhou Normal University, Hangzhou, Zhejiang, 310000, China , xcf628@163.com
Abstract: (76 Views)
Background:To explore the radiomics features of osteoporotic and malignant neoplastic vertebral compression fractures (VCFs), and to analyze the application value of radiomics in differential diagnosis of osteoporotic and malignant neoplastic VCFs. Materials and Methods: Fifty-one patients with VCFs caused by malignant tumors and forty-nine patients with osteoporosis-induced VCFs treated in the Xiaoshan Hospital Affiliated to Hangzhou Normal University from January 2020 to June 2023 were retrospectively collected into a training set (70 cases) and a verification set (30 cases) according to a stratified random sampling design and a 7:3 ratio. The radiomics parameters of T2WI images of the diseased vertebral bodies were extracted, and the parameters with statistical differences were screened out by dimensionality reduction, so as to build a prediction model. Receiver operating characteristic (ROC) curves were drawn to evaluate the differential diagnosis performance of radiomics for the etiology of vertebral fractures. Results: Eight radiomics features were obtained after dimensionality reduction using the LASSO algorithm. The constructed model was effective in differentiating osteoporotic and malignant neoplastic VCFs, with an area under the ROC curve (AUC) of 0.95; while the AUC for the validation set was 0.84. Conclusions: The radiomics features of T2WI images of vertebral fractures have high efficiency in the differential diagnosis of fracture etiology.