Objective To explore the diagnostic value of multi-sequence MRI radiomics models in benign and malignant spinal fractures. Methods Retrospectively collected MRI examination data of patients with pathologically confirmed malignant spinal fractures,and compared with those of benign spinal vertebral fractures.T1WI and T2WI radiomics features were extracted and randomly divided into the training and testing sets at a ratio of 8:2.The training set used recursive feature elimination and LASSO to select variables,and logistic regression was used to construct three models for T1WI,T2WI,and the combination of both sequences.The testing set was to validate the diagnostic efficacy of each model.The area under the receiver operating characteristic curve(AUC)was used to evaluate the diagnostic efficacy of the models.Results A total of 97 patients with 111 vertebral fractures were included,with 55 benign and 56 malignant vertebral fractures.In the training set,the AUC,sensitivity,and specificity of the T1WI model were 0.892,88.9％,and 81.4％,respectively.For the T2WI model,they were 0.924,82.2％,and 95.3％,respectively.The multi-sequence model scored 0.934,90.0％,and 87.5％,respectively.In the testing set,the AUC,sensitivity,and specificity of the T1WI model were 0.877,80.0％,and 92.3％.For the T2WI model,they were 0.923,90.0％,and 92.3％,respectively,and for the combined model of both sequences,they were 0.933,100.0％,and 87.5％,respectively.Conclusion The diagnostic efficacy of the combined model of T1WI and T2WI sequences is higher than that of single-sequence models.The combination of multi-sequence radiomics features can provide richer texture information,which is helpful in improving the diagnostic accuracy and efficiency of malignant spinal fractures.
万方数据
维普
