The application value of CT radiomics in assessing acute traumatic bone marrow edema in the knee joint

YUAN Hu; SHENG Mao

doi:10.20258/j.cnki.1006-9011.2026.01.024

Abstract：Objective To explore the utility of CT radiomics in assessing acute traumatic bone marrow edema in the knee.Methods A total of 86 patients with acute traumatic bone marrow edema were enrolled,and all underwent non-contrast CT scans of the knee.In this paired-design study,the knees with traumatic bone marrow edema on the affected side taken as the ob-servation group provided radiomic signatures of bone marrow edema lesions,whereas,the contralateral healthy knees served as controls to establish normative imaging baselines,effectively controlling for inter-individual anatomical and physiological con-founders.The 3D-Slicer software was utilized to delineate the region of interest(ROI)on both axial and coronal CT images of the bone marrow edema area and the contralateral healthy side.Texture features were extracted in bulk using a radiomics texture ex-tractor programmed in Python.The 86 patients with acute traumatic bone marrow edema were divided into a training set(n=60)and a validation set(n=26)in a ratio of 7∶3.Following this,a multivariate Logistic regression model was constructed using the validation set,after performing an independent sample t-test,Mann-Whitney U test,and employing CT image texture parameter reduction with LASSO regression.Results For each knee,1595 feature parameters were acquired.Following dimension reduc-tion,18 axial parameters and 16 coronal parameters were chosen.In the training set,the multivariate Logistic regression model exhibited an ROC area under the curve(AUC)of 0.91 and 0.84,with predictive accuracies of 82.0%and 79.0%,sensitivities of 82.0%and 81.0%,specificities of 82.0%and 77.0%,positive predictive values of 82.0%and 78.0%,and negative predictive values of 82.0%and 80.0%,respectively.Additionally,the ROC AUC for the model in the training set was 0.76 and 0.72,pre-diction accuracies were 72.0%and 71.0%,sensitivities were 75.0%and 69.0%,specificities were 68.0%and 73.0%,positive predictive values were 70.0%and 72.0%,and negative predictive values were 73.0%and 70.0%,respectively.Conclusion CT-based radiomics can be applied in knee joint trauma and demonstrates good diagnostic value for acute traumatic bone marrow edema,with the axial model being superior.This holds promise as a potential tool for auxiliary diagnosis.