The Value Of Magnetic Resonance Imaging T2WI Texture In The Diagnosis Of Cervical Lymph Nodes

Purpose: Aimed to explore the feasibility of identifying the nature of cervical swollen lymph nodes based on magnetic resonance imaging T2 WI texture analysis.Methods: This study retrospectively enrolled 152 patients with cervical lymphadenopathy who were treated in the First People’s Hospital of Chenzhou from January 2015 to December 2020.All patients received neck MR examination and swollen lymph node biopsy.According to pathology results,the study cases were divided into metastatic head and neck squamous cell carcinoma(HNSCC)group(62 cases),lymphoma group(42 cases),and non-tumor diseased lymph node group(48 cases).Two radiologists applied 3D-slicer software to semi-automatically segment lymph node regions of interest(ROI)layer by layer on axial T2 WI.3D-slicer software was used to extract texture parameters from the ROI.Through logistic polynomial regression analysis,statistically significant texture features were screened out,and a predictive model based on texture features was established to distinguish metastatic HNSCC,lymphoma,and non-tumor diseased lymph nodes.Results: In this study,the Lasso polynomial regression analysis method was used to reduce the dimensionality of the texture features of metastatic HNSCC,lymphoma,and non-tumor lesions.The best texture features were screened out,and Logistic polynomial regression analysis was used.This study establish a classification model based on T2 WI texture features,predict and classify the properties of cervical lymph nodes,draw receiver operating characteristic curve(ROC),and verify the analytical power of the established model through the area under the curve(AUC).The ROC curve predicted to be the lymphoma group,the area under the curve was 0.763,and the accuracy is 66.67%;the ROC curve predicted to be the non-tumor diseased lymph node group,AUC was 0.905,and the accuracy was 89.6%;the prediction was metastatic head and neck in the ROC curve of the HNSCC group,AUC was 0.908,and the accuracy was 90.3%.The AUC of each group was more than 70%,which proves that the analytical power of the established model had clinical significance.Conclusion: Texture analysis based on MRI T2 WI images was helpful to identify the nature of cervical swollen lymph nodes,and had certain reference value for the diagnosis of metastatic HNSCC,lymphoma,and non-tumor lesions.