Predictive value of multimodal MRI histology for mediastinal lymph node metastasis in non-small cell lung cancer
Objective:To construct radiomic models utilizing conventional MRI sequences to assess and compare their effectiveness in predicting mediastinal lymph node metastasis in non-small cell lung cancer(NSCLC).Materials and Methods:Preoperative MRI data from 90 patients diagnosed with NSCLC at the First People's Hospital of Nantong between October 2012 and May 2022 were retrospectively collected.Based on the surgical pathology results,these patients were categorized into two groups:lymph node metastasis-positive(52 cases)and negative(38 cases).The patients were allocated into a training set and a test set using a complete randomization method,with a ratio of 7∶3.Additionally,data from 31 patients at the First People's Hospital of Yancheng were used for external validation,consisting of 9 positive cases and 22 negative cases.Radiologists used layer-by-layer semi-automated delineation of the primary lesions,radiomics features were extracted from axial T1WI,T2WI,high b-value diffusion weighted imaging(DWI),and apparent diffusion coefficient(ADC)images,and selected the best method for dimensionality reduction among the feature screening methods,such as ANOVA F-value,Linear models penalized with the L1 norm,Tree-Based,etc.,by the hyperparameter search.Eleven models were established,including logistic regression(LR),Gaussian naive Bayes(Gaussian NB),random forest(RF),and Support vector machine(SVM),decision tree(DT),etc.The receiver operating characteristic(ROC)curves were calculated for each model.Results:DT,LR,and SVM models all performed well in different sequences.The SVM model based on T2WI images had the best performance,with area under the curve(AUC)of 0.98,0.98,and 0.72 for the training,test,and external validation sets,respectively,and with accuracies of 96%,67%,and 61%,sensitivities of 88%,67%,and 55%,and specificities of 100%,67%,and 78%,respectively.Conclusions:MRI-based radiomics is valuable in identifying mediastinal lymph node metastasis in NSCLC,with the SVM model based on T2WI images showing the best performance.
国家科技期刊平台
NSTL
万方数据
维普
