Preoperative Prediction Of Central Lymph Node Metastasis In Patient With Papillary Thyroid Carcinoma

Objectives:To establish a radiomics algorithm based on grayscale ultrasound images and to make preoperative prediction of lymph node metastasis in papillary thyroid carcinoma patient.Methods:In our retrospective study,totally 354 cases of surgical histopathological confirmed lesions papillary thyroid carcinoma were included.Which included 181 consecutive PTCs with lymph node metastasis and 173 PTCs without central lymph node metastasis.Randomly divided into training set and test set according to the ratio of 7:3.We extracted ultrasound radiomics features derived from the first order feature,shape-based feature,texture feature,wavelet-based features using US imaging,A radiomics signature was generated using the univariate analysis and the least absolute shrinkage and selection operator(LASSO)and was used to train a decision tree to predict central lymph node metastasis with pathological results as the gold standard.Results:We extracted a total of 851 radiomics features,The radiomics features models of training set and test set showed the area under the receiver operating characteristic curve(AUC)values respectively of 0.823(95%CI,0.617-0.888),0.780(95%CI,0.702-0.798);The accuracy,sensitivity,specificity,positive predictive value and negative predictive value of the training set and test set are 0.798、0.854、0.744、0.760、0.843;0.727、0.702、0.755、0.730、0.724,respectively.The accuracy,sensitivity,specificity,positive predictive value and negative predictive value of preoperative thyroid ultrasonography for diagnosis of central lymph node metastasis was 0.455、0.227、0.694、0.436、0.461,respectively.Conclusion:The accuracy,sensitivity,specificity of radiomic signature based on thyroid grayscale ultrasound images is much higher than conventional thyroid and lymph node ultrasound in predicting central lymph node metastasis of PTC,and can be used for non-invasive prediction of central lymph node metastasis in patients with PTC,lower than the radiomics signature.