Combined Application Of Two-Dimensional Ultrasound And VTI In The Estimation Of Lymph Node Metastasis Of Papillary Thyroid Carcinoma Based On Radiomics

Objective : Construction of a radiomics model based on two-dimensional ultrasound and virtual touch tissue imaging(VTI)of single nodules in papillary thyroid carcinoma(PTC)to predict cervical lymph node metastasis in PTC,and to explore the clinical application value of radiomics models in the prediction of cervical lymph node metastasis in PTC.Materials and Methods: A total of 70 patients with single nodules of PTC who were treated surgically in Liaoning Cancer Hospital were collected,including 31 cases with cervical lymph node metastasis and 39 cases without cervical lymph node metastasis.Preoperatively,two-dimensional ultrasound and VTI ultrasound elastic images of PTC nodules were collected,and evaluated by two ultrasonographers of associate chief physician or higher for bilateral cervical lymph node metastasis.Then,the included cases were randomly divided into training and validation groups according to the ratio of 7:3,and statistical methods such as ANOVA and correlation analysis were adopted to select features in the three feature sets of two-dimensional ultrasound radiomics,VTI ultrasound elastic radiomics,and two-dimensional + VTI ultrasound elastic radiomics of PTC nodules,respectively.Using random forest classifier model for machine learning,three ultrasound radiomic models of PTC nodes were established,and the sensitivity,specificity,accuracy,and AUC of the three radiomics models for predicting metastasis of cervical lymph nodes were calculated separately.The AUCs of the three models were compared by Delong test to evaluate the predictive efficacy of the above models.Finally,the ultrasound radiomic characteristics in the three groups were included in the logistic regression analysis respectively,obtained the sensitivity,specificity,and AUC of the three models for predicting cervical lymph nodes metastasis,and compared with the sensitivity and specificity of preoperative diagnosis by ultrasound to evaluate the clinical application value of the radiomics models.Results:1.In this study,a total of 214 radiomics features were extracted based on two-dimensional ultrasound and VTI ultrasound elastic images of PTC nodules,of which 107 features were extracted from each of two-dimensional ultrasound and VTI ultrasound elastic images.After statistical analysis,eight,seven,and thirteen features were selected in the three feature sets of two-dimensional ultrasound radiomics,VTI ultrasound elastic radiomics and two-dimensional ultrasound + VTI ultrasound elastic radiomics,respectively.And machine learning was performed using the random forest method to obtain three radiomics models,namely,two-dimensional ultrasound radiomics model,VTI ultrasound elastic radiomics model,and two-dimensional ultrasound + VTI ultrasound elastic radiomics model.The sensitivity of the three models for predicting cervical lymph node metastasis was 80.0%,61.2%,and 80.0%,respectively;the specificity was 83.3%,74.6%,and 75.0%,respectively;the accuracy was 81.8%,63.6%,and 77.3%,respectively;and the AUC was 0.788,0.671,and 0.808,respectively.2.The Delong test was used to compare the predictive efficacy of three random forest radiomics models on cervical lymph node metastasis,and the results showed that the predictive efficacy of the two-dimensional ultrasound + VTI ultrasound elastic radiomics model was the highest(AUC was 0.858),followed by the two-dimensional ultrasound radiomics model(AUC was 0.833),and the predictive efficacy of the two models was close,the difference was not statistically significant(p > 0.05);however,the predictive efficacy of both models was significantly better than that of the VTI ultrasound elastic radiomics model(AUC was 0.598),and the difference was statistically significant(p < 0.05).3.The sensitivity and specificity of two-dimensional ultrasound + VTI ultrasound elastic radiomics model for predicting the cervical lymph node metastasis were 71.4%,97.3%,respectively,which were higher than the sensitivity and specificity of preoperative ultrasound visual diagnosis results,which were 71.0% and 84.6%,respectively.Conclusion: Compared with the ultrasonographer diagnoses metastasis of papillary thyroid carcinoma cervical lymph node based on sonographic image features,the prediction model established by combining the two-dimensional ultrasound radiomics features and VTI ultrasound elastic radiomics features of papillary thyroid carcinoma nodes is more convenient,efficient and objective,with high clinical utility value.