The Application Of Spectral CT Quantitative Analysis And Radiomics In The Diagnosis Of Thyroid Papillary Carcinoma And Prediction Of Cervical Lymph Node Metastasis

Part One Diagnostic Value of CT Features in Papillary Thyroid Carcinoma with Multivariate AnalysisObjective:To investigate the value of multiple Computed Tomography(CT)features in the diagnosis and differential diagnosis of papillary thyroid carcinoma.Methods:This was a retrospective project enrolling 82 patients with 101 thyroid nodulesin neck CT plain scan and enhanced scan.49 papillary thyroid carcinoma(PTC)and 52 nodular goiter(NG)were histologically confirmed.Eight CT descriptors,including tumor location,size,shape,tiny calcification,cystic change,irregularring sign,margin defects sign and enhancement blur sign were observed and measured in all patients.The independent risk factor signs of PTC were analyzed by univariate and multivariate logistic regression,and the sensitivity,specificity,accuracy,Youden index and likelihood ratio of each independent positive sign and the combination of multiple signs in the diagnosis of PTC were calculated.Results:In univariate analysis,four CT features differed significantly between PTC and NG,including tiny calcification,irregular ring sign,margin defects sign and enhancement blur sign.The results of multivariate logistic regression model showed that irregular ring sign,margin defects sign and enhancement blur sign were associated with PTC and the odds ratio(OR)were 27.374(95% CI 5.871~127.636),28.587(95% CI 4.139~197.460)and4.315(95% CI 0.858~21.694),respectively.The diagnostic efficiency of predict model value in PTC was the sensitivity of 87.8%,the specificity of 94.2%,the coincidence rate of 91.1%,Youden index of 0.82 and likelihood ratio of 15.1.Conclusion:The study results showed that irregular ring sign,margin defects sign and enhancement blur sign may be helpful in PTC diagnosis and the combination of multiple signs can improve the accuracy of PTC diagnosis.Part Two The Diagnostic Accuracy of Ultrasound and Computed Tomography in Detecting of Cervical Lymph Node Metastasis in Patients with Papillary Thyroid Carcinoma: a Meta-analysisObjective:To determine the diagnostic accuracy of ultrasound(US),computed tomography(CT)and their combination in detecting cervical lymph node metastasis(CLNM)in patients with papillary thyroid carcinoma(PTC).Methods:Medline(via Pubmed),Web of Science and Embase were searched to identify studies that used both US and CT to detect CLNM in patients with PTC.Primary outcomes were sensitivity,specificity,and diagnostic odds ratios(DORs)in level-by-level or patient-based analysis.Secondary outcomes were sensitivity,specificity and DORs in central and lateral compartments.Results:Fourteen studies involving 6167 patients with 11601 neck lymph nodes met the inclusion criteria.Based on level-by-level analysis,the pooled sensitivity,specificity,DORs for US were 0.35(95% confidence interval [CI]0.34-0.37),0.95(95% CI 0.94-0.95),13.94(95% CI 9.34-20.82),for CT were 0.46(95% CI 0.44-0.47),0.88(95% CI 0.87-0.89),7.24(95% CI 5.46-9.62),for the combination of US and CT were 0.51(95% CI 0.49-0.52),0.85(95% CI 0.84-0.86),6.01(95% CI 3.84-9.40),respectively.The pooled estimates of sensitivity,specificity,and diagnostic OR for US were 0.41(95% CI 0.36-0.46),0.92(95%CI 0.89-0.94),7.56(95% CI 4.08-14.01),for CT were 0.49(0.44-0.54),0.91(0.89-0.94),9.40(5.79-15.27),for combination of US and CT were 0.64(95% CI 0.57-0.71),0.83(95%CI 0.77-0.88),8.59(95% CI 5.37-13.76)on patient-based analysis,respectively.Conclusion:These findings suggest that preoperative CT plays a complementary role to US in detecting CLNM in patients with PTC,especially at lateral compartment and patient-based analysis.Part Three The Value of Spectral CT Quantitative Parameters For Preoperative Diagnosis of Cervical Lymph Node Metastasis in Patients with Papillary Thyroid CarcinomaObjective:To compare the value of double-layer detector spectral computed tomography(CT)quantitative parameters and routine CT scanning qualitative parameters in preoperative evaluation of cervical lymph node metastasis in patients with papillary thyroid carcinoma(PTC).Methods:A total of 50 cases of PTC underwent spectral dual-energy CT scan and dual-phase enhanced CT scan before operation.According to the selection criteria of lymph nodes,the corresponding lymph nodes with horizontal axial short diameter >5 mm were selectedAccording to the histopathological results,they were divided into three groups,34 cases with lymph node metastasis confirmed by PTC pathology,95 metastatic lymph nodes(group 1),and 55 non-metastatic lymph nodes(group 2).No lymph node metastasis was confirmed by pathology in 16 cases of PTC,with 63 lymph nodes(group 3).Another 15 patients with benign thyroid lesions were selected as the benign hyperplasia group,with73 lymph nodes(group 4).The morphological characteristics of routine CT of lymph nodes in each group were analyzed,and the differences of normalized iodine concentration(NIC),slope of energy spectrum curve(? HU)and normalized effective atomic order value(Zeff-N)of plain scan,arterial phase,venous phase in metastatic lymph nodes,non-metastatic lymph nodes and benign lymph nodes were compared.The critical values of NIC,? HU and Zeff-N,and their diagnostic efficiency were obtained from the receiver operating characteristic(ROC)curve,and compared with the diagnostic efficiency of qualitative parameters.Results:1.The ? HU,NIC and Zeff-N of metastatic lymph nodes in the arterial and venous phases in PTC were significantly higher than those in non-metastatic lymph nodes.2.The best parameters for detecting metastatic lymph nodes were arterial phase ? HU and arterial phase NIC.When ? HU >2.945 in arterial phase,the sensitivity,specificity,positive predictive value(PPV),negative predictive value(NPV)and Youden index of metastatic lymph nodes were 0.811,0.906,0.875,0.855,0.717,respectively.When NIC >0.215 in arterial phase,the sensitivity,specificity,PPV,NPV and Youden index of diagnosing metastatic lymph nodes were 0.642,0.96,0.939,0.769,0.608,respectively.Compared with qualitative analysis,? HU in arterial phase had higher sensitivity(0.811 vs.0.789)and significantly higher specificity(0.966 vs.0.521).The comprehensive diagnostic performance of NIC in arterial phase was higher(0.608 vs.0.310).Conclusion:The quantitative parameters of spectral dual-energy CT showed high diagnostic value in preoperative diagnosis of metastatic cervical lymph nodes in patients with PTC.Part Four Application of CT Radiomic Analysis for Predicting Cervical Lymph Node Metastasis in Patients with Papillary Thyroid CarcinomaObjective:To investigate the value of radiomic features for predicting cervical lymph node metastasis in patients with papillary thyroid carcinoma(PTC).To explore the value of combining radiomic features,computed tomography(CT)subjective diagnosis and clinical risk factors for predicting cervical lymph node metastasis in patients with PTC,in order to help clinicians to prepare personalized treatment plans and evaluate prognosis.Methods:1.A total of 147 patients with PTC confirmed by pathology,who underwent plain CT scan and dual-phase enhanced CT scan at our hospital between December 2017 and December 2019,were retrospectively included in this study.All PTCs were solitary lesions,and randomly divided at a ratio of 7:3(training sets: n=103,validation sets:n=44).The clinical data of the patients,such as age,gender,Hashimoto's thyroiditis,adenoma and nodular goiter,were collected.According to the pathology,the patients were divided into the lymph node metastasis group and non-metastasis group.2.Three doctors,with 3 years,7 years and 15 years experience in head and neck imaging diagnosis,independently evaluated whether or not PTC had lymph node metastasis according to the CT features,which were divided into three categories:metastatic,suspected and non-metastatic.3.The ITK-SNAP software was used to manually delineate CT plain scan,and the ROIs of arterial and venous phases of PTC foci layer-by-layer.The high-throughput,multi-level imaging features(including first-order intensity features,shape features,texture features and high-dimensional wavelet features,with a total of 851 features in each phase,and 2553 features per patient)were extracted.Upsampling,Z-score,mean normalization and Pearson's correlation coefficient method were used to preprocess the data,and homogenize the image.Analysis of variance and Kruskal-Wallis test were used to screen the features.Support vector machine(SVM),multiple logistic regression and LASSO regression were used to establish the radiomic model,CT subjective diagnosis model and joint model,and the diagnostic efficiencies of the three models were compared.4.A joint diagnosis model was constructed after combining the radiomic characteristics,subjective diagnosis of CT and clinical risk factors.The calibration,identification and clinical application of the line map were evaluated.Results:1.General data and lesion characteristics: age,location,capsule invasion,anterior and posterior diameters,left and right diameters,aspect ratio,CT diagnosis 1(physician with 15 years experience)and CT diagnosis 2(physician with seven years experience)of the PTC patients were recorded,and there were significant differences between the lymph node metastasis group and non-metastasis group(The chi-square values were 12.520,7.554,15.650,18.693,15.506,13.924,30.093,11.552,respectively,p<0.05).Three clinical models were constructed based on the characteristics with statistical differences.2.The radiomic model showed good performance in predicting cervical lymph node metastasis of PTC(training and validation cohorts,p < 0.01).In the validation group,the predictive performances of plain scan,arterial phase,venous phase and three-phase radiomic models were better than that of physicians' subjective diagnosis(AUC,0.788,0.788,0.783,0.792 vs.0.765).The combination of radiomics and clinical model was better than the radiologist's subjective diagnosis and other clinical models in predicting lymph node metastasis(AUC,0.848 vs.0.765,0.735,0.719).3.Multivariate logistic regression results: CT diagnosis 1,age,anterior-posterior diameter and aspect ratio were independent predictors of PTC lymph node metastasis(the odds ratio were 1.455,-1.301,2.544,1.456,respectively,p <0.0l),and a nomogram was constructed using multivariable logistic regression.4.The prediction model showed that the addition of radiomic features to the nomogram constructed with clinical factors improved the predicted value in the training group(the area under the curve(AUC),0.857 to 0.907),which was confirmed in the validation group(AUC,0.783 to 0.835).The calibration curve in the two sets fit well with the standard curve.Decision curve analysis showed that the combined model provided the greatest net benefit in predicting the status of lymph node metastasis in patients with PTC.Conclusion:The model with the combination of imaging and clinical factors was found to be superior to the diagnostic model of radiologists,and the addition of radiomic features to the nomogram improved the preoperative prediction of cervical lymph node metastasis in patients with PTC.