Predicting Extra-thyroidal Extension And Cervical Lymph Node Metastasis Of Papillary Thyroid Carcinoma Using Radiomics

PART ONE: Constructing a Clinical-Ultrasound Radiomics Nomogram for the Prediction of Extra-Thyroidal Extension in Papillary Thyroid Carcinoma Research Background and Purpose: Papillary Thyroid Carcinoma(PTC)is the most typical type of well-differentiated thyroid carcinoma.Some PTCs have highly invasive biological behaviors and extra-thyroidal extension(ETE)is one of the important biohistological features.In this study,a clinical-ultrasound radiomics nomogram was constructed for noninvasive preoperative prediction of ETE in patients with PTC,to select highly invasive PTC and improve treatment and prognosis evaluation.Research Materials and Methods: Between January 2016 and January 2020,161 patients with PTC who underwent preoperative ultrasound(US)examination in the Jiangsu University Affiliated People’s Hospital were enrolled in this retrospective study.Based on the postoperative pathological results,the enrolled patients were assigned to an ETE group(n=93)and a non-ETE group(n=63).All patients were divided into a training cohort(n=97)and a validation cohort(n=64)randomly.A total of 479 radiomics features of tumor areas in US images were extracted.The radiomics signature was developed using the least absolute shrinkage and selection operator algorithms(LASSO)after feature selection using the minimum redundancy maximum relevance(m RMR)method.The clinical-ultrasound radiomics nomogram model was constructed by multivariable logistic regression analysis using the radiomics signature and clinical risk factors.The clinical usefulness,discrimination and calibration of the nomogram model were evaluated in the training and validation cohorts respectively.Research Results: The radiomics signature consisted of six radiomic features from US images.The radiomics nomogram included the radiomics signature,the parameters tumor location and radiological ETE that had statistical significance selected from all the clinical factors.AUC values confirmed good discrimination of this nomogram in the training cohort[AUC,0.837;95% confidence interval(CI),0.756–0.919] and the validation cohort(AUC,0.824;95% CI,0.723-0.925).The decision curve analysis(DCA)showed that the nomogram has good clinical application value.Conclusion: The constructed clinical-ultrasound radiomics nomogram model is a noninvasive,reliable and accurate tool for predicting ETE in PTC patients.PART TWO: Constructing Dual-Modal(US/Dual-Energy Computed Tomography)Radiomics for Predicting Cervical Lymph Node Metastasis in Papillary Thyroid CarcinomaResearch Background and Purpose: Cervical Lymph Node Metastasis(CLNM)is more likely to occur with highly invasive PTC.Current imaging examinations cannot identify all CLNM because of the enormous number of cervical lymph nodes,and CLNM is an independent risk factor for postoperative recurrence.Therefore,preoperative prediction of CLNM in patients with PTC is of great significance for surgical decision-making.The purpose of this study was to construct a dual-modal radiomics(DMR)model based on grayscale ultrasound(GSUS)and dual-energy computed tomography(DECT)for noninvasive preoperative prediction of CLNM in PTC.Research Materials and Methods: Between January 2021 and February 2022,348 patients with pathologically confirmed PTC at Jiangsu University Affiliated People’s Hospital who completed preoperative US and DECT examinations were enrolled and assigned to training(n=261)and test(n=87)cohorts randomly.The enrolled patients were divided into two groups based on pathology findings: CLNM(n=179)and CLNM-Free(n=169).Radiomics features were extracted from GSUS images(464 features)and DECT images(960 features)respectively.Pearson correlation coefficient(PCC)and LASSO regression with 10-fold cross-validation were then used to choose CLNM-related features.Based on the selected features,GSUS,DECT,and dual-modal(GSUS/DECT)radiomics models were constructed by using a linear support vector machine(L-SVM)and kernel ensemble support vector machine(KE-SVM).Research Results: Three radiomics models were constructed using L-SVM,including GSUS,DECT,and dual-modal(GSUS/DECT),with AUC 0.700(95% CI,0.662-0.706),AUC 0.721(95% CI,0.683-0.727),and AUC 0.760(95% CI,0.728-0.762)in the training dataset,AUC 0.643(95% CI,0.582-0.734),AUC 0.680(95% CI,0.623-0.772),and AUC 0.744(95% CI,0.686-0.784)in the test dataset,respectively.The dual-modal(GSUS/DECT)radiomics prediction model outperformed both the GSUS and the DECT predictive models separately.To further improve the performance of models,the KE-SVM classifier was used to construct three radiomics models,including GSUS,DECT and dual-modal(GSUS/DECT).The AUC of the training dataset was 0.728(95% CI,0.694-0.731),AUC 0.777(95% CI,0.739-0.781)and AUC 0.786(95% CI,0.745-0.789),respectively.The AUC of the test dataset was 0.688(95%CI,0.634-0.723),AUC 0.748(95%CI,0.681-0.814),and AUC 0.769(95%CI,0.701-0.832),respectively.The performance of the three radiomics models constructed by the KE-SVM classifier outperformed that of the L-SVM classifier,and the dual-modal(GSUS/DECT)radiomics model outperformed the prediction model of the GSUS and DECT single-modal radiomics models.Conclusion: The constructed dual-modal(GSUS/DECT)radiomics model may be able to predict CLNM in PTC patients and help in surgery planning.