Study On Predicting Invasiveness And Lymph Node Status Of Thyroid Cancer Using Machine Learning Based On Ultrasound Images

Part I Nomograms Combining Ultrasonic Features with Clinical and Pathological Features for Estimation of Delphian Lymph Node Metastasis Risk in Papillary Thyroid CarcinomaObjectives:The significance of Delphian lymph node(DLN)in papillary thyroid carcinoma(PTC)is still unclear.This work explores the clinical significance of DLN.At the same time,a nomogram is constructed based on clinical,pathological and ultrasonic(US)features to evaluate the possibility of DLN metastasis(DLNM)in PTC patients.Materials and Methods:1.Patients surgically diagnosed with PTC between February 2017 and June2021,all of whom underwent thyroidectomy,were retrospectively included in the study.Using the clinical,pathological and US information of patients,the related factors of DLNM were retrospectively analyzed.2.Randomly selected 80% of patients as the training set,and the remaining20% of patients as the validation set.The risk factors associated with DLNM were identified through Univariate and multivariate logistics analyses.3.According to clinical + pathology,clinical + US and clinical + US +pathology,the predictive nomograms for DLNM were established and validated.Results:1.Of the 485 patients with DLN,98(20.2%)exhibited DLNM.The DLNM positive group had higher positive rates of central lymph node metastasis(CLNM),lateral lymph node metastasis(LLNM)and T3b-T4 b thyroid tumors than the negative rates.The number of CLNM and LLNM lymph nodes in DLNM + group was higher as compared to that in DLNM-group.2.In training set,multivariate analysis demonstrated that the common independent risk factors of the three prediction models were male,bilaterality,located in the isthmus.Age≥45 years,located in the lower pole,and nodural goiter were protective factors.In addition,the independent risk factors were classified as follows:(I)P-extrathyroidal extension(ETE)and CLNM based on clinical + pathological characteristics;(II)US-ETE and US-CLNM based on clinical + US characteristics;(III)US-ETE and CLNM based on clinical +US + pathological features.3.In validation set,better diagnostic efficacy was reported with clinical +pathology + US diagnostic model than that of clinical + pathology diagnostic model[AUC 0.88(95% CI:0.79,0.97)vs.0.79(95% CI:0.67,0.91),P=0.040].However,there was no significant difference between clinical + pathology + US diagnostic model and clinical + US diagnostic model[AUC 0.88(95% CI:0.79,0.97)vs.0.87(95% CI:0.79,0.94),P=0.724].Conclusions:This study found that DLNM may be a sign that PTC is more invasive and has extensive lymph node metastasis.By exploring the clinical,pathology and US characteristics of PTC progression to DLNM,three prediction nomograms,established according to different combinations of features,can be used in different situations to evaluate the transfer risk of DLN.Part II Ultrasound Image-based peri-nodular radiomic nomogram for preoperative prediction of cervical lymph node status in papillary thyroid carcinomaObjectives:This study aimed to develop a model based on ultrasound(US)radiomic features extracted from the peri-nodular region to predict cervical lymph node(LN)status in papillary thyroid carcinoma(PTC)patients.Materials and Methods:1.A total of 402 PTC patients(247,107,and 48 in the derivation,internal,and external validation set,respectively)from two institutions between January 2019 and March 2021 were enrolled in the retrospective study.2.Radiomic features quantitatively extracted from ultrasound(US)images of intra-and peri-nodular regions.In the derivation set,through the selection and machine learning classifier training of radiomic features,developed the intra-,peri-nodular,and gross radiomic signature scores(rad-scores).3.Using the peri-nodular rad-score and clinical factors to construct a radiomic nomogram,and testing its predictive performance in both internal and external validation sets,including discrimination,calibration,and clinical usefulness.Results:1.The peri-nodular rad-score showed good discrimination with areas under the receiver operating characteristic curve(AUCs)of 0.80(95%CI:0.71,0.88)and 0.83(95%CI:0.72,0.95)in the internal and external validation set.Which were similar to the AUCs of gross rad-score [0.82(95% CI: 0.74,0.89)and 0.79(95% CI: 0.66,0.92)],and significantly higher than the AUCs of intra-nodular rad-score [0.73(95%CI: 0.63,0.82)and 0.69(95% CI: 0.53,0.84)].2.The radiomic nomogram based on the peri-nodular rad-score showed good discrimination [AUC 0.87(95% CI: 0.80,0.94)and 0.85(95% CI: 0.74,0.96)in the internal and external validation set,respectively] and calibration capabilities,and also has excellent performance in decision curve analysis.3.Nomogram has demonstrated excellent prediction for lateral,US-reported negative(c N0),and large-volume central LN metastasis [AUCs 0.78(95%CI:0.71,0.85),0.80(95%CI:0.75,0.86)and 0.80(95%CI:0.74,0.85)] indicated that the nomogram was clinically valuable.Conclusions:Radiomic features extracted from the peri-nodular region show a high predictive value for cervical LN metastasis.The radiomic nomogram can be an accurate tool for individualized preoperative prediction of LN staging in patients with PTC.Part III A Retrospective Multicenter Study of Ultrasound Image-based Deep Learning to Assist in Diagnosing Gross Extrathyroidal Extension Thyroid CancerObjectives:The presence of gross extrathyroidal extension(ETE)in thyroid cancer will affect the prognosis of patients,but imaging examination cannot provide a reliable diagnosis for it.This study was conducted to develop a deep learning(DL)model for localization and evaluation of thyroid cancer nodules in ultrasound images before surgery for the presence of gross ETE.Materials and Methods:1.Grayscale ultrasound images of 806 thyroid cancer nodules(4451 images)from 4 medical centers were retrospectively analyzed,including 517 no gross ETE nodules and 289 gross ETE nodules.2.283 no gross ETE nodules and 158 gross ETE nodules were randomly selected from the internal dataset to form a training set and validation set(2914images),and a multitask DL model was constructed for diagnosing gross ETE.In addition,the clinical model and the clinical and DL combined model were constructed.3.In the internal test set [974 images(139 no gross ETE nodules and 83 gross ETE nodules)]and the external test set [563 images(95 no gross ETE nodules and 48 gross ETE nodules)],the diagnostic performance of DL model was verified based on the pathological results.And then,compared the results with the diagnosis by 2 senior and 2 junior radiologists.Results:1.In the internal test set,DL model demonstrated the highest AUC [0.91(95%CI: 0.87,0.96)],which was significantly higher than that of two senior radiologists[(AUC 0.78(95% CI: 0.71,0.85),P<0.001)and(AUC 0.76(95% CI:0.70,0.83),P<0.001)]and two juniors radiologists [AUC 0.65(95% CI:0.58,0.73),P<0.001 and AUC 0.69(95% CI:0.62,0.77),P<0.001].DL model was significantly higher than clinical model [AUC 0.84(95% CI: 0.79,0.89),P=0.019],but there was no significant difference between DL model and clinical and DL combined model [AUC0.94(95% CI: 0.91,0.97),P = 0.143].2.In the external test set,DL model also demonstrated the highest AUC[0.88(95% CI: 0.81,0.94],which was significantly higher than that of one of senior radiologists [(AUC 0.75(95% CI: 0.66,0.84),P = 0.008)and(AUC,0.81(95%CI:0.72,0.89),P=0.152)]and two junior radiologists[(AUC 0.72(95%CI:0.62,0.81),P=0.002)and(AUC 0.67(95% CI:0.57,0.77),P<0.001)].There was no significant difference between DL model and clinical model [AUC 0.85(95% CI: 0.79,0.91),P=0.516] and clinical+DL model [AUC 0.92(95% CI: 0.87,0.96),P=0.093].3.Using DL model,the diagnostic ability of two junior radiologists was significantly improved.Conclusions:The DL model based on ultrasound imaging is a simple and helpful tool for preoperative diagnosis of gross ETE thyroid cancer,and its diagnostic performance is equivalent to or even better than that of senior radiologists.