Predictive value of ultrasound risk model combined with CT for central district lymph node metastasis of papillary thyroid carcinoma
10.16066/j.1672-7002.2023.12.002
- VernacularTitle:超声风险模型联合CT检查对甲状腺乳头状癌中央区淋巴结转移的预测价值
- Author:
Xiaofen YE
1
,
2
;
Qiong CHEN
;
Yuegui WANG
;
Ling LI
;
Haolin SHEN
Author Information
1. 福建医科大学附属漳州市医院超声医学科,福建 漳州 363000
2. 福建医科大学临床医学部,福建 福州 320000
- Keywords:
Thyroid Neoplasms;
Ultrasonography;
Forecasting;
central district lymph node;
computed tomography
- From:
Chinese Archives of Otolaryngology-Head and Neck Surgery
2023;30(12):753-757
- CountryChina
- Language:Chinese
-
Abstract:
OBJECTIVE To construct a predictive model of ultrasound(US),analyze and compare with its diagnostic efficacy in different forms of combination with CT of the neck in predicting central district lymph node metastasis(CLNM)of papillary thyroid carcinoma.METHODS Lesions confirmed as PTC by surgical pathology in our hospital from January 2021 to December 2021 were included for study,The lesions were scored according to the American college of radiology thyroid imaging reporting and data system(ACR TI-RADS),the risk model to predict CLNM was constructed based on the training set and verified internally in the testing set.The model was combined with CT to diagnose CLNM using both serial and parallel modes.The receiver operating characteristic(ROC)curves of CT,model,and model combined with CT for the diagnosis of CLNM was drawn separately,and then calculated and compared the area under the curve(AUC).RESULTS A total of 470 lesions were included in 440 patients.The model can be presented as Y=-4.664+0.171 ×maximum diameter+0.685×gender+0.600×multifoca lity+0.251×ACR TI-RADS score.After ROC curves analysis,the optimal diagnostic cut-off value of the model was 0.407.When Y≥ 0.407(optimal diagnostic cut-off point),CLNM was considered to be positive.In the training set,the C-index of model was 0.780(95%CI:0.661-0.756).In the testing set,the C-index was 0.778(95%CI:0.682-0.874).The Homsmer-Lemeshow goodness-of-fit test showed that the calibration of the model was good(P=0.294,P=0.879).In the testing set,compared with CT,model and the serial mode,the sensitivity(77.5%),diagnostic coincidence rate(80.6%),and negative predictive value(84.6%)of parallel mode were higher,whereas specificity(83.0%)was relatively lower.On diagnostic CLNM,the parallel mode had a higher AUC than the series mode(0.803 vs.0.669,Z=-2.931,P=0.003).CONCLUSION The model combined with CT in parallel mode can improve the clinical accuracy of diagnosis in CLNM and compensate for the shortcomings of traditional imaging techniques such as US and CT,which has specific clinical applicability.