构建基于超声特征列线图预测甲状腺乳头状癌滤泡亚型的侵袭性  

Prediction of invasiveness in follicular variant of papillary thyroid carcinoma using nomograms based on ultrasonic features

作　　者：郑雨欣 张雅娇 陈丽羽[2] 鲁科峰[3] 楼江燕[4] 刘俊平[2,5] 徐栋 Zheng YuXin;Zhang Yajiao;Chen Liyu;Lu Kefeng;Lou Jiangyan;Liu Junping;Xu Dong(Second Clinical Medical College,Zhejiang Chinese Medical University,Hangzhou 310022,China;Department of Diagnostic Ultrasound Imaging&Interventional Therapy,Zhejiang Cancer Hospital,Hangzhou 310022,China;Ultrasound Medicine Department,Zhejiang Provincial People′s Hospital,Hangzhou 310014,China;Department of Pediatrics,Zhejiang Provincial People′s Hospital,Hangzhou 310014,China;Hangzhou Institute of Medical Research,Chinese Academy of Sciences,Hangzhou 310018,China;Taizhou Key Laboratory of Minimally Invasive Intervention and Big Data Artificial Intelligence,Taizhou 317502,China)

机构地区：[1]浙江中医药大学第二临床医学院,杭州310022 [2]浙江省肿瘤医院超声医学科,杭州310022 [3]浙江省人民医院超声医学科,杭州310014 [4]浙江省人民医院儿科,杭州310014 [5]中国科学院杭州医学研究所,杭州310018 [6]台州市微创介入和大数据人工智能重点实验室,台州317502

出　　处：《中华超声影像学杂志》2024年第9期800-806,共7页Chinese Journal of Ultrasonography

基　　金：国家自然科学基金(82071946);国家卫生健康委能力建设和继续教育中心(CSJRZC2021JJSJ001);浙江省"尖兵""领雁"研发攻关计划(2023C04039);浙江省自然科学基金(LTGY24H180012);浙江省医药科技一般项目(2023KY592,2023KY030,2024KY685,2024KY832)。

摘　　要：目的:探讨甲状腺乳头状癌滤泡亚型(FVPTC)超声特征与侵袭性的关系,并利用列线图整合多个超声参数可视化评估预测结果。方法:回顾性收集2013年1月至2023年12月于浙江省肿瘤医院和浙江省人民医院经外科病理确诊的312例FVPTC患者的资料,依据定义将FVPTC分为高侵袭性组和低侵袭性组,并以7∶3比例分成训练集和验证集,收集患者的临床信息和超声特征参数。对训练集进行单因素和多因素Logistic回归分析,基于超声特征构建FVPTC侵袭性预测模型,在验证集中评价模型的区分度和校准度,并绘制列线图。结果:训练集共纳入218例FVPTC患者,其中131例为高侵袭性;验证集有94例患者,其中53例FVPTC为高侵袭性。训练集多因素Logistic回归分析显示,肿瘤多灶性(OR=6.505,P=0.016)、低回声(OR=3.235,P=0.103)、形态(OR=0.521,P=0.049)和微钙化(OR=2.479,P=0.004)是预测FVPTC侵袭性的独立影响因素。训练集中,超声预测模型的曲线下面积(AUC)为0.704(95%CI=0.634~0.771),验证集的AUC为0.650(95%CI=0.531~0.770),体现了良好的区分度,校准曲线与理想曲线吻合表现出较好的校准度。结论:超声特征能为评估FVPTC的侵袭性提供有价值的信息,结合超声特征构建的模型对FVPTC侵袭性具有较好的预测效能。ObjectiveTo explore the relationship between ultrasound characteristics and invasiveness in the follicular variant of papillary thyroid carcinoma(FVPTC),and to integrate multiple ultrasound parameters for visual assessment of predictive outcomes by using Nomogram.MethodsA total of 312 FVPTC patients who were pathologically confirmed through surgery in Zhejiang Cancer Hospital and Zhejiang Provincial People′s Hospital from January 2013 to December 2023 were retrospectively collected.Based on defined criteria,FVPTC patients were categorized into high-invasion and low-invasion groups.The dataset was divided into a training set and a validation set in a ratio of 7 to 3.Clinical information and ultrasound feature parameters were collected.Univariate and multivariate Logistic regression analyses were performed on the training set.A predictive model for FVPTC invasiveness was constructed based on ultrasound features.The model′s discriminative ability and calibration were evaluated in the validation set,and a nomogram was generated.ResultsThe training set included a total of 218 patients with FVPTC,among which 131 were classified as high invasive.The validation set consisted of 94 patients,with 53 cases of high invasive FVPTC patients.Multivariate logistic regression analysis on the training set revealed that tumor multifocality(OR=6.505,P=0.016),hypoechoic(OR=3.235,P=0.103),shape(OR=0.521,P=0.049),and microcalcifications(OR=2.479,P=0.004)were independent influencing factors for predicting invasiveness in FVPTC.In the training set,the area under the curve(AUC)of the ultrasound predictive model was 0.704(95%CI=0.634-0.771),and in the validation set,the AUC was 0.650(95%CI=0.531-0.770),indicated good discriminative ability.The calibration curve showed good alignment with the ideal curve,demonstrating favorable calibration performance.ConclusionsUltrasound features provide valuable information for assessing the invasiveness of FVPTC,and the model constructed by combining ultrasound features demonstrates good predictive

关 键 词：超声检查 甲状腺乳头状癌滤泡亚型 列线图 预测模型 

分 类 号：R445.1[医药卫生—影像医学与核医学] R736.1[医药卫生—诊断学]