中华普通外科杂志2024,Vol.39Issue(12) :929-934.DOI:10.3760/cma.j.cn113855-20240908-00576

超声影像组学联合术前临床超声语义特征预测甲状腺髓样癌侧颈部淋巴结转移的风险

Ultrasound radiomics combined with clinical and ultrasound semantic features to predict lateral cervical lymph node metastasis in medullary thyroid carcinoma

赵静 赵利辉 张蕾 邓艳 毛怡然 郑向前 魏玺
中华普通外科杂志2024,Vol.39Issue(12) :929-934.DOI:10.3760/cma.j.cn113855-20240908-00576

超声影像组学联合术前临床超声语义特征预测甲状腺髓样癌侧颈部淋巴结转移的风险

Ultrasound radiomics combined with clinical and ultrasound semantic features to predict lateral cervical lymph node metastasis in medullary thyroid carcinoma

赵静 1赵利辉 1张蕾 1邓艳 1毛怡然 1郑向前 2魏玺1
扫码查看

作者信息

  • 1. 天津医科大学肿瘤医院超声诊疗科 国家恶性肿瘤临床医学研究中心 天津市肿瘤防治重点实验室 天津市恶性肿瘤临床医学研究中心,天津 300060
  • 2. 天津医科大学肿瘤医院 甲状腺颈部肿瘤科 国家恶性肿瘤临床医学研究中心 天津市恶性肿瘤临床医学研究中心天津市肿瘤防治重点实验室,天津 300060
  • 折叠

摘要

目的 探讨甲状腺髓样癌超声影像组学特征联合临床超声语义特征对侧颈淋巴结转移的预测价值.方法 回顾性分析2015年1月至2023年1月于天津医科大学肿瘤医院初次手术并且术后病理证实为甲状腺髓样癌271例患者的临床及超声资料,其中侧颈淋巴结转移114例,无侧颈淋巴结转移157例.应用ITKSNAP在肿瘤最大横切面勾画感兴趣区,使用Pyradiomics提取影像组学特征.将患者按7:3比例随机分为训练集和验证集.采用最大相关最小冗余筛选最优特征,使用随机森林算法构建临床超声语义模型、超声影像组学模型以及融合模型3个模型.应用受试者工作特征(ROC)曲线评估3个模型的预测性能,应用Delong检验对比模型的预测性能.结果 临床超声语义模型、超声影像组学模型和融合模型曲线下面积(AUC)、敏感性、特异性及准确性分别为0.78,65.8%,78.9%,73.2%;0.75,60.8%,69.8%,65.8%;0.84,70.1%,81.6%.77.0%.Delong检验显示,甲状腺髓样癌侧颈淋巴结转移预测的ROC曲线,临床超声语义模型与超声影像组学模型的AUC差异无统计学意义(Z=0.731,P=0.465),而融合模型AUC优于临床超声语义模型(Z=-2.126.P=0.034)和超声影像组学模型(Z=-2.318,P=0.021).结论 超声影像组学联合临床超声语义特征对甲状腺髓样癌侧颈淋巴结转移风险具有较好的术前预测价值.

Abstract

Objective To investigate the value of a model combined ultrasound radiomics with clinical and ultrasound image semantic features in predicting lateral cervical lymph node metastasis of medullary thyroid carcinoma(MTC)Methods The clinical and ultrasound imaging data of 271 patients with MTC admitted at Tianjin Medical University Cancer Institute and Hospital from Jan 2015 to Jan 2023 were retrospectively analyzed.Among those there were 114 cases with lateral cervical lymph node metastasis and 157 cases.The region of interest was delineated by ITKSNAP in the maximum transverse section of the tumor and the radiomics features were extracted by Pyradiomics.The patients were randomly divided into training set and validation set in a ratio of 7:3.Maximum correlation minimum redundancy was screened for stable radiomics features.The random forest algorithm was applied to construct clinical ultrasound semantics model,ultrasound radiomics model,and combined model,The predictive performance of each model was evaluated by the area under the ROC curve(AUC).The predictive ability of the three models was compared by Delong test.Results The AUC,sensitivity,specificity,and accuracy of clinical ultrasound semantics model,ultrasound radiomics model,and combined model were 0.78,65.8%,78.9%,73.2%;0.75,60.8%,69.8%,65.8%;0.84,70.1%,81.6%,77.0%respectively.Delong Test showed that the difference of AUC between clinical ultrasound semantics model and ultrasound radiomics model was no statistical difference(Z=0.731,P=0.465).The AUC of combined model was better than clinical ultrasound semantics model(Z=-2.126,P=0.034)and ultrasound radiomics model(Z=-2.318,P=0.021).Conclusion The model combined clinical characteristics,biochemical indexes,ultrasound semantic findings and ultrasound radiomics features has good performance in predicting lateral cervical lymph node metastasis of MTC.

关键词

甲状腺肿瘤/淋巴转移/超声检查/预测

Key words

Thyroid neoplams/Lymphatic metastasis/Ultrasonography/Forecasting

引用本文复制引用

出版年

2024
中华普通外科杂志
中华医学会

中华普通外科杂志

CSTPCDCSCD北大核心
影响因子:1.11
ISSN:1007-631X
段落导航相关论文