双期^(18)F-FDG PET/CT原发灶影像组学对非小细胞肺癌淋巴结转移的预测价值  

Predictive value of dual phase ^(18)F-FDG PET/CT radiomics of primary lesions for lymph node metastasis in non-small cell lung cancer

作　　者：李雨奇[1] 罗晓燕[1] 张鹏博 韩婷婷 盖立俊 LI Yu-qi;LUO Xiao-yan;ZHANG Peng-bo;HAN Ting-ting;GAI Li-jun(Department of Nuclear Medicine,Chifeng Municipal Hospital,Chifeng Inner Mongolia 024000,China)

机构地区：[1]内蒙古赤峰市医院核医学科,内蒙古赤峰024000

出　　处：《局解手术学杂志》2023年第2期142-145,共4页Journal of Regional Anatomy and Operative Surgery

基　　金：内蒙古自治区卫生健康委医疗卫生科技计划项目(202202330);赤峰市自然科学基金(SZR21161)。

摘　　要：目的探讨双期^(18)F-FDG PET/CT原发灶影像组学特征对非小细胞肺癌(NSCLC)淋巴结转移的预测价值。方法收集我院经手术病理证实为NSCLC并明确有无淋巴结转移的101例患者的术前双期^(18)F-FDG PET图像,自动勾画并提取原发灶常规及延迟显像影像组学特征各117个。按常规显像模式(a)、延迟显像模式(b)及常规+延迟显像模式(c)分别构建影像组学模型,采用t检验与LASSO回归筛选最优子集模型,建立支持向量机(SVM)模型与受试者工作特征(ROC)曲线,Delong检验不同模型曲线下面积(AUC)差异。结果常规+延迟显像模式(c)影像组学模型对淋巴结转移的预测效能最优,其灵敏度、特异度、AUC、训练组预测准确率、测试组预测准确率分别为88.06%、94.12%、0.967、92.86%和80.65%,高于常规显像模式(a)的86.57%、79.41%、0.882、80.00%和70.97%和延迟显像模式(b)的80.60%、88.24%、0.872、85.71%和70.97%(a vs.b Z=0.202,P=0.8399;a vs.c Z=2.385,P=0.0171;b vs.c Z=2.273,P=0.0230)。结论双期^(18)F-FDG PET/CT原发灶影像组学特征模型可用于预测NSCLC淋巴结转移,可指导临床治疗和预后判断。Objective To investigate the predictive value of dual phase ^(18)F-FDG PET/CT radiomics features of primary lesions for lymph node metastasis in non-small cell lung cancer(NSCLC).Methods The preoperative dual phase ^(18)F-FDG PET images of 101 patients who were confirmed as NSCLC by surgery and pathology and identified the presence or absence of lymph node metastasis in our hospital were collected,and 117 conventional and delayed imaging radiomics features of primary lesions were automatically delineated and extracted.The radiomics models were constructed according to conventional imaging mode(a),delayed imaging mode(b)and conventional+delayed imaging mode(c).The t-test and LASSO regression were used to screen the optimal subset model,the support vector machine(SVM)model and the receiver operating characteristic(ROC)curve were established,and Delong was used to test the difference of area under the curve(AUC)of different models.Results The conventional+delayed imaging mode(c)radiomics model had the best prediction efficiency for lymph node metastasis,with the sensitivity,specificity,AUC,prediction accuracy of training group and prediction accuracy of test group of 88.06%,94.12%,0.967,92.86%and 80.65%respectively,which were higher than 86.57%,79.41%,0.882,80.00%and 70.97%of the conventional imaging mode(a)and 80.60%,88.24%,0.872,85.71%,and 70.97%of delayed imaging mode(b)(a vs.b Z=0.202,P=0.8399;a vs.c Z=2.385,P=0.0171;b vs.c Z=2.273,P=0.0230).Conclusion Dual phase ^(18)F-FDG PET/CT radiomics features of primary lesions can be used to predict lymph node metastasis in NSCLC,and guide clinical treatment and prognosis judgement.

关 键 词：非小细胞肺癌 淋巴结转移 氟代脱氧葡萄糖 正电子发射断层显像技术 影像组学 

分 类 号：R817.4[医药卫生—影像医学与核医学]