摘要
目的探讨术前CT增强静脉期影像组学对胃癌患者脉管侵犯的预测价值。方法选取徐州医科大学经手术治疗病理证实为胃癌的患者409例,依据术后病理结果分为脉管侵犯阳性组(267例)及脉管侵犯阴性组(142例),将其随机(7:3)分为训练组(286)和验证组(123),分析两组临床、CT影像特征,使用logistic回归分析筛选出脉管侵犯的独立危险因素,勾画ROI,提取并筛选出最优特征子集,建立影像组学模型及临床联合模型,应用受试者工作特征曲线下面积(AUC)值及决策分析法(DCA)比较模型间的诊断效能。结果共筛选出7个最优特征构建影像组学标签,训练组联合模型ACU值为0.830(95%CI:0.781~0.879),验证组联合模型ACU值为0.853 (95%CI:0.781~0.925),效能均高于临床模型及组学模型,两组间的差异具有统计学意义,P<0.05。结论基于CT增强影像组学模型可以较好的在术前预测胃癌脉管侵犯,影像组学标签联合临床标签建立的联合模型具有更好的预测效能。
Objective To investigate the value of preoperative CT enhanced venous phase imaging in predicting lymphovascular invasion in patients with gastric cancer.Methods A total of 409 patients with gastric cancer confirmed pathologically by surgical treatment in Xuzhou Medical University were selected and divided into positive group and negative group of vascular invasion(267 cases)and negative group of vascular invasion(142 cases),according to postoperative pathological results.The two groups were divided into the training group(286)and the verification group(123).The clinical and CT image features of the two groups were analyzed,and the independent risk factors of lymphovascular invasion were screened out by logistic regression analysis.Outlined the ROI and the optimal feature subset was extracted and screened,and the imaging model were established.The area under receiver operating characteristic curve(AUC)value and decision analysis method(DCA)were used to compare the diagnostic efficiency of the models,and the Delong test was used to compare the AUC differences among the model,and P<0.05 was statistically significant.Results A total of 7 optimal features were selected to construct radiomics model.The AUC value of the combined model of the training group was 0.830(95%CI:0.781~0.879),the verification group was 0.853(95%CI:0.781~0.925),which was higher than the clinical model and the radiomics model,and the difference between the two groups was statistically significant(P<0.05).Conclusion CT enhanced radiomics model can better predict the lymphovascular invasion of gastric cancer before operation,and the combined model established by radiomics label and clinical label has better prediction efficiency.
作者
董玉莹
荣玉涛
张银洋
胡春峰
张心然
窦莉
DONG Yu-ying;RONG Yu-tao;ZHANG Yin-yang;HU Chun-feng;Zhang Xin-ran;Dou li(Department of Medical imaging,The Affiliated Hospital of Xuzhou Medical University,Xuzhou 221000,Jiangsu Province,China)
出处
《中国CT和MRI杂志》
2025年第8期150-153,共4页
Chinese Journal of CT and MRI
关键词
胃癌
影像组学
脉管侵犯
Gastric Cancer
Radiomics
Lymphovascular Invasion
