摘要
目的:探讨CT纹理分析联合临床病理信息在预测进展期胃癌术后复发中的价值。方法:回顾性搜集经手术病理确诊的胃癌患者162例,随机分为训练集(n=100)及验证集(n=62)。从术前静脉期CT图像中提取纹理特征,先对训练集和验证集特征进行U检验及组间相关系数检验,得到一致性较好的特征(n=48),再采用LASSO回归方法筛选出与复发相关的特征(n=8),进而构建纹理标签。通过筛选临床病理指标构建临床模型,再与纹理标签构建组合模型。采用受试者工作特征(ROC)曲线和决策曲线分析(DCA)评价不同模型的预测效果。最后以纹理标签中位数值为界将病例分为两组,并进行无瘤生存率分析(DFS)。结果:胆汁酸返流、TNM分期、脉管侵犯等3个指标纳入临床模型,而组合模型包含胆汁酸返流及纹理标签两个因子。临床模型、纹理标签及组合模型预测胃癌术后复发的曲线下面积分别为0.748、0.809及0.841。阈值概率为0.31~0.87时,组合模型预测胃癌复发的效果更佳。小于纹理标签中位数与大于中位数患者的5年DFS差异具有统计学意义(P<0.05)。结论:CT纹理分析可用于胃癌患者术后复发风险分层,联合纹理标签及胆汁酸返流的组合模型预测效果更佳,有助于评估患者预后。
Objective:To explore the value of CT texture analysis combined with clinicopathological information in predicting postoperative recurrence of advanced gastric cancer(GC).Methods:A total of 162 patients with GC confirmed by surgical pathology were retrospectively collected and randomly divided into training set(n=100)and validation set(n=62).Texture features were extracted from preoperative venous CT images.The U test and inter-group correlation coefficient(ICC)test were tested to evaluate the features of training set and validation set,and the features with good consistency(n=48)were obtained.The texture signature were constructed with the features related to recurrence(n=8)screened by LASSO regression.The clinical model was constructed by screening the clinicopathological indexes,and the combined model was constructed by the texture signatures and selected clinicopathological indexes.The prediction effect of different models were analyzed using receiver operating characteristic(ROC)curve and decision curve analysis(DCA).Finally,the patients were divided into two groups based on the median value of texture signature,and the disease-free survival rate(DFSR)was analyzed.Results:Bile acid reflux,TNM staging and vascular invasion were enrolled in the clinical model,while for the combined model,bile acid reflux and texture signature were enrolled.The area under the curve(AUC)of clinical model,texture signature and combined model for predicting postoperative recurrence of GC were 0.748,0.09 and 0.841,respectively.The combined model was more effective in predicting GC recurrence with the threshold probability of 0.31~0.87.There was a statistically significant difference in 5-year DFSR between patients with less than the median value of texture signature and those with more than the median value of texture signature(P<0.05).Conclusion:CT texture analysis can be used to stratify the risk of postoperative recurrence in patients with GC.The model combined texture signature with bile acid reflux has a better prediction performance,which is helpful to evaluate the prognosis of patients.
作者
黄列彬
刘昱
黄文斯
陈钦贤
薛慧敏
冯宝
龙晚生
李荣岗
HUANG Lie-bin;LIU Yu;HUANG Wen-si(Department of Radiology,Jiangmen Central Hospital,Guangdong 529030,China)
出处
《放射学实践》
CSCD
北大核心
2022年第2期209-214,共6页
Radiologic Practice
基金
江心医[2020]331号(J202005)。
关键词
纹理分析
体层摄影术
X线计算机
胃肿瘤
复发
模型构建
Texture analysis
Tomography,X-ray computed
Gastric tumor
Recurrence
Model building
