摘要
目的探讨影像组学对胰腺导管腺癌(pancreatic ductal adenocarcinoma,PDAC)患者术后胰瘘、感染及长期生存状况的预测价值。方法回顾性选择2014年1月至2020年12月于复旦大学附属中山医院接受胰腺癌根治性切除手术且术后病理证实为PDAC的患者206例,均具有完整的手术资料及长期随访资料。采用pyradiomics分析所有患者的增强CT图片,提取影像组学特征。采用LASSO降维联合logistic回归分析构建PDAC术后胰瘘和腹腔感染的预测模型,并用ROC曲线评估模型效能;采用LASSO降维联合Cox回归分析构建PDAC患者长期生存预测模型并计算患者风险评分,按中位数分成高风险组和低风险组,比较生存曲线从而评估模型效能。将权重最高的影像组学特征按照中位数分成高表达组和低表达组,比较预后差异及临床特征。结合影像组学特征和临床特征,分析长期预后的影响因素并构建临床-影像综合模型。结果共提取1595个影像组学特征。成功构建PDAC术后胰瘘和感染预测模型,其AUC值分别为0.81和0.79;成功构建PDAC长期生存预测模型,且高风险组较低风险组预后差(P<0.001)。影像组学特征“log-sigma-5-mm-3D_glszm_ZonePercentage”的权重高达59.557,高表达组的CA19-9高于低表达组(P=0.017),且两组生存曲线差异有统计学意义(P=0.021)。临床-影像组学综合模型提示年龄、AJCC分期、淋巴浸润、CA19-9水平和影像组学特征是PDAC患者长期预后的危险因素(HR=1.028、4.084、2.566、1.232和2.536)。结论基于影像组学的预测模型对PDAC术后胰瘘、感染及长期预后具有良好的预测效能,影像组学特征“log-sigma-5-mm-3D_glszm_ZonePercentage”高表达的患者面临更差的预后。
Objective To explore the predictive value of radiomics for postoperative pancreatic fistula,infection,and long-term survival in patients with pancreatic ductal adenocarcinoma(PDAC).Methods 206 patients who received radical resection of pancreatic cancer in Zhongshan Hospital,Fudan University from January 2014 to December 2020 and were pathologically confirmed as PDAC after surgery were retrospectively selected,all of whom had complete surgical data and long-term follow-up data.Pyradiomics was used to analyze the enhanced CT images of all patients and extract radiomics features.LASSO dimensionality reduction combined with logistic regression analysis was used to construct a predictive model for pancreatic fistula and abdominal infection after PDAC surgery,and evaluating the model’s effectiveness using ROC curves.A long-term survival prediction model for PDAC patients was constructed using LASSO dimensionality reduction combined with Cox regression analysis,and patient risk scores were calculated.The patients were divided into high-risk and low-risk groups based on the median,and the survival curves were compared to evaluate the effectiveness of the model.The imaging omics features with the highest weight were divided into high expression group and low expression group according to the median,and the prognostic differences and clinical features were compared.Radiomics and clinical features were combined to analyze the influencing factors of long-term prognosis and construct a clinical imaging comprehensive model.Results A total of 1595 radiomics features were extracted.A predictive model for pancreatic fistula and infection after PDAC surgery was constructed,with AUC values of 0.81 and 0.79,respectively.The PDAC long-term survival prediction model was successfully constructed,and the prognosis of the high-risk group was worse than that of the low-risk group(P<0.001).The weight of the radiomics feature“log-sigma-5-mm-3D_glszm_Zone Percentage”was 59.557.The CA19-9 level in the high expression group is higher than that in the low expression group(P=0.017),and there is a statistically significant difference in survival curves between the two groups(P=0.021).The comprehensive clinical imaging model suggested that age,AJCC stage,lymph infiltration,CA19-9 level and imaging characteristics were risk factors for long-term prognosis of PDAC patients(HR=1.028,4.084,2.566,1.232 and 2.536).Conclusions The predictive model based on radiomics has good predictive performance for pancreatic fistula,infection,and long-term prognosis after PDAC surgery.Patients with high expression of the radiomics feature“log-sigma-5-mm-3D_glszm_ZonePercentage”face poorer prognosis.
作者
韩思阳
许耀麟
徐天威
戎叶飞
楼文晖
王单松
HAN Siyang;XU Yaolin;XU Tianwei;RONG Yefei;LOU Wenhui;WANG Dansong(Department of Emergency Surgery,Zhongshan Hospital,Fudan University,Shanghai 200032,China;Department of Pancreatic Surgery,Zhongshan Hospital,Fudan University,Shanghai 200032,China)
出处
《中国临床医学》
2025年第6期973-981,共9页
Chinese Journal of Clinical Medicine
基金
国家自然科学基金(8217110888)。
关键词
胰腺导管腺癌
影像组学
术后并发症
预测
pancreatic ductal adenocarcinoma
radiomics
postoperative complications
prediction
