摘要
目的 探究治疗前基线^(18)F-脱氧葡萄糖(FDG)PET/CT影像组学特征对中晚期非小细胞肺癌(NSCLC)患者疗效的预测价值。方法 回顾性收集2017年1月至2021年12月在杭州市肿瘤医院接受治疗前^(18)F-FDG PET/CT检查的107例中晚期NSCLC患者资料,其中男性70例、女性37例,年龄(63.6±9.9)岁。治疗6个月后依据疗效评价分为临床获益(CB)组(61例)和疾病进展(PD)组(46例)。按7∶3比例采用留出法将患者分为训练集(74例)及测试集(33例)。使用PyRadiomics工具包(Python 3.9.7)提取PET与CT双模态影像组学特征。经Mann-Whitney U检验(P<0.05)初步筛选后,通过LASSO回归(λ=0.01)构建影像组学标签评分(Rad-score)。采用单因素及多因素Logistic逐步回归筛选与疗效相关的PET代谢参数及临床病理因素,构建临床模型。最终整合Rad-score、PET代谢参数及临床变量构建多参数复合预测模型。采用受试者工作特征(ROC)曲线评估各模型的区分效能,并通过Delong检验比较模型间AUC值差异。结果 共筛选出6个影像组学特征(PET与CT各3个)用于构建Rad-score。Rad-score在训练集与测试集的AUC值分别为0.839(95%CI:0.743~0.935)和0.789(95%CI:0.625~0.953),高于临床模型(基于肿瘤长径与SUL_(min)建立)的AUC值(训练集0.670,95%CI:0.544~0.796;测试集0.714,95%CI:0.531~0.897),但差异无统计学意义(P>0.05)。综合Rad-score、PET传统代谢参数及临床病理因素构建的复合模型显示出最佳的预测效能,训练集AUC值=0.852(95%CI:0.759~0.945),测试集AUC=0.793(95%CI:0.630~0.956)。其训练集灵敏度、特异性、准确性分别为66.22%、86.49%与77.03%,测试集相应为57.58%、78.79%与69.70%。然而,复合模型与Rad-score、临床模型之间的AUC值差异均无统计学意义(均P>0.05)。结论 融合影像组学标签、PET代谢参数及临床病理因素的复合模型能够有效预测中晚期NSCLC患者的治疗疗效,对临床识别治疗获益人群具有一定的指导意义。
Objective To investigate the predictive value of pretreatment baseline 18F-fluorodeoxyglucose(FDG)positron emission tomography/computed tomography(PET/CT)radiomic features for treatment efficacy in patients with advanced non-small cell lung cancer(NSCLC).Methods A retrospective analysis was conducted on 107 patients with advanced NSCLC who underwent pretreatment 18F-FDG PET/CT scans at Hangzhou Cancer Hospital between January 2017 and December 2021.The cohort included 70 males and 37 females,with a mean age of(63.6±9.9)years.Six months posttreatment,patients were classified into a Clinical Benefit(CB)group(n=61)and a Progressive Disease(PD)group(n=46)based on efficacy evaluation.Patients were divided into a training set(n=74)and a test set(n=33)using a hold-out method in a 7:3 ratio.Radiomic features from both PET and CT modalities were extracted using the PyRadiomics toolkit(Python 3.9.7).After initial screening via the Mann-Whitney U test(P<0.05),a radiomics signature score(Rad-score)was constructed using LASSO regression(λ=0.01).Univariate and multivariate Logistic stepwise regression were employed to screen PET metabolic parameters and clinico-pathological factors associated with treatment efficacy,leading to the construction of a clinical model.Finally,a multiparameter composite prediction model was built by integrating the Rad-score,PET metabolic parameters,and clinical variables.The discriminatory performance of each model was evaluated using receiver operating characteristic(ROC)curve analysis,and differences in the area under the curve(AUC)between models were compared using DeLong's test.Results A total of 6 radiomic features(3 from PET and 3 from CT)were selected to construct the Rad-score.The AUCs for the Rad-score were 0.839(95%CI:0.743~0.935)in the training set and 0.789(95%CI:0.625~0.953)in the test set,which were higher than those of the clinical model(based on tumor longest diameter and SULmin)with AUCs of 0.670(95%CI:0.544~0.796)in the training set and 0.714(95%CI:0.531~0.897)in the test set,although the differences were not statistically significant(P>0.05).The composite model,integrating the Rad-score,traditional PET metabolic parameters,and clinicopathological factors,demonstrated the best predictive performance,with an AUC of 0.852(95%CI:0.759~0.945)in the training set and 0.793(95%CI:0.630~0.956)in the test set.Its sensitivity,specificity,and accuracy were 66.22%,86.49%,and 77.03%in the training set,and 57.58%,78.79%,and 69.70%in the test set,respectively.However,no statistically significant differences in AUC values were observed between the composite model and either the Rad-score or the clinical model(all P>0.05).Conclusion A composite model integrating the radiomics signature,PET metabolic parameters,and clinicopathological factors can effectively predict treatment efficacy in patients with advanced NSCLC,offering certain guiding significance for clinically identifying patients likely to benefit from treatment.
出处
《浙江临床医学》
2026年第2期184-187,共4页
Zhejiang Clinical Medical Journal
基金
杭州市医药卫生科技计划项目(A20230958)。
关键词
肺肿瘤
非小细胞肺癌
正电子发射断层显像
影像组学
疗效预测
Lung neoplasms
Non-small cell lung cancer
Positron-emission tomography
Radiomics
Efficacy prediction
