基于动脉期增强CT影像组学模型预测小细胞肺癌无进展生存期  

Arterial phase contrast-enhanced CT radiomics model for predicting progression-free survival of patients with small cell lung cancer

作　　者：雷洁 杨阳 邱晓辉[1] LEI Jie;YANG Yang;QIU Xiaohui(Department of Radiology,Bozhou People's Hospital,Bozhou 236000,China)

机构地区：[1]亳州市人民医院放射科,安徽亳州236000

出　　处：《中国医学影像技术》2024年第11期1698-1703,共6页Chinese Journal of Medical Imaging Technology

摘　　要：目的 观察基于动脉期增强CT影像组学模型预测小细胞肺癌(SCLC)患者无进展生存期(PFS)的价值。方法 回顾性纳入210例SCLC患者,按7∶3比例将其随机分为训练集(n=147)与测试集(n=63),以Cox比例风险回归分析获取PFS率的临床独立影响因素并构建临床模型;提取肿瘤影像组学特征,于训练集遴选与PFS率最为相关者构建影像组学模型,计算影像组学评分(Radscore)并据以对疾病进展风险进行分层,比较不同层次PFS率;联合临床独立影响因素及影像组学特征构建临床-影像组学模型。于测试集评估并比较各模型预测SCLC患者PFS率的区分度、校准度及临床净收益。结果 广泛期为SCLC患者PFS缩短的临床独立危险因素[HR=1.841,95%CI(1.288,2.633),P=0.001]。基于训练集选出5个与PFS率最为相关的影像组学特征,根据Radscore区分疾病进展风险低(Radscore<0.235)或高(Radscore≥0.235),后者PFS率低于前者(P<0.001)。临床模型、影像组学模型及临床-影像组学模型预测测试集6个月内PFS率的受试者工作特征曲线下面积(AUC)分别为0.646、0.920及0.931,预测12个月内PFS率分别为0.591、0.917及0.919;临床模型的一致性指数(C-index)为0.595,影像组学模型及临床-影像组学模型的C-index分别为0.878和0.884,均高于临床模型。相比临床模型,影像组学模型净重新分类指数(NRI)为75.82%(P<0.001)、综合判别改善指数(IDI)为59.76%(P<0.001),临床-影像组学模型依次为78.94%(P<0.001)及61.13%(P<0.001),后二模型间差异均无统计学意义(P均>0.05)。DCA结果显示影像组学模型和临床-影像组学模型的临床净收益均高于临床模型。结论 基于动脉期增强CT影像组学模型可预测SCLC患者PFS率,有助于临床制定个体化治疗方案。Objective To observe the value of arterial phase contrast-enhanced CT radiomics model for predicting progression-free survival(PFS)of patients with small cell lung cancer(SCLC).Methods A total of 210 patients with pathology confirmed SCLC were retrospectively enrolled and randomly divided into training set(n=147)and test set(n=63)at the ratio of 7∶3.Clinical independent influence factors of PFS rate were selected with Cox proportional hazards regression.The radiomics features of tumors were extracted from arterial phase contrast-enhanced CT,and those being most relevant to PFS rate selected in training set were used to construct the radiomics model.Then Radscores were calculated,and the risks of disease progression were stratified,and PFS rates of patients with different risks were compared.A clinical-radiomics model was constructed combining clinical independent influence factors and radiomics features.The differentiation,calibration,and net benefit of each model for predicting PFS rates of SCLC patients were evaluated and compared in test set.Results Extensive stage was an independent risk factor for shorter PFS of SCLC patients(HR=1.841,95%CI[1.288,2.633],P=0.001).Five radiomics features which relevant to PFS rate were selected based on training set.The patients were categorized as low-risk(Radscore<0.235)or high-risk(Radscore≥0.235)for disease progression,and those with high-risk had lower PFS rates than the low-risk ones(P<0.001).In test set,the area under the curve(AUC)of receiver operating characteristic curves of the clinical model,radiomics model and clinical-radiomics model for predicting PFS rate within 6 months was 0.646,0.920 and 0.931,respectively,while within 12 months was 0.591,0.917 and 0.919,respectively.The concordance index(C-index)of the radiomics model was 0.878,of the clinical-radiomics model was 0.884,both higher than that of the clinical model(C-index=0.595).Compared with clinical model,the radiomics model had net reclassification index(NRI)of 75.82%(P<0.001)and integrated discriminat

关 键 词：癌 非小细胞肺 体层摄影术 X线计算机 影像组学 无病生存 

分 类 号：R734.2[医药卫生—肿瘤] R445.3[医药卫生—临床医学]