摘要
目的探讨基于双层探测器光谱CT(DLCT)参数图影像组学模型在预测非小细胞肺癌PD-L1表达中的价值。方法回顾性搜集行DLCT胸部增强扫描并经病理证实且获得PD-L1检测数据的非小细胞肺癌患者110例,选取88例为训练集,22例为测试集,分析其常规影像学征象并结合外周血免疫细胞建立临床模型。基于单能40 keV图像勾画容积感兴趣区(VOI),将此VOI匹配至单能40 keV图像、碘密度(ID)图、有效原子序数(Z-eff)图及电子密度(ED)图中进行影像组学特征提取,采用主成分分析法(PCA)进行降维,采用Kruskal-Wallis检验和多变量方差分析(ANOVA)进行特征选择,建立影像组学模型。影像组学结合外周血免疫细胞及影像学征象建立预测PD-L1表达的联合模型。使用受试者工作特征(ROC)曲线及曲线下面积评价模型的预测效能。结果患者年龄、白细胞计数、中性粒细胞数、血小板数、中性粒细胞与淋巴细胞计数比例,病灶的形态及是否有空洞在PD-L1高表达组和低表达组之间存在显著性差异(P<0.05)。临床模型、影像组学模型及联合模型在训练集的ROC曲线下面积分别为0.867、0.976、0.984,在测试集的曲线下面积分别为0.547、0.958、0.979。结论基于DLCT多参数图影像组学联合形态学特征及外周血免疫细胞构建模型能有效地预测非小细胞肺癌PD-L1表达水平。
Objective To explore the value of radiomics for predicting the expression of PD-L1 in non-small cell lung cancer based on Dual-layer detector CT(DLCT)parameter Images.Methods 110 patients with pathologically confirmed of non-small cell lung cancer who underwent DLCT chest enhancement scanning and PD-L1 expression testing were retrospectively enrolled for this study.88 patients were selected as the training dataset and 22 patients as the testing dataset.Their conventional imaging findings were analyzed and the clinical model was established by combining peripheral blood immune cells.The volume area of interest(VOI)was outlined by the Mono-E 40keV image.Radiomics features were extracted from the Mono-E 40keV image,iodine density(ID)map,Z-effective map,and electron density map.The principal component analysis was used to descend dimension.Multivariate analysis of variance(ANOVA)and Kruskal-Wallis testing were used for feature selection,then the radiomics model was built.Combined peripheral blood immune cells with imaging findings,the combined prediction model of PD-L1 expression was established.Receiver operating curve(ROC)and areas under curve(AUC)were used to evaluate the prediction efficiency of the model.Results There were significant differences in age,white blood cell counts,neutrophil counts,platelet counts,the ratio of neutrophil to lymphocyte counts,lesion shape and with or without cavities between PD-L1 high expression group and low expression group(P<0.05).The AUC of the clinic model,radiomics model and the combined model in the training set were 0.867,0.976 and 0.984,respectively,and the AUC in the testing set were 0.547,0.958 and 0.979,respectively.Conclusion The radiomics model based on the multi-parameter images of DLCT can predict the expression level of PD-L1 in non-small cell lung cancer.The combination of clinical model and radiomic model can obtain higher prediction efficiency and provide a good proof for clinical immunotherapy and prognosis evaluation.
作者
郑小霞
马娅琼
崔雅琼
陈杏彪
郑文霞
黄刚
ZHENG Xiaoxia;MA Yaqiong;CUI Yaqiong(The First Clinical Medical College of Gansu University of Chinese Medicine,Lanzhou,Gansu Province 730030,P.R.China)
出处
《临床放射学杂志》
北大核心
2023年第7期1129-1138,共10页
Journal of Clinical Radiology
基金
甘肃省自然科学基金项目(编号:21JR7RA605)
中国红十字基金会“ICON科研基金”(编号:XM_HR_ICON_2021_05)
甘肃省卫生健康行业科研计划项目(编号:GSWSKY2020-15)。
关键词
非小细胞肺癌
影像组学
免疫治疗
Non-small cell lung cancer
Radiomics
Immunotherapy
