目的 本研究旨在阐明具有放射基因组特征的基于CT图像的活检以预测非小细胞肺癌(NSCLC)患者仅同源域蛋白同源盒(homeodomain-only protein homeobox,HOPX)基因表达状态和预后.方法 根据HOPX的表达将患者标记为HOPX阴性或阳性,并将其分为训练数据集(n=92)和测试数据集(n=24).在对116例患者进行基因与图像特征的相关性分析中,从1218个图像特征中选出了8个与HOPX表达相关的显著特征作为放射基因组特征候选,以预测HOPX的表达状态和预后.结果 通过叠加集成学习模型建立具有放射基因组特征的影像活检模型,在测试数据集中,模型显示出对HOPX表达的预测能力,ROC曲线下的面积为0.873,Kaplan-Meier曲线的预测能力(P=0.0066).结论 具有放射基因组特征的基于CT图像的活检可以帮助医生预测HOPX在非小细胞肺癌中的表达状况和预后.
Biopsy Based on CT Images Can Help Predict HOPX Expression Status and Prognosis of Patients with Non-small Cell Lung Cancer
Objective This study aimed to elucidate CT image-based biopsies with radiogenomic features to predict homeodomain-only protein homeobox(HOPX)gene expression status and prognosis in patients with non-small cell lung cancer(NSCLC).Methods Patients were labeled as HOPX-negative or HOPX-positive based on HOPX expression and divided into a training dataset(n=92)and a test dataset(n=24).In the analysis of correlation between gene and image features in 116 patients,8 significant features related to HOPX expression were selected from 1218 image features as candidates for radiogenomic features to predict the expression status and prognosis of HOPX.Results An image biopsy model with radiogenomic features was established by superposition ensemble learning model.In the test data set,the model showed the predictive ability of HOPX expression,the area under ROC curve was 0.873,and the predictive ability of Kaplan-Meier curve was 0.0066(P=0.0066).Condusion CT image-based biopsies with radiogenomic signatures can help physicians predict HOPX expression status and prognosis in non-small cell lung cancer.
CT ImageNon-small Cell Lung CancerHomeodomain Protein Homeobox
万方数据
维普
