定量CT特征预测非小细胞肺癌的ALK重排及克唑替尼疗效

Obtaining precise molecular phenotype is the basis for target therapy of the non-small cell lung cancer (NSCLC). However, because of heterogeneity of NSCLC, the genetic mutations of the tumor may not be accurately obtained from small biopsy samples. CT examination, as a routine modality for NSCLC, its findings may comprehensively reflect the histopathological and molecular phenotypes of the tumor, and may provide a broad prospect in prediction of ALK rearrangements and target therapy response. However, the conventional qualitative imaging methods cannot systematically extract the features of the tumor. Our previously published research has demonstrated that combining different scales of semantic features and semi-automatic computerized features can predict prognosis of NSCLC patients. In this project, we propose to extract pre- and post-contrast quantitative CT features of NSCLC tumor and the surrounding pulmonary tissue using the similar methods. We will analyze the association between these features and ALK rearrangements and the therapy effect of crizotinib. We aim to find the CT imaging biomarkers for ALK-rearranged NSCLC, and furthermore, to explore those features or feature changes which can predict the therapy effect of crizotinib. This study will be helpful for guiding specific biopsy for ALK rearrangements, and also providing guidance for clinical decision making for the personalized precision therapy.

获得准确的分子表型是指导非小细胞肺癌（NSCLC）靶向治疗的前提，但因NSCLC的不均质性，活检所获的小块组织不能完全反映肿瘤的基因突变。CT作为NSCLC的常规检查手段，其表现能全面反映肿瘤的组织病理学及分子生物学基础，在ALK重排和靶向治疗疗效预测中具有广阔的应用前景。然而，传统CT定性方法具有无法系统提取肿瘤信息的局限性。申请者已发表的前期工作证明，结合放射科医师阅片和计算机半自动化的方法提取CT特征可预测NSCLC患者的预后。本课题拟结合放射科医师阅片和计算机半自动化的方法提取NSCLC肿瘤和周围肺组织增强前后的定量CT影像特征，分析定量CT影像特征与ALK重排及克唑替尼疗效的关系，寻找能预测NSCLC患者ALK重排的CT影像标志物，并进一步探索能预测克唑替尼疗效的CT特征或特征变化。该研究有助于指导临床针对性活检行ALK重排检测，并可指导精准个体化治疗方案的制定。

非小细胞肺癌（NSCLC）的靶向治疗为患者带来了生存获益。获得准确的分子表型是指导NSCLC靶向治疗的前提，但因NSCLC的不均质性，活检所获的小块组织不能完全反映肿瘤的基因突变。由于靶向药物耐药性的存在，克唑替尼治疗ALK重排NSCLC患者疗效和预后的早期判定对临床治疗方案的选择具有重要的意义。本研究结合放射科医师阅片和计算机半自动化的方法提取NSCLC肿瘤和周围肺组织增强前后的定量CT影像特征，寻找能预测NSCLC患者ALK重排的CT影像标志物，并进一步探索能预测克唑替尼疗效和预后的CT特征。研究发现，分叶、实性和较大的最大淋巴结短径是预测ALK重排的重要CT特征，该预测模型的AUC为0.794。影像组学特征original GLCM Correlation与克唑替尼疗效关系密切。疗效预测模型训练集AUC=0.734，验证集AUC=0.679。胸膜牵拉征和original first-order Maximum对患者无进展生存期（PFS）具有预测潜能。PFS预测模型训练集平均C-index=0.734，验证集平均C-index=0.649。根据预测值中位数将患者分为高危组和低危组，两组PFS具有显著差异（Log-rank P<0.001）。研究结果表明，CT特征可预测NSCLC患者的ALK重排，有助于指导临床针对性活检行ALK重排检测，用于肺癌患者ALK重排的筛查。晚期ALK重排NSCLC患者的CT特征对克唑替尼疗效和预后具有一定的预测价值，有助于指导个体化靶向治疗。