基于影像基因组学的肝动脉化疗栓塞术联合索拉非尼治疗中晚期肝癌预后模型研究

Transarterial chemoembolization (TACE) combined with sorafenib is a common approach for the treatment of intermediate-advanced hepatocellular carcinoma (HCC). Nevertheless, the prognosis of such approach is diverse and the general treatment outcome is unsatisfied. It is mainly due to high heterogenicity and lack of specific prognostic-related biomarkers on HCC. Radiomics, genomics, and radiogenomics are now developing rapidly and are commonly applied to the prediction of prognosis on several diseases. Our preliminary results established a radiomics-based prognostic model for the prediction of survival after TACE combined with sorafenib on intermediate-advanced HCC with high accuracy. To this end and based on the advantages of radiomics and genomics, we propose the following hypothesis: Based on the noninvasive radiogenomics signatures, we are able to stratify heterogenicity of the tumor and establish prognostic models to guide individualized treatment. To test such hypothesis, our prospective trial use MaZda to analyze and find out radiomics signatures significantly associated with survival based on pre-treatment computed tomography (CT). Then we extract exosomes and miRNAs from peripheral blood successively. High-throughput sequencing and data mining are then carried out to identify genomics signatures significantly associated with survival. These identified radiomics and genomics signatures are then fused to noninvasive radiogenomics signatures. Finally, prognostic models are established based on radiogenomics signatures and clinical biomarkers by artificial intelligence. The models provide a new approach and idea to assess and screen intermediate-advanced HCC who are sensitive to TACE combined with sorafenib treatment individually.

肝动脉化疗栓塞术（TACE）联合索拉非尼是治疗中晚期肝细胞肝癌（HCC）常用的方式，但由于HCC的高度异质性及缺乏特异性治疗和预后预测靶点，该方式预后个体化差异大、总体疗效欠佳。近年来影像及基因组学发展迅速，已应用于多种疾病的预后预测。我们研究发现：基于影像组学标签建立的模型能准确预测该类患者预后。结合影像及基因组学各自优势，我们提出假说：基于预后相关无创性影像基因组学指标，可对中晚期HCC进行异质性分层，建立疗效预测模型并个体化指导治疗。为验证该假说，本前瞻性研究基于临床病例，用MaZda行CT影像组学分析；于外周血中行外泌体及其miRNAs提取并行高通量测序及数据挖掘，分别找出与预后显著相关的影像组学、基因组学标签并融合成无创性影像基因组学标签，再结合预后相关临床指标，运用人工智能建立能治疗前对患者进行预后分层的模型，为该治疗于中晚期HCC的高敏感病例筛选提供新的个体化评价方法和思路。

肝细胞癌（Hepatocellular Carcinoma, HCC）患者接受经动脉化疗栓塞术（Transarterial chemoembolization, TACE）联合索拉非尼在理想状态下能起到协同作用，有效控制肿瘤病灶，理论上能起到1+1>2的效果。然而，目前该联合治疗方案缺乏准确性高、临床推广便捷的疗效预测靶点及模型。本研究开展系列临床研究，从影像组学及其他预后预测标志物为切入点，探讨接受TACE及联合索拉非尼治疗的HCC患者系列预后标志物，并基于该类高敏感性及特异性的标志物建立能个体化精准预测预后的模型，从而达到于治疗前即能个体化、精准化预测及筛选最适合HCC患者的有效治疗方式，最终提高该类患者的长期生存及生活质量。本研究首先基于卷积神经网络算法，探索能个体化、精确化预测接受TACE联合索拉非尼治疗的HCC患者生存时间的影像组学标签集，并结合其他临床标志物建立生存时间预测模型。随后基于线性判别分析法，探索能于首次TACE治疗之前便可筛选出对TACE治疗不敏感的HCC患者。最后分别基于人工智能算法建立如列线图模型、随机生存森林模型等系列预后预测模型。研究结果表明，基于卷积神经网络算法的影像组学标签集可作为TACE联合索拉非尼治疗HCC患者生存时间预测的标志物，基于该标签集及其他临床标志物建立的列线图预测模型可有效筛选对该联合治疗敏感的HCC患者。其次，基于线性判别分析法证实获得的影像组学标签及预测模型能有效筛选出对于TACE治疗不敏感的HCC患者。最后，基于人工智能算法建立了能有效预测接受TACE联合索拉非尼治疗伴有门静脉癌栓的HCC患者生存时间的列线图模型，以及能有效预测接受TACE联合索拉非尼治疗的HCC患者能否获得长期疾病控制的随机生存森林模型。该类影像组学标志物及预后预测模型已可以准确有效应用于临床诊疗工作，对于HCC患者预后预测及筛选最合适的治疗方案从而提高该类患者长期预后及生活质量具有指导作用。