肝癌的术前病理组织学分级评价是函待解决的临床难题，肝癌的分化程度影响患者的手术预后和肝移植策略的选择。本课题组前期研究及文献发现：基于影像数据的影像组学被证实可用于肿瘤的各种表型的评价，且与肿瘤的基因、分子分型密切相关；近期，本课题组在研究中发现影像组学特征与肿瘤的病理分级密切相关，提示影像组学可以用来评估肝癌组织分化，并可能与肝癌相关基因/蛋白质的表达相关。据此，本项目拟通过肝癌影像数据研究纹理特征、梯度特征和深度特征提取方法；基于特征过滤和包裹法选择最佳的特征子集；构建集成学习影像组学模型用于术前评估肝癌病理分级。同时，本项目将基于稀疏匹配的多组学数据关联分析方法开展影像组学特征与差异基因/蛋白质的关联分析，揭示肝癌分化相关影像组学特征背后的生物学基础。本项目的开展将为肝癌患者提供术前病理分级信息，指导治疗策略，也为影像组学研究提供重要的理论基础。

Preoperative evaluation of hepatocellular carcinoma histopathological grading is a clinical problem that requires a solution. The degree of differentiation of hepatocellular carcinoma will influence a patient’s surgical prognosis and the choice of liver transplantation strategy. Previous studies by our group and others have found that radiomics-based imaging data are useful for evaluation of tumors with various phenotypes and radiomics results correlate closely with those based on tumor gene and molecular typing. We also previously found that radiomics data are closely associated with tumor pathological grade, suggesting that this methodology could be used to assess hepatocellular carcinoma histological grade, and may be related to the expression of hepatocellular carcinoma-related genes/proteins. Accordingly, in this project we intend to retrospectively collect image data from patients with hepatocellular carcinoma and study extraction methods of texture features, gradient features and deep feature. The best feature subsets will be selected based on feature filtering and the wrapper method, and an Ensemble Learning radiomics model constructed for preoperative evaluation of hepatocellular carcinoma histological grade. Simultaneously, we will analyze the associations between radiomics features and differentially expressed genes/proteins related to hepatocellular carcinoma differentiation, using the sparse matching multi-omics data association analysis method, to reveal the biological characteristics underlying radiomics features. This project will provide preoperative pathological grade information for patients with hepatocellular carcinoma, guide treatment strategies, and provide an important theoretical basis for the study of radiomics.