基于CT和病理WSI的深度学习量化肝细胞癌免疫微环境预测术后复发风险的研究

肝功能良好且无明显门脉高压的肝细胞癌（HCC）首选手术切除，但多达50-70％的患者术后5年内发生肿瘤复发；精准预测术后复发风险对制定治疗方案有重要意义，但目前临床缺乏精准预测的方法。研究证实肿瘤免疫微环境是独立的预后预测指标，且通过深度学习挖掘肿瘤影像特征可表征肿瘤免疫微环境，为精准预测HCC术后复发风险带来机遇。本项目拟：1、基于HCC术后病理数字化全切片图像（WSI）检测免疫细胞数量及空间拓扑关系，建立免疫微环境评分，量化肿瘤免疫微环境状态；2、基于HCC术前CT图像，采用深度学习等方法提取并筛选与肿瘤免疫状态相关的CT影像特征，构建免疫相关标签；3、融合免疫相关CT影像组学标签和临床信息，构建复发风险预测模型并验证。本项目的研究结果，将构建量化HCC免疫微环境的影像组学模型，实现个体化复发风险精准量化预测，指导临床决策。

Surgical resection is the preferred treatment option for the hepatocellular carcinoma patients with good liver function and without obvious portal hypertension. Nevertheless, 50-70% of resected hepatocellular carcinoma patients have disease recurrence within 5 years after surgery. Accurate prediction of the risk of recurrence after surgery is of great significance for the treatment option. However, there remains an unmet clinical need for the accurate identification of resected hepatocellular carcinoma patients who are at high risk for disease recurrence. Many studies have indicated that the tumor immune microenvironment is an independent predictor of prognosis. Moreover, deep learning, through mining of data from images and subsequent implementation in assessment of tumor immune microenvironment, offers new opportunities for quantitative stratification of risk of disease recurrence for hepatocellular carcinoma. This project is to: (1) detect the number and spatial distribution of immune cells based on pathological whole slide image after surgery, follow by establishing the score of immune microenvironment to quantify immune status of tumors; (2) extract and select the key radiomics features highly related to the status of tumor immune based on preoperative CT images using deep learning approach, then develop an immune-based radiomics signature; (3) integrate the immune-based radiomics signature and clinical variables to build the prognostic prediction model for predicting risk of disease recurrence of resected hepatocellular carcinoma, and the prediction model will be validated. The result of this project will develop an immune-based radiomics prediction model facilitating precise quantification of the risk of disease recurrence of hepatocellular carcinoma after resection, which will aid in clinical decision-making in precise medicine for the patient management.