乳腺癌早期筛查、风险人群管理及预警模型构建

Breast cancer is one of the most common cancer in women. Early screening of breast cancer has been widely accepted as the most effective prevention measure to improve survival and reduce mortality in malignant tumors. Due to the fact that Chinese women have smaller and more dense breasts than those in the West and the age of breast cancer onset is also younger, therefore, the effects of various screening methods are not the same as those in Western women’s breast cancer censuses. From a cost-effectiveness point of view, the direct introduction of western women breast cancer screening models may not meet the actual situation with relatively tense health resources in China. Our study will conduct early screening of breast cancer in health physical examination population. Through long-term follow-up, intervention and monitoring, health management will be conducted in screened high-risk population to achieve breast cancer early prevention, early diagnosis and early treatment and to reach “medical and prevention synergy”, which has very important practical significance. According to clinical parameters of breast cancer acquired in early stage of this study, we will build a natural history progress model of breast cancer. Then based on the combination of breast cancer patients with imaging features, traditional risk predictors and tumor markers, we will establish a three-dimensional early warning model of breast cancer for Chinese women, which will provide a new way and a new tool to prevent and diagnose breast cancer and to monitor high-risk population.

乳腺癌是女性最常见的恶性肿瘤之一，而乳腺癌早期筛查已被广泛接受是恶性肿瘤中最能有效改善患者生存率和降低死亡率的群防措施。由于我国女性乳腺较西方女性小而致密，发病年龄也更年轻化，因此各种筛查方法效果也与其在西方女性乳腺癌普查中的效果不尽相同，直接引入西方女性乳腺癌筛查模式，从成本效果角度讲也很难符合我国相对紧张的卫生资源实际情况。因此，本研究将在健康体检人群中开展乳腺癌早期筛查，对筛选的高危人群进行健康管理，通过长期随访追踪与干预监测，实现健康人群的乳腺癌预防与早诊早治，达到“医防协同”，实现全方位、全生命周期的健康管理，具有十分重要的现实意义。根据本研究中获取的乳腺癌发病特点等临床参数，构建乳腺癌自然史进展模型，在此基础上结合乳腺癌患者影像学图像特征及筛选的传统危险因素预测因子和肿瘤标志物，建立乳腺癌早期风险三维预警模型，为我国女性乳腺癌的预防及诊断、高危人群的监测提供一种新思维和新工具。

乳腺癌早期筛查已被广泛接受是恶性肿瘤中最能有效改善患者生存率和降低死亡率的群防措施。由于我国女性乳腺癌发病率、女性乳腺结构、大小、腺体致密度及生育史等都与欧美国家女性不同，因此直接引入西方基于人群为基础的乳腺癌普查模式，从成本效果角度讲也很难符合我国相对紧张的卫生资源实际情况。参与年度健康体检人群由于其人群相对固定、连续性好、重复性高、信息稳定，可密切监测检出高危肿块的进展变化情况。本研究已建立基于健康体检人群的个性化乳腺癌早期筛查模式，从检前风险评估到检中血清学检测及不同乳腺检查方法实施，已建立基于体检人群的超声检出高危肿块检后随访管理流程，并依托医院信息管理系统，建立乳腺疾病检后回访信息系统。项目组已构建13状态连续时间变化乳腺癌自然史进展模型，并初步评估其临床前期逗留时间约7.04年。项目组已建立包括乳腺超声检出高危肿块个体、乳腺癌患者、良性乳腺疾病手术史个体健康管理档案，由专人专项针对性提供健康管理指导。项目组筛选乳腺癌传统危险因素（包括乳腺癌家族史、初潮年龄、初产年龄、流产史、口服避孕药史、雌激素替代治疗史）、肿瘤标志物（CA125、CA153）及超声图像影像学特征，建立乳腺癌早期风险三维预警模型，基于深度卷积神经网络建模的预测模型性能较优，AUC为0.858，其次为XGBoost建模，预测模型AUC为0.743。本项目共计发表学术论文13篇，其中英文期刊收录6篇，中华系列论文1篇，中文核心期刊论文6篇，已接收中文核心期刊论文1篇。参加国内学术会议4次；协助培养硕士研究生1人，指导本科生5人。