乳腺癌是威胁我国女性健康主要的癌症之一，基于钼靶影像的早期筛查是乳腺癌防治的重要手段，有助于预警患癌风险。现有的乳腺癌风险评估模型大多以人群流行病学数据为风险因子，缺乏个体化信息，从而在个体层面预测准确率较低。同时由于人种和生活方式不同，基于西方女性的风险预测模型难以直接适用于中国女性。本课题旨在研究针对中国女性人群的基于钼靶和乳腺容积超声影像的个体化乳腺癌风险预测模型。我们前期基于钼靶的影像生物标记物研究了乳腺癌风险预测模型，在此基础上，针对我国女性致密型乳腺比例大的特点，通过融合多模态的影像信息，运用乳腺实质密度量化分析、影像组学、深度学习等多种机器学习方法从钼靶和乳腺容积超声的多模态影像数据中提取个体化影像特征并构建乳腺癌风险量化的影像生物标记物，并结合现有北京协和医学院主持研究的流行病学中国女性乳腺癌发病相关危险因素及激素水平等多维度潜在风险因子建立个体化乳腺癌风险预测模型。

Breast cancer is the most common type of cancers among women in China, and its incidence and mortality are increasing rapidly in recent years. Full-field digital mammography (FFDM) is the most cost effective screening tool for early detection of breast cancer because earlier detection increases survival rate with a possibility of complete cure. Great improvements in breast cancer treatment have been made over the past decade, which has contributed to a significant increase in overall survival. On the other hand, prevention of breast cancer has been an issue that has not been given adequate attention. Currently, there were several western breast cancer risk prediction models for designing intervention trials (predicts the numbers of women who will develop breast cancer in various subsets of the population). However, their prospects in personalized medicine are limited due to their poor discriminatory powers at the personal level. In additional, epidemiological studies indicate that lifestyle and genetic factors play an important role in the development of breast cancer, which makes the most of the existing western models inapplicable to Chinese women. Aiming to the early prevention of breast cancer, this study is to develop a personalized prediction model for Chinese female population based on FFDM and breast volume ultrasound. We propose to develop an imaging biomarker for personalized breast cancer risk prediction while we will integrate multi-modality imaging information in Chinese female population. We plan to obtain quantitative imaging features from FFDM and 3D breast ultrasound with multiple machine learning methods including automatic assessment of breast parenchyma, radiomics, deep learning et al. The proposed biomarker would be further integrated with other important lifestyle factors observed in the Chinese women breast cancer research conducted by Peking Union Medical College (PUMC) as well as hormone measurement. At the end of this project, we will establish a personalized breast cancer risk prediction model, which is able to identify the Chinese individuals with high risk of breast cancer for early prevention and personalized treatment of breast cancer.