Hierarchical Cancer Heterogeneity Analysis Based On Histopathological Imaging Features

时间:2022-09-16         阅读:

光华讲坛——社会名流与企业家论坛第6227

主题:Hierarchical Cancer Heterogeneity Analysis Based On Histopathological Imaging Features

讲人:耶鲁大学 Shuangge Ma教授

主持人:统计学院 林华珍教授

时间:9月21日 9:30-10:30

直播平台及会议ID:腾讯会议,会议ID:426 359 881

主办单位:统计研究中心 统计学院 科研处

主讲人简介:

Dr. Shuangge Ma obtained Ph.D. in Statistics from University of Wisconsin, Madison. He was a Postdoctoral Associate at University of Washington between 2004 and 2006. He is now Professor of Biostatistics at Yale University. His research interests include high-dimensional data analysis, cancer biostatistics, health economics, and others.

Shuangge Ma从威斯康星大学麦迪逊分校获得统计学博士学位。2004年至2006年,他在华盛顿大学做博士后。他现在是耶鲁大学生物统计学教授。他的研究兴趣包括高维数据分析、癌症生物统计学、健康经济学等。

内容提要:

In cancer research, supervised heterogeneity analysis has important implications. Such analysis has been “traditionally” based on clinical/demographic/molecular variables. Recently, histopathological imaging features, which are a “byproduct” of biopsy, have been shown as effective for modeling cancer outcomes, and a handful of supervised heterogeneity analysis has been conducted based on such features. There are two types of histopathological imaging features, which are extracted based on specific biological knowledge and using automated imaging processing software, respectively. In this study, using both types of histopathological imaging features, our goal is to conduct thefirst supervised cancer heterogeneity analysis that has a hierarchical structure. That is, thefirst type of imaging features defines a “rough” structure, and the second type defines a nested and more refined structure. This objective can be achieved using either a penalization or Bayesian approach. Simulation shows satisfactory performance of the proposed analysis. In the analysis of lung adenocarcinoma data, it identifies a heterogeneity structure significantly different from the alternatives and has satisfactory prediction and stability.

在癌症研究中,监督异质性分析具有重要意义。这种分析“传统”是基于临床/人口统计学/分子变量。最近,组织病理学成像特征(活检的一个“副产品”)已被证明对癌症预后建模有效,并基于这些特征进行了少量有监督的异质性分析。组织病理学成像特征分为两类,分别是基于特定的生物学知识和利用自动成像处理软件提取的。在本研究中,使用两种类型的组织病理学成像特征,我们的目标是进行第一个有监督的分级结构的癌症异质性分析。也就是说,第一种类型的成像特征定义了一个“粗略的”结构,而第二种类型定义了一个嵌套的和更精细的结构。这一目标可以使用惩罚或贝叶斯方法来实现。仿真结果表明,所提出的分析方法具有较好的性能。在对肺腺癌数据的分析中,它识别出了与替代方法显著不同的异质性结构,具有令人满意的预测和稳定性。