Kun Huang, Ph.D.

Professor &Assistant Dean for Data, Indiana University School of Medicine

黄昆

2004年起在美国俄亥俄州立大学医学院生物医学信息学系任教，2010年获评终身教职，并先后担任综合癌症中心生物信息共享资源主任，计算生物学与生物信息学部主任，医学院副院长等职务。2017年加入印第安纳大学医学院参与领导精准健康计划，担任数据科学与信息学主任，同时任基因组数据科学讲席教授，医学院主管数据科学副院长。其主要研究方向包括生物信息学，医学图像分析，医疗大数据，机器学习及其在癌症研究及神经科学等方面的应用。2018年当选美国医学与生物工程学院(AIMBE)会士。

Machine Learning Methods in Cancer Integrative Genomics

Histopathology images of solid cancer specimens are essential for diagnosis and prognosis by pathologists. With the advancement of imaging technology, computer vision, and machine learning, various quantitative features can be automatically extracted from whole-slide images (WSIs) using advanced algorithms including deep learning methods. We have recently developed a pipeline for extracting multiple quantitative features from H&E stained histopathology image for capturing nucleic morphology, density, and tissue organization. In addition, a convolution neural network based method was developed to accurately segment stromal and epithelial tissues in the tumor. The pipeline provided a foundation for further integration of tissue morphological and genomic data, for which we have developed multiple machine learning algorithms to identify relationships between morphological features and genomic information as well as integrating the two types of data to predict cancer patient outcomes. These algorithms have been successfully applied to studies on breast, lung, and kidney cancers. Currently we are expanding these studies to a pan-cancer study covering 32 types of solid tumors.