During the past decades,many studies have led to biomarkers for cancer outcome predictions,which assist clinicians on selecting the effective treatment strategy.These biomarkers include both histopathological attributes and various types of omic data.However,there is a lack of a unified means for patient stratification that can effectively integrate the heterogeneous types of molecular and clinical data and improve accuracy on patient outcome prediction.In this presentation,we will discuss the progress and challenges on integrating multiple types of omic data with imaging,especially histophathology images for cancer,patient stratification.We identify three types of integration strategies including correlative analysis,sparse modeling methods,and consensus learning.The applications of these approaches in cancer studies lead to new hypothesis on gene functions and molecular basis of cell and tissue morphology as well as potential new markers for cancer patient precision medicine.