CVPR Daily - Wednesday

histopathology images . Their method not only alleviates the burden on pathologists but achieves state-of-the-art performance compared to existing methods. “ This is not the first time people have adopted graph methods in histopathology images, ” Howard tells us. “ However, not many works have focused on advanced graphs . We study how heterogeneous graphs can be applied to histopathology image analysis and explore which kind of heterogeneous graph best suits our problem. We also explore causal inference methods in our framework. ” Fernando adds: “ We use causal inference to give some explanation about our machine learning model and to perform cancer localization, which is one of the real-life medical applications. ” Although the team’s work on this aspect is still in its early stages and has only been evaluated qualitatively, they are highly optimistic about extending their causal inference framework in the future. Their goal is to enhance cancer localization and gain a deeper understanding of the underlying causes of cancer. Could this be an early preview of their paper for next year? Howard laughs: “ Yes, we will try! ” Howard was born in Hong Kong, while Fernando hails from Indonesia. “ This is my first CVPR paper and my first time working on medical images, ” Fernando tells us. “ I’ve been very fortunate to work with a really good researcher like Howard and his supervisors, Guosheng Yin and Lequan Yu, as well as my supervisor from my previous company, Ruby Ma. ” 15 DAILY CVPR Wednesday Histopathology Whole Slide Image... Our attempt to apply causal inference to graph localization