Oregon State University

Can’t find an event? We’re busy migrating to a new event calendar. Try looking at the new calendar



Event Details

MS Final Examination – Thi Kim Phung Lai

Monday, April 23, 2018 9:00 AM - 11:00 AM

Multiple Instance Learning for Histopathological Image Classification

In histopathological image analysis, image classification as well as pattern detection play a crucial role in the diagnosis and treatment process since the goal is to not only differentiate cancer types but also identify cancerous manifestations. Fully supervised learning strategies tend to address these problems using manually annotated cancerous regions and labeled cancer-type images. The success of these approaches heavily depends on manual segmentation from pathologists/experts. However, the manual process is challenging due to two major issues of histopathological images: (i) manual segmentation process over the entire image is time-consuming and labor-intensive and (ii) boundaries of different cancerous regions in the image are naturally ambiguous, which may create inter- and intra-observation variations among experts. Therefore, weakly supervised learning approaches solely based on the label of images are well-suited for the data. Multiple instance learning (MIL), one of the weakly supervised learning methods, is recently considered as a machine learning paradigm to analyze histopathological images. Based on image labels/cancer types, MIL approaches learn to predict a cancer type as well as detect and localize cancerous regions in the image. In this report, existing strategies for modeling histopathological image analysis as MIL problems are reviewed. Recent trends and future directions are also discussed.

Major Advisor: Raviv Raich
Committee: Jinsub Kim
Committee: Xiao Fu
Committee: Molly Megraw

Kelley Engineering Center (campus map)
Calvin Hughes
1 541 737 3168
Calvin.Hughes at oregonstate.edu
Sch Elect Engr/Comp Sci
This event appears on the following calendars: