A Framework for Automatic Analysis of WSI using Deep Learning

Peter Bokor

Supervisor(s): Lukáš Hudec

Slovak Technical University


Abstract: Early diagnosis of breast cancer using microscopic analysis of tissue is a prerequisite for its treatment. This process is time-consuming and difficult due to the inter-intraclass variance of histopathology data, the vast size of histopathology slides, and environmental influences on pathologists. A robust, unbiased diagnostic assistance tool, resistant to outside influences would be of utmost help for pathologists performing the diagnosis. We propose an end-to-end framework for processing Whole Slide Images (WSI) and their analysis using methods of computer vision and deep learning. Multiple deep learning models are used to apply their knowledge to new WSIs, creating complex analyses, which may be used to assist with diagnosis. Our system is fast enough for everyday use and may speed up the workflow and ease up the workload of medical experts performing the analysis of tissue.
Keywords: Computer Vision, Image Processing
Full text:
Year: 2021