Computer generative method on brain tumor segmentation in MRI images

Li, Yanye

Permalink

https://hdl.handle.net/2142/107243 Copy

Description

Title

Computer generative method on brain tumor segmentation in MRI images

Author(s)

Li, Yanye

Contributor(s)

Liang, Zhi-Pei

Issue Date

2020-05

Keyword(s)

• magnetic resonance imaging
• biomedical imaging
• signal processing
• machine learning
• generative model

Date of Ingest

2020-06-11T18:50:44Z

Abstract

Computer generative method has been used for a long time in brain tumor segmentation tasks on magnetic resonance images. The popularity of machine learning also prompts people to explore the use of generative methods to better train their segmentation models. At the early stage, brain tumor segmentation competitions like BraTS 2012 used computer synthetic MR images with tumor to solve the lack of enough data in the training set, and now, with the rise of computer generative models in deep learning, more researchers have started to work on this track to find a better solution for the task. This thesis addresses the implementation and analysis of some existing methods, specifically a tumor synthetic tool called TumorSim and a competition winning deep learning model that incorporates variational auto-encoder as a generative model. This thesis also reports on an experiment that uses imperfect segmented tumors from simple models as the input to a generative adversarial network to generate a better result.

Type of Resource

text

Genre of Resource

other

Language

en

Permalink

http://hdl.handle.net/2142/107243

Owning Collections