Exploring image segmentation methods to segment tumors by training over a dataset marked by skilled professionals

Somani, Varun

Permalink

https://hdl.handle.net/2142/97883 Copy

Description

Title

Exploring image segmentation methods to segment tumors by training over a dataset marked by skilled professionals

Author(s)

Somani, Varun

Contributor(s)

Forsyth, D. A.

Issue Date

2017-12

Keyword(s)

• image segmentation
• machine learning
• lung CT
• supervised learning
• unsupervised learning

Date of Ingest

2017-08-28T18:08:07Z

Abstract

This research uses chest CT scan images of lung cancer patients to examine current methods in image segmentation in the context of tumor segmentation. Potential benefits of the research include faster processing and detection time for patients as well as allowing doctors to rapidly proceed with the requisite procedures. We use both supervised and unsupervised methods to segment images. In terms of supervised methods, we use neural networks and SVMs with various “kernel tricks.” In terms of unsupervised methods, we use K-means clustering and Otsu’s method. Neural network gave the best result while other methods tended to have inferior performance. The results suggest that there is a possibility of further developing neural networks to conclusively solve the problem.

Type of Resource

text

Genre of Resource

Other

Language

en

Permalink

http://hdl.handle.net/2142/97883

Owning Collections