NORMA eResearch @NCI Library

Lung Cancer Classification from Histologic Images using Capsule Networks

Roy Medhi, Bedanga Bikash (2020) Lung Cancer Classification from Histologic Images using Capsule Networks. Masters thesis, Dublin, National College of Ireland.

[img]
Preview
PDF (Master of Science)
Download (1MB) | Preview
[img]
Preview
PDF (Configuration manual)
Download (1MB) | Preview

Abstract

Cancer is one of the most dangerous and invasive disease in the human body. Lung cancer has the highest mortality rate amongst the other cancers. Therefore, detection and classification has become critical for the diagnosis of lung cancers. The manual determination of cancer from histology slides require expert supervision and takes time. Computer vision systems using Convolutional Neural Network (CNN) have shown remarkable performance in automated detection of cancers. But the current CNN based system have some limitations. In this study the state-of the art Capsule Networks are used for lung cancers classification from histopathology images. Capsule Network does have the capability to preserve the orientation, pose and texture which is achieved by the vector transformation of the extracted features. For pre-processing the images were stain normalized. The classification performance of the model is evaluated using Matthew’s correlation coefficient (MCC), specificity, sensitivity, false negative and false positive rate and accuracy. The Capsule Network implemented in the study achieved 0.99 MCC score and 99% accuracy for lung cancer classification.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science

T Technology > T Technology (General) > Information Technology
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Dan English
Date Deposited: 25 Jan 2021 14:23
Last Modified: 25 Jan 2021 14:23
URI: http://norma.ncirl.ie/id/eprint/4465

Actions (login required)

View Item View Item