Alsubaie, Najah Mohammed (2018) Automatic analysis of lung adenocarcinoma histology whole slide images. PhD thesis, University of Warwick.

PDF WRAP_Theses_Alsubaie_2018.pdf - Submitted Version Embargoed item. Restricted access to Repository staff only until 25 January 2021. Contact author directly, specifying your specific needs. - Requires a PDF viewer. Download (128Mb)

Request Changes to record.

Abstract

Histology is the backbone in the diagnosis and prognosis pipeline of most types of cancer, especially lung adenocarcinoma (LUAD). However, a pathologist's assessment of histology slides is often subjective, semi-quantitative, and limited to selected regions of the tumour. Recently, automatic tools are being widely employed by utilising digital slide scanners. These tools exploit a large amount of information captured from the tumour which could not be achieved by human observers. In addition, the automatic tools enable an objective and reproducible way of conducting the clinical experiments, and provide the opportunity to discover new potential image-based features. Utilising these features could provide a second opinion for the pathologists in different tasks such as distinguishing between various types of lung cancer, predicting possible metastases and others. Therefore, a solid foundation is provided towards objective, comparable histology assessments, and personalised treatment of the disease.

In this thesis, we examine different stages of LUAD automatic analysis. We begin with preprocessing of the WSI, then cell analysis, and finally tumour morphology analysis. We use scanned slides of tissue sections stained with Haematoxylin and Eosin (H&E). In the first part of the thesis, we start with preprocessing of histology images. We propose two methods for stain deconvolution: First, a supervised method where a classifier is trained to distinguish between different stain colours. We transfer the stain colours into another chromatic colour space such that the intensity variations between pixels of one stain colour are minimised. Second, we present an unsupervised method which directly estimates the separation between different stain colours by filtering the image such that the dependency between the stain colours is reduced.

In the second part of the thesis, we propose two novel automatic tools to quantify the heterogeneity of tumour cells and tissue morphology. The first tool performs WSI analysis of cellular features by extracting statistics from the WSI heat map and then examining these statistics to find the correlation with survival of LUAD patients. In the second tool, we propose an automatic method for quantification of LUAD growth patterns in WSI using a deep-learning-based method. The proposed framework is applied to automatically locate and classify tumour growth patterns within the WSI. Finally, we quantify the percentages of each pattern and analyse the impact of these percentages on the survival of LUAD.

Item Type:	Thesis or Dissertation (PhD)
Subjects:	Q Science > QA Mathematics > QA76 Electronic computers. Computer science. Computer software R Medicine > RC Internal medicine > RC0254 Neoplasms. Tumors. Oncology (including Cancer)
Library of Congress Subject Headings (LCSH):	Histology -- Technique, Lungs -- Cancer -- Diagnosis, Lungs -- Cancer -- Histology, Diagnostic imaging
Official Date:	September 2018
Dates:	Date Event September 2018 UNSPECIFIED
Institution:	University of Warwick
Theses Department:	Department of Computer Science
Thesis Type:	PhD
Publication Status:	Unpublished
Supervisor(s)/Advisor:	Rajpoot, Nasir M. (Nasir Mahmood)
Sponsors:	Jāmiʻat al-Malik ʻAbd al-ʻAzīz
Format of File:	pdf
Extent:	xxix, 152 leaves : illustrations, charts
Language:	eng

Request changes or add full text files to a record

Repository staff actions (login required)

View Item