Trahearn, Nicholas (2017) Registration and multi-immunohistochemical analysis of whole slide images of serial tissue sections. PhD thesis, University of Warwick.

Preview

PDF WRAP_Theses_Trahearn_2017.pdf - Submitted Version - Requires a PDF viewer. Download (108Mb) | Preview

Request Changes to record.

Abstract

The identification and classification of tissue abnormalities for the purpose of disease diagnosis have been greatly served by the discipline of histopathology, and Immunohistochemistry (IHC) in particular. The advent of digital slide scanners and computerised slide viewing software have opened the door for introducing automated algorithms into what has traditionally been a predominantly manual discipline. Multi-IHC analysis is one potential area of interest for automation, which will be discussed in detail in this work.

Analysis occurs on serial sections of tissue, which must be realigned before their IHC marker expressions can be compared directly. This requires a robust method of serial section registration. Two methods of automated serial section registration are present, which are each designed to align a particular tissue type: breast core biopsy sections or resected colorectal cancer sections.

Automated multi-IHC analysis is presented from the perspective of two case studies: Scoring of Oestrogen Receptor and Progesterone Receptor (ER/PR) on breast core biopsies and IHC scoring and colocalisation of resected colorectal cancer (CRC) sections. For each case study the background of the problem is introduced, followed by a discussion of how each type of analysis is performed in clinical practice, and it is then explained how this is implemented as an automated algorithm. For the scoring of ER/PR, it is shown that the algorithm can achieve good agreement with a pathologist on a sample of 50 cases, which suggests that automated ER/PR scoring is suitable for clinical practice. For the analysis of CRC, the results of scoring and colocalisation are shown in the form of localised maps with a discussion into how they may be used for further analysis.

As part of this framework a number of additional steps must be carried out before the goal of multi-IHC analysis can be realised. Two pre-processing steps, both of which are key to ensuring that the end results are of the highest quality, are presented: Tissue Segmentation and Out of Focus Area detection. A complete Out of Focus Area detection system is presented, which has led to the development of a Windows software that is currently being used in a local hospital. In addition, we present an automated method of Stain Separation, based around Independent Component Analysis, which allows us to extract and process the IHC marker expressions directly. This method includes a novel correction process to improve any faults in the primary analysis.

Item Type:	Thesis (PhD)
Subjects:	Q Science > QR Microbiology
Library of Congress Subject Headings (LCSH):	Immunohistochemistry, Image processing -- Digital techniques, Cancer -- Histopathology, Tissue slices
Official Date:	February 2017
Dates:	Date Event February 2017 Submitted
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:	Engineering and Physical Sciences Research Council ; Omnyx LLC
Format of File:	pdf
Extent:	xiv, 181 leaves : illustrations, charts
Language:	eng

Request changes or add full text files to a record

Repository staff actions (login required)

View Item

Downloads