Use this URL to cite or link to this record in EThOS:
Title: Digital image processing for prognostic and diagnostic clinical pathology
Author: Maddison, John
Awarding Body: University of Huddersfield
Current Institution: University of Huddersfield
Date of Award: 2005
Availability of Full Text:
Access from EThOS:
Access from Institution:
When digital imaging and image processing methods are applied to clinical diagnostic and prognostic needs, the methods can be seen to increase human understanding and provide objective measurements. Most current clinical applications are limited to providing subjective information to healthcare professionals rather than providing objective measures. This Thesis provides detail of methods and systems that have been developed both for objective and subjective microscopy applications. A system framework is presented that provides a base for the development of microscopy imaging systems. This practical framework is based on currently available hardware and developed with standard software development tools. Image processing methods are applied to counter optical limitations of the bright field microscope, automating the system and allowing for unsupervised image capture and analysis. Current literature provides evidence that 3D visualisation has provided increased insight and application in many clinical areas. There have been recent advancements in the use of 3D visualisation for the study of soft tissue structures, but its clinical application within histology remains limited. Methods and applications have been researched and further developed which allow for the 3D reconstruction and visualisation of soft tissue structures using microtomed serial histological sections specimens. A system has been developed suitable for this need is presented giving considerations to image capture, data registration and 3D visualisation, requirements. The developed system has been used to explore and increase 3D insight on clinical samples. The area of automated objective image quantification of microscope slides presents the allure of providing objective methods replacing existing objective and subjective methods, increasing accuracy and rsducinq manual burden. One such existing objective test is DNA Image Ploidy which seeks to characterise cancer by the measurement of DNA content within individual cell nuclei, an accepted but manually burdensome method. The main novelty of the work completed lies in the development of an automated system for DNA Image Ploidy measurement, combining methods for automatic specimen focus, segmentation, parametric extraction and the implementation of an automated cell type classification system. A consideration for any clinical image processing system is the correct sampling of the tissue under study. VVhile the image capture requirements for both objective systems and subjective systems are similar there is also an important link between the 3D structures of the tissue. 3D understanding can aid in decisions regarding the sampling criteria of objective tests for as although many tests are completed in the 2D realm the clinical samples are 3D objects. Cancers such as Prostate and Breast cancer are known to be multi-focal, with areas of seeming physically, independent areas of disease within a single site. It is not possible to understand the true 3D nature of the samples using 2D micro-tomed sections in isolation from each other. The 3D systems described in this report provide a platform of the exploration of the true multi focal nature of disease soft tissue structures allowing for the sampling criteria of objective tests such as DNA Image Ploidy to be correctly set. For the Automated DNA Image Ploidy and the 3D reconstruction and visualisation systems, clinical review has been completed to test the increased insights provided. Datasets which have been reconstructed from microtomed serial sections and visualised with the developed 3D system area presented. For the automated DNA Image Ploidy system, the developed system is compared with the existing manual method to qualify the quality of data capture, operational speed and correctness of nuclei classification. Conclusions are presented for the work that has been completed and discussion given as to future areas of research that could be undertaken, extending the areas of study, increasing both clinical insight and practical application.
Supervisor: Tian, Gui Yun ; Danielsen, Havard ; Thorn, Mike Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Q Science (General) ; RB Pathology