Use this URL to cite or link to this record in EThOS:
Title: Automated analysis of liver tissue
Author: Chomphuwiset, Phatthanaphong
Awarding Body: University of Leeds
Current Institution: University of Leeds
Date of Award: 2012
Availability of Full Text:
Access from EThOS:
Conventional histological diagnostic practise largely relics on visual interpretation, which suffers from subjectivity and is prone to poor reproducibility. The development of computer technology offers new perspectives in information extraction from histological images. With the development of the digitalisation of complete glass slides and the development of automatic classification algorithms, automated analysis of tissue images has potential to assist pathologists in their diagnosis. By providing machines with knowledge based capabilities for image understanding, important tissue structures can be identified, segmented and subjected to measurement for the purpose of automated analysis. There is a number of practical medical benefits that may result from automated analysis. The primary advantage of automated analysis (which is limited in human analysis) is the quantitative description that is obtained from histological images such as object arrangement and texture descriptions. This quantitative description is not prone to the subjectivity of human observation. Therefore, comparison (e.g. between patients, diseases, tissues and cellular objects) is possible and the results of the comparison are objective. The difficulty in developing automated analysis is, however, caused by the diversity of the cellular structures contained in the images and variations in the colour of the histopathological image that can be caused by, for example, the slide preparation processes. This Ph.D. thesis explores the development of automated analysis techniques for characterising liver disease. The characterisation of liver disease encapsulates a number of low-level processing to methods, i.e. (i) tissue region classification (ii) nuclear detection and classification, (iii) bile duct detection, (iv) fat detection. As detecting cellular objects in tissues is considered as a preliminarily step in histopatholgy image analysis, this work addresses an issue of detecting cellular objects based on object-based properties and proposes a technique to additionally determine the relationships between cellular objects/tissue regions. By integrating the information between cellular objects and regions, this work shows that the performance of cellular objects/tissue region is improved. After detecting cellular objects/ tissue regions, quantitative descriptors - based on (i) quantity (ii) morphology, (iii) texture and (iv) structure of cellular objects/tissue regions - can be extracted to objectively describe the cellular objects, which facilitate the analysis of liver diseases and thus liver disease characterisation can be carried out.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available