Added benefits of computer-assisted analysis of Hematoxylin-Eosin stained breast histopathological digital slides
Access status:
Open Access
Type
ThesisThesis type
Doctor of PhilosophyAuthor/s
Gandomkar, ZibaAbstract
This thesis aims at determining if computer-assisted analysis can be used to better understand pathologists’ perception of mitotic figures on Hematoxylin-Eosin (HE) stained breast histopathological digital slides. It also explores the feasibility of reproducible histologic nuclear ...
See moreThis thesis aims at determining if computer-assisted analysis can be used to better understand pathologists’ perception of mitotic figures on Hematoxylin-Eosin (HE) stained breast histopathological digital slides. It also explores the feasibility of reproducible histologic nuclear atypia scoring by incorporating computer-assisted analysis to cytological scores given by a pathologist. In addition, this thesis investigates the possibility of computer-assisted diagnosis for categorizing HE breast images into different subtypes of cancer or benign masses. In the first study, a data set of 453 mitoses and 265 miscounted non-mitoses within breast cancer digital slides were considered. Different features were extracted from the objects in different channels of eight colour spaces. The findings from the first research study suggested that computer-aided image analysis can provide a better understanding of image-related features related to discrepancies among pathologists in recognition of mitoses. Two tasks done routinely by the pathologists are making diagnosis and grading the breast cancer. In the second study, a new tool for reproducible nuclear atypia scoring in breast cancer histological images was proposed. The third study proposed and tested MuDeRN (MUlti-category classification of breast histopathological image using DEep Residual Networks), which is a framework for classifying hematoxylin-eosin stained breast digital slides either as benign or cancer, and then categorizing cancer and benign cases into four different subtypes each. The studies indicated that computer-assisted analysis can aid in both nuclear grading (COMPASS) and breast cancer diagnosis (MuDeRN). The results could be used to improve current status of breast cancer prognosis estimation through reducing the inter-pathologist disagreement in counting mitotic figures and reproducible nuclear grading. It can also improve providing a second opinion to the pathologist for making a diagnosis.
See less
See moreThis thesis aims at determining if computer-assisted analysis can be used to better understand pathologists’ perception of mitotic figures on Hematoxylin-Eosin (HE) stained breast histopathological digital slides. It also explores the feasibility of reproducible histologic nuclear atypia scoring by incorporating computer-assisted analysis to cytological scores given by a pathologist. In addition, this thesis investigates the possibility of computer-assisted diagnosis for categorizing HE breast images into different subtypes of cancer or benign masses. In the first study, a data set of 453 mitoses and 265 miscounted non-mitoses within breast cancer digital slides were considered. Different features were extracted from the objects in different channels of eight colour spaces. The findings from the first research study suggested that computer-aided image analysis can provide a better understanding of image-related features related to discrepancies among pathologists in recognition of mitoses. Two tasks done routinely by the pathologists are making diagnosis and grading the breast cancer. In the second study, a new tool for reproducible nuclear atypia scoring in breast cancer histological images was proposed. The third study proposed and tested MuDeRN (MUlti-category classification of breast histopathological image using DEep Residual Networks), which is a framework for classifying hematoxylin-eosin stained breast digital slides either as benign or cancer, and then categorizing cancer and benign cases into four different subtypes each. The studies indicated that computer-assisted analysis can aid in both nuclear grading (COMPASS) and breast cancer diagnosis (MuDeRN). The results could be used to improve current status of breast cancer prognosis estimation through reducing the inter-pathologist disagreement in counting mitotic figures and reproducible nuclear grading. It can also improve providing a second opinion to the pathologist for making a diagnosis.
See less
Date
2018-11-13Licence
The author retains copyright of this thesis. It may only be used for the purposes of research and study. It must not be used for any other purposes and may not be transmitted or shared with others without prior permission.Faculty/School
Faculty of Health SciencesDepartment, Discipline or Centre
Medical Imaging and Radiation SciencesAwarding institution
The University of SydneyShare