Use this URL to cite or link to this record in EThOS:
Title: Algorithms for the analysis of bone marrow cancer histology images
Author: Song, Tzu-Hsi
ISNI:       0000 0004 7227 6030
Awarding Body: University of Warwick
Current Institution: University of Warwick
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Automated computer-aided systems and approaches are widely required to investigate and analyze histology images for improving the accuracy of cancer diagnosis and effective treatment decision making. Quantitative analysis has immense potential to investigate and analyze the tissue and cellular characteristics of histology images in cancer research. It is based on accurate cellular, morphological, and tissue features. Automated approaches not only make the feature extraction and analysis more objective and more reproducible, but they can also help pathologists look for useful potential clues from a vast amount of hidden information in cancer tissues, whose clinical value may not be fully realized and visualized. This entails the automated computer algorithms with a key role of quantitative analysis of histology images for different cancers. In this thesis, I concentrate on bone marrow cancers and develop automated computer algorithms to extract and realize cellular and texture characteristics of bone marrow biopsies for efficiently characterizing different types of bone marrow cancers in further investigation and analysis. We focus on the development of automated algorithms for identifying various types of cells in bone marrow trephine biopsies, which are tiny cores of bone marrow tissues. All the algorithms are specifically designed for histological sections stained by a standard hematoxylin and eosin (H&E) stain. Firstly, we propose an automated framework with a novel segmentation model for delineating and segmenting megakaryocytes. Secondly, we create a novel deep learning network that processes the nuclear detection with irregular shape for various types of bone marrow stem cells. Then we construct another synchronized deep learning approach to simultaneously do detection and classification. We demonstrate the effectiveness of the network of detection and classification at same time and the training time consumed in this synchronized network.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: RC0254 Neoplasms. Tumors. Oncology (including Cancer)