Use this URL to cite or link to this record in EThOS:
Title: Learning to extract tumour vasculature : techniques in machine learning for medical image analysis
Author: Bates, Russell
ISNI:       0000 0004 7229 7835
Awarding Body: University of Oxford
Current Institution: University of Oxford
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Cancer is a leading cause of death worldwide with an estimated 14 million new cases occurring yearly and approximately 8 million deaths. Although much progress has been made in the understanding and treatment of cancer, there are still many mechanisms that remain poorly understood. The development of vasculature is known to be a key element in facilitating the growth of a tumour. Modern imaging modalities such as multi-photon fluorescence microscopy allow unprecedented opportunities to examine and quantify this vasculature in vivo. However, the appearance of vascular networks, imaged at these scales, can be extremely complex and the automatic delineation of such large, tortuous and chaotic vascular networks is a non-trivial task. In this thesis we develop a number of methods for the automatic delineation of tumour vasculature, imaged using in vivo microscopy. Recent developments in machine learning have provided a powerful set of techniques for the automated analysis of complex structures in images. Leveraging these, it is possible to develop algorithms, capable of learning from human annotations, which are able to analyse extremely large images quickly and with a high degree of accuracy. The key contributions of this thesis are as follows: we present a novel supervoxel algorithm for use in a lightweight machine learning framework for segmentation. We adapt the current state-of-the-art in segmentation using 2D deep fully convolutional neural networks for use in 3D vascular segmentation. We further demonstrate the use of hybrid convolutional-recurrent networks for extracting 3D vessel centrelines. We propose the use of Conditional Adversarial Networks for refining the extraction of vessel centrelines directly. Finally, we demonstrate the ability of the developed methods to make quantitative observations on longitudinal changes to in vivo tumour vasculature development.
Supervisor: Grau, Vicente ; Kersemans, Veerle ; Schnabel, Julia A. Sponsor: EPSRC
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available