Complex spatial structures in biomedical images: investigating the interaction of cells with their microenvironment

Gilbert, Sebastian George Belton ORCID: 0000-0002-7502-8409 (2024). Complex spatial structures in biomedical images: investigating the interaction of cells with their microenvironment. University of Birmingham. Ph.D.

Preview

Gilbert2024PhD.pdf
Text - Accepted Version
Available under License Creative Commons Attribution.
Download (155MB) | Preview

Abstract

Imaging modalities can resolve biological systems with a precision that can spatially locate individual cells of multiple cell phenotypes. Identifying meaningful patterns in heterogeneous, multi-agent and high-dimensional images grows in importance as high-throughput versions of this technology develop. Mathematical constructs offer simplifications which are then able to interpret a myriad of complex spatial structures, including point patterns, spatial complex networks and homological features. These constructs provide quantification which enables predictive power for patient disease progression and to infer optimal biological function.

Before the characterisation of images into mathematical representations, ascertaining a reliable signal amidst noise is a foundational step in describing the features of images accurately. Recent developments in machine learning have provided a powerful set of techniques for the automated analysis of complex structures in images. Leveraging these, we integrate the use of advanced image processing techniques, combining tools to construct effective processing workflows. For instance, in the identification of cell boundaries and their respective phenotypes from 2-dimensional multiplex immunofluorescence images or blood vessel structures from 3-dimensional confocal fluorescence microscopy.

Spatial analyses then encapsulate pairwise interactions, multi-phenotypic neighbourhoods and structured cellular neighbourhoods, a novel concept introduced in this research. In this thesis, we apply our techniques to 2-dimensional and 3-dimensional imaging modalities investigating features in three biological systems: Colorectal carcinoma, Sjogren's syndrome and self-organised microvascular models.

In colorectal carcinoma, we identify several immune cell phenotypes of interest for patient prognosis, such as M2 macrophages with cancerous epithelial cells. In Sjogren's syndrome we automatically identify tertiary lymphoid structures, otherwise known as ectopic lymphoid organs, a prognostic predictor of patient progression in this disease as well as in many cancers. Finally, the characterisation of self-organised microvascular models as complex spatial networks improved the classification of samples according to the medium environment and related subsequent morphology to analytical simulations of oxygen and nutrient flow.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):

Supervisor(s)	Email	ORCID
Spill, Fabian	UNSPECIFIED	orcid.org/0000-0001-8462-5080
Styles, Iain B	UNSPECIFIED	orcid.org/0000-0002-6755-0299
Dennis, Mark	UNSPECIFIED	UNSPECIFIED
Boyer, Vincent	UNSPECIFIED	UNSPECIFIED