Automated Quantification of Human Islet Structure and Geometry

Abstract

This thesis presents a novel toolkit for image analysis that seeks to improve our understanding of the structures of the islets of Langerhans within the human Pancreas. Our aim is to improve understanding of structural changes associated with age and diabetes. The islets of Langerhans are structures within the pancreas which are the metabolic control centres for blood sugar (glycaemic) homeostasis, amongst other things. These islets contain several endocrine cell types including insulin producing beta cells and in diabetes this composition changes. One of the major problems with our understanding of what happens in human diabetes is the lack of available samples for study. In the University of Exeter medical school, we are able to access approximately 160 diabetic pancreases, making this the world’s largest collection of such material. Due to the progress of diseases such as diabetes being associated with changes in cellular composition of the islets, we examine the structure of islets both as a totality and at a cellular level. We have developed algorithms to automate the mapping of islets from microscopy images to spatially informative false colour map at a cellular resolution. We begin this analysis with a systematic analysis of the relevant cell numbers in the pancreas in ex vivo human subjects, with and without type 1 or type 2 diabetes. We subsequently investigate this structure by constructing metrics that characterise the geometry of the islets (such as the isoparametric ratio) and how diabetes may change these. We analyse the topological properties of the islet structure using recently developed techniques in computational homology and in particular persistent homology upon individual islets. Specifically, this involves finding generating sets for the first and second Betti numbers for particular mapped microscopy images and investigating Betti number stability under perturbations of parameters used for the automated image analysis. Finally, we explore cellular network structures within the islets, and how diabetes may change the pattern of intracellular connections within the islets. We do this via network metrics which characterise some aspects of the structure of the networks such as connectedness. We do this considering the graph as an overall islet network and the subnetworks between different cell types.