Use this URL to cite or link to this record in EThOS:
Title: Learning to study the developing brain : characterising brain development with machine learning, image analysis and ex vivo culture
Author: Hailstone, Martin
ISNI:       0000 0004 7966 2276
Awarding Body: University of Oxford
Current Institution: University of Oxford
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Brain development, and indeed development in general, is complex. Correct development requires tight regulation of cell numbers, movements and fates. However, decisions are taken by individual cells, integrating spatial and temporal cues. Understanding development, therefore, necessitates studying cells in the native tissue context. In this thesis, I focus on neurogenesis, studying neural stem cell division in cultured whole Drosophila brains. To facilitate this, I develop an approach based on live imaging, image analysis and the use of a novel software tool that is applicable to studying development in a range of systems. I start with optimised imaging, culture and image processing to study cell divisions in Drosophila brains. Building on previous work by L. Yang, I develop a pipeline allowing high resolution, medium-throughput imaging of multiple brains. Using patch-based denoising, and gradient-descent-based image registration, I follow divisions of stem cells and their daughters in an intact brain, in a way that was previously unachievable. This allowed me to record cell behaviours, such as cell cycle length, which I show are consistent with published results. However, analysing the resulting datasets was highly time-consuming, and impractical on larger scales. Existing automated analysis tools proved to be inadequate, likely due to signal-to-noise ratio and complex structures in these images. To overcome this, I developed CytoCensus, image analysis software for complex 4D data, in collaboration with D. Waithe. Unlike other software, which aims to identify whole cells and their boundaries, CytoCensus identifies only cell centres, allowing it to cope with neighbouring cells and low signal-to-noise. I show that CytoCensus identifies different cell types in the brain, performing better than competing approaches. Combining the aforementioned live imaging and culture methods with CytoCensus analysis, provided insights into normal and aberrant development. I characterise changes in the development mutant syp, in which both neuroblasts and daughter cells exhibit an increased division rate, accounting for the observed increase in brain size. I further demonstrate that this approach is generally applicable to studying developmental processes in other contexts by quantifying transcription factors in mouse embryos, counting organelles in spermatogenesis, and comparing cell distributions in zebrafish retinal organoids. Taken together, it is clear that this approach, and these tools can be used to study the development of different organisms, by studying the behaviour of cells in a relevant context.
Supervisor: Davis, Ilan ; Parton, Richard M. Sponsor: Engineering and Physical Sciences Research Council ; Medical Research Council
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Development ; Image analysis ; Neurobiology ; Machine learning