Recent advances in cellular microscopy show the capability of gathering large volumes of data for examining small structures in biology. This thesis describes algorithms for locating, segmenting and measuring structures in microscope images. Collagen fibres form important structures in tissue, and are essential for force transmission, scaffolding and cell addition. Each fibre is long and thin, and large numbers group together into complex networks of bundles, which are little studied as yet. Serial block-face scanning electron microscopy (SBFSEM) can be used to image tissues containing the fibres, but analysing the images manually is almost impossible - there can be over 30,000 fibres in each image slice, and many hundreds of individual image slices in a volume. In this thesis we describe a system for automatically identifying and reconstructing the individual fibres, allowing analysis of their paths, how they form bundles and how individual fibres weave from one bundle to another. We also describe and evaluate a method for segmenting cell nuclei from SBFSEM, an important task for many studies. We use a Convolutional Neural Networks to locate the boundary of the nuclei in each image slice. Geometric constraints are used to discard false matches. The full 3D shape of each nucleus is reconstructed by linking the boundaries in neighbouring slices. We demonstrate the system on several large image volumes.
|Date of Award||1 Aug 2019|
- The University of Manchester
|Supervisor||Karl Kadler (Supervisor) & Timothy Cootes (Supervisor)|
- Fibres Reconstruction - Nuclei Segmentation - Nuclei Reconstruction - CNN