Biomedical imaging techniques play an important role in visualisation of e.g., biological structures, tissues, diseases and medical conditions in cellular level. The techniques bring us enormous image datasets for studying biological processes, clinical diagnosis and medical analysis. Thanks to recent advances in computer technology and hardware, automatic analysis of biomedical images becomes more feasible and popular. Although computer scientists have made a great effort in developing advanced imaging processing algorithms, many problems regarding object analysis still remain unsolved due to the diversity of biomedical imaging. In this thesis, we focus on developing object analysis solutions for two entirely different biomedical image types: uorescence microscopy sequences and endometrial histology images. In uorescence microscopy, our task is to track massive uorescent spots with similar appearances and complicated motion pattern in noisy environments over hundreds of frames. In endometrial histology, we are challenged by detecting different types of cells with similar appearance and in terms of colour and morphology. The proposed solutions utilise several novel locality sensitive models which can extract spatial or/and temporal relational features of the objects, i.e., local neighbouring objects exhibiting certain structures or patterns, for overcoming the difficulties of object analysis in uorescence microscopy and endometrial histology.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:693243 |
| Date | January 2016 |
| Creators | Li, Guannan |
| Publisher | University of Warwick |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://wrap.warwick.ac.uk/81399/ |
Page generated in 0.0138 seconds