The automated analysis and statistical characterisation of subcellular biological images
<p>Biomedical research has been revolutionised by the development of high throughput systems which enable thousands of experiments to be carried out simultaneously, leading to the generation of vast amounts of microscopy data. At the same time smaller scale experiments are increasingly depende...
| Main Authors: | , |
|---|---|
| 其他作者: | |
| 格式: | Thesis |
| 語言: | English |
| 出版: |
2012
|
| 主題: |
| _version_ | 1826315973261524992 |
|---|---|
| author | Li, S Mr Simon W. Li |
| author2 | Noble, J |
| author_facet | Noble, J Li, S Mr Simon W. Li |
| author_sort | Li, S |
| collection | OXFORD |
| description | <p>Biomedical research has been revolutionised by the development of high throughput systems which enable thousands of experiments to be carried out simultaneously, leading to the generation of vast amounts of microscopy data. At the same time smaller scale experiments are increasingly dependent on accurate quantification of relatively small changes since the intricate network of interactions within a cell often counteracts attempts to perturb the function of any single protein or pathway. This has motivated the development of algorithms and software packages to assist with the analysis of biological imaging data.</p> <p>This thesis develops and applies image analysis tools to multiple biological problems, and also illustrates the whole process required to label live samples so they can be imaged in the first place, thus giving an appreciation of the practical difficulties of biomedical research. We begin with the development of a kymograph tool, which is used for the spatio-temporal subcellular quantification of the chromosomal passenger complex, a complex of proteins involved in regulating mitosis. Following this we start to consider inter-cellular signalling interactions in a study of the IGF2 pathway in clumped cells. A random forest classification framework using cell-based image features is developed, and shown to be able to distinguish between cells with different levels of activation. Importantly the random forest feature importance measures are shown to be useful in helping to identify biologically relevant differences. Finally we apply the random forest classifier to the analysis of immunofluoresence images of Ewing's sarcoma tumour biopsies, which are associated with clinical outcome data, allowing us to develop a framework for integrating imaging data with patient prognosis data. The samples originate from multiple sources, making this a very challenging data set due to a lack of normalisation, but we are still able to obtain clinically useful results, in addition to identifying areas where further research is required. </p> |
| first_indexed | 2024-03-06T20:00:31Z |
| format | Thesis |
| id | oxford-uuid:27218e5e-f40b-4e2b-94e9-a3536fbd6d28 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-12-09T03:35:48Z |
| publishDate | 2012 |
| record_format | dspace |
| spelling | oxford-uuid:27218e5e-f40b-4e2b-94e9-a3536fbd6d282024-12-01T19:23:01ZThe automated analysis and statistical characterisation of subcellular biological imagesThesishttp://purl.org/coar/resource_type/c_db06uuid:27218e5e-f40b-4e2b-94e9-a3536fbd6d28Cell Biology (see also Plant sciences)TumoursBioinformatics (life sciences)Image understandingMicroscopyEnglish2012Li, SMr Simon W. LiNoble, JWakefield, J<p>Biomedical research has been revolutionised by the development of high throughput systems which enable thousands of experiments to be carried out simultaneously, leading to the generation of vast amounts of microscopy data. At the same time smaller scale experiments are increasingly dependent on accurate quantification of relatively small changes since the intricate network of interactions within a cell often counteracts attempts to perturb the function of any single protein or pathway. This has motivated the development of algorithms and software packages to assist with the analysis of biological imaging data.</p> <p>This thesis develops and applies image analysis tools to multiple biological problems, and also illustrates the whole process required to label live samples so they can be imaged in the first place, thus giving an appreciation of the practical difficulties of biomedical research. We begin with the development of a kymograph tool, which is used for the spatio-temporal subcellular quantification of the chromosomal passenger complex, a complex of proteins involved in regulating mitosis. Following this we start to consider inter-cellular signalling interactions in a study of the IGF2 pathway in clumped cells. A random forest classification framework using cell-based image features is developed, and shown to be able to distinguish between cells with different levels of activation. Importantly the random forest feature importance measures are shown to be useful in helping to identify biologically relevant differences. Finally we apply the random forest classifier to the analysis of immunofluoresence images of Ewing's sarcoma tumour biopsies, which are associated with clinical outcome data, allowing us to develop a framework for integrating imaging data with patient prognosis data. The samples originate from multiple sources, making this a very challenging data set due to a lack of normalisation, but we are still able to obtain clinically useful results, in addition to identifying areas where further research is required. </p> |
| spellingShingle | Cell Biology (see also Plant sciences) Tumours Bioinformatics (life sciences) Image understanding Microscopy Li, S Mr Simon W. Li The automated analysis and statistical characterisation of subcellular biological images |
| title | The automated analysis and statistical characterisation of subcellular biological images |
| title_full | The automated analysis and statistical characterisation of subcellular biological images |
| title_fullStr | The automated analysis and statistical characterisation of subcellular biological images |
| title_full_unstemmed | The automated analysis and statistical characterisation of subcellular biological images |
| title_short | The automated analysis and statistical characterisation of subcellular biological images |
| title_sort | automated analysis and statistical characterisation of subcellular biological images |
| topic | Cell Biology (see also Plant sciences) Tumours Bioinformatics (life sciences) Image understanding Microscopy |
| work_keys_str_mv | AT lis theautomatedanalysisandstatisticalcharacterisationofsubcellularbiologicalimages AT mrsimonwli theautomatedanalysisandstatisticalcharacterisationofsubcellularbiologicalimages AT lis automatedanalysisandstatisticalcharacterisationofsubcellularbiologicalimages AT mrsimonwli automatedanalysisandstatisticalcharacterisationofsubcellularbiologicalimages |