ABOUT

OVERVIEW INFO GALLERY FUNDING
ABOUT PROJECT

Tristan Henser-Brownhill

Tristan Henser-Brownhill

BBSRC-CASE PhD Student in Computer Vision and Bioimaging

About Me


I am an interdisciplinary PhD Student in Computer Vision and Bioimaging based at the University of Manchester. My doctoral research combines machine learning (ML) approaches with advanced bioimaging, with a particular emphasis on deep convolutional neural networks (CNNs). I use a form of quantitative phase imaging (QPI) called ptychography to extract label-free information from cells in order to characterise their specific phenotypic states.

I am primarily based in the Division of Informatics, Imaging & Data Sciences where I am supervised by Prof. Tim Cootes (Professor of Computer Vision). I also work with two groups in the Wellcome Centre for Cell-Matrix Research where I am supervised by Dr. Christoph Ballestrem and Dr. Patrick Caswell. My project is funded by a CASE industrial partnership between the BBSRC and Phase Focus Ltd.

Before starting my PhD I completed a Masters in Post-Genomic Biology at the University of York, where I first used ptychography during my work with Dr. Peter O'Toole in the Imaging and Cytometry labs. Throughout my Masters research, I employed an array of bioinformatics and bioimaging methods to study cell behaviour and identity, and to investigate self-renewal associated epigenetic states. I subsequently spent several years in active research postitions, including two years in Dr. Paola Scaffidi's lab (Cancer Epigenetics) at the London Research Institute and Francis Crick Institute, where I was responsible for the design, generation, and validation of CRISPR libraries targeting epigenetic factors; and one year at the CRUK Manchester Institute, where I investigated potential prognostic prostate cancer biomarkers.

I am an interdisciplinary PhD Student in Computer Vision and Bioimaging based at the University of Manchester. My doctoral research combines machine learning (ML) approaches with advanced bioimaging, with a particular emphasis on deep convolutional neural networks (CNNs). I use a form of quantitative phase imaging (QPI) called ptychography to extract label-free information from cells in order to characterise their specific phenotypic states.

I am primarily based in the Division of Informatics, Imaging & Data Sciences where I am supervised by Prof. Tim Cootes (Professor of Computer Vision). I also work with two groups in the Wellcome Centre for Cell-Matrix Research where I am supervised by Dr. Christoph Ballestrem and Dr. Patrick Caswell. My project is funded by a CASE industrial partnership between the BBSRC and Phase Focus Ltd.

Before starting my PhD I completed a Masters in Post-Genomic Biology at the University of York, where I first used ptychography during my work with Dr. Peter O'Toole in the Imaging and Cytometry labs. Throughout my Masters research, I employed an array of bioinformatics and bioimaging methods to study cell behaviour and identity, and to investigate self-renewal associated epigenetic states. I subsequently spent several years in active research postitions, including two years in Dr. Paola Scaffidi's lab (Cancer Epigenetics) at the London Research Institute and Francis Crick Institute, where I was responsible for the design, generation, and validation of CRISPR libraries targeting epigenetic factors; and one year at the CRUK Manchester Institute, where I investigated potential prognostic prostate cancer biomarkers.

Academic Background:

MRes, Distinction, University of York

BSc, First Class (with Distinction), University of York

Contact


University of Manchester, UK

tristan.henser-brownhill(at)postgrad.manchester.ac.uk

LinkedIn

Google Scholar

ORCID: 0000-0002-6505-6123



Published Works


1. Henser-Brownhill, T., Monserrat, J., and Scaffidi, P. (2017). Generation of an arrayed CRISPR-Cas9 library targeting epigenetic regulators: from high-content screens to in vivo assays. Epigenetics, 12(12).

2. Morales Torres, C., Biran, A., Burney, M. J., Patel, H., Henser-Brownhill, T., Shapira Cohen, A-H., Li, Y., Hamo, R. B., Nye, E., Spencer-Dene, B., Chakravarty, P., Efroni, S., Matthews, N., Misteli, T., Meshorer, E., and Scaffidi, P. (2016). The linker histone H1.0 generates epigenetic and functional intratumor heterogeneity. Science, 353(6307).

3. Suman, R., Henser-Brownhill, T. J., Langley, K., & O'Toole, P. (2014). Ptychography: Label free imaging of the cell cycle and mitosis. In MOLECULAR BIOLOGY OF THE CELL (Vol. 25). 8120 WOODMONT AVE, STE 750, BETHESDA, MD 20814-2755 USA: AMER SOC CELL BIOLOGY.

Deep Learning for the Analysis of Ptychographic Imaging Data

Deep Learning for the Analysis of Ptychographic Imaging Data

Tristan Henser-Brownhill and Tim Cootes

The aim of this project is to improve and automate the analysis of quantitative phase imaging (QPI) data from live cell experiments using machine learning approaches. It is hypothesised that the detection and characterisation of different cellular phenotypes from live time-lapses can be achieved computationally using deep-learning methods without the need for conventional fluorescent labels.



Project


Cells are mostly transparent, and morphological alterations in response to transcriptional changes can be relatively subtle. Therefore, fluorescent labelling is usually required to differentiate different cell states under a standard light microscope1. The use of fluorescent markers, such as conjugated antibodies; or the genetic manipulation of cells with reporter molecules, such as fluorescent proteins and peptides, can significantly perturb normal cellular function2-4. They can also be difficult to employ effectively when using sensitive samples such as primary patient cells.

Ptychography is a lens-free quantitative phase imaging (QPI) technique used to study the composition of live cell populations5. QPI methods can retrieve vast amounts of information without the need for labelling, as phase measurements can be used to quantify attributes such as dry mass, thickness, volume, granularity, and morphology6. Time-lapse experiments also provide information about cell division and motility, along with a multitude of other phenotypic features7. Hence, the potential to facilitate the classification of different cell types and disease states from primary populations, without the need to perturb the sample, is a major advantage of ptychography.

Despite their advantages, QPI methods generate enormous amounts of data, especially when used to interrogate large cell numbers across multiple time points. Hence, analysis of ptychographic experiments remains a significant challenge. Current approaches rely on semi-automated cell-segmentation and tracking pipelines using conventional image processing algorithms. This information can then be used to extract frame-by-frame phase-related information for each segmented cell. Although under continuing improvement, these methods require manual re-adjustment between experiments and become less effective across longer time courses.

In recent years, deep-learning methods, primarily in the form of convolutional neural networks (CNNs), have revolutionised many areas of computer vision, owing largely to the increased availability of large-scale GPU-based computation8. Furthermore, the introduction of standardised open-source deep-learning frameworks has allowed these powerful machine learning techniques to be adapted to various areas of biomedical imaging9-10. In this project CNNs will be deployed against QPI data in order to improve initial cell segmentation and tracking, and then to help classify healthy from diseased cells in live co-cultures using entirely label free information.



1. Day, R. N., & Davidson, M. W. (2009). The fluorescent protein palette: tools for cellular imaging. Chemical Society Reviews, 38(10), 2887-2921.

2. Agbulut, O., Coirault, C., Niederländer, N., et al. (2006). GFP expression in muscle cells impairs actin-myosin interactions: implications for cell therapy. Nature Methods, 3(5), 331-331.

3. Huang, W. Y., Aramburu, J., Douglas, P. S., & Izumo, S. (2000). Transgenic expression of green fluorescence protein can cause dilated cardiomyopathy. Nature medicine, 6(5), 482-483.

4. Krestel, H. E., Mihaljevic, A. L., Hoffman, D. A., & Schneider, A. (2004). Neuronal co-expression of EGFP and β-galactosidase in mice causes neuropathology and premature death. Neurobiology of Disease, 17(2), 310-318.

5. Marrison, J., Räty, L., Marriott, P., & O'toole, P. (2013). Ptychography–a label free, high-contrast imaging technique for live cells using quantitative phase information. Scientific Reports, 3.

6. Kasprowicz, R., Suman, R., & O’Toole, P. (2017). Characterising live cell behaviour: Traditional label-free and quantitative phase imaging approaches. The International Journal of Biochemistry & Cell Biology, 84, 89-95.

7. Suman, R., Smith, G., Hazel, K. E., et al. (2016). Label-free imaging to study phenotypic behavioural traits of cells in complex co-cultures. Scientific Reports, 6, 22032.

8. Schmidhuber, J. (2015). Deep learning in neural networks: An overview. Neural Networks, 61, 85-117

9. Litjens, G., Kooi, T., Bejnordi, B. E., et al. (2017). A survey on deep learning in medical image analysis. arXiv Preprint arXiv:1702.05747.

10. Van Valen, D. A., Kudo, T., Lane, K. M., et al. (2016). Deep learning automates the quantitative analysis of individual cells in live-cell imaging experiments. PLoS Computational Biology, 12(11), e1005177

Gallery


Coming soon...