Use this URL to cite or link to this record in EThOS:
Title: Quantitative analysis of cell function and death in label-free high content screening
Author: Nketia, Thomas
ISNI:       0000 0004 6496 1759
Awarding Body: University of Oxford
Current Institution: University of Oxford
Date of Award: 2016
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Time-lapse data is increasingly being used to conduct more detailed high throughput studies. Quantitative information that is being derived from vast and complex image data sets is essential in our understanding of basic cellular processes. Advances in cell culturing methods combined with sophisticated molecular probes allow the monitoring of a broad array of cellular functions in vivo. These experiments provide a wealth of multi-channel time-lapse data that renders traditional manual interpretation infeasible. The work outlined in the thesis is aimed at quantitative analysis of such time-lapse data resulting in statistical analysis methods and software tools that offers an improvement in the efficiency and impact of phenotypic screening experiments. The research contributions towards the stated aim is outlined in three (3) main areas: cell morphology and lineage, population-based summaries for cell function, and cell state labelling. Firstly, cell morphology and lineage tasks address the two fundamental tasks in time-lapse quantitative biological image analysis; segmentation and tracking. Cell segmentation presents a key challenge in any image-based single cell analysis as most further analysis is heavily dependent on this step. Details of a segmentation approach based on an existing light phase retardation model and results on sample data of phase contrast HeLa cervical cells is presented. Using phase retardation feature extraction to precondition images for deep learning is also explored. Cell shape and texture features from segmentation are then used in the tracking to obtain a lineage metric to assess the progression of cell morphology changes and motion over time. Details of a tracking scheme based on coupled minimum cost ow network is presented. Secondly, metrics that quantify cell function of a population such as proliferation, viability and migration based on a summary of single cell measurements are common. Such metrics however generally do not account for segmentation errors resulting from cell crowding and overlapping cell boundaries. Details of an approach that allows incorporating the confidence in segmentation accuracy for each single cell in the population metric is presented. Analysis on simulated data shows that the proposed method provides better summary that is representative of the cell population and hence could improve conclusions made from quantitative analysis of cell populations. Determining the variability in the mode of death of cells is important in multiple live cell phenotypic toxicity screens. The mechanism associated with the mode a cell progresses towards death can be observed as a sequence of morphological events associated with cell death. As these events are observed morphologically, changes in shape and texture features can be used to model the time series process. Here, such temporal evolution is modelled as hierarchical dirichlet process of a hidden markov model (HDP-HMM). The model eliminates the limit on the number of states representing morphological events and hence allows new states to be discovered as more cell data from new screens are added. Details also include sequence analysis methods used to group similar cells based on the temporal changes of their morphological features. Overall, the proposed work provides methods and software tools that enable the efficient interpretation of patterns and groups in large amounts of data obtained from high throughput phenotypic screens that otherwise would be infeasible to obtain manually. Hence, the work improves efficiency of analysis as well as help obtain objective and repeatable conclusions from live cell experiments.
Supervisor: Rittscher, Jens ; Noble, Alison ; Zisserman, Andrew Sponsor: RCUK Digital Economy Programme
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available