Label-Free Modelling of Cell Cycle Stages Using Quantitative Phase Imaging and Machine Learning

  • Tristan Henser-Brownhill

Student thesis: Phd


Study of the cell cycle is vital to furthering our understanding of human development and disease. Tracking stages of the cell cycle in living cells is primarily performed using fluorescent labels and dyes such as the dual-marker fluorescence ubiquitination cell cycle indicator (FUCCI) system. Marker-based approaches can perturb normal cellular function and introduce phototoxicity. They can also be relatively time-consuming to set up and are not suitable for all cell types. The recent introduction of commercially available quantitative phase imaging (QPI) has enabled researchers to embrace practical label-free imaging as an alternative. Several studies have shown that stages of the cell cycle appear to correlate with QPI-specific features such as intracellular dry mass. However, the data generated by these information-rich approaches is often difficult to interpret manually using standard analysis techniques, which has hindered mainstream uptake of QPI despite its significant potential. In this research, computational models were constructed to simulate the biphasic trajectories of FUCCI cell cycle markers from single images of individually segmented cells captured using QPI. These estimates permit label-free cell cycle stage sorting. It was demonstrated that conventional machine learning techniques could generate estimations based on a few hand-crafted ptychographic features. Moreover, using deep learning, it was shown that cell cycle states could be estimated directly from QPI images through learned features. Model performance was evaluated on examples comprising fully tracked cell cycle trajectories, and at the population level using large images comprised of many cells. The capacity of models to generalise across multiple human cancer cell lines was also assessed. Finally, the capability of the marker-free computational approach to identify cell cycle arrest induced by chemotherapeutic agents was evaluated with promising results.
Date of Award31 Dec 2022
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorTimothy Cootes (Supervisor), Christoph Ballestrem (Supervisor) & Patrick Caswell (Supervisor)


  • quantitative phase imaging
  • QPI
  • cell cycle
  • machine learning

Cite this