Image-Based Cell Profiling and Data Analysis

Analysing large image datasets to generate insights into biological systems

I work with large microscopy datasets and image-based cell profiling tools to automate the process of analysing hundreds of thousands of cells. I work extensively in the R programming language to develop data analysis pipelines that can clean, transform and model large datasets to generate meaningful outcomes.

Quantifying the response of cells to high-aspect-ratio nanostructures

Illustration of an image analysis pipeline for cell profiling
Illustration of cell-based image profiling approach.

To uncover the complexity of cellular response to high-aspect-ratio nanostructures, I construct image analysis pipelines. These can process single-cell measurements from hundreds of thousands of cells from each experiment.

Assessing many cells at once allows us to understand how the distributions of morphological features and staining intensities vary across a heterogeneous cell population, revealing subtle effects.1

Download my workshop on working with CellProfiler data in R

Analysing the response of progenitor cells in a neural tube model

A pixel prediction map
Example pixel prediction map for cell colonies.

Working as part of the PhD project of Daniel Boland, and in collaboration with Dr Despina Stamataki of the Briscoe Lab, The Francis Crick Institute, I have constructed an image analysis pipeline to assess the state of neural mesodermal progenitor cells in a neural tube model.

I combine the machine-learning based pixel classification software Ilastik, with the segmentation and measurement tool CellProfiler to assess the ventral/dorsal state of the progenitor cells. This approach allows us to separate background and density effects from our model system.

Analysing the geometry of lysosomes from STORM images

A segmentation map of lysosomes
Example of a lysosome segmentation map.

I worked to analyse the impact of a new therapeutic on the geometry of lysosomes, supporting the work of Jeremy Bost and Dr Miina Ojansivu at the Karolinska Institutet, Sweden. Stochastic Optical Reconstruction Microscopy (STORM) images were analysed using a combination of Ilastik / CellProfiler pipleline, to overcome the limitations of thresholding based pixel classification. The analysis revealed subtle effects of the therapeutic on lysosome geometry.2


  1. Adapted under terms of CC-BY license from: Seong H, Higgins SG, Penders J, Armstrong JPK, Crowder SW, Moore AC, Sero JE, Becce M, Stevens MM, Size-Tunable Nanoneedle Arrays for Influencing Stem Cell Morphology, Gene Expression, and Nuclear Membrane Curvature, 2020, ACS Nano. DOI: 10.1021/acsnano.9b08689 ↩︎

  2. Adapted under terms of CC-BY license from: Bost JP, Ojansivu M, Munson MJ, Wesén E, Gallud A, Gupta D, Gustafsson O, Saher O, Rädler J, Higgins SG, Lehto T, Holme MN, Dahlén A, Engkvist O, Strömstedt P-E, Andersson S, Smith CIE, Stevens MM, Esbjörner EK, Collén A, El Andaloussi S, Novel Endosomolytic Compounds Enable Highly Potent Delivery of Antisense Oligonucleotides, 2022, Commun. Biol. DOI: 10.1038/s42003-022-03132-2 ↩︎