Cell Neighbourhood Analyses
AI related methods become more important in various areas of research, especially in Biology and Life science. We use them to investigate the cell neighbourhood in mouse embryos and organoids. To analyze the patterns in cell fate decisions, we trained machine learning models on experimental and simulated data to classify organoids and predict the cell fate of single cells.
Previous work
We developed a multiscale image analysis pipeline to investigate the changes in internal morphology of spheroids upon drug treatment. It combines image segmentation with graph theory and computational topology to obtain quantitative features of the individual cells, the local cell neighbourhood and the global topology of the spheroid. As an output we obtain over 30 features that are relevant in multicellular systems. Our pipeline works very well for spheroids of different sizes (5,000 - 40,000 cells).
