Characterisation of cell motility through a bioimage analysis perspective

In the tumour Extracellular Matrix (ECM) stroma, cancer cells follow a 3D mesenchymal migration mode driven by cellular dendritic protrusions. Together with cellular adhesions to the 3D ECM, cells use elongated protrusions to exert forces and make contractile displacements. Therefore, it is possible to detect varying mechanical patterns closely related to the invasive behaviour of the cells.To study the relationship between cell motility and the morphology and dynamics of cell protrusions we acquired long (~16 h) phase contrast microscopy time-lapse videos of cells embedded in 3D collagen type I matrices are acquired. Label-free microscopy supports a more sample- friendly acquisition at the expense of reduced image quality: high presence of artefacts, low signal-to-noise ratio, and a weak contrast between the cell membrane and the collagen matrix. Thus, we propose an advanced image processing pipeline based on deep learning to automatically segment cells, quantify their protrusions and track both over time. We use recurrent units and pre-trained encoders to improve the time consistency of the results. Automatic image-processing pipelines support the extraction of a large sets of data that can be hardly analysed with the classical statistical null hypothesis test (NHT) approaches, as the p-value depends on the sample size. We conceive a new mathematical method that exploits the relationship between the sample size and the p-value to characterise the differences among the compared groups of data.The proposed analytical pipelines supported the characterisation of cell motility and cellular protrusions dynamics, establishing a starting point to discover new hallmarks for enhanced cell migration and formulate new hypotheses on cell morpho-dynamics.
Host: Ignacio Arganda Carreras

Hybrid Seminar: Donostia International Physics Center


Estibaliz Gomez de Mariscal, Instituto Gulbenkian de Ciencia

Source Name