Brain Development: Machine Learning Analysis Of Individual Stem Cells In Live 3D Tissue
Hailstone M., Yang L., Waithe D., Samuels T., Arava Y., Dobrzycki T., Parton R., Davis I.
Brain malformations often result from subtle changes in neural stem cell behaviour, which are difficult to characterise using current methods on fixed material. Here, we tackle this issue by establishing optimised approaches for extended 3D time-lapse imaging of living explanted Drosophila brains and developing QBrain image analysis software, a novel implementation of supervised machine learning. We combined these tools to investigate brain enlargement of a previously difficult to characterise mutant phenotype, identifying a defect in developmental timing. QBrain significantly outperforms existing freely available state-of-the-art image analysis approaches in accuracy and speed of cell identification, determining cell number, location, density and division rate from large 3D time-lapse datasets. Our use of QBrain illustrates its wide applicability to characterise development in complex tissue, such as tumours or organoids, in terms of the behaviour in 3D of individual cells in their native environment.