Machine Learning In Dynamic Microscopy

The ability to rapidly extract information about immune cells from large numbers of microscope images is an on-going computational challenge. In this thesis, deep neural networks were used to track and segment T cells from microscope movies. An innovative 3D convolution neural network approach was developed, allowing us to simplify the tracking by conducting detection and association phases simultaneously. These results will help us to characterize decision-making in immune cells, improving our understanding of the vaccination response and allowing us to better treat immunity-related diseases.