We have a large database of time-course microscopy images showing individual neural stem cells as they grow, move, and change within hundreds of micro-environments. We would like to develop a fully automated computational pipeline based on convolutional neural networks (CNNs) that can segment the cells from the still images, and then track the cells over the timelapses as they move around, divide, and die. Once we have this tracking data, our aim is to relate the cells’ history (lineage, migration, time spent in each part of the microenvironment) with their eventual fate (differentiation or proliferation), which we can relate to the timelapse images with a corresponding end-point imaging dataset. We have already annotated a large training set for CNN segmentation training, and we have identified at least two candidate CNN segmentation frameworks.
