Fully automated extraction of neural signals from imaging data
State of the art methods to extract fluorescence traces from calcium imaging data require manual validation of every single cell identified. This becomes tedious and labor intensive when recordings over several experiments can involve tens of thousands of cells. The goal of this project is to create new and/or improve existing methods such that they provide a stronger guarantee of successful cell extraction.
To reach that goal the aim is to bring together expertise in calcium imaging and machine learning. Our methodology is based on new approaches in generative modelling that allow for the automated extraction of underlying latent variables that govern a complex dynamical system.
Experience with Tensorflow (and/or other equivalent deep learning libraries) and good programming skills are necessary.
Benjamin Grewe, Hafsteinn Einarsson