Title: Deep Learning for Neuron Tracing (2 positions)
Joel Welling, PSC Senior Scientific Specialist
The tiny brain of the larval zebrafish has been scanned in great detail via serial electron microscopy (Hildebrand et al., Nature, 2017). It’s neural network has been mapped out by a combination of automation and painstaking human effort. As methods for scanning the brains of larger organisms are developed much better automated methods will be needed to map the huge, spectacularly complex networks of neurons involved.
We are exploring a possible approach to neuron tracking using deep neural networks. We have a large dataset of training and testing examples collected from Hildebrand et al.’s zebrafish, courtesy of Art Wetzel, one of their collaborators and a PSC scientific specialist. In a previous internship project by Richard Zhao a preliminary convolutional neural network was developed and trained using TensorFlow. The goal of this project is to improve on that network. Two students will participate in this project, one coding and one running experiments.
The coder student must have significant experience with Python programming. The experimenter must have some coding experience in any language and a desire to advance their data analysis skills.
Experience with deep learning packages, ideally TensorFlow. Experience or interest in biology and brain anatomy.
The coding student will deepen their skills with Python and deep neural networks. The experimenting student will gain experience with large scale computation and data management, and will expand their Python skills along the way.
Students gain experience with deep learning, and learn methods for quantitative analysis of the results of numerical experiments.
The coding student would be a major in Computer Science or a related discipline. The experimenter might be from any background as long as they have some computing experience.
Students in this position will receive a stipend or course credit.