Brainchop in Brainforge using Meshnet

Proposal Abstract

MeshNet is an efficient neural network architecture for image segmentation, which can provide results comparable to freesurfer output in a fraction of the time. This project involves integrating an end-to-end computation for MeshNet segmentation in docker, and integrating this computation into the BrainForge backend and frontend.

About the Project

Umbrella Project

• BrainForge

Emphasis:

• Engineering

Expected Background

• Undergraduate friendly. Familiarity with python is recommended.

Primary Point of Contact

• Brad Baker (bbaker43@gsu.edu)

Supervisor

• Jeremy Bockholt (jbockholt@gsu.edu)

References and External Resources

• brainforge.rs.gsu.edu

Estimated Timelines

• 2-3 months (summer project)

Possible Deliverables

• Singularity container which wraps MeshNet
• Demonstration of the container on a small study + comparison with freesurfer