LSI Seminar Series: Hanchuan Peng, Ph.D., Allen Institute for Brain Science

Despite substantial advancement in the automatic tracing of brain cells' 3D morphology in recent years, it is challenging to apply existing algorithms to very large image datasets containing billions or more voxels, especially for applications such as morphometry of single neurons at the whole-brain scale. We have developed a new platform combining several newly developed technologies including Vaa3D, TeraFly, UltraTracer, and TeraVR (Nature, 2019), to attempt this challenge. Particularly, we have used TeraFly to invoke Vaa3D to quickly visualize the whole mouse brain image volume and manage the thousands of billions of voxels in each of the brain volume. We then used UltraTracer to wrap several efficient base tracers to trace such massive data volumes. Finally, we combined virtual reality and machine learning into a tool called TeraVR for efficient proofreading and editing of such reconstructed neuron morphology. We are further improving the integration of these tools for more scalable and accurate single neuron morphometry.