We implemented a fully functional prototype workflow for mapping cytoarchitectonic areas across an unregistered series of consecutive histological sections. It is a web-based tool, which combines editing and display of 2D annotations (using microdraw) with interactive configuration and remote monitoring of Deep Learning training processes on a supercomputer. The supercomputing worfklows need to be accessible via ssh from the server running the web frontend. Based on this software, we successfully mapped several brain areas precisely in thousands of consecutive tissue sections by providing expert annotations only in approximately every 100^th section. This allowed us to publish the first ultrahigh-resolution full 3D maps of cytoarchitectonic areas in the BigBrain, which are shared in the knowledge graph (Area hOc2, area hOc1, several others in preparation).

The prototype software codes is available as a zip file in the drive of this collab.