SegEM: Efficient Image Analysis for High-Resolution Connectomics

Abstract

Progress in electron microscopy-based high-resolution connectomics is limited by data analysis throughput. Here, we present SegEM, a toolset for efficient semi-automated analysis of large-scale fully stained 3D-EM datasets for the reconstruction of neuronal circuits. By combining skeleton reconstructions of neurons with automated volume segmentations, SegEM allows the reconstruction of neuronal circuits at a work hour consumption rate of about 100-fold less than manual analysis and about 10-fold less than existing segmentation tools. SegEM provides a robust classifier selection procedure for finding the best automated image classifier for different types of nerve tissue. We applied these methods to a volume of 44 × 60 × 141 μm(3) SBEM data from mouse retina and a volume of 93 × 60 × 93 μm(3) from mouse cortex, and performed exemplary synaptic circuit reconstruction. SegEM resolves the tradeoff between synapse detection and semi-automated reconstruction performance in high-resolution connectomics and makes efficient circuit reconstruction in fully-stained EM datasets a ready-to-use technique for neuroscience.