Neuron crawler: An automatic tracing algorithm for very large neuron images

Automatic 3D neuron reconstruction for very large 3D light microscopy images remains to be a challenge in neuroscience. Few existing neuron tracing algorithms can be used with commonly available computers (laptops, desktops, or workstations) to efficiently and accurately reconstruct a neuron in image stacks that are tens of gigabytes or greater. We introduce a new automatic tracing algorithm called Neuron Crawler, which works by first tracing a region of interest (e.g., around the soma), and then iteratively tracing in adjacent image tiles to grow the neuron structure in 3D to its termination point within the image. Our experimental results show that Neuron Crawler can achieve reconstruction accuracy that is comparable to several state-of-the-art algorithms, but with much less computational cost.