TOWARDS AUTOMATED NEURON TRACING VIA GLOBAL AND LOCAL 3D IMAGE ANALYSIS

The reconstruction of the neural network is essential in computational neuroscience. Here, we present an automatic algorithm to trace single neuron projections based on two core algorithmic ideas: a global step segmenting all neuron bodies and their projections and a local growing phase that accommodates to the nonuniform illumination and to the noise of the sample. We tested our algorithm on two 3D stacks of two-photon images acquired from a human dysplastic brain sample. The results show that the traces produced are statistically equivalent to the ground truth, according to the Friedman and Li tests. Furthermore, we found that our algorithm outperforms other state-of-the-art methods.