Semi-Automatic Neuron Tracing in Electron Microscope Images: Preliminary Results

Deli Wang, Weijie Liu, Nenggan Zheng, Ting Zhao

Abstract: Reconstruction of high-resolution connectomes has been a key approach towards understanding how the brain functions. However, current methods of reconstructing neurons from electron microscope images are either computationally expensive or labor intensive. In this paper, we proposed a novel framework using interactive segmentation as a region extraction method and automatic skeletonization as a morphology fitting method to accelerate neuron tracing, especially for researchers with limited resources. Preliminary user experiments showed that our method could be at least six times more efficient than the manual method, while the cost of computation remains affordable to a small lab.