Neuron segmentation with level sets

We propose a variational scheme for neuron segmentation. The growing demand for an automated image analysis tool for efficient segmentation of neurites has led to significant focus and advancement in the field of neuro-image analysis. However, the complex structure of the neurites and imaging artifacts pose considerable challenge for automated image analysis. Analysis of the neuronal morphology is important to determine the cause of certain neurological diseases (e.g. Alzheimer´s disease). Such applications demand accurate segmentation of the neuron that respect the geometric structure as well as the branching patterns. Our proposed method relies on level set evolution to perform segmentation. The propagation of the level sets is primarily governed by a medial flow vector field, which encourages the growth of the level set along the width and the axis of the vascular neurite structures. Experimental results on both 2D and 3D images suggest that the method is robust to imaging artifacts and is efficient in capturing the complex branching structures of the neurons.