Abstract
Dendritic spines are membranous protrusions from the neuronal surface. They play an important part in the nervous system.They are very small and have various shapes, so it is challenging to detect them in neuron images. This paper presents a novel method to detect dendritc spines and sepa- rate spines from dendrites in 2D images. A new backbone ex-traction method which is based on directional morphological filter, Hessian filter and thinning iteratively to remove spines and obtain a smooth backbone is developed, and shortest path method along backbone is applied here to extract the boundary of the dendrites. Spines are then separated and detected from the dendrites after the boundaries are obtained. We test our algorithm on real images and show good results.