Abstract
The direction of signal transmission is crucial for neural networks. Therefore, the direction of signal flow, which could be provided by identifying neuronal polarity, should be included when we analyze neural networks. However, neuronal polarity is usually identified by biochemical method which is time consuming and might not be an appropriate way to deal with large-scale neural networks. To solve this problem, we proposed the algorithm for skeleton-based neuronal polarity identification (ASNeuPI). In ASNeuPI, we first morphologically divide a neuron into several substructures, and then extract their morphological features. By applying methods in machine learning, we got an optimal axis providing highest accuracy to serve as the discriminant feature for polarity classification. We tested this method on neurons innervating protocerebral bridge (PCB) or medulla (MED) in Drosophila. The data were obtained from Brain Research Center, National Tsing Hua University. On average, the polarity of above 85% terminal points in a neuron could be correctly identified. Among all the morphological features tested, the distance to soma is the most useful one. Our results show that ASNeuPI is workable and has the potential to provide a computer-based semi-automatic procedure to predict neuronal polarity from skeleton data.