Deep feature extraction of neuronal reconstruction data using tree-structured sequence neural network

Abstract:
In the past, researches on neuronal reconstruction data mostly focused on the simple morphological features. With the explosive growth of complex brain-wide data, they can no longer meet the requirements of exploring the relationship between morphological patterns and physiological functions. This paper proposes a tree-structured sequence neural network that simulates the biological pattern of neurons to extract deep features from reconstruction data. In both tasks of classification and retrieval, satisfactory results can be achieved using the network as backbone. This network provides great usability and inspiration for future researches.