Multilayer Neural Networks with Receptive Fields as a Model for the Neuron Reconstruction Problem

• Published in: Artificial Intelligence and Soft Computing pp. 242 - 250
• Main Author: Czarnecki, Wojciech
• Format: Book Chapter
• Language: English
• Published: Berlin, Heidelberg Springer Berlin Heidelberg
• Series: Lecture Notes in Computer Science
• Subjects: artificial neural network; computer vision; neuron reconstruction; receptive fields
• ISBN: 3642293492; 9783642293498
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-642-29350-4_29

The developed model consists of a multilayer neural network with receptive fields used to estimate the local direction of the neuron on a fragment of microscopy image. It can be used in a wide range of classical neuron reconstruction methods (manual, semi-automatic, local automatic or global automatic), some of which are also outlined in this paper. The model is trained on an automatically generated training set extracted from a provided example image stack and corresponding reconstruction file. During the experiments the model was tested in simple statistical tests and in real applications, and achieved good results. The main advantage of the proposed approach is its simplicity for the end-user, one who might have little or no mathematical/computer science background, as it does not require any manual configuration of constants.