Neuronal Network Signal Process Soft Tool Integration And Analysis Of The Correlation Algorithms

Nowadays, studying the neural function in network lever is one of the hot research topics in neuroscience. Multi-Electrode Arrays (MEA) can non-invasively record continuous spiking activities of neuronal network in real time for a long time. It's a powerful tool to research the characters of neuronal networks. Processing the signal collected by MEA is a prerequisite to reveal the character and the mechanism of the neuronal networks. Correlation analysis is an important purpose of the neuronal signal process. Using stimulus responding signal of neuronal network for correlation analysis can reveal the relation of the neurons in the network. To wipe off the stimulus-induced correlation is a matter that must be considered to analyze the correlation of the stimulus responding signal.To process the MEA-collected signal, we developed a soft tool to process the neuronal network signal collected by MEA in Java called Neuronal Network Signal Processing Tools (NNSPT). This soft tool is integrated as an important functional module in BioLAB that is a soft platform developed by our laboratory. The soft tool involves 20 algorithms that are commonly used to process the neuronal signal and corresponding visualize program to show the analysis result. The soft tool can be easily used and is extendable. It's a powerful tool to process the neuronal network signal.In this paper, the follows are introduced in detail: the usage of the soft tool, the development of the interface program and the picture display program, the techniques used in the development process.In this thesis, two commonly used correlation analysis algorithms: Crosscorrelogram and Joint peristimulus time histogram (JPSTH) are used to analyze the stimulus responding signal of the neuronal network cultured on MEA. The effect of the two algorithms'corrected algorithms to wipe off the stimulus-induced relation is the main point of discussion. To wipe off the stimulus-induced correlation, Crosscorrelogram has an corrected method called shift predictor corrected crosscorrelogram, JPSTH has two corrected method: normalized JPSTH and shift-corrected JPSTH. The result shows that 1) No matter how to construct the shift predictor corrected cross-correlogram, the effect of wiping off the stimulus-induced correlation has no difference.2) Though normalized JPSTH and shift-corrected JPSTH both can wipe off the stimulus-induced relation, the analysis result of the former isn't influenced by the detail of the data while the later reverse.In this paper, an improved method is brought forward to solve the problem shift-corrected JPSTH: Normalized shift-corrected JPSTH. The improved method is used to analyze the neuronal network signal collected by MEA. The result shows that, when the neural network keeps up a constant link state but has different experiment data, the proposed method can embody exactly the link state of the neural network.