Automation of Synchronicity Identification in Hippocampal Neurons through Intelligent Data Clustering Approach

Neuronal synchronicity is central in controlling the cognitive functions and disruption in neuronal synchronicity may lead to diseased state. Since the neurons show significant heterogeneity in firing pattern in case of in vitro and in vivo cell recordings, automated identification of synchronous and asynchronous neurons in a large population remains challenging. In this context, an efficient data analytics approach is proposed where the time-lapse data is primarily obtained from imaging of intracellular $\mathrm{C}\mathrm{a}^{2+}$ in primary cultures of hippocampal neurons. Here, F1uo-4 is used as the fluorescent indicator for measuring for cytosolic calcium through imaging using confocal microscope. To categorize synchronous response from a set of heterogeneous $\mathrm{C}\mathrm{a}^{2+}$ spiking data, an efficient artificial neural networks based clustering algorithm has been proposed, which proceeds through a variable reduction approach. This algorithm further enables the usage of evolutionary optimizers to solve the optimization problem of Fuzzy C-means (FCM) clustering. Moreover, the novel algorithm estimates the optimal number of cluster and optimal artificial neural network topology, which remains to be a longstanding issue. In order to validate the obtained clustering solution, the cross-correlation coefficient and spiking pattern is measured for the clustered neuron cell data. The obtained solution is compared with that of conventional FCM algorithm such that the efficiency of proposed approach could be tested.