Gaussian Mixture Modeling of Single-Neuron Responses Obtained from Confocal-Calcium-Imaging of Dissociated Rat Hippocampal Neurons

Advances in microscopy enable monitoring a broad spectrum of heterogeneity in calcium (Ca2+)-spike train in dissociated cultures at a higher-resolution. The resulting dataset requires reproducible analytics that is robust, automated, and scalable to large datasets. Here, we present the monitoring of Ca2+ -activity in rat hippocampal-neurons using spinning-disk confocal microscopy. Moreover, we propose a clustering framework based on Gaussian mixture modeling (GMM) that can be used for the identification of functional subgroups. Specifically, we propose an approach for validation of the clusters through fitting appropriate probability density function to the spiking train with minimum Akaike information criterion (AIC). Here we show that a dataset of 118 neurons obtained from 1-2 day old mice pup can be grouped in 6 clusters. We demonstrate that the proposed approach can be used to isolate the dormant cells and active cells of various types with limited user intervention. The proposed pipeline for analysis can be used for the grouping of neurons that follow a similar distribution of activated states