Neuronal calcium imaging signals modeling and analysis

The advent of two-photon calcium imaging in vivo has presented a new arena to detect neuronal action potentials and identify neuron types based on their fluorescence signatures. However, despite the growing popularity, reconstructing spike patterns from the fluorescence traces still remains a major challenge. Also, not much is usually said about how the calcium waveforms corresponding to a spike (calcium kernel) should be estimated. In this thesis, we present a novel approach for calcium kernel estimation from slopes of a fluorescence trace by combining the Savitzky-Golay filter with an iterative algorithm for fitting a nonlinear model (Levenberg-Marquardt). We also present a new method for spike detection, which employs deconvolution and greedy optimization. First we test these methods on synthesized calcium signals, and then we apply them to experimental traces from wild-type and transgenic mice expressing human α- synuclein (model of Parkinson’s disease). We show longer calcium response in the somatosensory cortex neurons of the transgenic mice, read-out both spontaneous and evoked activities as well as follow the hierarchy in fluorescence transient elevation arrivals when mice whiskers were stimulated electrically