Sparse-Prony for spike detection in two-photon calcium imaging

Abstract

The spatiotemporal activity of neural networks can be observed in the brain using two-photon calcium imaging (TCI). It is important to precisely identify spikes from TCI that belong to individual neurons before this activity can be examined. The finite rate of innovation (FRI) framework is utilized successfully in the past to estimate spike trains from noisy TCI data. Conventionally, matrix-pencil and Prony's methods are proposed to estimate spikes. These techniques can breakdown at high noise environment due to subspace swap. In this work, we propose a polynomial root-free sparse-Prony approach for detecting spikes. The method produces perfect estimation of spikes in a noise-free environment. Simulations are run to compare sparse-Prony method's performance with that of matrix pencil and Prony's techniques for noisy case. The results demonstrate that sparse-Prony performs better in the breakdown region. © 2024 IEEE.