Efficient position decoding methods based on fluorescence calcium imaging in the mouse hippocampus

Abstract

Large-scale and long-term calcium imaging has widely been used for decoding positions from hippocampal place cells in rodents. Developing efficient neural decoding methods for reconstructing the animal's position in real or virtual environments based on calcium imaging can provide a real-time readout of spatial information in closed-loop neuroscience experiments. Spike deconvolution, a procedure to infer the underlying spike trains from calcium imaging data, presents computational challenges in the processing of calcium imaging data for subsequent decoding analysis and hinders the progress of real-time decoding. Here, we developed an efficient strategy to extract features from fluorescence calcium imaging traces that sidestepped the computationally slow spike deconvolution and further decoded animal's positions from these features. We validated our proposed decoding method in multiple in vivo calcium imaging recordings of the mouse hippocampus and simulated data, based on both supervised and unsupervised decoding analysis. We systematically investigated the decoding performance of our proposed method with respect to the number of neurons and signal-to-noise ratio. Our analysis pipeline is ultrafast and robust and therefore promising for online decoding of animal's positions in closed-loop calcium imaging experiments.