Rapid deep widefield neuron finder driven by virtual calcium imaging data

Abstract

Widefield microscope provides optical access to multi-millimeter fields of view and thousands of neurons in mammalian brains at video rate. However, calcium imaging at cellular resolution has been mostly contaminated by tissue scattering and background signals, making neuronal activities extraction challenging and time-consuming. Here we present a deep widefield neuron finder (DeepWonder), which is fueled by simulated calcium recordings but effectively works on experimental data with an order of magnitude faster speed and improved inference accuracy than traditional approaches. The efficient DeepWonder accomplished fifty-fold signal-to-background ratio enhancement in processing terabytes-scale cortex-wide recording, with over 14000 neurons extracted in 17 hours in workstation-grade computing resources compared to nearly week-long processing time with previous methods. DeepWonder circumvented the numerous computational resources and could serve as a guideline to massive data processing in widefield neuronal imaging.