Calcium imaging-based brain-computer interface for investigating long-term neuronal code dynamics

Brain-computer interfaces (BCI) have important applications both in medicine and as a research tool. In recent years the field of calcium imaging based BCIs has been accelerating1–4. Calcium imaging has the advantages of recording hundreds of cells simultaneously, and tracking the same neurons over many days, allowing a longitudinal analysis of neural code development and dynamics5. The collected data usually requires post experiment offline processing to extract single cells’ activity. Yet, emerging techniques for real-time processing of imaging data offer the potential for novel BCI experiments, in which closed-loop feedback is triggered at short latency in response to neural population activity3,6. Here, we present a proof-of-concept for a closed-loop BCI that allows longitudinal readout of neuronal activity from tens to hundreds of neurons in freely behaving mice, and enables the activation of external hardware based on real-time detection of specified activity patterns. Our system is modular, comprising several hardware and software components, which allows customization according to experimental needs. The hardware components are a head-mounted miniature fluorescence microscope, an overhead camera for tracking the mouse behavior, and an Arduino-controlled system for feedback within the behavioral arena. The software components include code for online processing of acquired recordings, aligning region of interest (ROI), fluorescence trace extraction and event detection, matching activity patterns, and animal’s tracking. With this system, we trained mice to run back and forth along a linear track, to attain water rewards at the track’s edges, upon activation of the same neuronal population activity pattern in hippocampal CA1. Mice maintained a stable level of reward across 6 recording days, demonstrating the ability to track and activate a BCI with the same set of cells across days. In the future, such a system could facilitate longitudinal interrogation of neural code dynamics.