Extracting individual neural activity recorded through splayed optical microfibers

Abstract

Previously introduced bundles of hundreds or thousands of microfibers have the potential to extend optical access to deep brain regions, sampling fluorescence activity throughout a three-dimensional volume. Each fiber has a small diameter (8 μm) and follows a path of least resistance, splaying during insertion. By superimposing the fiber sensitivity profile for each fiber, we model the interface properties for a simulated neural population. Our modeling results suggest that for small (<200) bundles of fibers, each fiber will collect fluorescence from a small number of nonoverlapping neurons near the fiber apertures. As the number of fibers increases, the bundle delivers more uniform excitation power to the region, moving to a regime where fibers collect fluorescence from more neurons and there is greater overlap between neighboring fibers. Under these conditions, it becomes feasible to apply source separation to extract individual neural contributions. In addition, we demonstrate a source separation technique particularly suited to the interface. Our modeling helps establish performance expectations for this interface and provides a framework for estimating neural contributions under a range of conditions.