Next-generation head-mounted microscopes for large-scale neural recordings

Miniature head-mounted microscopes have enabled in vivo optical measurements of neural activity in freely moving animals, allowing neuroscientists to study the neural basis of behaviors that are difficult to recapitulate in head-fixed preparations like social interaction and navigation. Existing head-mounted microscopes for small animals are limited to small fields of view (< 1mm2), which limit the number of neurons that can be recorded simultaneously to at most several hundred. However, understanding how behavior is controlled by large neural populations across brain regions requires simultaneous recording in thousands of neurons, which until now has only been possible in small animals using head-fixed preparations.
We address these limitations with a novel optical pathway that allows for microscopes with an order of magnitude increase in field-of-view (FOV) over previous designs while also weighing less than any previously published technology, enabling recordings in thousands of neurons while preserving naturalistic behaviors. Most existing epifluorescence microscopes couple excitation light using a dichroic filter placed between the objective lens and tube lens, a design that adds weight and severely limits the types of optics that can be used to focus light onto the image sensor. In contrast, our new optical path uses a light-guide and coupling prism to introduce excitation light, making the optical assembly significantly more compact while allowing the use of cutting-edge lens assemblies designed for smartphone cameras. Using this design, we have recorded calcium activity, for the first time, in thousands of cortical neurons simultaneously in freely moving mice across a 4 mm diameter region of cortex. Using this activity, we can decode the position of the mouse in a circular maze and quantify the relative amount of position information contained in different cortical regions. This order-of-magnitude increase in the size of recordable populations opens the door to testing computational models of cognition across large neural populations and between multiple brain regions.