Decoding with Calcium Signals from Layer 2/3 Motor Cortex during A Pressing Movement

Abstract

Brain-machine interfaces (BMIs) have been promising for not only neuroprosthesis research but also brain function investigation. Electrophysiological recording commonly used in traditional BMIs is spatially sparse and lack of information about neuron types and spatial organization. However, optical imaging methods might avoid these limitations by providing dense, spatially organized and annotated with genetic information over a large field of view. Here, we tried to demonstrate the potential of calcium imaging signals obtained through the one-photon microscope in neural decoding. When mice were trained to perform a lever press task to obtain water as rewards, the calcium signals of neurons in their layer 2/3 motor cortex were recorded by microscope. With the calcium signals, we analyzed the neural activity at both single individual neuron and neuronal population level. We found two typical classes of pressing-related neurons and distinct ensemble activity patterns between a pressing movement and baseline. The decoding results further demonstrated that the movement-related information could be more completely specified by population response structure. Our results suggested that neural signals from more types and a larger amount of neurons, are crucial for accurate decoding in BMI applications.