High-accuracy Decoding of Complex Visual Scenes from Neuronal Calcium Responses

Abstract

The brain contains billions of neurons defined by diverse cytoarchitectural, anatomical, genetic, and functional properties. Sensory encoding and decoding are popular research areas in the fields of neuroscience, neuroprosthetics and artificial intelligence but the contribution of neuronal diversity to these processes is not well understood. Deciphering this contribution necessitates development of sophisticated neurotechnologies that can monitor brain physiology and behavior via simultaneous assessment of individual genetically-defined neurons during the presentation of discrete sensory cues and behavioral contexts. Neural networks are a powerful technique for formulating hierarchical representations of data using layers of nonlinear transformations. Here we leverage the availability of an unprecedented collection of neuronal activity data, derived from [~]25,000 individual genetically-defined neurons of the parcellated mouse visual cortex during the presentation of 118 unique and complex naturalistic scenes, to demonstrate that neural networks can be used to decode discrete visual scenes from neuronal calcium responses with high ([~]96%) accuracy. Our findings highlight the novel use of neural networks for sensory decoding using neuronal calcium imaging data and reveal a neuroanatomical map of visual decoding strength traversing brain regions, cortical layers, neuron types, and time. Our findings also demonstrate the utility of feature selection in assigning contributions of neuronal diversity to visual decoding accuracy and the low requirement of network architecture complexity for high accuracy decoding in this experimental context.