De Luca, DanielaMoccia, SaraLupori, LeonardoMazziotti, RaffaelePizzorusso, TommasoMicera, Silvestro2023-02-272023-02-272023-02-272022-01-0110.1109/SENSORS52175.2022.9967250https://infoscience.epfl.ch/handle/20.500.14299/195268WOS:000918629700232Optic nerve stimulation holds great potential for visual prostheses. Its effectiveness depends on the stimulation protocol, which can be optimized to achieve cortical activation similar to that evoked in response to visual stimuli. To identify a target cortical activation, it is necessary to characterize the cortical response. We here propose a convolutional neural network (CNN) to do it exploiting widefield calcium brain images, which allow large-scale visualization of cortical activity with high signal-to-noise ratio. A mouse was presented with 10 different visual stimuli, and the activity from its primary visual cortex (V1) was recorded. The CNN was trained to predict the visual stimulus, with an accuracy of 78.46%+/- 3.31% on the test set, showing it is possible to automatically detect what is present in the visual field of the animal.Engineering, Electrical & ElectronicRemote SensingEngineeringdeep learningwide-field imagingvisual cortexvisual prosthesestext::conference output::conference proceedings::conference paper