UNRAVELING AUDITORY CORTEX ENCODING OF COMPLEX SOUNDS

Despite major advances in cochlear implants, many people with hearing loss still have difficulty perceiving complex sounds, such as music, and understanding speech in noisy environments. Drawing on the success of cortical stimulation, an auditory cortex implant (ACI) could offer a personalized solution for patients with profound deafness. Indeed, the auditory cortex (AC) is a particularly promising target as preliminary results demonstrate that cortical stimulation activates auditory neural circuits and reproduces dynamics close to natural auditory processing.
Yet, designing such an ACI requires a detailed understanding of how complex auditory information is represented at the cortical level. Defining effective stimulation strategies requires elucidating the neural code of the AC and the receptive fields of its neurons, in other words, how the brain encodes rich and dynamic sound experiences. Here, we are contributing to unraveling the neural code of the auditory cortex in processing complex sounds. With this perspective, we are fine-tuning a deep learning model based on contrastive learning, aimed at decoding self-supervised representations of complex sounds from non-invasive electroencephalography (EEG) in human and electrocorticography (ECoG) in rats.
We show that EEG can be used to decode certain complex auditory information, suggesting that the relevant cortical representations are accessible at the macroscopic level. Using EcoG, we are able to capture neural dynamics with sufficient accuracy to subsequently generate rich auditory perceptions through cortical stimulation, while remaining minimally invasive. These decoding capabilities pave the way for new cortical stimulation strategies and the development of biomimetic cortical auditory implants.