Modelling the neural coding of natural sounds in the auditory cortex

Thesis

When neurons fire in the auditory cortex what do they represent? What is the form of transformation that sound stimuli go through in order to produce activity in the auditory cortex? Studying how neural activity changes in response to a stimulus parameter, e.g. the level or frequency of sound, has been a way to address such questions in the past. In more recent times, an alternative approach has been to measure the responses of cortical neurons to natural stimuli. Here, we investigate the responses of neurons in ferret auditory cortex, recorded using extracellular electrode probes, to a diverse collection of natural sounds using models that aim to capture many aspects of the information transformation process from the ear to the cortex. We use the method of encoding - whereby we build functional transformations that predict neural responses from the sound stimulus, and also of decoding - whereby we build functional transformations that estimate the stimulus from neural responses. Our primary goal has been to focus our modelling effort to produce mechanistic and interpretive insights into auditory processing in the brain. Firstly, we examine to what extent biological details matter in making a model of the auditory periphery that provides input to the cortex. Secondly, we study possible biological mechanisms underlying the temporal receptive fields of auditory cortical neurons. Thirdly, we ask whether neural populations in the auditory cortex represent the past or future of ongoing sound stimuli. We found that simple models are well suited for models of auditory encoding in the brain, and that stimulus history in linear receptive fields of auditory neurons can be explained by a dynamic network model, and also that a population of neurons might contain non-linearities that enable them to predict the future. These results will improve our understanding of some of the fundamental questions relating to neural coding in the auditory system.

Authors: Rahman, M

• DOI: 10.5287/ora-vgqknnkpx
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: decoding stimuli from neural responses; models of the auditory periphery; encoding models of neural responses; neural network model of neural response; Marr’s levels of analysis; predicting neural responses; predicting the future inputs; auditory cortex; predicting responses to natural sounds; predictive information in the brain
• Subjects: Neurosciences