Final Report Summary - NONSTATENCODING (Intensity and timing encoding of naturalistic sounds in auditory brainstem neurons of cats and owls)
The auditory system has evolved efficient and robust processing of natural sounds in adverse conditions. For instance, humans can detect, localize, and identify speech embedded in background noise. For mammals and birds in general detecting and localizing preys or predators is critical for survival. Many studies demonstrated the high degree of precision that the auditory system achieves in encoding sound properties. For instance, brainstem neurons are sensitive to certain stimulus properties, such as the interaural time difference between both ears, or to the temporal structure of the sound envelope. In order to assess this neural sensitivity, such experiments involves the variation of one stimulus parameter, while keeping the others fixed. Our acoustical inputs are nevertheless multi-dimensional, e.g. a sound with a certain frequency can have different intensities, with the consequence that neural sensitivity to a certain property does not equate with the information contained about this property. In this project, we studied whether the sound encoding capacity of brainstem neurons is robust against changes in stimulus parameters, such as overall intensity and how this encoding depends on stimulus history. Recording and analysing realistic stimuli is crucial for the long-term goal of understanding auditory processing in real environments, as they introduce additional constraints for efficient sound encoding. We also used a computational approach to test different neural mechanisms, both at the network or cellular levels, that could implement the neural processing we characterized.
At the cellular level we investigated the functional aspects of spike threshold dynamics. Neurons encode information through spikes, which are triggered when their membrane potential crosses a threshold. In vivo, the spiking threshold displays large variability. The first goal of this research was to prove that this variability could be explained by threshold adaptation (as opposed to noise). This adaptation is therefore a part of information processing of the cell, leading to a shorter time constant and a higher signal to noise ratio. We then showed that the temporal encoding in the cochlear nucleus of cat was almost invariant with respect to input level and demonstrated that threshold adaptation could possibly be the underlying mechanism. Finally, we showed that the effect of sound intensity on the band-pass filtering performed on the sound envelope can be explained by threshold adaptation. These results are of fundamental importance as they demonstrate that a spike generation non-linearity can influence neural processing in a network, without invoking additional ionic channels, network or synaptic mechanisms. Those results are also important for engineering applications as they show how simple spiking neuron models, that can be efficiently implemented, embedded in a network, can implement functions relevant for sound processing.
We then collected neural responses in the mammalian cochlear nucleus to study the effect of acoustic context on neural encoding. In particular we sought neural correlates of a psycho-physical effect called auditory enhancement; a tone complex presented in isolation is perceived as a single auditory object. If it is preceded by a sound with a notch in its spectrum, the tone at the notch frequency will pop out of the complex. In the great majority of the neurons recorded in the dorsal cochlear nucleus, neural correlates, such as an increase in firing rate, were measured.
Finally, we undertook a comparative study of the early auditory processing in birds and mammals. The avian basilar papilla shares a common origin with (i.e. is homologous to) the mammalian cochlea but has evolved salient specializations independently from its mammalian counterpart to achieve equivalent functions. Surprisingly, clear morphological differences can still result in auditory nerve responses with similar characteristics such as group delays and frequency tuning Therefore a better understanding of the avian cochlea can also help gain insight into the mammalian cochlea, by comparing similarities and differences in morphology and physiology. For the first time in any avian species, we studied the impulse responses of auditory-nerve fibers of barn owls, derived from reverse correlation of broadband noise responses. Using linear filters, we studied the relationship between properties of the transfer functions like sharpness of frequency tuning and response delays, which can be informative about the underlying cochlear mechanisms. We showed qualitative agreement with mammalian data to a large extent, but also some intriguing quantitative differences, e.g. a much smaller frequency glide slope and a bimodal impulse response for the barn owl.
The outputs of this project have important consequences, both fundamental and practical. They have an impact on the understanding of the auditory processing of realistic stimuli. Besides bringing new insights in such fundamental issues, the results could help designing technological and medical applications. Indeed, engineers creating innovative concepts in signal processing and in biomedical devices need to interact with theoretical and experimental neuroscientists. The research output of this project can be directly transferable to companies and research centers involved in auditory prosthetic devices design, in particular, in the translation of acoustic information to neuronal impulses.
At the cellular level we investigated the functional aspects of spike threshold dynamics. Neurons encode information through spikes, which are triggered when their membrane potential crosses a threshold. In vivo, the spiking threshold displays large variability. The first goal of this research was to prove that this variability could be explained by threshold adaptation (as opposed to noise). This adaptation is therefore a part of information processing of the cell, leading to a shorter time constant and a higher signal to noise ratio. We then showed that the temporal encoding in the cochlear nucleus of cat was almost invariant with respect to input level and demonstrated that threshold adaptation could possibly be the underlying mechanism. Finally, we showed that the effect of sound intensity on the band-pass filtering performed on the sound envelope can be explained by threshold adaptation. These results are of fundamental importance as they demonstrate that a spike generation non-linearity can influence neural processing in a network, without invoking additional ionic channels, network or synaptic mechanisms. Those results are also important for engineering applications as they show how simple spiking neuron models, that can be efficiently implemented, embedded in a network, can implement functions relevant for sound processing.
We then collected neural responses in the mammalian cochlear nucleus to study the effect of acoustic context on neural encoding. In particular we sought neural correlates of a psycho-physical effect called auditory enhancement; a tone complex presented in isolation is perceived as a single auditory object. If it is preceded by a sound with a notch in its spectrum, the tone at the notch frequency will pop out of the complex. In the great majority of the neurons recorded in the dorsal cochlear nucleus, neural correlates, such as an increase in firing rate, were measured.
Finally, we undertook a comparative study of the early auditory processing in birds and mammals. The avian basilar papilla shares a common origin with (i.e. is homologous to) the mammalian cochlea but has evolved salient specializations independently from its mammalian counterpart to achieve equivalent functions. Surprisingly, clear morphological differences can still result in auditory nerve responses with similar characteristics such as group delays and frequency tuning Therefore a better understanding of the avian cochlea can also help gain insight into the mammalian cochlea, by comparing similarities and differences in morphology and physiology. For the first time in any avian species, we studied the impulse responses of auditory-nerve fibers of barn owls, derived from reverse correlation of broadband noise responses. Using linear filters, we studied the relationship between properties of the transfer functions like sharpness of frequency tuning and response delays, which can be informative about the underlying cochlear mechanisms. We showed qualitative agreement with mammalian data to a large extent, but also some intriguing quantitative differences, e.g. a much smaller frequency glide slope and a bimodal impulse response for the barn owl.
The outputs of this project have important consequences, both fundamental and practical. They have an impact on the understanding of the auditory processing of realistic stimuli. Besides bringing new insights in such fundamental issues, the results could help designing technological and medical applications. Indeed, engineers creating innovative concepts in signal processing and in biomedical devices need to interact with theoretical and experimental neuroscientists. The research output of this project can be directly transferable to companies and research centers involved in auditory prosthetic devices design, in particular, in the translation of acoustic information to neuronal impulses.
