Stephen V. David, Ph.D. - US grants

[unreadable] DESCRIPTION (provided by applicant): Mammals have evolved selective attention to extract useful sensory information from a complex world. Most previous neurophysiology research has modeled sensory representation and selective attention as independent, hierarchical processes: First, sensory systems passively filter input stimuli, and then attention systems select relevant information from the filtered output. The goal of this study is to test the alternative hypothesis that these two processes are tightly coupled at intermediate stages of sensory processing. If the filtering properties of sensory neurons are modulated by attention, then the number of possible computational strategies employed by attention will expand dramatically. This work will focus on characterizing the functional properties of neurons in primary auditory cortex under varying attention conditions. Findings using simple stimuli will be tested for their generality in natural behavior tasks. Patients with partial hearing loss and cochlear implants often complain of difficulty in auditory scene segmentation and hearing in noisy environments. This study will provide insight into the cortical processes required for performing these tasks successfully. [unreadable] [unreadable] [unreadable]

DESCRIPTION (provided by applicant): Humans and other mammals are able to recognize and discriminate sounds even when masked by substantial irrelevant noise. Although this process is often effortless for animals, common sources of environmental noise severely confound automatic speech processors and distort the output of hearing aids and prosthetics. Understanding how complex noisy sounds are processed in central brain areas can provide critical insights into how to address these ongoing challenges. The goal of this project is to study cortical responses to naturalistic noisy auditory stimuli in order to understand neurophysiological mechanisms for the robust perception of noisy signals. Initial experiments will study automatic enhancement of natural signals in neural representations during passive listening. These experiments will focus specifically on environmental noise that challenges engineered auditory processing systems. Further experiments will study how neuronal mechanisms facilitate this process when selective attention is directed to auditory and multisensory audio-visual features. Computational analysis will be used to understand the algorithms employed by single neurons and neural populations to enhance the representation of important signals. In addition to revealing basic neural mechanisms of sensory processing, these experiments will provide insight into how sound processors can be improved for hearing-impaired patients. The benefits of hearing aids and prosthetics are often limited by common environmental noise, which can severely distort their outputs. In contrast, normal-hearing humans and other mammals are exquisitely adept at recognizing complex sounds, even in very noisy conditions. We propose to study how the brain processes noisy sounds in order to understand the neural mechanisms underlying this remarkable ability and to learn how sound processors might be improved for hearing-impaired patients.

Humans and other mammals are able to recognize and discriminate sounds even when masked by substantial irrelevant noise. Although this process is often effortless for animals, common sources of environmental noise severely confound automatic speech processors and distort the output of hearing aids and prosthetics. Understanding how complex noisy sounds are processed in central brain areas can provide critical insights into how to address these ongoing challenges. The goal of this project is to study cortical responses to naturalistic noisy auditory stimuli in order to understand neurophysiological mechanisms for the robust perception of noisy signals. Initial experiments will study automatic enhancement of natural signals in neural representations during passive listening. These experiments will focus specifically on environmental noise that challenges engineered auditory processing systems. Further experiments will study how neuronal mechanisms facilitate this process when selective attention is directed to auditory and multisensory audio-visual features. Computational analysis will be used to understand the algorithms employed by single neurons and neural populations to enhance the representation of important signals. In addition to revealing basic neural mechanisms of sensory processing, these experiments will provide insight into how sound processors can be improved for hearing-impaired patients.

Humans and other mammals are able to recognize and discriminate sounds even when masked by substantial irrelevant noise. Although this process is often effortless for animals, common sources of environmental noise severely confound automatic speech processors and distort the output of hearing aids and prosthetics. Understanding how complex noisy sounds are processed in central brain areas can provide critical insights into how to address these ongoing challenges. The goal of this project is to study cortical responses to naturalistic noisy auditory stimuli in order to understand neurophysiological mechanisms for the robust perception of noisy signals. Initial experiments will study automatic enhancement of natural signals in neural representations during passive listening. These experiments will focus specifically on environmental noise that challenges engineered auditory processing systems. Further experiments will study how neuronal mechanisms facilitate this process when selective attention is directed to auditory and multisensory audio-visual features. Computational analysis will be used to understand the algorithms employed by single neurons and neural populations to enhance the representation of important signals. In addition to revealing basic neural mechanisms of sensory processing, these experiments will provide insight into how sound processors can be improved for hearing-impaired patients.

Mentorship provides an important influence on how scientific researchers develop professionally. Most researchers spend several years training under just one or two graduate and/or postdoctoral mentors, suggesting that this small number of relationships can have large impact on subsequent careers. Mentorship can have both direct intellectual benefits to the trainee through the learning of new skills and concepts and indirect social benefits through engagement with the social network of the mentor. Networks of mentors and trainees can be represented by a directional graph resembling a traditional family tree. This project develops a large database of academic mentorship relationships and tools for analyzing the impact of mentorship on scientific careers. Development of this database addresses the mission of the Science of Science & Innovation Policy program. The data will be made open-access for general use by the public, providing a new resource for studying the dynamics of academic research fields. More generally, the Academic Family Tree represents a successful crowdsourcing project and provides an example for other efforts to engage the public in scientific research projects.

This project uses semantic and graph theoretic analyses to characterize the impact of mentorship networks on professional success and on the evolution of research fields. This project uses crowdsourcing to expand the Academic Family Tree, a public, web-based database of formal mentoring relationships (doctoral and postdoctoral) across all fields of academic research. Researchers in the mentorship database are linked to publication metadata in multiple publication databases and to funding data in the Federal RePORTER database. Semantic and graph theoretic analyses are used to determine what features of the mentorship network influence the subsequent careers of trainees. Focusing on trainees with at least two mentors, these analyses evaluate direct intellectual impact of mentor expertise, measured by latent semantic analysis of publications, and indirect social impact, measured by the similarity of the lineage of the two mentors in the larger academic genealogy. Two measures of trainees' professional success are considered: the number of trainees they themselves mentor and their ability to obtain research funding.

Project Summary Throughout life, humans and other animals learn statistical regularities in the acoustic environment and adapt their hearing to emphasize the elements of sound that are important for behavioral decisions. Using these abilities, normal-hearing humans are able to perceive important sounds in crowded noisy environments and understand the speech of individuals the first time they meet. However, patients with peripheral hearing loss or central processing disorders often have problems hearing in these challenging settings, even when sound is amplified above perceptual threshold. This study seeks to characterize how two major areas in the brain's auditory network, auditory cortex and midbrain inferior colliculus, establish an interface between incoming auditory signals and the internal brain states that select information appropriate to the current behavioral context. Single-unit neural activity will be recorded from both of these brain areas in awake ferrets during the presentation of complex naturalistic sounds that mimic the acoustic environment encountered in the real world. Internal brain state will be controlled by selective attention to specific sound features in these complex stimuli. Changes in stimulus-evoked neural activity as attention shifts among sound features will be measured to identify interactions between internal state and incoming sensory signals in these different areas. Previous work has identified a large corticofugal projection from auditory cortex to inferior colliculus that could produce task-dependent changes in selectivity in inferior colliculus. This study will test the role of these corticofugal projections by optogenetic inactivation of auditory cortex during recordings from inferior colliculus. Selective inactivation of specific pathways will characterize how the network of brain areas works together to produce effective auditory behaviors. Computational modeling tools will be used to determine, from an algorithmic perspective, how neurons encode information about the natural stimuli and how this encoding changes as attention is shifted between features. Data collected during behavior will be used to develop models that combine bottom-up sensory processing and top-down behavioral control. This computational approach builds on classic characterizations of neural stimulus-response relationships using spectro-temporal receptive field models. New models will be developed that incorporate behavioral state variables and nonlinear biological circuit elements into established model frameworks. Together, these studies will provide new insight into the computational strategies used by the behaving brain to process complex sounds in real-world contexts.

Project Summary Throughout life, humans and other animals learn statistical regularities in the acoustic environment and adapt their hearing to emphasize the elements of sound that are important for behavioral decisions. Using these abilities, normal-hearing humans are able to perceive important sounds in crowded noisy environments and understand the speech of individuals the first time they meet. However, patients with peripheral hearing loss or central processing disorders often have problems hearing in these challenging settings, even when sound is amplified above perceptual threshold. This study seeks to characterize how two major areas in the brain's auditory network, auditory cortex and midbrain inferior colliculus, establish an interface between incoming auditory signals and the internal brain states that select information appropriate to the current behavioral context. Single-unit neural activity will be recorded from both of these brain areas in awake ferrets during the presentation of complex naturalistic sounds that mimic the acoustic environment encountered in the real world. Internal brain state will be controlled by selective attention to specific sound features in these complex stimuli. Changes in stimulus-evoked neural activity as attention shifts among sound features will be measured to identify interactions between internal state and incoming sensory signals in these different areas. Previous work has identified a large corticofugal projection from auditory cortex to inferior colliculus that could produce task-dependent changes in selectivity in inferior colliculus. This study will test the role of these corticofugal projections by optogenetic inactivation of auditory cortex during recordings from inferior colliculus. Selective inactivation of specific pathways will characterize how the network of brain areas works together to produce effective auditory behaviors. Computational modeling tools will be used to determine, from an algorithmic perspective, how neurons encode information about the natural stimuli and how this encoding changes as attention is shifted between features. Data collected during behavior will be used to develop models that combine bottom-up sensory processing and top-down behavioral control. This computational approach builds on classic characterizations of neural stimulus-response relationships using spectro-temporal receptive field models. New models will be developed that incorporate behavioral state variables and nonlinear biological circuit elements into established model frameworks. Together, these studies will provide new insight into the computational strategies used by the behaving brain to process complex sounds in real-world contexts.

Mentorship plays a critical role in the careers of scientific researchers. Most researchers spend several years training under just one or two graduate and/or postdoctoral mentors, suggesting that these few relationships can have a large impact on scientific careers. Mentorship can provide both direct intellectual benefits to the trainee, through the learning of new skills and concepts, and indirect social benefits, through engagement with the social network of the mentor. Given the importance of mentorship in scientific career development, this aspect of training may play an important role in determining access of underrepresented groups to scientific careers. The purpose of this study is to characterize the impact of demographic variables--such as gender, ethnicity and socio-economic background--on the outcome of scientific training.

Networks of mentors and trainees can be represented by a directional graph resembling a traditional family tree. This project develops a large crowdsourced database of academic mentorship relationships, and links that data to databases that measure scientific productivity (publications and grants) and demographic variables. Graph theoretic and semantic tools will be used to determine if and how demographic variables, associated with both of the mentor and trainee, impact scientific productivity. A preliminary analysis of gender replicates previous reports of bias toward representation male researchers, especially at more senior career stages. Accurately modeling effects of demographic variables requires accounting for other variables that impact scientific productivity metrics, namely differences between fields and long-term temporal trends. This project will use semantic analysis of publication data to develop the concept of the "intellectual neighborhood" of mentors. and incorporate this into the modeling of career outcomes. Data will be made open-access for general use by the public, providing a new resource for studying the dynamics of research fields.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Project Summary Throughout life, humans and other animals learn statistical regularities in the natural acoustic environment. They adapt their hearing to emphasize the features of sound that are important for making behavioral decisions. Normal-hearing humans are able to perceive important sounds in crowded noisy scenes and to understand the speech of individuals the first time they meet. However, patients with peripheral hearing loss or central processing disorders often have problems hearing in these challenging settings, even when sound is amplified above perceptual threshold. A better understanding of the function of the healthy and impaired auditory system will support new treatments for these deficits. This project will develop computational tools to study central auditory processing. A software library will support fitting and evaluating a large number of encoding models to describe the functional relationship between a time-varying natural auditory stimulus and the corresponding neural response. Many such models have been proposed, but relatively few direct comparisons have been made between them. This project will enable their comparison, allowing identification of the key features that contribute positively to their performance. The system will have a modular design so that useful elements from different models can be combined into comprehensive models with even greater explanatory power. The software will be open source and will support data from multiple recording modalities, including small-scale single unit electrophysiological and calcium imaging data, as well as large-scale local field and magnetoencephalography data. In addition to building on existing hypotheses about neural coding, the system will support machine learning methods for fitting artificial neural network models using the same datasets. These large, data-driven models have proven valuable for wide ranging signal processing problems, but their value and relation to existing models for neural sensory processing remain to be explored. Sensory processing involves coherent activity of large neural populations. To study coding at the population level, the system will support models that characterize the simultaneous activity of multiple neural signals and identifies latent subspaces of population activity related to sound encoding. Sensory coding is also influenced by behavioral context, reflecting changes in behavioral demands and the more general environment. The system will incorporate behavioral state variables into models, where encoding properties can be modulated by changes in behavioral context.

Project Summary Throughout life, humans and other animals adapt their hearing to perceive features of sound that are important for successful behavioral decisions. Normal-hearing humans are able to detect and discriminate important sounds in crowded noisy scenes and to understand the speech of individuals the first time they meet. However, patients with peripheral hearing loss or central processing disorders often have problems hearing in these challenging settings. Even when they can perceive sounds accurately, the additional listening effort required negatively impacts other cognitive functions. A better understanding of how the healthy auditory system operates in cognitively challenging contexts will support new treatments for these deficits. This project will study how the auditory system represents sound information as it operates in challenging acoustic environments. There are three specific aims. First, high-density microelectrode arrays will be used to record the simultaneous activity of neural populations in auditory cortex during behaviors that require detecting sounds masked by noise or learning new sound-reward associations. Recording from multiple neurons will enable characterizing how information is encoded by the simultaneous activity of neural populations. These experiments will test the hypothesis that population activity in auditory cortex generates representations that are invariant to irrelevant distracting sounds. Second, optogenetic tools will be used to identify distinct neuronal cell types (excitatory versus inhibitory) in cortex. This study will test the hypothesis that tonic activation of inhibitory neurons can explain changes in population activity during behavior. Third, machine learning tools will be used to model the simultaneously recorded neural activity. These experiments will test the hypothesis that neurons in the same local anatomical circuit in auditory cortex encode information about a relatively small domain in the space of all possible auditory stimuli. Models fit to experimental data will also describe how changes in behavioral state shift the way neurons encode sounds and describe sources of correlated population activity that impact neural discriminability during behavior. Together these experiments will establish new links between neural representation of sound and the cognitive processes that extract important information from sound for successful behavior.