Spectro-Temporal Processing in a Two-Stream Computational Model of Auditory Cortex

Auditory hemispheric asymmetry for actions and objects

What is the function of auditory hemispheric asymmetry? We propose that the identification of sound sources relies on the asymmetric processing of two complementary and perceptually relevant acoustic invariants: actions and objects. In a large dataset of environmental sounds, we observed that temporal and spectral modulations display only weak covariation. We then synthesized auditory stimuli by simulating various actions (frictions) occurring on different objects (solid surfaces). Behaviorally, discrimination of actions relies on temporal modulations, while discrimination of objects relies on spectral modulations. Functional magnetic resonance imaging data showed that actions and objects are decoded in the left and right hemispheres, respectively, in bilateral superior temporal and left inferior frontal regions. This asymmetry reflects a generic differential processing-through differential neural sensitivity to temporal and spectral modulations present in environmental sounds-that supports the efficient categorization of actions and objects. These results support an ecologically valid framework of the functional role of auditory brain asymmetry.

Spectral-temporal processing of naturalistic sounds in monkeys and humans

Human speech and vocalizations in animals are rich in joint spectrotemporal (S-T) modulations, wherein acoustic changes in both frequency and time are functionally related. In principle, the primate auditory system could process these complex dynamic sounds based on either an inseparable representation of S-T features or, alternatively, a separable representation. The separability hypothesis implies an independent processing of spectral and temporal modulations. We collected comparative data on the S-T hearing sensitivity in humans and macaque monkeys to a wide range of broadband dynamic spectrotemporal ripple stimuli employing a yes-no signal-detection task. Ripples were systematically varied, as a function of density (spectral modulation frequency), velocity (temporal modulation frequency), or modulation depth, to cover a listener's full S-T modulation sensitivity, derived from a total of 87 psychometric ripple detection curves. Audiograms were measured to control for normal hearing. Determined were hearing thresholds, reaction time distributions, and S-T modulation transfer functions (MTFs), both at the ripple detection thresholds and at suprathreshold modulation depths. Our psychophysically derived MTFs are consistent with the hypothesis that both monkeys and humans employ analogous perceptual strategies: S-T acoustic information is primarily processed separable. Singular value decomposition (SVD), however, revealed a small, but consistent, inseparable spectral-temporal interaction. Finally, SVD analysis of the known visual spatiotemporal contrast sensitivity function (CSF) highlights that human vision is space-time inseparable to a much larger extent than is the case for S-T sensitivity in hearing. Thus, the specificity with which the primate brain encodes natural sounds appears to be less strict than is required to adequately deal with natural images.NEW & NOTEWORTHY We provide comparative data on primate audition of naturalistic sounds comprising hearing thresholds, reaction time distributions, and spectral-temporal modulation transfer functions. Our psychophysical experiments demonstrate that auditory information is primarily processed in a spectral-temporal-independent manner by both monkeys and humans. Singular value decomposition of known visual spatiotemporal contrast sensitivity, in comparison to our auditory spectral-temporal sensitivity, revealed a striking contrast in how the brain encodes natural sounds as opposed to natural images, as vision appears to be space-time inseparable.

Resting-state functional connectivity in an auditory network differs between aspiring professional and amateur musicians and correlates with performance

Auditory experience-dependent plasticity is often studied in the domain of musical expertise. Available evidence suggests that years of musical practice are associated with structural and functional changes in auditory cortex and related brain regions. Resting-state functional magnetic resonance imaging (MRI) can be used to investigate neural correlates of musical training and expertise beyond specific task influences. Here, we compared two groups of musicians with varying expertise: 24 aspiring professional musicians preparing for their entrance exam at Universities of Arts versus 17 amateur musicians without any such aspirations but who also performed music on a regular basis. We used an interval recognition task to define task-relevant brain regions and computed functional connectivity and graph-theoretical measures in this network on separately acquired resting-state data. Aspiring professionals performed significantly better on all behavioral indicators including interval recognition and also showed significantly greater network strength and global efficiency than amateur musicians. Critically, both average network strength and global efficiency were correlated with interval recognition task performance assessed in the scanner, and with an additional measure of interval identification ability. These findings demonstrate that task-informed resting-state fMRI can capture connectivity differences that correspond to expertise-related differences in behavior.

Data-driven multiscale model of macaque auditory thalamocortical circuits reproduces in vivo dynamics

We developed a detailed model of macaque auditory thalamocortical circuits, including primary auditory cortex (A1), medial geniculate body (MGB), and thalamic reticular nucleus, utilizing the NEURON simulator and NetPyNE tool. The A1 model simulates a cortical column with over 12,000 neurons and 25 million synapses, incorporating data on cell-type-specific neuron densities, morphology, and connectivity across six cortical layers. It is reciprocally connected to the MGB thalamus, which includes interneurons and core and matrix-layer-specific projections to A1. The model simulates multiscale measures, including physiological firing rates, local field potentials (LFPs), current source densities (CSDs), and electroencephalography (EEG) signals. Laminar CSD patterns, during spontaneous activity and in response to broadband noise stimulus trains, mirror experimental findings. Physiological oscillations emerge spontaneously across frequency bands comparable to those recorded in vivo. We elucidate population-specific contributions to observed oscillation events and relate them to firing and presynaptic input patterns. The model offers a quantitative theoretical framework to integrate and interpret experimental data and predict its underlying cellular and circuit mechanisms.

Visual information processing through the interplay between fine and coarse signal pathways

Object recognition is often viewed as a feedforward, bottom-up process in machine learning, but in real neural systems, object recognition is a complicated process which involves the interplay between two signal pathways. One is the parvocellular pathway (P-pathway), which is slow and extracts fine features of objects; the other is the magnocellular pathway (M-pathway), which is fast and extracts coarse features of objects. It has been suggested that the interplay between the two pathways endows the neural system with the capacity of processing visual information rapidly, adaptively, and robustly. However, the underlying computational mechanism remains largely unknown. In this study, we build a two-pathway model to elucidate the computational properties associated with the interactions between two visual pathways. Specifically, we model two visual pathways using two convolution neural networks: one mimics the P-pathway, referred to as FineNet, which is deep, has small-size kernels, and receives detailed visual inputs; the other mimics the M-pathway, referred to as CoarseNet, which is shallow, has large-size kernels, and receives blurred visual inputs. We show that CoarseNet can learn from FineNet through imitation to improve its performance, FineNet can benefit from the feedback of CoarseNet to improve its robustness to noise; and the two pathways interact with each other to achieve rough-to-fine information processing. Using visual backward masking as an example, we further demonstrate that our model can explain visual cognitive behaviors that involve the interplay between two pathways. We hope that this study gives us insight into understanding the interaction principles between two visual pathways.

Modelling homeostatic plasticity in the auditory cortex results in neural signatures of tinnitus

Tinnitus is a clinical condition where a sound is perceived without an external sound source. Homeostatic plasticity (HSP), serving to increase neural activity as compensation for the reduced input to the auditory pathway after hearing loss, has been proposed as a mechanism underlying tinnitus. In support, animal models of tinnitus show evidence of increased neural activity after hearing loss, including increased spontaneous and sound-driven firing rate, as well as increased neural noise throughout the auditory processing pathway. Bridging these findings to human tinnitus, however, has proven to be challenging. Here we implement hearing loss-induced HSP in a Wilson-Cowan Cortical Model of the auditory cortex to predict how homeostatic principles operating at the microscale translate to the meso- to macroscale accessible through human neuroimaging. We observed HSP-induced response changes in the model that were previously proposed as neural signatures of tinnitus, but that have also been reported as correlates of hearing loss and hyperacusis. As expected, HSP increased spontaneous and sound-driven responsiveness in hearing-loss affected frequency channels of the model. We furthermore observed evidence of increased neural noise and the appearance of spatiotemporal modulations in neural activity, which we discuss in light of recent human neuroimaging findings. Our computational model makes quantitative predictions that require experimental validation, and may thereby serve as the basis of future human studies of hearing loss, tinnitus, and hyperacusis.

Distinct neuronal types contribute to hybrid temporal encoding strategies in primate auditory cortex

Studies of the encoding of sensory stimuli by the brain often consider recorded neurons as a pool of identical units. Here, we report divergence in stimulus-encoding properties between subpopulations of cortical neurons that are classified based on spike timing and waveform features. Neurons in auditory cortex of the awake marmoset (Callithrix jacchus) encode temporal information with either stimulus-synchronized or nonsynchronized responses. When we classified single-unit recordings using either a criteria-based or an unsupervised classification method into regular-spiking, fast-spiking, and bursting units, a subset of intrinsically bursting neurons formed the most highly synchronized group, with strong phase-locking to sinusoidal amplitude modulation (SAM) that extended well above 20 Hz. In contrast with other unit types, these bursting neurons fired primarily on the rising phase of SAM or the onset of unmodulated stimuli, and preferred rapid stimulus onset rates. Such differentiating behavior has been previously reported in bursting neuron models and may reflect specializations for detection of acoustic edges. These units responded to natural stimuli (vocalizations) with brief and precise spiking at particular time points that could be decoded with high temporal stringency. Regular-spiking units better reflected the shape of slow modulations and responded more selectively to vocalizations with overall firing rate increases. Population decoding using time-binned neural activity found that decoding behavior differed substantially between regular-spiking and bursting units. A relatively small pool of bursting units was sufficient to identify the stimulus with high accuracy in a manner that relied on the temporal pattern of responses. These unit type differences may contribute to parallel and complementary neural codes.

Neurons in auditory cortex show highly diverse responses to sounds. This study suggests that neuronal type inferred from baseline firing properties accounts for much of this diversity, with a subpopulation of bursting units being specialized for precise temporal encoding.

Predicting neuronal response properties from hemodynamic responses in the auditory cortex

Recent functional MRI (fMRI) studies have highlighted differences in responses to natural sounds along the rostral-caudal axis of the human superior temporal gyrus. However, due to the indirect nature of the fMRI signal, it has been challenging to relate these fMRI observations to actual neuronal response properties. To bridge this gap, we present a forward model of the fMRI responses to natural sounds combining a neuronal model of the auditory cortex with physiological modeling of the hemodynamic BOLD response. Neuronal responses are modeled with a dynamic recurrent firing rate model, reflecting the tonotopic, hierarchical processing in the auditory cortex along with the spectro-temporal tradeoff in the rostral-caudal axis of its belt areas. To link modeled neuronal response properties with human fMRI data in the auditory belt regions, we generated a space of neuronal models, which differed parametrically in spectral and temporal specificity of neuronal responses. Then, we obtained predictions of fMRI responses through a biophysical model of the hemodynamic BOLD response (P-DCM). Using Bayesian model comparison, our results showed that the hemodynamic BOLD responses of the caudal belt regions in the human auditory cortex were best explained by modeling faster temporal dynamics and broader spectral tuning of neuronal populations, while rostral belt regions were best explained through fine spectral tuning combined with slower temporal dynamics. These results support the hypotheses of complementary neural information processing along the rostral-caudal axis of the human superior temporal gyrus.

A bio-inspired geometric model for sound reconstruction

The reconstruction mechanisms built by the human auditory system during sound reconstruction are still a matter of debate. The purpose of this study is to propose a mathematical model of sound reconstruction based on the functional architecture of the auditory cortex (A1). The model is inspired by the geometrical modelling of vision, which has undergone a great development in the last ten years. There are, however, fundamental dissimilarities, due to the different role played by time and the different group of symmetries. The algorithm transforms the degraded sound in an 'image' in the time-frequency domain via a short-time Fourier transform. Such an image is then lifted to the Heisenberg group and is reconstructed via a Wilson-Cowan integro-differential equation. Preliminary numerical experiments are provided, showing the good reconstruction properties of the algorithm on synthetic sounds concentrated around two frequencies.