Combining ultra-high-field fMRI adaptation and computational modelling to estimate neuronal frequency selectivity in human auditory cortex

Abstract

Measuring neuronal frequency selectivity in human auditory cortex may be crucial for understanding common auditory deficits such as speech-in-noise difficulty. Non-invasive methods measure aggregate responses of large populations of neurons and therefore overestimate single-neuron tuning selectivity. Here we explore whether cortical frequency selectivity can be estimated through fMRI adaptation. Using ultra-high-field (7T) BOLD-fMRI and individualized functional parcellation of auditory cortex, we measured the suppression (or adaptation) of primary and non-primary cortical responses to a high-frequency (3.8 kHz) probe sound as a function of the frequency of a preceding adaptor sound (ranging from 0.5 to 3.8 kHz). The degree of frequency tuning of the adaptation effect strongly depended on the temporal structure of the adaptor. Suppression by a single 200-ms adaptor showed little or no tuning, despite clear frequency tuning of the responses to the different adaptors. In contrast, suppression by multiple (four) 50-ms adaptor bursts was clearly tuned, with greater frequency selectivity than the adaptor response tuning, suggesting that fMRI adaption to multiple adaptors may reflect the frequency tuning of the underlying neuronal response. Importantly, adaptation tuning differed between primary and non-primary regions, suggesting a local suppression effect, rather than inheritance from upstream subcortical structures. Using a computational model of fMRI adaptation in a tonotopically-organized neuronal array, we identify key factors determining the relationship between observed fMRI adaptation tuning and the frequency selectivity of the underlying neuronal response. Using this model, we derive a plausible range for the frequency selectivity of individual neurons in each region of auditory cortex.