Towards the detection of tonotopic reorganisation in the human auditory cortex

Peripheral hearing loss not only causes a reduction in hearing sensitivity (increase in hearing threshold) but also leads to problems in suprathreshold hearing – such as difficulty understanding speech in noise – and is often associated with tinnitus. Animal research has suggested that such impairments may be related to changes in central auditory processing arising as a consequence of a reduction in peripheral auditory output (Norena and Farley 2013). The auditory cortex is tonotopically organised, with neighbouring parts of the cortex responding to neighbouring frequencies. In animals, peripheral hearing loss has been shown to cause large-scale tonotopic reorganisation in auditory cortex, whereby neurons responsive to frequencies affected by the hearing loss (typically higher frequencies), shift their tuning towards less affected (lower) frequencies. Tonotopic reorganisation in auditory cortex has been investigated in several studies using functional magnetic resonance imaging (fMRI) to non-invasively assess cortical frequency tuning properties in humans. However, it is unclear whether these measurements are sufficiently sensitive to detect hearing loss induced reorganisation. This thesis presents three complementary studies concerned with developing fMRI methods for effective measurement of tonotopic properties of the human auditory cortex. Voxel-wise measures of cortical frequency preference and selectivity, and summary measures of organisation are evaluated in normal-hearing subjects. The impact of hearing loss on these measurements is assessed. To estimate cortical frequency tuning properties, blood oxygen level-dependent (BOLD) fMRI responses to sounds with a wide range of frequencies were recorded from normal-hearing participants in three studies to derive the preferred frequency and selectivity of discrete cortical areas. The three studies presented serve to: 1) optimise the fMRI tonotopic mapping procedure, 2) measure tonotopic cortical magnification, and 3) develop methods for measuring tonotopic reorganisation in subjects with hearing loss. The first study investigated sources of error from experimental factors and suggests how procedures can be optimised for tonotopic mapping fMRI experiments to maximise the validity of tonotopic estimates. The impact of acoustic scanner noise on estimates of tonotopic properties was assessed by comparing tonotopic property estimates derived from data collected using either a sparse acquisition protocol, in which acoustic stimuli are presented in silence but fewer data points are acquired, or a continuous acquisition protocol, where stimulus are presented during data acquisition with typical acoustic scanner noise and more data points are acquired. Frequency preference and selectivity were also estimated using different statistical methods of fitting models to BOLD activity, measured in response to narrowband noise (NBN) stimuli from 8 normal-hearing subjects at 7T; population receptive field modelling (pRF) or calculating the centroid and spread of frequency response profiles estimated using general linear modelling (GLM). Acoustic scanner noise was found to systematically bias preferred frequency estimates, presumably because it masked some frequencies of the acoustic stimuli. The distribution and range of preferred frequencies derived from the GLM analysis were more biased than those derived from the pRF analysis but the spatial location of high- and low- frequency regions, relative to the frequency range afforded by the analysis, appeared to be robust to these biases. The results suggest that using a sparse acquisition protocol and pRF analysis achieves preferred frequency estimates that best reflect the true neuronal frequency tuning properties. It has been shown in other sensory systems, that the cortical surface allocated to a given region in sensory space is proportional to its behavioural importance. For example, visual studies have demonstrated that proportionally more cortical area dedicated to the centre of the visual field than to the periphery (Wandell, Dumoulin, and Brewer 2007). This cortical magnification function has been well characterised in visual cortex using fMRI (Dumoulin and Wandell 2008). Tonotopic magnification in the human auditory cortex (the cortical distance spanned by a given frequency) is still unknown but we expect it to follow behavioural relevance as in the visual system. The second study of this thesis aimed to measure and describe tonotopic cortical magnification in normal-hearing subjects. The auditory cortical magnification function was calculated in 8 normal-hearing subjects from voxel-wise estimates of preferred frequency derived using a pRF analysis from fMRI data acquired using a sparse acquisition protocol at 7T. Cortical magnification was estimated for two mirror-symmetric tonotopic gradients located on Heschl's gyrus in each hemisphere by characterising the relationship between cortical distance and preferred frequency with the best fitting of four functions: linear, logarithmic, cochlear frequency tuning and behavioural frequency discrimination. Preferred frequency transformed by a behavioural frequency discrimination scale provided the best fit, suggesting that the cortical tonotopic magnification function directly reflects frequency discrimination. Estimates of the cortical magnification function provide a global measure and characterisation of cortical organisation of the auditory cortex which may be able to detect and summarise tonotopic reorganization in hearing loss subjects. Understanding the retinotopic organisation and cortical magnification of the human visual cortex was critical to investigate retinotopic reorganisation following localised periphery deprivation (Wandell and Smirnakis 2009; Baseler et al. 2011). Likewise, for auditory research, understanding tonotopic magnification will benefit studies of tonotopic reorganisation and its possibility as a generation mechanism for chronic tinnitus. The third study of this thesis develops a better control for future studies of tonotopic reorganisation in subjects with hearing loss, and its potential as a mechanism for supra-threshold hearing deficits and chronic tinnitus, which is not addressed directly in the current literature. In participants with hearing loss, BOLD responses to frequencies affected by the hearing loss are expected to be reduced, potentially skewing the shape of the measured frequency response functions and biasing cortical preferred frequency estimates. Two methodological procedures were proposed to address these issues. The first is to simulate hearing loss in normal-hearing controls in order to replicate the biases expected for hearing-impaired subjects. The second, tested here in normal-hearing controls with simulated hearing loss, is a novel data analysis method that accounts for variations in sensation level due to hearing loss to alleviate the biases of hearing loss on estimates of preferred frequency. Tonotopic properties of auditory cortex were estimated in seven normal-hearing subjects using a sparse acquisition sequence at 3T and a pRF analysis, either in the presence or absence of a continuous threshold-elevating noise simulating a steeply sloping high-frequency hearing loss. Steeply sloping high-frequency hearing loss has been suggested to lead to tonotopic reorganisation (Sereda, Hall, et al. 2011), providing a realistic case for assessment of the proposed methods. In addition, a modified population receptive field modelling method was implemented that attempted to minimize the effect of hearing loss on tonotopic property estimates by taking into account the hearing thresholds in its fitting procedure. The results show that using the unmodified pRF method for simulated hearing loss data results in expected biases in preferred frequency and tuning width estimates, and that the modified method reduces these biases and results in tonotopic maps closer to ones estimated from normal-hearing data. These results represent a key step towards studying the effects of hearing loss on the tonotopic organisation in human auditory cortex using BOLD fMRI, and understanding their potential perceptual consequences. The work presented in this thesis provides actionable methodologies that could potentially improve the study of hearing-loss-induced tonotopic reorganisation in human auditory cortex