Developing Diffuse Optical Tomography (DOT) for Neuroimaging of Speech Perception in People with Cochlear Implants

Optical neuroimaging has shown a great potential to be used as a surrogate for fMRI, wherever fMRI is impractical, such as in patients with metal implants or in children who cannot be sedated. The momentum to use optical methods in systems neuroscience has increased recently with developing new techniques that provide fMRI comparable image quality. These techniques consist of a relatively dense array of source and detectors and are called high-density diffuse optical tomography (HD-DOT). Imaging arrays may cover the whole head and image the whole cortex. While these optical systems are evolving, techniques for extracting information about the neuronal activities from complex signals that they record have yet to be developed and tested. My research has been: (1) to develop fundamental temporal and spatial statistical tools for HD-DOT data analysis. These tools enable event-related study designs with faster (10x) presentation rates (compared to previous simple block averaging) that are critical for mapping higher order cognitive functions, (2) to demonstrate the first HD-DOT mapping of the cortical network for single words and sentences, and (3) to translate HD-DOT into studies of speech perception in patients with cochlear implants (CIs). Studies of speech processing in CI recipients are interesting because while implantation has been successful in improving their speech perception, there are enormous individual differences in the speech performance outcomes. Neuroscientists who are interested in learning about the neurobiological reasons behind this variance desire a reliable neuroimaging tool that allows non-invasive longitudinal monitoring of brain activity in this population. MRI scanner instrument poses a critical limitation on the studies of patients with cochlear implants due to contraindications of metallic components of the implanted device. In addition, even studies of subjects with poor hearing but no cochlear implants have been challenging with fMRI due to substantial scanner noise that can interfere with the presentation of auditory stimuli and can add additional perceptual and cognitive demands to the experimental task. Such auditory task demands are likely to differentially affect participants with hearing impairment or reduced cognitive capacity. HD-DOT has the potential to fill this gap. Because it combines fMRI\u27s ability to construct cortical maps with EEG\u27s quiet and cap-based system, and is compatible with implanted devices and does not interfere with its electronics. Since speech processing involves different regions of the brain concurrently and consecutively, it results in a distributed and highly complex hemodynamic response. In order to record this response and extract information about the underlying neural activity we needed to develop techniques for processing these complex signals. Based on understanding of physics that underlies the HD-DOT signals, and following a similar approach used in fMRI, I incorporated linear models for un-mixing and extracting brain response patterns. I developed and optimized ways of quantifying statistical significance of results and established a method for investigating the functional connections between brain areas. For validation, brain responses to mixed visual stimuli were recorded. The algorithms successfully un-mixed the signals and provided rigorous quantitative statistics. I then extended the approach to hierarchical language tasks including both words and sentences. Building on this infrastructure, I examined the sensitivity of HD-DOT in detecting effects of subtle changes in the speech content on the brain response and cortical connections. Finally, I recorded cortical activity from several CI patients during different visual and auditory language tasks to show feasibility of neuroimaging in the CI population with HD-DOT