Dynamic Features of Auditory Bistable Perception Extracted from Human Electrocorticography Recordings
Author | : Pake Melland |
Publisher | : |
Total Pages | : 0 |
Release | : 2021 |
ISBN-10 | : OCLC:1380764711 |
ISBN-13 | : |
Rating | : 4/5 (11 Downloads) |
Download or read book Dynamic Features of Auditory Bistable Perception Extracted from Human Electrocorticography Recordings written by Pake Melland and published by . This book was released on 2021 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: Since the 1950s, there has been continued advancement in technology for obtaining long-term neural recordings in mammals, including present-day techniques yielding data sets with increasing spatial and temporal resolution. Discovering dynamical patterns encoded in high fidelity neural measurements is a challenging, yet rewarding, task equipped with techniques applicable to data obtained with various methods. Emerging approaches employ data-driven methods to identify low-dimensional structures from large-scale recordings, which encapsulate key characteristics of the measured neural populations. In this thesis, we consider electrocorticography recordings in primary auditory cortex from human patients performing an auditory streaming task designed to elicit bistable perception. Listeners are presented with 5-minute sequences of alternating pure tones and self-report changes in their perception of the stimulus. Our goal is to analyze the neural recordings obtained concurrently during stimulus presentation for correlates with subject-reported perception. To this end, we employ the data-driven methods of (1) time-delay coordinates, (2) diffusion maps, and (3) the extended dynamic mode decomposition to yield a low-dimensional collection of Fourier-like features, which are adapted to intrinsic dynamics encoded in the neural data. We find two sets of features, one which correlates with the stimulus, and one with the reported percept. In particular, we characterize a slow rhythm correlated with switches in perception that manifests and modulates the phase of a subset of the derived features. By leveraging the dynamical techniques, we identify neural characteristics associated with perception, which may remain undetected by traditional static methods from data analysis. The data-driven methods implemented in this thesis can be applied in contexts outside of neuroscience and are part of the growing movement towards equation-free modeling and analysis. This emerging paradigm is highly applicable to physical processes or systems that exhibit rich dynamics but are accompanied by incomplete, or entirely unknown, governing equations.