CNS*2020 Online has ended
Welcome to the Sched instance for CNS*2020 Online! Please read the instruction document on detailed information on CNS*2020.
Back To Schedule
Sunday, July 19 • 8:00pm - 9:00pm
P31: Exploring fast and slow neural correlates of auditory perceptual bistability with diffusion-mapped delay coordinates

Log in to save this to your schedule, view media, leave feedback and see who's attending!

Feedback form is now closed.
Link for poster: https://meet.google.com/fra-scrk-ymb

Pake Melland
, Rodica Curtu

****Perceptual bistability is a phenomenon in which an observer is capable of perceiving identical stimuli with two or more interpretations. The auditory streaming task has been shown to produce spontaneous switching between two perceptual states [1]. In this task a listener is presented a stream of tones, called triplets, with the pattern ABA\-- where A and B are tones with different frequencies and '--' is a brief period of silence. The listener can alternate between two perceptual states: 1-stream in which the stimulus is integrated into a single stream, and 2-stream in which the stimulus is perceived as two segregated streams. In order to study the localization and dynamic properties of neural correlates of auditory streaming we collected electrocorticography (ECoG) data from neurosurgical patients while they listened to sequences of repeated triplets and self-reported switching between the two perceptual states.

****It is necessary to find meaningful ways to analyze ECoG recordings, which are noisy and inherently high dimensional. Diffusion Maps is a non-linear dimensionality reduction technique which embeds high dimensional data into low dimensional Euclidean space [2]. The Diffusion Map method leverages the creation of a Markov matrix from a similarity measure on the original data. Under reasonable assumptions, the eigenvalues of the Markov matrix are positive and bounded above by 1. The d largest eigenvalues along with their respective eigenvectors provide coordinates for an embedding of the data into d-dimensional Euclidean space. In [3] Diffusion Maps were used for a group level analysis of neural signatures during auditory streaming based on subject reported perception. We extend this approach by taking into account the time ordered property of the ECoG signals. For data that has a natural time ordering, it is beneficial to structure the data to emphasize its temporal dynamics; in [4] the authors develop the Diffusion- Mapped Delayed Coordinates (DMDC) algorithm. In this algorithm, time-delayed data is first created from general time series data; this initial step projects the data onto its most stable sub-system. The stable sub-system may remain in a high dimensional space, so they next apply Diffusion Maps to the time-delayed data which projects the (potentially high dimensional) stable sub-system onto a low dimensional representation adapted to the dynamics of the system.

We apply the DMDC algorithm to ECoG recordings from Heschl’s Gyrus in order to explore and reconstruct the underlying dynamics present during the auditory steaming task. We find that the eigenvalues obtained through the DMDC algorithm provide a way to uncover multiple time scales present in the underlying system. The corresponding eigenvectors form a Fourier-like basis that is adapted both to the fast properties of ECoG signal encoding the physical properties of the stimulus as well as a slow mechanism that corresponds to perceptual switching reported by subjects.

Acknowledgments: National Science Foundation, CRCNS grant 1515678, and The Human Brain Research Lab, University of Iowa, Iowa (Matthew A. Howard & Kirill Nourski).


[1] van Noorden et al. Temporal coherence in the perception of tone sequences, volume 3. Institute for Perceptual Research Eindhoven, the Netherlands, 1975.

[2] Coifman & Lafon. Applied and Comp Harmonic Anal, 21(1):5–30, 2006.

[3] Curtu et al J Neurosci, 2019.

[4] Berry et al SIADS 12(2):618–649, 2013.


Pake Melland

PhD Student, University of Iowa
I am a PhD student in the Applied Mathematical & Computational Sciences at The University of Iowa.  I am interested in data driven methods for the analysis and modeling of dynamical processes.

Sunday July 19, 2020 8:00pm - 9:00pm CEST
Slot 07