| Start Date: | 3/20/2017 | Start Time: | 1:00 PM |
| End Date: | 3/20/2017 | End Time: | 3:00 PM |
|
Event Description
BIOMED Master's Thesis Defense
Title:
Signal Separation Algorithm for Combinatoric Multi-Electrode Probe Design Enabling Increased Neural Recording Yield
Speaker:
Joanna Wycech, MS Candidate, School of Biomedical Engineering, Science and Health Systems, Drexel University
Advisors:
Simon F. Giszter, PhD, Professor of Neurobiology and Anatomy, College of Medicine, Drexel University
Hualou Liang, PhD, Professor, School of Biomedical Engineering, Science and Health Systems, Drexel University
Abstract:
Neural signal recordings relay important information about the neuronal networks in the Central Nervous System (CNS), which allow the physician or an investigator to gain an understanding about the system’s connections. Current computational approaches usually limit recordings to a single site on a wire. We seek instead to exploit innovative work in the design of combinatoric multi-site electrode probes (Giszter et al.) using multisite wires.
In this project, we utilize independent component analysis with reference (ICA-r) incorporated into a MATLAB algorithm to efficiently separate neural recording yields from each wire in the probe in a combinatoric design. First, we present a 3C2 combinatoric design with and without additive noise, with multiple neurons per recording-site. Additionally, we look at the signal-to-noise ratio of the noise carrying data, which is seen to vary depending on the type of noise introduced. Finally, we present an expanded combinatorics algorithm in a 9C4 design, once again with and without additive noise. In both simple and expanded, pure signal scenarios, as well as in scenarios with added noise, our algorithm allows for achieving of 97% of original signal reconstruction. Additionally, we are able to identify the component neurons at each of the sites on the braided electrode structure.
We have identified our opportunity statement as a chance to develop a novel signal processing tool to fully leverage the new multisite wire recording probe. Our project was driven by the claim that expanding the complexity of the design and test data will allow us to increase neural recording yield by at least an order of magnitude, while maintaining the precision of combinatoric multi-electrode probe signal separation. |
|
Location:
Bossone Research Center, Room 709, located at 32nd and Market Streets. |
Audience:
Undergraduate StudentsGraduate StudentsFacultyStaff |
|