- Author: Hongbiao Chen
- Supervisor: Dr. Michel Besserve
- Date: January 31, 2020
- Place: Neural Information Processing, Graduate Training Center for Neuroscience, University of Tübingen
Abstract
The simultaneous acquisition of functional magnetic resonance imaging (fMRI) and electrophysiology (Ephy) data is a promising experiment technique in neuroscience because it allows us to study the same neural system from two different levels concurrently. However, recording fMRI-Ephy together induces fMRI gradient interference on Ephy data and denoising Ephy signals is still challenging. Here we review methods of reducing gradient artifacts, including hardware-based and software-based, and present a new denoising method, Non-negative Matrix Factorization with Itakura-Saito divergence (IS-NMF). Our results demonstrate that IS-NMF approach can effectively remove residual artifacts to obtain high quality Ephy data after applying hardware compensation and standard averaged artifact subtraction (AAS).
Summary
In summary, when we record fMRI and Ephy signals together, the Ephy data contains not only the original neural electrical signals but also fMRI-induced gradient artifacts and residual noise. After using hardware circuits to compensate the fMRI interference, the rest of noise can be eliminated by performing IS-NMF in addition to AAS method to get the denoised Ephy data.