Denoising Algorithm for Sets of Evoked Potentials

Denoising Algorithm for Sets of Evoked Potentials

A mathematical algorithm that denoises separately measured evoked potentials.

Institute Reference: 2-23107

Background

Evoked potentials are measured brain responses to some stimulus typically measured on the scalp using electroencephalography (EEG) electrodes. Measuring evoked potential serves many purposes, such as measuring auditory brainstem responses (ABR), a specific kind of evoked potential, to diagnose hearing loss in infants or other patients who cannot complete behavioral tests of hearing loss. A problem with measuring evoked potentials is the noise produced from additional unrelated signals measured by the EEG. Evoked potentials have a lower amplitude compared to these larger unrelated signals picked up by the EEG, making them harder to isolate. The relationship between the amplitude of evoked potential to noise is represented by the signal-to-noise ratio (SNR), a ratio of the energy of the signal (E­s) to the energy of the noise (En). To improve the speed and quality of diagnoses that can be made using evoked potential measurements, a noise reduction method is required.

Technology Overview

Researchers offer a mathematical algorithm that allows separately measured evoked potentials, such as the auditory brainstem response to stimuli at various frequencies and sound levels, to be denoised. The principal innovation of this invention is to use the information present to improve the SNR of responses by leveraging neighboring responses to similar stimuli. This is accomplished by computing a dual‐tree complex wavelet transform of the set of responses along with regularization that penalizes the size of the coefficients and the size of the differences between corresponding coefficients of responses at neighboring stimulus frequencies and levels.

Benefits

Current research shows that the denoising algorithm offers a substantial SNR improvement even with minimal data collection. Early pilot experiments resulted in SNR improvements of 4 or 5 dB. Such SNR improvements reduce the amount of data needed by two thirds, meaning denoised parallel auditory brainstem response (pABR) recording sessions may be 3 times faster than standard pABR recording sessions, which are 2 to 4 times faster than traditional ABR methods currently used in the clinic. Thus, combining this denoising method with previously published methods could allow full sets of diagnostic ABR waveforms to be measured on the order of 10× faster than traditional methods

Applications

Although this algorithm was initially designed for auditory brainstem responses (ABR), its application extends to other types of evoked potentials. This includes, but is not limited to, the multi-focal electroretinogram and other signal sets sharing similar characteristics, such as head-related impulse responses making this technology applicable in augmented and virtual reality endeavors.

URV Reference Number: 2-23107
Patent Information:
Category(s):
Research Tools
Diagnostic
For Information, Contact:
McKenna Geiger
Licensing Manager
University of Rochester
585-276-6600
mckenna_geiger@urmc.rochester.edu
Inventors:
Ross Maddox
Keywords: