Event Related Potential Signal Capture Can Be Enhanced through Dynamic SNR-Weighted Channel Pooling

Autor: Xiaowei Song, Careesa C. Liu, Ryan C.N. D'Arcy, Shaun D. Fickling, Sujoy Ghosh Hajra, Gabriela Pawlowski
Jazyk: angličtina
Rok vydání: 2021
Předmět:
Adult
Male
Computer science
Pooling
neural signal processing
TP1-1185
Signal-To-Noise Ratio
Electroencephalography
Biochemistry
Signal
050105 experimental psychology
Article
Analytical Chemistry
03 medical and health sciences
0302 clinical medicine
Signal-to-noise ratio
channel pooling
medicine
Humans
Attention
0501 psychology and cognitive sciences
EEG
Electrical and Electronic Engineering
Evoked Potentials
Instrumentation
Language
medicine.diagnostic_test
Noise (signal processing)
business.industry
signal to noise ratio
Chemical technology
05 social sciences
Search engine indexing
Signal Processing
Computer-Assisted
Pattern recognition
Atomic and Molecular Physics
and Optics
N400
signal augmentation
Female
Artificial intelligence
business
030217 neurology & neurosurgery
ERP
Communication channel
Zdroj: Sensors (Basel, Switzerland)
Sensors, Vol 21, Iss 7258, p 7258 (2021)
Sensors
Volume 21
Issue 21
ISSN: 1424-8220
Popis: Background: Electroencephalography (EEG)-derived event-related potentials (ERPs) provide information about a variety of brain functions, but often suffer from low inherent signal-to-noise ratio (SNR). To overcome the low SNR, techniques that pool data from multiple sensors have been applied. However, such pooling implicitly assumes that the SNR among sensors is equal, which is not necessarily valid. This study presents a novel approach for signal pooling that accounts for differential SNR among sensors. Methods: The new technique involves pooling together signals from multiple EEG channels weighted by their respective SNRs relative to the overall SNR of all channels. We compared ERP responses derived using this new technique with those derived using both individual channels as well as traditional averaged-based channel pooling. The outcomes were evaluated in both simulated data and real data from healthy adult volunteers (n = 37). Responses corresponding to a range of ERP components indexing auditory sensation (N100), attention (P300) and language processing (N400) were evaluated. Results: Simulation results demonstrate that, compared to traditional pooling technique, the new SNR-weighted channel pooling technique improved ERP response effect size in cases of unequal noise among channels (p’s <
0.001). Similarly, results from real-world experimental data showed that the new technique resulted in significantly greater ERP effect sizes compared to either traditional pooling or individual channel approach for all three ERP components (p’s <
0.001). Furthermore, the new channel pooling approach also resulted in larger ERP signal amplitudes as well as greater differences among experimental conditions (p’s <
0.001). Conclusion: These results suggest that the new technique improves the capture of ERP responses relative to traditional techniques. As such, SNR-weighted channel pooling can further enable widespread applications of ERP techniques, especially those that require rapid assessments in noisy out-of-laboratory environments.
Databáze: OpenAIRE
Externí odkaz: