Extracting information from the shape and spatial distribution of evoked potentials. J Neurosci Methods 2018 Feb 15;296:12-22
Date
12/27/2017Pubmed ID
29277720Pubmed Central ID
PMC5840508DOI
10.1016/j.jneumeth.2017.12.014Scopus ID
2-s2.0-85039751518 (requires institutional sign-in at Scopus site) 7 CitationsAbstract
BACKGROUND: Over 90 years after its first recording, scalp electroencephalography (EEG) remains one of the most widely used techniques in human neuroscience research, in particular for the study of event-related potentials (ERPs). However, because of its low signal-to-noise ratio, extracting useful information from these signals continues to be a hard-technical challenge. Many studies focus on simple properties of the ERPs such as peaks, latencies, and slopes of signal deflections.
NEW METHOD: To overcome these limitations, we developed the Wavelet-Information method which uses wavelet decomposition, information theory, and a quantification based on single-trial decoding performance to extract information from evoked responses.
RESULTS: Using simulations and real data from four experiments, we show that the proposed approach outperforms standard supervised analyses based on peak amplitude estimation. Moreover, the method can extract information using the raw data from all recorded channels using no a priori knowledge or pre-processing steps.
COMPARISON WITH EXISTING METHOD(S): We show that traditional approaches often disregard important features of the signal such as the shape of EEG waveforms. Also, other approaches often require some form of a priori knowledge for feature selection and lead to problems of multiple comparisons.
CONCLUSIONS: This approach offers a new and complementary framework to design experiments that go beyond the traditional analyses of ERPs. Potentially, it allows a wide usage beyond basic research; such as for clinical diagnosis, brain-machine interfaces, and neurofeedback applications requiring single-trial analyses.
Author List
Lopes-Dos-Santos V, Rey HG, Navajas J, Quian Quiroga RAuthor
Hernan Gonzalo Rey PhD Assistant Professor in the Neurosurgery department at Medical College of WisconsinMESH terms used to index this publication - Major topics in bold
Auditory PerceptionBrain
Computer Simulation
Electrodes, Implanted
Electroencephalography
Evoked Potentials
Humans
Information Theory
Pattern Recognition, Automated
Pattern Recognition, Physiological
Signal Processing, Computer-Assisted
Supervised Machine Learning
Visual Perception
Wavelet Analysis
