Related Experiment Videos
Nonlinear adaptive filtering of stimulus artifact.
R Grieve1, P A Parker, B Hudgins
1Department of Electrical and Computer Engineering, University of New Brunswick, Fredericton, Canada.
IEEE Transactions on Bio-Medical Engineering
|April 1, 2000
Summary
This study introduces a novel neural network-based adaptive noise cancelling filter to effectively reduce electrical artifact in noninvasive somatosensory evoked potential measurements, improving signal accuracy.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Signal Processing
Background:
- Noninvasive somatosensory evoked potentials (SEPs) are crucial for clinical and research applications.
- Electrical stimulus artifact interferes with SEP signals, complicating analysis and timing estimation.
- Conventional ensemble averaging is ineffective against stimulus-synchronous artifact.
Purpose of the Study:
- To investigate the efficacy of an adaptive noise cancelling (ANC) filter utilizing a neural network for stimulus artifact reduction in SEPs.
- To develop and evaluate a segmented neural network training approach for improved artifact cancellation.
Main Methods:
- Implementation of an ANC filter with a neural network as the adaptive element.
- In vivo median nerve measurements were used to assess performance.
- A segmented neural network training strategy was employed, training on artifact-free signal segments.
Main Results:
- The neural network demonstrated effective generalization from training segments to segments containing evoked potentials.
- Significant reduction in stimulus artifact was quantitatively and qualitatively confirmed.
- The proposed method successfully addressed interference in noninvasive SEP recordings.
Conclusions:
- Neural network-based ANC is a promising technique for mitigating stimulus artifact in SEPs.
- Segmented training enhances the adaptability and effectiveness of the neural network filter.
- This approach improves the reliability of noninvasive SEP measurements for clinical and research use.