Automated Labeling of Movement- Related Cortical Potentials Using Segmented Regression

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Movement-related cortical potential (MRCP) signals are crucial for understanding motor planning and execution.
• Key MRCP features include early and late Bereitschaftspotential (BP1, BP2) and negative peak (PN).
• Current methods for extracting these features rely on manual labeling or fixed time points, limiting their accuracy.

Purpose of the Study:

• To develop and validate an automated method for labeling MRCP features.
• To improve the quantification of neurophysiological changes in motor tasks.
• To overcome the limitations of manual feature extraction in MRCP analysis.

Main Methods:

• Proposed a novel approach combining segmented regression with a local peak method for automated feature labeling.
• Evaluated regression techniques (bounded segmented regression, change point, MARS) using root-mean-square error on simulated MRCPs.
• Applied the best method to simulated and experimental electroencephalography (EEG) data, comparing results with expert manual labeling.

Main Results:

• Bounded segmented regression demonstrated the lowest error on simulated data.
• The automated method showed comparable performance to expert manual labeling on experimental EEG data.
• No statistically significant bias was observed in the modeled MRCP features using the proposed method.

Conclusions:

• The developed automated method provides robust estimates of MRCP features.
• This technique offers a reliable and accurate alternative to manual feature extraction.
• The findings support the use of this automated method in neurophysiological research and motor control studies.