Comparison of Classifier Calibration Schemes for Movement Intention Detection in Individuals with Cerebral Palsy for...

Area of Science:

• Neuroscience
• Rehabilitation Engineering
• Biomedical Engineering

Background:

• Brain-computer interfaces (BCIs) are established tools for stroke rehabilitation.
• BCI training shows potential benefits for individuals with cerebral palsy (CP).
• Accurate detection of movement intentions from single-trial electroencephalography (EEG) is crucial for developing BCI applications in CP.

Purpose of the Study:

• To assess the feasibility of detecting movement intentions in individuals with CP using EEG.
• To compare BCI performance in CP patients versus able-bodied participants.
• To evaluate different classification scenarios for BCI training in CP.

Main Methods:

• EEG data were recorded from five CP patients and fifteen able-bodied participants performing wrist and ankle movements.
• Movement intention and idle epochs were classified using a random forest algorithm.
• Features included temporal, spectral, and template matching characteristics of the EEG signals.

Main Results:

• Classification accuracies of 75% for CP participants and 85% for able-bodied participants were achieved when classifiers were trained and tested on the same day and participant.
• Performance decreased by 5-15% when classifiers were trained on data from different days or participants.
• These findings indicate reliable detection of movement intentions from single-trial EEG.

Conclusions:

• Movement intentions can be detected from single-trial EEG in individuals with CP.
• This technical feasibility supports the development of BCI-based motor training for CP.
• BCIs offer a potential avenue for enhancing rehabilitation outcomes in CP patients.