TY - GEN
T1 - Subject-Independent Detection of Movement-Related Cortical Potentials and Classifier Adaptation from Single-Channel EEG
AU - Jochumsen, Mads Rovsing
N1 - This work was funded by VELUX FONDEN (project no. 22357).
PY - 2022
Y1 - 2022
N2 - Brain-computer interfaces have been proposed for stroke rehabilitation, but there are some impeding factors for them to be translated into clinical practice. One of them is the need for calibration. In this study it was investigated if subject-independent calibration is possible for detecting movement-related potentials associated with hand movements, and what the optimal number of movement epochs is to maximize the detection performance. Twelve healthy subjects performed 100 palmar grasps while continuous EEG was recorded. Template matching was performed between movement and idle epochs. 72 ± 10% of all epochs were correctly classified using the subject-independent approach while 78 ± 9% of the epochs were correctly classified using the individualized approach. The highest classification accuracies were obtained when using 54 ± 23 movement epochs for calibration. In conclusion, it is possible to use a subject-independent approach for detecting movement-related cortical potentials, but the performance is slightly lower compared to individualized calibration.
AB - Brain-computer interfaces have been proposed for stroke rehabilitation, but there are some impeding factors for them to be translated into clinical practice. One of them is the need for calibration. In this study it was investigated if subject-independent calibration is possible for detecting movement-related potentials associated with hand movements, and what the optimal number of movement epochs is to maximize the detection performance. Twelve healthy subjects performed 100 palmar grasps while continuous EEG was recorded. Template matching was performed between movement and idle epochs. 72 ± 10% of all epochs were correctly classified using the subject-independent approach while 78 ± 9% of the epochs were correctly classified using the individualized approach. The highest classification accuracies were obtained when using 54 ± 23 movement epochs for calibration. In conclusion, it is possible to use a subject-independent approach for detecting movement-related cortical potentials, but the performance is slightly lower compared to individualized calibration.
UR - http://www.scopus.com/inward/record.url?scp=85116926513&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-70316-5_13
DO - 10.1007/978-3-030-70316-5_13
M3 - Article in proceeding
SN - 978-3-030-70315-8
SN - 978-3-030-70318-9
T3 - Biosystems and Biorobotics
SP - 77
EP - 81
BT - Converging Clinical and Engineering Research on Neurorehabilitation IV
A2 - Torricelli, Diego
A2 - Akay, Metin
A2 - L. Pons, Jose
PB - Springer
T2 - 5th International Conference on NeuroRehabilitation
Y2 - 13 October 2020 through 16 October 2020
ER -