Universal matched-filter template versus individualized template for single trial detection of movement intentions of different tasks

Abstract

Brain-computer interfaces (BCIs) have been proposed for neurorehabilitation after stroke by inducing cortical plasticity. To transfer the technology from the controlled settings in the lab to the clinic several issues must be addressed. In this study, it was investigated how the performance was affected by using a universal task template to detect movement intentions associated with movements performed with two different levels of force and speed. The performance of the universal template was compared to an individualized template constructed for each task. Twelve healthy subjects performed four types of dorsi-flexions while continuous electroencephalography (EEG) was recorded from ten channels. The movement intentions were detected (∼200-300 ms before the movement onset) from the continuous EEG using a matched-filter approach. The true positive rate was significantly higher (P = 0.001) when using the individualized template where 68-76% of the movements were correctly detected on the contrary to 65-70% when using the universal template. The number of false positive detections per 5 min was lower (P = 0.036) when using the universal template (∼13) compared to the individualized template (∼14). Despite the lower performance when using the universal detection template, the performance of the detector is in the range of what has been reported previously for inducing cortical plasticity.