Efeito da prática mental no eletroencefalograma durante tarefas visuo-motoras

Abstract

The brain computer interface (BCI) is a communication system that translates brain signals recorded by means of the electroencephalogram (EEG) into commands to control external devices. The motor mental practice (MP) has been used as training method in which an internal representation of an action is repeated in order to increase the performance of a specific task. It is well known that the EEG recorded during repetitive movement execution or movement imagination (MI) has a random component (spontaneous EEG) and a stimulation response component, named motor event-related potential (ERP). The recognition of this pattern (ERP) is one of the main steps of the BCI-MI. Therefore, the extraction of relevant characteristics of the response is needed. This work aims at investigating the contingent negative variation (CNV) – the ERP readiness component – along the MP sessions. Furthermore, verify the effects of the MP in the hit rate classification between two classes: visual-motor task (index imagination movement) and visual task (without specific imagination). EEG signals were collected from 17 electrodes in the somatosensory and motor regions from 7 healthy individuals during 6 MP sessions (Se1-Se6). The Support Vector Machine (SVM) was employed as a classification algorithm, using the Magnitude- Squared Coherence (MSC) values within the delta band (0, 0,33, 0,67, 1 Hz) as input characteristic data. The MSC values were estimated for EEG signals from 4 volunteers, separately. First, the Se1, Se3, Se4 and Se5 sessions were used for machine training and Se6 for validation (case 1). Second, the machine training was performed using Se1 and the validation with Se3, Se4, Se5 and Se6 (case 2). In case 1, the algorithm achieved classification rates of 0.95, 0.7, 0.93 and 0.98 for the 4 subjects, respectively. In case 2, a trend of increasing in the hit rate percentage along the sessions was observed, mainly in the second and the fifth sessions. Subjects S1 and S3 showed a progressive improvement in the classification rates along all sessions. These results suggest that MP leads to an increase in the motor task execution/imagination performance, being a potential tool for BCI-MI improvement.