Analysis and Classification of EEG Signals from Passive Mobilization in ICU Sedated Patients and Non-sedated Volunteers

Electroencephalography (EEG) has been the focus of research and advances for many years, yet there are several tasks to be explore and methods to be tested to improve analysis and classification. Event-Related Potential (ERP) is one of the brain responses measured with EEG, resulting from motor tasks usually are related to motor imagery or real movement. This study aims to analyze and classify event-related desynchronization (ERD) and event-related synchronization (ERS) occurred in tasks involving passive mobilization in Intensive Care Unit (ICU) sedated patients and non-sedated volunteers. Our main goal is to provide preliminary analysis and comparisons between sedated and non-sedated groups based on signal visualization and a classifier. Common Spatial Pattern filtering (CSP) and visual inspection of best band and time were used to verify signal and phenomena. From that, specific features (i.e., Root Mean Square, standard deviation, mean of Welch periodogram and differential entropy) were extracted based on time and frequency to apply a Linear Discriminant Analysis (LDA) classifier. Once the two Intensive Care Unit sedated patients and the two volunteers were analyzed, it was possible to observe the proposed phenomena. Mean accuracy in the best scenario and best person for each group (two people in each group) was found higher than 80 and 77% to sedated and non-sedated participants, respectively. Preliminary results, based on four participants (i.e., two sedated and two non-sedated patients), suggested lateralization in tasks performed with passive mobilization and provided accuracy comparable to previous studies involving motor tasks.