Study On Mis-triggering Of MI-BCI System For Stroke Rehabilitation

Stroke(Stroke)patients often have limbs movement dysfunction after onset,timely and scientific rehabilitation training can restore the limb movement function.Recently,Motor Imagery-Brain Computer Interface(MI-BCI)has been introduced into the clinical trails as a new rehabilitation therapy.Comparing to the traditional therapy,MI-BCI has better neural plasticity due to the direct decoding of the movement intention and the real-time feedback,which strengthens the active participation of patients during rehabilitation training.Some physicians and patients indicated that there was so-called "Mis-triggering" phenomenon(External devices were activated even there was no target behavior inputing)during training.Given mis-triggering may cause a series of negative effects of treatment,two stages of studies were carried out in order to solve this problem.First,experiments about mis-triggering factors were designed in this study.Classifier with right arm imagery data and relax data was built.Then,five kinds of factors data from 11 people were sent into it in order to test the performance of the classifier.The results indicate that all the five factors can mis-trigger MI-BCI system,with mis-triggering rate from 48% to 63%.Apparently,mis-triggering of MI-BCI system has serious impact.Next,further experiments for the main mis-triggering factors were carried out.EEG data from 15 people were collected in this stage.A double-classification strategy were proposed so that the mis-triggering problem could be solved.Experimental results prove the validity of this method.The mis-triggering rate of MI-BCI system decreases significantly.The AUCs of double-stage classifiers with different modeling condition increase 11%-14% and the mis-triggering rates fall 40%-70%.A universal model and a new ClusterCSP algorithm were also developed in this thesis in order to explore the development direction of mis-triggering problem for MI-BCI system.Mis-triggering factors and the solution for this problem have not been reported in literature yet.This study enables the improvement of the performance of rehabilitation MI-BCI and clinical application,as well as provides a direction for the development of MI-BCI.