Classification of patient's reaction in language assessment during awake craniotomy

Surgical video recording is widely used in operation rooms in order to analyze such as surgical procedures and intraoperative incident detection. Therefore, a number of useful operation video records are stored in the hospitals. It is considered that these video records contain significant information, so it is needed to utilize these video data. In awake craniotomy, which is one of the advanced neurological surgery, surgeon peforms direct electrical stimulation to patient's brain area during linguistic tasks(such as, naming objects or generating verbs) in order to detect brain functional areas. The electrical stimulation of the cortical speech area causes temporary speech arrest. Hence, video segments which speech arrest is caused are significant in terms of surgical video analysis. The electrical stimulation timings are obtained from sound information, however that segments are not tagged speech arrest or not. In this paper, we report on the performance of a classification method for classifying patient's response for linguistic tasks just after electrical stimulation. In order to extract patient's speech features, we used melfrequency cepstrum coefficient(MFCC) and its delta parameters which are often used in speech recognition. We used Relevance Vector Machine(RVM) and Support Vector Machine(SVM) for classification and compared their results. We applied RVM and SVM for extracted patient's speech features and evaluated in F-measure. The classifier achieves in classification rates about 80[%] in 10-fold cross validation. The result shows that speech features are effective for classifying patient's responses.