Study On The Algorithm For The Intraoperative Localization Of Brain Functional Areas Based On Clustering And Binary Classification Of Rest State ECoG

The classification model-based intraoperative functional mapping is expected to achieve accurate,rapid,and non-invasive neurosurgical brain function localization.However,due to the specific differences of functional ECoG between individuals,there is a problem of poor generalization performance of the functional area localization algorithm.This article envisaged the use of a combination of clustering and classification to improve the generalization performance of the localization algorithm,and developed an intraoperative localization algorithm of brain functional areas based on clustering and binary classification of rest state ECoG.Firstly,the acquisition method and preprocessing algorithm of the rest state ECoG were optimized.This paper designed a kind of intelligent detector to analysis and indicate the three states of contact between cortical electrode and the cerebral cortex,including good contact,suspended,and short-circuited electrodes,which could ensure accurate ECoG collection when good contact is detected.Then this paper combined 1 Hz high-pass filter,the 49-51 Hz notch filter with independent component analysis filter to filter out power frequency interference.Secondly,an algorithm for the intraoperative localization of brain functional areas based on clustering and binary classification of rest state ECoG was studied.First,a clustering algorithm of functional area's boundary was used,this clustering algorithm used db3 wavelet to perform six-layer wavelet decomposition on the rest state ECoG and reconstructed seven single sub-band data,then extracted the energy ratio of these seven sub-band data as feature of clustering.And this clustering algorithm designed a cohesive hierarchical clustering model to obtain two clusters after clustering;Second,a classification algorithm of functional area's attribute was designed,This classification algorithm used 6-dimensional time-domain statistics(sum,mean,variance,standard deviation,maximum,minimum)and the energy of5-dimensional rhythm(?(1-3 Hz),?(4-8 Hz),?(9-12 Hz),?(13-30 Hz),?(60-120 Hz))of the samples as feature of the classification,and principal component analysis was used to reduce the dimensionality of the feature.Then,a support vector machine based on cross-validation and radial basis kernel function was used to obtain two classifications after classify;Last,an integrated algorithm of intraoperative functional mapping based on ECoG was established based on the above two algorithms.This integrated algorithm uses the similarities between clusters and classifications to determine the attribute of the clusters and regard the clusters with attribute as the result of intraoperative functional mapping.Three patient?s rest state ECoG were collected and pre-processed to train and test model.The test results showed that the recognition accuracy of this paper?s algorithm is 90.91%,and the false detection rate is9.09%,while the generalization performance is good,it has the potential value of application in neurosurgery clinical.Finally,an intraoperative functional mapping software that based on clustering and binary classification of rest state ECoG was designed.First,the software module for clustering algorithm of functional area's boundary which has three steps of data import,preprocessing and feature extraction of clustering was designed to get two clusters;Second,the software module for classification algorithm of functional area's attribute which has three steps of loading classification model,feature extraction of classification and attribute recognition was designed to get two classifications;Last,the software module for integrated algorithm of intraoperative functional mapping which has two steps of functional area positioning table and functional area map was designed.This integrated algorithm extracted the electrode coordinates and combine the functional area positioning table to draw a functional area map.Then,this functional area map could be used to help the doctors clarify the functional areas around the lesions,to guide the doctors to remove the lesions in neurosurgery.