The Research Of Automatic Method Based On Hybrid Level Set And Parameter Setting For Brain Extraction

Nowadays, with the improvement of medical technology, Magnetic ResonanceImaging is playing an increasingly important part in measurement of brain. MRIprovides both clear anatomical morphology and display of contrast between soft tissues.The extraction of brain tissue from cerebral T1-Weighted MRI volume is an importantpre-procedure for determination of brain volume, segmentation of brain tissues,quantification of brain atrophy.The brain extraction methods are divided into manual and automatic methods. Themanual method is time-consuming and subjective, requiring the operator withprofessional ability. The automatic method is more and more popular, because only aftersetting some related parameters and the following process can be done automatically.Usually, the parameters of the automatic method are hard to determine. To address thisproblem, this paper uses the random forest classifier to automatically set parameters.A new Active Contour Neighborhood (ACN) Model based hybrid level set methodwas presented in this article. It used the improved BET (Brain Extraction Tool)algorithm to get a rough brain boundary. Then, the initial ACN was obtained byexpanding the rough contour. After which the renewed contour line and ACN wereobtained by the hybrid level set step by step. Finally, the accurate cerebral contour wasobtained. The method used a nonlinear speed function in the hybrid level set model toeliminate boundary leakage.In order to solve the parameter setting problem, this paper proposed a method ofautomatic setting brain extraction parameter based on the Random Forest. This methodfirstly performed brain extraction on training samples with above proposed method andobtained the corresponding parameters. After feature extraction of training samples, thispaper used random forest classifier to classify the training samples and then performeddiscriminant for testing samples to obtain the parameters of testing samples.