基于阈值分割的K-最近邻规则的磁共振脑图像分割算法的研究

目的:提取脑组织中的白质(WhiteMatter)、灰质(GreyMatter)和脑脊髓液(CSF)。方法:针对去除颅骨和肌肉等非脑部组织的磁共振脑图像,根据解剖学知识采用最大类别方差法(Otsu法)自动寻找阈值,以此为依据进行分类标记,再通过K-最近邻(K-NearestNeighbor,简称KNN)规则对大脑组织结构进行划分。结果:在脑部T1加权像中分割算法分别提取脑组织中的白质、灰质和脑脊髓液。结论:结果表明该算法具有较好的自动性和稳定性。

Study of segmentation algorithm of brain MRI based on K-nearest neighbour rule with threshold segmentation

Objective:To divided the structures of the brain to white matter (WM),grey matter (GM),cerebral spinal fluid (CSF).Methods:The objects that were going to be processed were MR images of the brain from which the non-brain tissues, such as muscle and skull, had been removed. This algorithm based on the knowledge of anatomy adopted Ostu method to search the thresholds automatically, and then according to thresholds we classified and taken pixels of the MR image, after that we adopted K-nearest neighbour rule to classify the structures of the brain.Results:The segmentation method separates white matter, grey matter, and cerebral spinal fluid from T_1-weighted of brain MR images.Conclusion:The result proved that this method had a good automaticity and stability.