基于数学形态学人脑MR图像感兴趣区域的提取

Interested area extraction of human brain MR image based on mathematical morphology

背景:在人脑MRI图像中感兴趣区域提取中,应用数学形态学方法取得了较好的效果,但是在抗噪性能和结构元素选取时存在一些不足之处,使得提取效果有缺陷。目的:在数学形态学的基础上,采用一系列改进的数学形态学方法,以期清晰完整地提取人脑MR图像中的感兴趣区域如脑脊液部位,为医学诊断提供准确信息。方法:首先采用复合形态学滤波去除脉冲和高斯噪声,用高低帽变换进行图像增强,然后用形态分水岭阈值分割提取脑部各成分,对分割出的脑脊液图像进行形态开闭滤波、边缘跟踪和灰度填充后,运用抗噪型边缘检测算子检测出清晰完整的脑脊液区域边缘,最后在原图像中用彩色标定,突出感兴趣区域。结果与结论:综合应用多种数学形态学算法,清晰完整地提取了人脑MRI图像中的感兴趣区域——脑脊液部位。经验证,该方法具有简单、快速、精度高、适用性强等特点。

BACKGROUND：The applying of mathematical morphological algorithm has achieved good result in extracting the interested area of the brain MR image.However,it is has limitation in anti-noise property and structuring element selection. OBJECTIVE：To extract the interested area of the brain MR image clearly and fully based on mathematical morphology,to provide accurate information for clinical medical diagnosis. METHODS：Firstly,a compound mathematical morphological algorithm was used to filter the pulse and gauss noise and a hip-top cap transform was used to buildup the image.Then,the brain compositions were extracted based on watershed threshold segmentation method.After morphological filter,tracking edge and filling in gray degree,the edge of the interested area was detected clearly by the anti-noise edge detectors.At last,in order to stand out the physician,interested area,it was demarcated colorful in original image. RESULTS AND CONCLUSION：It introduced a combination utilization of multi-mathematical morphology algorithms to realize the interested area extraction of brain clearly and fully.The experimental results show that the proposed algorithm is characterized by simple,fast,high precision and strong applicability.