利用分区处理和水平集算法分割序列三维乳腺MRI

3-D breast MRI sequence segmentation based on region and level set algorithm

对比增强的磁共振成像技术是检测乳腺肿瘤的新方法,但需处理大量随时间变化的三维影像序列。为了从这种四维影像中分割出乳房组织,该文提出了自动分区域分割方法。先分割乳房与空气,并统计乳房部分的灰度;再分割乳房与胸腔,将前面统计出的乳房灰度作为先验知识设定初始轮廓和特征图像,采用基于阈值区间的三维水平集算法,并将前一时间点的结果作为后一时间点分割的初始轮廓,从而实现自动的四维分割。临床数据实验表明,该方法可有效地分割四维乳腺磁共振影像,且自动化程度和分割准确度均较高。

Contrast-enhanced magnetic resonance imaging （MRI） is used to detect breast tumors by analyzing multiple 3-D image sequences that change with time. A multistage processing procedure was developed to segment the breast tissue in these 4-D MRI. The images are first divided into different regions. Then the breast area in the anterior part of the image is segmented from the background to estimate the mean intensity of the breast tissue. The breast area is segmented from the chest by using the estimated mean breast intensity as initial guess of the grey level of the contour between the breast and the chest. A threshold-based 3-D level set algorithm was developed to identify the boundary surface between the chest and the breast by using the estimated breast intensity as a feature image. The segmentation of one 3-D image was used for the initial estimate of the boundary on the following image to achieve automatic 4-D segmentation. Clinical trials show that this processing procedure is effective for automatic segmentation of 4-D breast MRI.