MRI大脑图像灰质与白质的分割

Segmentation for brain grey matter and white matter of a MRI brain image

目的利用小波变换对MRI大脑图像进行多尺度下的自动阈值处理,实现大脑灰质与白质的分割。方法首先将MRI大脑图像去噪,接着进行预分割以去除非脑组织,余下的脑实质部分选择sym4小波函数对其一维直方图信号进行不同层次的小波系数的分解,实现多尺度下的自动阈值分割,从而提取脑实质中的灰质和白质。经过图像的后处理,以错误分割的百分比作为分割结果的评判标准。结果该方法能正确分离白质和灰质,对多幅MRI大脑图像重复实验,计算得到的像素差异百分比不超过3.7%,错误分割的百分比在允许范围内。结论该方法对于MRI大脑的灰白质分割具有一定的有效性,且操作简单、快速,分割效果理想。但由于小波阈值分割法的单一性,分割过程仍有人工干涉,分割结果也存在一定的过分割现象,应在此方法的基础上进一步研究和完善。

Objective To realize the segmentation of brain grey matter and white matter of a MRI brain image accurately by means of wavelet transformation which could automatically gain the threshold with multi-scale decomposition. Methods Firstly,The MRI brain image was denoised and pre-segmented to remove non- brain tissue. Then onedimensional histogram signal of the rest brain tissue was decomposed by different levels of wavelet coefficients to realize automatically threshold segmentation under multi-scale, thus the brain grey matter and white matter were extracted. Finally, the results were evaluated by the percentage of error segmentation after the post-treatment. Results The expriment results showed that brain grey matter and white matter were correctly segmented, and the max percentage of error segmentation was 3.7% ,which was within the allowable range. Conclusions The method is effective, easy and fast in segmentation for brain grey matter and white matter of a MRI image. However, due to the simplicity of the threshold segmentation with wavelet transformation, the whole process does not work without human intervention and there is some over-segmentation in the results. Further study and improvement should be made based on this method.