基于代价函数掩模法的局部脑损伤fMRI数据处理

The Method of Processing Brain fMRI Data with Focal Lesions Based on Cost Function Masking

功能磁共振数据空间预处理中,个体图像到标准空间的归一化一般采用基于强度最小化的自动归一化标准方法。在处理有明显脑结构局灶异常的数据时,由于病灶区与模板的不匹配,该算法会导致图像的严重扭曲和变形,造成数据失真和功能激活区定位错误。采用代价函数掩模方法,归一化处理有明显脑结构局灶异常功能数据,可减少传统归一化方法因脑损伤而扭曲、变形对数据造成的影响。本研究对4例顶叶主运动区有明显脑结构异常(2例脑膜瘤,2例脑胶质瘤),手动功能评价的fMRI数据进行处理。首先采用代价函数掩模法对病灶区进行掩盖,然后再进行标准模板的归一化处理,归一化后的结果用SPM软件分析脑激活区。实验表明,基于代价函数掩模法的归一化方法可以有效地减少局灶脑结构异常fMRI数据归一化过程中的扭曲,提高脑功能区定位的准确性,数据处理的结果更符合脑病理情况下脑区活动的真实情况。

The spatial normalization is usually based on automated algorithms, which minimize the intensity values between source images and template. In spatial pre-processing of functional MRI data with focal lesions, if the images are mismatched with the template at the site of the lesion, the algorithms can lead to significant inappropriate image distortion and inaccurate functional location. In this paper, we used cost function masking （CFM） to reduce the influence. The function of hand movement was evaluated via the fMRI data for the four patients who had parietal lobe lesion in the main motor area （two cases of meningioma and two cases of brain glioma）. First, we used cost function masking to deal with the area of the lesion in spatial normalization, then normalized images with the template. We evaluated the final functional activated areas using SPM. The experimental results showed that cost function masking reduced the mismatch in the spatial normalization of brain fMRI data with focal lesions, and improved the accuracy of the brain functional location.