摘要
医学图像中病变信息的计算机自动提取是实现计算机智能辅助诊断的关键与难点,本研究的目的就是提出一个解决该难题的算法,称之为PATHOINFER。该算法的基本过程是首先选择一幅具有代表性的模板图像M0和一系列与其相应的正常图像样本Mi,利用非刚性配准分别建立表示“正常图像”灰度变化的灰度均值图谱,表示正常变异的统计概率图谱和反映其解剖结构空间关系的分割模板,以实现对“正常图像”的计算机描述。再通过M0与目标图像S的配准,达到“正常图像”与S在空间关系上的一致,然后通过S与“正常图像”的比较,利用模糊逻辑推理,自动检出S中的病变区域,并实现对其病变特征信息的自动提取。实验结果表明,PATHOINFER算法可自动地检出并分割病变区域,并能够自动地提取包括病变发生部位在内的特征信息,实现了计算机智能辅助诊断研究中病变信息自动提取的难题。
The key process and difficulty to realize computer intelligent auxiliary diagnosis of medical image is how to automatically acquire the pathological information. Our goal in this paper is to put forward a method called PATHOINFER to solve this problem. The basic procedure of this method is as follows. First, a representative normal image M and a series of randomly chosen normal images Mi corresponding to M0 was selected. By non-rigid registration from Mi to M, computerized description of ＂Normal Images＂ related to M0 was created, including an average intensity image to describe the intensity variability, a probability atlas to describe structure physiological variability and a segmented and labeled image to describe the relationship of different anatomical structures. Second, ＂Normal Images＂ were transformed to the same coordinate systems of pathology image S by registration from M to S. Finally by comparing the S with those registered ＂Normal Images＂ utilizing fuzzy logic inferrenee system, the pathologic region was segmented and pathological information was extracted. The experimental results showed that PATHOINFER could automatically segment pathology regions and could easily extract pathological information about pathology and be used to solve the problem of computer intelligent auxiliary diagnosis of medical image.
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2006年第1期19-24,共6页
Chinese Journal of Biomedical Engineering
基金
安徽省教委自然科学基金资助项目(2003kj238)
关键词
非刚性配准
病变分割
病变信息提取
计算机智能辅助诊断
non-rigid registration
pathological segmentation
pathological information extraction
computer intelligent auxiliary diagnosis
