一种快速提取连续CT脑肿瘤病灶区的算法

A Rapid Extraction Algorithm in Continuous CT Brain Lesions

脑肿瘤图像提取就是将肿瘤病灶区域(水肿、坏死、癌变)从正常的脑部组织(灰质、白质、脑脊液)分开,精确的脑肿瘤分割对脑瘤的诊断、研究和治疗有重要的临床意义;针对传统脑部CT肿瘤病灶提取的缺点,即需要耗费大量时间并且分割精度不高的问题,提出一种综合了形态学重建、分水岭分割和改进的区域生长算法;先用形态学重建进行去噪,再用结合多尺度梯度分水岭分割提取整个图像的边界,然后在肿瘤病灶区域内选取种子点进行区域生长,提取肿瘤区域轮廓,滤除其他封闭区域,得到的图像作为改进的区域生长法的初始分割区域,使用改进的区域生长法,滤除过分割区域;实验结果显示该算法分割出的结果有效区域大,分割精度高;该算法提高了分割精度,由于不用匹配结构参数,加快了分割速度,具有一定的临床价值。

Brain tumor image extraction is separate the tumor lesion area （edema,necrosis,cancer） from normal brain tissue （gray matter,white matter and cerebrospinal fluid）,It has important clinical significance that accurate segmentation of brain tumors in the diagnosis,the research and treatment of brain tumor.Based on the shortcomings of traditionally extracting brain tumor lesion area,which requires a lot of time and image segmentation accuracy is not high,proposed a kind of comprehensive morphological reconstruction,a watershed segmentation and the improved region growing algorithm.Using morphological reconstruction to remove noise firstly,and then using watershed segmentation combined with multi-scale gradient to extract the the border of the whole image,then selecting seed points in the area of the tumor lesion to be region growing,extracted the tumor area outline,filter out other closed area.The image as the initial segmented regions of the improved region growing method.Finally using the improved region growing method,filtering excessive segmentation region.The experiment results show that the algorithm segmented effective area is larger and higher segmentation accuracy.Conclusion：the algorithm improved the precision of segmentation,and need not frequent matching structure parameters of the watershed algorithm,speed up the segmentation,has a certain clinical value.