基于改进区域生长法分割CT图像肝肿瘤的研究

Research on segmentation of liver tumor in CT image based on improved region growing algorithm

针对腹部内存在组织器官多、解剖结构复杂、组织器官边界模糊等特点,该文提出改进区域生长法分割CT图像肝肿瘤。算法根据肝肿瘤CT值与正常肝组织CT值差异遍历整幅图像预判种子点位置。种子区域的生长结合局部区域像素的均值和局部区域像素梯度均值。通过对不同类型的CT图像肝肿瘤进行分割,并以面积迭代度和误分率评判分割效果。研究结果表明,改进算法的面积迭代度是0.98±0.16、误分率是0.08±0.03,明显高于传统区域分割法(面积迭代度是0.78±0.32、误分率是0.35±0.18)。因此,改进区域生长法可以较好地实现肝组织中感兴趣区域的提取,并保持肝组织的完整,为后续肝脏肿瘤的三维重构和外科手术治疗提供准确的临床数据。

There are many tissues and organs in the abdomen,the anatomical structure is complex,and the boundaries of tissues and organs are fuzzy.This paper presents an improved region growing algorithm for liver tumor segmentation.The algorithm traverses the whole image to predict the seed point position according to the difference between the CT value of liver tumor and the CT value of normal liver tissue.The growth of the seed region is combined with the average value of the pixels in the local region and the average value of the pixel gradient in the local region.Different types of CT images of liver tumors were segmented,and the segmentation effect was evaluated by area overlap measure and misclassified error.The results show that the area overlap measure of the improved algorithm is 0.98±0.16 and the misclassified error is 0.08±0.03,which are significantly higher than the traditional area segmentation method(the area overlap measure is 0.78±0.32 and the misclassified error is 0.35±0.18).Therefore,the improved algorithm can better achieve the extraction of regions of interest in liver tissue and maintain the integrity of liver tissue,which provides accurate clinical data for the three⁃dimensional reconstruction and surgical treatment of liver tumors in the future.