脑部MRI图像的半自动实时轮廓提取方法

摘要

Study semi - automatic real - time contour extraction problem. Because of high resolution and large a-mount of information in Brain MRI image, it is difficult to guarantee real - time accurate contour extraction for traditional methods of segmention and contour extractio. To solve above problem, a semi - automatic real - time contour extraction medthod was proposed based on domin decomposition and mark image, the intrest region was decomposed into a number of sub - region, the contour of every sub - region was extracted and finally extraction of the intrest region was completed. Mark image which noted marker status of the corresponding points on medical image was used for boundary tracking, after acquiring the contour of sub - region, the marker status of points on mark image was changed at the same time, then the merged contour of medical image was obtained through neighborhood search of mark image. Experimental results show that the method can extract contour of intrest region instantaneously and fluently , and the satisfactory extracition efficiency of narrow - connected region and outer surface proves its strong robustness.%研究脑部MRI图像的实时轮廓提取问题.由于脑部MRI图像分辨率高且信息量大,传统方法对整幅图像做图像分割和轮廓搜索,难以保证实时精确的轮廓提取.为解决上述问题,提出一种用区域分解和图像标记的实时轮廓提取方法,将感兴趣区手动分解为若干个自定义子区域的集合,依次提取各子区域轮廓并与当前图像轮廓合并,最终实现感兴趣区的轮廓提取.标记图像记录医学图像上各点的标记状态用于边界跟踪,提取当前子区域的轮廓后,标记图像上各点状态随之改变,对其做邻域搜索可得到合并后的医学图像轮廓线.实验结果表明,改进方法可实时流畅地提取感兴趣区域轮廓,鲁棒性强,对狭窄连通区域和外表面轮廓也有很好的提取效果.