一种基于有偏场的医学图像自动分割算法

临床应用对医学图像分割的准确度和算法的速度要求较高,但是由于图像本身受噪声、偏移场效应等的影响,使得分割算法很难达到理想的效果。人工分割方法可在原始图像上直接画出期望的边界,然而费时费力,分割结果完全依赖于分割者的解剖知识和经验且分割结果难以再现。针对磁共振(MR)图像存在偏移场的问题,以及避免人为参与,将一种基于有偏场的适配聚类算法与活动形状模型算法(ASM)相结合,提出了一种将基于MR图像的自动分割方法。该算法能够校正扫描间和扫描内部的灰度不均匀性,避免了图像不同程度地丢失原图像的信息,并达到全自动分割图像数据,该方法较传统算法更精确和稳健,可在医学图像处理和分析中发挥作用。

A medical image auto-segmentation algorithm in the presence of intensity inhomogeneities

Accuracy and speed of medical image segmentation algorithm are technical requirement for clinical application.However,the medical images are always affected by noise,intensity inhomogeneities,etc,which make the segmentation algorithm is difficult to achieve the desired result.Manual segmentation can directly draw the expected boundaries on the original image,however,segmentation result is entirely depended on the anatomical knowledge and experience of dissector and the segmentation results are difficult to ...