Tanizaki N, Okamura T, Senda M, Toyama H, Ishii K
Research and Development Center, Sumitomo Heavy Industries, Ltd.
Kaku Igaku. 1994 Oct;31(10):1157-66.
We have developed a semi-automatic ROI setting system for brain PET images. It is based on the elastic network model that fits the standard ROI atlas into individual brain image. The standard ROI atlas is a set of segments that represent each anatomical region. For transformation, the operator needs to set only three kinds of district anatomical features: manually determined midsagittal line, brain contour line determined with SNAKES algorithm semi-automatically, a few manually determined specific ROIs to be used for exact transformation. Improvement of the operation time and the inter-operator variance were demonstrated in the experiment by comparing with the conventional manual ROI setting. The operation time was reduced to 50% in almost all cases. And the inter-operator variance was reduced to one seventh in the maximum case.
我们开发了一种用于脑PET图像的半自动感兴趣区域（ROI）设置系统。它基于弹性网络模型，该模型将标准ROI图谱适配到个体脑图像中。标准ROI图谱是一组代表每个解剖区域的分割。对于变换，操作人员仅需设置三种区域解剖特征：手动确定的正中矢状线、使用蛇形算法半自动确定的脑轮廓线、用于精确变换的少数手动确定的特定ROI。通过与传统手动ROI设置进行比较，实验证明了操作时间和操作人员间差异有所改善。几乎在所有情况下，操作时间都减少到了50%。在最大的情况下，操作人员间差异减少到了七分之一。
