基于水平集人机交互模型的医学图像分割

AN INTERACTIVE MODEL FOR SEGMENTATION OF MEDICAL IMAGE

医学图像中要分割的对象常比较复杂,通常的图像自动分割方法很难得到理想的结果.本文在曲线演化图像分割方法的基础上,提出了一种基于水平集方法的人机交互模型.该模型不仅继承了水平集方法对拓扑变化的自适应性,而且还有良好的人机交互性能.在分割过程中,医生只要在图像的适当位置上加入少许几个标记点,本方法就可以在医生的监督指导下对复杂的对象进行准确的分割.实验表明,本文的交互模型具有良好的实用性.

In the medical image processing, the desired segmentation object is often very complicated. It is difficult to acquire a satisfied result using a fully automatic image segmentation method. The segmentation guided by experienced doctors is the only possible solution to have reasonable results. In the paper, an interactive model is presented based on the level set method, and combine it with geodesic active region segmentation method. The model inherits the ability of topology adaptability of the level set method. Furthermore, doctors are only needed to put few mark points on the suitable image positions, and then able to monitor the segmentation results. Experiments show that the model is not only practical, but also reliable.