Medical image segmentation plays a very key role in medical image processing.A segmentation algorithm that based on the topology correlation of regions which mainly depend on the adjacent regions and their degrees of discrimination's presented in the paper.The algorithm is designed by three parts:image iso-level regions marking,degrees of topology correlation discriminations calculation on these regions,marking homogeneous regions according with the degrees.Also,comparison with other region-based segmentation, numerical experiments are implemented and the results are proved the algorithm is better than others at visible perception.