A new evaluation method for medical image segmentation
Zhen-Sen Yang · 2008
An improved evaluation method for medical image segmentation is presented,by analyzing several widely used evaluation methods.The new method can not only reflect how much the segmentation result is deviated from the object,but also indicate the degree of fluctuation near the contour.Furthermore,this method can be used for both parallel comparison,comparing the results of different algorithms operating on the same object,and longitudinal comparison,comparing the results of the same algorithm operating on different objects.The experiment results show that the proposed evaluation method has good clinical value.