An interactive approach of 3D medical images segmentation was improved in this paper.Firstly

based on the Otsu′ single threshold segmentation method

a double-threshold segmentation method founded on class variance was proposed.It analyzed the histogram of volume data of 3D medical images

and then got two thresholds automatically that the result of segmentation had max variance between classes.This algorithm applying an iterative procedure was simple and fast

and ensured that the result included the goal tissue.Secondly

the 3D medical images segmented were processed by mathematical morphology operation and 3D region growing.Finally

the exact 3D segmentation and 3D visualization of goal tissue were achieved.The results of experiment show that the algorithm is feasible

and 3D reconstruction meets the goal.