Abstract: This paper presented a threshold-based segmentation framework for 3D medical volumetric images, in which two classical image segmentation algorithms, OSTU and Gradient based Segmentation, were re-designed and optimized to be suitable for 3D volumetric images. In order to evaluate the framework's performance, we defined two novel ctuanti- tative performance indicators: Segmentation Accuracy and Segmentation Balance, and use them to evaluate and analyze their performances. Experimental results show that both of the proposed segmentation algorithms can yield satisfied segmentation results for 3D medical volumetric images, and OShU segmentation achieves better performance than Gra- dicnt-based segmentation.

Key words: 3D medical image segmentation, Volumetric data, Medical image processing