Table 1. Application of GPU-based medical image segmentation.

Type	Approach	Application	Characteristics
2D CT MRI	Active Contour model (43)	Brain	Far from perfect for practice medical images because of the segmentation in only two regions.
3D CT	Active learning (18)	Pelvis	Reduce the required user input in interactive 3D image segmentation tasks.
3D CT MRI	Level set (44)	Kidneys, brain	The first and only GPU level set segmentation algorithm with linear work - complexity and logarithmic step-complexity.
	Point radiation technique (45)	Brain	Create high-quality real-time feedback of the segmented regions
3D MRI	Swarm-based level set (46)	Brain	The swarm-based level set is in the robustness to a noisy environment.
	Hybrid method (32)	Brain	An interactive hybrid segmentation technique which combines threshold-based and diffusion-based region growing.
	Seeded Region Growing (42)	Brain, skull	Easily extended to a number of applications including other point based systems, polygonal meshes, and irregular volume with changing topology
	Cellular automaton (47)	Kidney	Simple, efficient and straightforward.
	Level set (48)	Brain tumor	Interactivity enables users to produce reliable segmentation. Limitations are mostly in the speed function and the interface.
X-ray	Active Shape Model (49)	Vertebra	The initialization of the model is accomplished by the edge detection and the edge polygonal approximation.