Skip to main content
. Author manuscript; available in PMC: 2016 Dec 1.
Published in final edited form as: Med Image Anal. 2015 Aug 28;26(1):82–91. doi: 10.1016/j.media.2015.08.010

Figure 7.

Figure 7

Summary of the simulation and results. The flowchart shows the framework of the simulation: (1) 3 images were deformed to 90 simulated images and converted to MNI space by affine registration. (2) 10 of them were used as atlases for multi-atlas segmentation while 80 of them were used as training data for the MLF framework. (3) 3 images were deformed to 27 testing images for comparing the Multi-Atlas segmentation, small-scale model and big data model. The results demonstrate (1) the performance of the MLF framework is significantly improved when using big data model (3464 training images) and (2) the MLF framework under big data model provides the better performance than majority vote and SS even without using non-local information.