Table 1: Summary of pros and cons of supervised, unsupervised, and registration based segmentation.

Supervised segmentation [2]Unsupervised segmentation [3]Registration based segmentation [4]

ProsCan perform different task simply by changing the training set data. Learning is an essential component of this technique and when learning is completed then tumor will be detected automaticallyNo training requiredEasy to make the comparison of two images

ConsSpecial training required and, therefore, additional processing time requiredManual selection of region growing. Number of regions needs to be prespecified. Tumors have not clearly defined intensitiesRegistration may not be perfect due to different anatomical structure of the image and template