Here we investigate how regions in the Medial Temporal Lobe(MTL) in a dataset consisting of 13 di®erent people changes us- ing a Principal Component Analysis(PCA). The regions investigated are the Temporopolar, Parahippocampal, Entorhinal, Hippocampal, Perirhi- nal and Amygdalar regions. The MTL is located fairly deep in the brain where the contrast is quite low, and region-boundaries can be di±cult to ¯nd, which is why a shape guiding term would be helpful for a segmen- tation algorithm. An expert used an interactive tool to draw binary(1 inside and 0 out- side) Volumes Of Interests(VOI) for each of the 13 subjects. As the brain is symmetric, 12 VOIs has been drawn for each subject. A simultane- ous multi-shape rigid registration scheme, similar to the one used in [Tsai et al., 2004] was used on the training shapes to remove linearities. As these are binary shapes, a set-di®erence cost function is minimized be- tween all shapes. To represent shapes in the coupled shape model, signed distance maps(SDM) was used, [Tsai et al., 2004]. The eigen-problem was solved on the covariance matrix using svd, to ¯nd the eigenshapes and their magnitude. Seven modes of variation was extracted, represent- ing 75% of the total variance which each represents di®erent modes of variations. An interactive program was developed to investigate how the ¯rst seven modes changes the shapes. In ¯gures 1 to 3 the most signi¯cant mode is seen varying with §2¾ from the meanshape. Fig. 1. meanshape + 2¾ Fig. 2. meanshape Fig. 3. meanshape - 2¾ References [Tsai et al., 2004] Tsai, A.,Wells, W., Tempany, C., Grimson, E. og Willsky, A. (2004). Mutual information in coupled multi-shape model for medical image segmentation. Elsevier Science.