In this paper, we present the principal component analysis (PCA) of shape deformations of bilateral hippocampi in Alzheimer's disease (AD) as derived in the large deformation diffeomorphic metric mapping setting. We investigated the PCA patterns (the scores and loadings) of the bilateral hippocampi for 51 subjects, 28 of which had AD while 23 were normal aging. Student's t-tests were used to select the components displaying significant group difference for a more purposed analysis. Our findings revealed that the head part of the hippocampus in each hemisphere, to be specific the CA1 and subiculum subregions, contributed the most to the significant group differences. To further our analysis, we examined the classification accuracy yielded by these shape deformation patterns when either solely PCA or PCA followed by a Student's t-test was utilized as the approach to dimension reduction. Linear discriminant analysis (LDA) and support vector machine (SVM) were used as candidates for the classification technique. According to our leave-one-out cross-validation experiments, SVM had a much higher accuracy than LDA, with the best performance (overall accuracy: 94.1% (48/51); sensitivity: 92.9% (26/28); and specificity: 95.7% (22/23)) achieved by SVM when using both the left and the right hippocampal shape deformation patterns and employing solely PCA for dimension reduction.