PCA-based filtering for hypothesis testing on hippocampus shape differences between schizophrenics and normals

Statistical shape analysis has become of increasing interest to the neuroimaging community due to its potential to precisely locate morphological changes and thus potentially discriminate between healthy and pathological structures. A common shape description is based on the sampled boundary defined by a spherical harmonic SPHARM description. Shape analysis is computed by means of hypothesis testing of differences from an average template structure using multivariate group mean difference tests. The presence of noise in the landmark location or correspondence may obscure the statistical significance at certain locations of the structure. This work shows that statistical significance is improved when hypothesis testing is done on reconstructed shapes by means of PCA. Performance evaluation was done on a hippocampus study for schizophrenic patients, confirming deformations previously found in the head and tail regions of the right hippocampus.