Procedure of tensor-based morphometry using brain SPECT. First, an original SPECT image (A) is linearly transformed to the Talairach space. Subsequent non-linear spatial normalization using the high-dimension-warping algorithm deforms a linearly standardized brain (B) to match it to a 99mTc-ECD template (C). This step generates a high-dimensionally warped SPECT image (D) and a parametric image of Jacobian determinants (E) indicating a local volume change relative to the template brain. Then the natural logarithm of this parametric image (F) is separated into negative log J (G) and positive log J maps (H) representing contraction and expansion respectively. A conventionally warped SPECT image is also shown (I).