Activity Number:
|
17
|
Type:
|
Contributed
|
Date/Time:
|
Sunday, August 11, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Biometrics Section*
|
Abstract - #301258 |
Title:
|
Temporal-Spatial Pattern Extraction in PET Data Using Partial Least Squares
|
Author(s):
|
Lei Xu*+ and Julie Price and Sati Mazumdar
|
Affiliation(s):
|
University of Pittsburgh and University of Pittsburgh and University of Pittsburgh
|
Address:
|
Graduate School of Public Health, Pittsburgh, Pennsylvania, 15261, USA
|
Keywords:
|
Partial Least Squares ; Positron Emission Tomography ; Temporal-Spatial Pattern
|
Abstract:
|
Partial Least Squares (PLS) method was implemented to extract temporal-spatial pattern of cerebral blood flow (CBF) changes caused by d-amphetamine (AMPH) administration. Six dynamic [O-15] water Positron Emission Tomography (PET) scans were acquired in five baboons (2 pre-AMPH and 4 post-AMPH). The summed PET images were co-registered to an MR atlas of baboon brain to remove anatomic differences between baboons. PLS is a multivariate method for finding orthogonal rotations on variables that re-arranges the correlation matrix between two blocks of variables X and Y. We let X be a matrix with one column for each voxel and one row for each scan; Y a set of orthogonal contrasts for the temporal effect. PLS computes the coefficients of orthogonal rotations for both X and Y to generate two blocks of latent variables that best relate the blocks with maximal covariance. Compared with the conventional analysis for PET data, viz., Region of Interest (ROI) analysis and Statistical Parametric Mapping (SPM), we found PLS serves as an enlightening extension that provides temporal-spatial pattern information that is directly extracted from image data as a whole.
|