|
Activity Number:
|
174
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, July 30, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #307839 |
|
Title:
|
A GCV Approach for Bandwidth Selection in PET Image Reconstruction
|
|
Author(s):
|
Ranjan Maitra*+
|
|
Companies:
|
Iowa State University
|
|
Address:
|
Department of Statistics, Ames, IA, 50011-1210,
|
|
Keywords:
|
generalized deconvolution ; smoothed least squares ; filtered backprojection ; circulant matrix ; Fourier Transforms ; BCCB matrices
|
|
Abstract:
|
The problem of bandwidth estimation for filtered backprojection (FBP) in Positron Emission Tomography (PET) image reconstruction is studied in the context of generalized deconvolution. A generalized cross-validation (GCV) approach is presented. Results on eigendecomposition of symmetric one-dimensional and two-dimensional circulant matrices are derived and computational implications of the above in making reconstruction bandwidth estimation a simple extension of standard FBP are developed. The methodology is studied using numerical simulations on a class of idealized one-dimensional generalized deconvolution problems and PET phantom data. In each case, performance evaluations are excellent and in real-time, proving the potential viability of this approach in a practical setting. This research is supported in part by (NSF) CAREER DMS #0437555 and the NIH DC-006740.
|
