Premium
Bootstrap method for nonlinear filtering of EM‐ML reconstructions of PET images
Author(s) -
Coakley Kevin J.
Publication year - 1996
Publication title -
international journal of imaging systems and technology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.359
H-Index - 47
eISSN - 1098-1098
pISSN - 0899-9457
DOI - 10.1002/(sici)1098-1098(199621)7:1<54::aid-ima7>3.0.co;2-t
Subject(s) - computer science , nonlinear system , artificial intelligence , computer vision , algorithm , physics , quantum mechanics
Reconstructions of positron emission tomography images are obtained with the iterative expectation maximization (EM) algorithm. The EM algorithm is halted according to a cross‐validation procedure. For the cases studied, this method yields a reconstruction with high variability about its expected value. The variability of the reconstruction about its expected value is reduced by computing its bootstrap expectation. Based on the reconstruction computed from the observed projection data, synthetic projection data sets are simulated. Reconstructions of the synthetic projection data sets are averaged to yield the bootstrap expectation. This bootstrap procedure is a nonlinear filtering method. The procedure is automatic; no smoothing kernel or bandwidth parameter need be specified. For simulated data, the bootstrap method yielded somewhat sharper reconstructions than did an optimized linear approach. The method is applied to real data from a fluorodeoxiglucose study of the human brain. Near the boundaries, the resampling procedure yielded a sharper reconstruction. © 1996 John Wiley & Sons, Inc.