Premium
Improved motion‐compensated image reconstruction for PET using sensitivity correction per respiratory gate and an approximate tube‐of‐response backprojector
Author(s) -
Dikaios Nikolaos,
Fryer Tim D
Publication year - 2011
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.3611041
Subject(s) - imaging phantom , iterative reconstruction , sensitivity (control systems) , algorithm , mathematics , interpolation (computer graphics) , normalization (sociology) , computer vision , artificial intelligence , computer science , physics , nuclear medicine , optics , motion (physics) , engineering , electronic engineering , medicine , sociology , anthropology
Purpose: One limitation of positron emission tomography (PET) imaging of the torso is patient motion. Motion‐compensated image reconstruction (MCIR) is one method employed to reduce the deleterious effects of motion. Existing MCIR algorithms use a single sensitivity correction term, which provides inexact normalization for multigate data. Consequently, in this study, the authors derive and examine the performance of an MCIR algorithm with sensitivity correction per gate. In addition, they demonstrate an approximate tube‐of‐response (TOR) backprojector. Methods: Simulated data from the NCAT phantom with six lesions added were used to compare MCIR algorithms with and without the incorporation of sensitivity correction per gate and TOR backprojection to postreconstruction registration (PRR) and images reconstructed without motion correction. To make the simulations more realistic, intragate motion was included. Deformation fields were determined from NCAT anatomical images using a free‐form deformation approach with bending energy regularization. Results: Sensitivity correction per gate and TOR backprojection improved mean lesion contrast‐to‐noise ratio by 6%–8%, with the maximum increase (21%–23%) found for the smallest lesion. These increases were obtained despite a small increase (3%) in noise as measured by standard deviation in a uniform lung region. Sensitivity correction per gate comes at no extra computational cost, whilst replacing line‐of‐response backprojection with TOR backprojection increased the overall computation time by ∼20%. In addition, MCIR was found to be superior to PRR, with one factor contributing to this difference being the differential impact of interpolation following deformation. MCIR was also shown to exhibit super‐resolution. Conclusions: Replacing a single sensitivity correction term in MCIR with sensitivity correction per gate improves lesion detectability. For a small increase in computational expense, further improvements are achieved using an approximate TOR backprojector rather than line‐of‐response backprojection.