Abstract
670
Objectives: Respiratory gated acquisitions lead to poor quality images since only part of the available data is used. The objectives of this work were to develop and evaluate, using simulated and clinical datasets, a reconstruction algorithm combining all of the available respiratory synchronised data. Methods: In this work we present an algorithm that integrates the spatial deformation approach into the classical EM reconstruction framework. The deformation is incorporated in the reconstruction process by applying the transformations directly on the system matrix during the forward-projection and backprojection steps. The elastic deformations applied are calculated for each of the 4D CT or 4D PET respiratory gated frames using a spatiotemporal registration algorithm based on B-splines previously validated for use with 4D PET/CT images1. Different implementations were tested in order to evaluate the effect of accounting or not in the reconstruction process for deformed voxel shapes as a result of considering elastic transformations. The algorithm has been tested using simulated 4D NCAT PET/CT datasets including lesions (7 to 22mm) in the lung fields, while clinical validation was performed using three 4D PET/CT patient list mode datasets. Results: The incorporation of continuous B-spline interpolations into the system matrix eliminated artefacts present around the diaphragm or the lesion where large respiratory motion was encountered. An improved contrast was seen on the respiratory corrected images (simulated and clinical) in comparison to individual gated images. In addition, NCAT lesions’ size and location accuracy was significantly improved irrespective of their location in the lung field. Conclusions: An iterative reconstruction algorithm incorporating respiratory motion compensation without increasing the overall time of reconstruction was developed and validated using simulated and clinical 4D PET/CT datasets. 1. D Visvikis et al, J Nucl Med 2006, 47(5), 234P
- Society of Nuclear Medicine, Inc.