Simulation and reconstruction of PET images: application to attenuation correction using CT images
(2013) Belgian Hospital Physicists Association — Location: Mechelen, Belgium (1.February.2013)
Files
No attached file found for this publication.
Details
- Authors
- Abstract
- Introduction PET reconstruction algorithms process the information acquired in PET systems to provide an image of the radiotracer distribution into the patient. In this work we want to evaluate STIR (Software for Tomographic Image Reconstruction), which would be a good alternative to commercial systems for image reconstruction. STIR is an open source library in C++ that implements several tools for PET image reconstruction. To obtain a good quality image, it is necessary to apply several corrections to the data before reconstruction. In this study, we mainly focused on the attenuation correction using CT images, as a replacement of the correction initially available on our preclinical PET system. This method better estimates the attenuation of photons that pass through dense materials like bones, leading to significantly improved image quality. Finally, Monte Carlo simulations of our PET system allowed us to estimate some parameters that are not easily measurable. Materials and methods In this study, we concentrated on the Philips Mosaic PET camera, designed for small animal imaging. To reconstruct the image, raw PET data are first converted with a Matlab script and then corrected with tools provided by the STIR library. STIR supplies several methods for PET image reconstruction. We chose the ML-EM iterative algorithm. In contrast to analytical algorithms, iterative algorithms allow for a better modelling of the problem at the cost of longer computation times. Acquired data are corrected for several unwanted effects that may cause artifacts on the image. We focused on the attenuation correction. Photons that pass through denser materials have a higher probability of being scattered, which leads to an underestimation of the radiotracer distribution in some areas. In order to correct for this attenuation effect, it is necessary to get a map of the attenuation coefficients in the field of view of the PET camera. In the Mosaic system, this information is obtained by recording a transmission image with an external Cs-137 source that rotates around the field of view. The PET detectors are used to measure the transmitted photons. However, this method suffers from several disadvantages, including low spatial resolution and image artifacts. This motivated us to develop tools for PET attenuation correction using CT images. To do that, the CT data must first be processed with a Matlab script to convert Hounsfield units (HU) in attenuation coefficients. This conversion is based on the model described in [1]. When HU is less than zero, we use the “water-air assumption”. The material is considered as a uniform mixture of water and air. Otherwise, we use the “water-bone assumption”. Finally, we used GATE and GEANT4 to model and simulate the Mosaic PET camera. This Monte Carlo simulation allowed us to estimate some parameters that are not easily measurable experimentally, e.g. the effective radius of detection which takes into account the average depth of interaction of photons in the scintillator crystals. This parameter is necessary for the image reconstruction. Results The following results are obtained from the PET acquisition of a mouse in the Mosaic PET camera. If we compare the PET images corrected for attenuation with the two methods described above (Mosaic transmission imaging and CT imaging), we observe differences between the two activity profiles. For instance, correction with CT imaging allows a better estimation of the attenuation of photons that pass through the skull or spine. Monte Carlo simulations allow us to measure the depth of interaction of photons into the scintillator crystals of PET detector. Figure 2 shows the depth distribution of photon interactions. Taking the average depth of interaction, we can calculate the effective radius of our PET camera detectors. Reff = Rdet + Dint = 105 mm + 4,37 mm Conclusion Available open source software allows for image reconstruction and Monte Carlo simulation as an alternative to commercial systems. This software can be used to better understand the characteristics of an imaging system and to improve the quality of the reconstructed images. We know that the raw data must be corrected to obtain an exploitable PET image. Such corrections can be easily implemented in this software as we have shown by improving the attenuation correction using CT images. Some complex and expensive experiments can be easily modelled and simulated with Monte Carlo tools like the GATE/GEANT4 software. For instance, we have measured the depth of interaction of photons in the PET detectors. References [1] Chuanyong Bai et al. A generalized model for the conversion from CT numbers to linear attenuation coefficients. IEEE Transactions on Nuclear Science, 50, 2003. [2] Charbel Merheb et al. Full modelling of the mosaic animal PET system based on the GATE monte carlo simulation code. Physics in Medicine and Biology, 52, 2007. [3] Kris Thielemans et al. STIR : software for tomographic image reconstruction release 2. Physics in Medicine and Biology, 57, 2012.
- Affiliations
Citations
Souris, K., Bol, A., & Lee, J. (2013). Simulation and reconstruction of PET images: application to attenuation correction using CT images. Belgian Hospital Physicists Association, Mechelen, Belgium. https://hdl.handle.net/2078.5/218918