Aim: The IRIS PET/CT scanner (now distributed by Inviscan s.a.s., France) is a novel multimodal preclinical tomograph for high resolution PET/CT imaging of small animals. The PET component of the scanner features the unique possibility of performing both static and“step and shoot” rotational acquisitions. This latter modality allows the user to obtain quantitative high-quality images. However, in case of rotational scans, the image reconstruction becomes a very time-consuming task, thus hindering the use of this modality in clinical practice. To overcome this limit, we implemented a dedicated GPU CUDA-based 3D iterative reconstruction algorithm. In this work we present an assessment of the quantitative performance of both the static and rotational modalities obtained with the GPU algorithm. Materials and methods: The PET component of the scanner consists of 16 modular detectors arranged in two octagonal rings. Each detector module comprises a matrix of 702 LYSO:Ce crystals of 1.6 mm×1.6 mm×12 mm with a pitch of about 1.7mm. The scanner features a total of 24 million of Lines Of Responses (LOR). The standard 3D quantitative image reconstruction of the IRIS/PET is implemented with a dedicated multi-core Maximum Likelihood Estimation Maximization (MLEM) iterative algorithm. The system response matrix is pre-calculated using a multi-ray “Siddon” algorithm and stored on file. The size of the system matrix is reduced using symmetries.As image reconstruction exhibits a high degree of parallelism, to improve the performance of the standard reconstruction software we ported it on a Graphics Processing Unit (GPU) using the NVIDIA CUDA programming model. The GPU reconstruction implements an “on-the-fly”calculation of the system model. In order to perform a quantitative evaluation of the different modalities, four NEMA quality phantoms were acquired: one was acquired with a stationary scan, while the others were acquired in “step and shoot” mode using 2,4,8 views spanning over 45 degrees. All the acquisitions lasted 20 minutes. The image quality of 18F-FDG imaging on mice and rats was also evaluated. Results and Conclusions: We compared the results obtained with the GPU algorithm to those obtained with the reconstruction software included with the IRIS PET, obtaining no significant difference. Moreover, we achieved speedups in the reconstruction times up to 5x with respect to the original implementation.

Performance evaluation of the step and shoot acquisition modality of the IRIS PET scanner

CAMARLINGHI, NICCOLO';LUONGO, CARMELA;BELCARI, NICOLA;SPORTELLI, GIANCARLO;DEL GUERRA, ALBERTO
2016-01-01

Abstract

Aim: The IRIS PET/CT scanner (now distributed by Inviscan s.a.s., France) is a novel multimodal preclinical tomograph for high resolution PET/CT imaging of small animals. The PET component of the scanner features the unique possibility of performing both static and“step and shoot” rotational acquisitions. This latter modality allows the user to obtain quantitative high-quality images. However, in case of rotational scans, the image reconstruction becomes a very time-consuming task, thus hindering the use of this modality in clinical practice. To overcome this limit, we implemented a dedicated GPU CUDA-based 3D iterative reconstruction algorithm. In this work we present an assessment of the quantitative performance of both the static and rotational modalities obtained with the GPU algorithm. Materials and methods: The PET component of the scanner consists of 16 modular detectors arranged in two octagonal rings. Each detector module comprises a matrix of 702 LYSO:Ce crystals of 1.6 mm×1.6 mm×12 mm with a pitch of about 1.7mm. The scanner features a total of 24 million of Lines Of Responses (LOR). The standard 3D quantitative image reconstruction of the IRIS/PET is implemented with a dedicated multi-core Maximum Likelihood Estimation Maximization (MLEM) iterative algorithm. The system response matrix is pre-calculated using a multi-ray “Siddon” algorithm and stored on file. The size of the system matrix is reduced using symmetries.As image reconstruction exhibits a high degree of parallelism, to improve the performance of the standard reconstruction software we ported it on a Graphics Processing Unit (GPU) using the NVIDIA CUDA programming model. The GPU reconstruction implements an “on-the-fly”calculation of the system model. In order to perform a quantitative evaluation of the different modalities, four NEMA quality phantoms were acquired: one was acquired with a stationary scan, while the others were acquired in “step and shoot” mode using 2,4,8 views spanning over 45 degrees. All the acquisitions lasted 20 minutes. The image quality of 18F-FDG imaging on mice and rats was also evaluated. Results and Conclusions: We compared the results obtained with the GPU algorithm to those obtained with the reconstruction software included with the IRIS PET, obtaining no significant difference. Moreover, we achieved speedups in the reconstruction times up to 5x with respect to the original implementation.
2016
https://link.springer.com/content/pdf/10.1007/s00259-016-3484-4.pdf
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11568/850365
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? 0
social impact