Encrypted login | home

Program Information

Development of GPU-Based Fast Reconstruction Algorithm for Gamma Ray Imaging with Insufficient Conditions

no image available
M Kim

M Kim1*, J Jung2 , H Shin3 , S Kim4 , D Yoon5 , T Suh6 , (1) Catholic University of Korea, Seoul, Seoul, (2) University of Florida, Gainesville, FL, (3) Catholic University of Korea, Seoul, Seoul, (4) Catholic University of Korea, Seoul, Seoul, (5) Catholic University of Korea, Seoul, Seoul, (6) Catholic Univ Medical College, Seoul, SEOUL


TU-L-GePD-IT-5 (Tuesday, August 1, 2017) 1:15 PM - 1:45 PM Room: Imaging ePoster Theater

Purpose: The purpose of this study is to develop a graphic processing unit (GPU)-based fast reconstruction algorithm for nuclear medicine image under insufficient conditions, and verification of the developed algorithm is carried out to achieve the purpose.

Methods: Simple-pattern water phantoms containing isotopes were designed using Monte Carlo simulation. Then, single-photon emission computed tomography (SPECT) and positron emission tomography (PET) scanning processes were simulated to acquire the projection data. The image reconstruction was performed using the GPU-based algorithm. After the image acquisition, to verify the performance of the algorithm, an analysis of the image profile, signal-to-noise ratio (SNR), and reconstruction time was performed by comparing the reconstructed images under different conditions.

Results: The image reconstruction times for SPECT and PET using the GPU were 449 and 811 times faster than those using the central processing unit (CPU), respectively. The contrast and SNR results in three radioisotope uptake regions of the image using the GPU-based fast iterative reconstruction algorithm were clearly better than those using the CPU-based filtered back projection algorithm.

Conclusion: We confirmed the good performance of the developed GPU-based algorithm in that image reconstruction can be conducted in significantly short time and with relatively good image quality, compared with that using a conventional CPU-based algorithm under the same conditions.

Contact Email: