Investigation of Attenuation Correction Methods for Dual-Gated Single Photon Emission Computed Tomography (DG-SPECT).

BACKGROUND: Cardiac-respiratory dual gating in SPECT (DG-SPECT) is an emergent technique for alleviating motion blurring artifacts in myocardial perfusion imaging (MPI) due to both cardiac and respiratory motions. Moreover, the attenuation artifact may arise from the spatial mismatch between the sequential SPECT and CT attenuation scans due to the dual gating of SPECT data and non-gating CT images. OBJECTIVES: This study adapts a four-dimensional (4D) cardiac SPECT reconstruction with post-reconstruction respiratory motion correction (4D-RMC) for dual-gated SPECT. In theory, a respiratory motion-matched attenuation correction (MAC) method is expected to yield more accurate reconstruction results than the conventional motion-averaged attenuation correction (AAC) method. However, its potential benefit is not clear in the presence of practical imaging artifacts in DG-SPECT. In this study, we aim to quantitatively investigate these two attenuation methods for SPECT MPI: 4D-RMC (MAC) and 4D-RMC (AAC). METHODS: DG-SPECT imaging (eight cardiac gates and eight respiratory gates) of the NCAT phantom was simulated using SIMIND Monte Carlo simulation, with a lesion (20% reduction in uptake) introduced at four different locations of the left ventricular wall: anterior, lateral, septal, and inferior. For each respiratory gate, a joint cardiac motion-compensated 4D reconstruction was used. Then, the respiratory motion was estimated for post-reconstruction respiratory motion-compensated smoothing for all respiratory gates. The attenuation map averaged over eight respiratory gates was used for each respiratory gate in 4D-RMC (AAC) and the matched attenuation map was used for each respiratory gate in 4D-RMC (MAC). The relative root mean squared error (RMSE), structural similarity index measurement (SSIM), and a Channelized Hotelling Observer (CHO) study were employed to quantitatively evaluate different reconstruction and attenuation correction strategies. RESULTS: Our results show that the 4D-RMC (MAC) method improves the average relative RMSE by as high as 5.42% and the average SSIM value by as high as 1.28% compared to the 4D-RMC (AAC) method. Compared to traditional 4D reconstruction without RMC ("4D (MAC)"), these metrics were improved by as high as 11.23% and 27.96%, respectively. The 4D-RMC methods outperformed 4D (without RMC) on the CHO study with the largest improvement for the anterior lesion. However, the image intensity profiles, the CHO assessment, and reconstruction images are very similar between 4D-RMC (MAC) and 4D-RMC (AAC). CONCLUSIONS: Our results indicate that the improvement of 4D-RMC (MAC) over 4D-RMC (AAC) is marginal in terms of lesion detectability and visual quality, which may be attributed to the simple NCAT phantom simulation, but otherwise suggest that AAC may be sufficient for clinical use. However, further evaluation of the MAC technique using more physiologically realistic digital phantoms that incorporate diverse patient anatomies and irregular respiratory motion is warranted to determine its potential clinical advantages for specific patient populations undergoing dual-gated SPECT myocardial perfusion imaging.