Abstract
15
Objectives: Respiratory motion is a challenge in cardiac PET imaging, causing blurring, artifacts, and inaccuracies in clinical diagnosis. Many current studies require external devices to track respiratory motion before applying gating and correcting approaches to compensate for the motion. Data-driven gating methods provide an appealing alternative as they extract the respiratory signal directly from the acquired PET data and lead to comparable performance on respiratory motion correction with hardware gating methods in static cardiac PET imaging. In this study, we propose a data-driven respiratory motion correction method in dynamic cardiac PET imaging and evaluate its impact on image reconstruction and quantitation of kinetic parameters with clinical datasets.
Methods: The data-driven respiratory motion compensation consists of two steps: respiratory gating and motion correction. Respiratory motion signal is first extracted by computing the center-of-mass of all radioactive events in the heart region of the list mode data along with time, with a resolution of 100 msec. The respiratory gating is then performed by dividing the list mode data into 5 equal-amplitude respiratory phases based on the obtained motion signal. We reconstruct the list mode data in each static respiratory gate and estimate the motion vector fields (MVFs) between each gate and the end-expiration gate, which serves as a reference. Finally, we apply a 4D motion compensated image reconstruction algorithm in every respiratory gated dynamic frame, incorporating the estimated MVFs to result in motion-free dynamic images. Two Rb-82 PET imaging datasets were used to evaluate the proposed method, one is of a healthy control and the other is from a patient with the diagnosis of normal perfusion and normal LV function. The data were acquired in list mode for 7 min from the time of injection on a Siemens Biograph mCT PET/CT scanner. Eight dynamic image frames, with and without respiratory motion correction, were reconstructed from the first 2 min with equal durations (15 sec each), with normalization, attenuation, randoms, scatter, and gamma correction as applied clinically. The MCIR algorithm in the STIR package was used to reconstruct the motion-corrected images with the MVFs estimated using the ITK software. For comparison, the image without motion correction of each dynamic frame was reconstructed by the OSEM algorithm in the STIR. To evaluate the effectiveness of the proposed method, we used the quantitative dynamic analysis software Corridor4DM to analyze the reconstructed images. The kinetic parameters were estimated using the one compartment kinetic modeling according to Lortie et al. (2007). The K1 values were calculated on the standard parametric 5-segment polar map corresponding to the anterior, lateral, inferior, septal and apex walls. Results: The data-driven respiratory gating and correction method results in a visually noticeable reduction in respiratory blurring of the reconstructed image frames, especially in the inferior and apex regions of the hearts. The corresponding dynamic uptake polar maps with motion correction appear more uniform than their counterparts without motion correction. The mean and standard deviation values of the estimated K1 from the 5 polar map segments with and without motion correction are 1.38+/-0.10 and 1.10+/-0.15 for the volunteer, and 1.65+/-0.09 and 1.48+/-0.17 for the patient, respectively. Conclusions: We have demonstrated that the data-driven respiratory motion compensation method for dynamic cardiac PET imaging reduces the respiratory blurring and improves the uniformity of reconstructed polar map frames. The estimated K1 values with motion correction were shown to be more uniform over the polar map regions than those without motion correction in both a volunteer and a normal patient case. The proposed method has shown its promise for improving accuracy in clinical diagnosis.