A posteriori respiratory motion gating of dynamic PET images

The presence of patient physiological motion during imaging may cause significant artifacts in image quality. Proposed correction methodologies involve the use of gated acquisitions through simultaneous recording of an external signal. The purpose of our work is to determine the feasibility of post-acquisition synchronization of dynamically acquired PET images in the absence of any external signal. The principle of the technique is based on the assumption that although the amplitude of the motion may vary from pixel to pixel inside the same organ, the frequency of the periodic motion is the same. Under such conditions, the prerequisite for a posteriori gating is the ability to accurately estimate that frequency. We performed simulation studies using the NCAT phantom and a Monte Carlo simulation of the GE Advance PET system (3D mode of operation). A number of NCAT emission and the corresponding transmission frames were generated throughout a respiratory cycle. Time frames of 0.15, 0.45 and 0.62 seconds were simulated with variable count statistics (namely 30k, 70k and 120k of total simulated coincidences). Time activity curves were obtained, for each of the dynamic series formed, using different ROIs and Fourier transform was performed in order to estimate the frequency of the simulated motion. We were able to determine the frequency of motion (within 2% of the simulated frequency) for all three frame time durations evaluated. Using the estimated frequency we were able to calculate on a pixel by pixel basis the amplitude and the phase of the motion, allowing us to reconstruct an a posteriori gated time series.