Dynamic PET Reconstruction using the split Bregman formulation

In this work, we propose a novel 4D reconstruction method for PET which is based on spatiotemporal total variation (ST-TV). The ST-TV method takes advantage of image redundancies in 4D and was efficiently implemented using the split Bregman formulation, which has been shown to be optimal for decreasing noise while maintaining image quality. To evaluate the proposed approach we simulated data for a dynamic numerical phantom with different number of counts to mimic high and low Signal-to-Noise Ratio (SNR) scenarios.

Data were reconstructed with the proposed methodology and results were compared with the ones obtained with a reconstruction algorithm based only on spatial TV (i.e. without using the timing information), and with other well-known PET reconstruction methods (2D-OSEM and FBP algorithms). Our proposed 4D reconstruction led to major improvements in the recovery of time activity curves and image SNR while preserving spatial resolution. Furthermore, ST-TV also demonstrated to be less sensitive to noise in the data than any other method, suggesting that this approach holds the potential to provide significant improvements in image quality for dynamic PET.