Data-driven dual-gating for cardiac PET

Image quality in cardiac PET is impaired by both respiratory and cardiac motion. Dual-gating has been proposed as a means to obtain sinograms in various motion states. This data can then be used to produce dual-gated images or as a first step in processing that corrects for one or both types of motion. The gating is normally performed based on external signals such as an ECG signal and a camera or belt for monitoring respiration. However, it could be advantageous if the gating could be performed without ancillary equipment. It has been shown by various authors that it is possible to extract a respiratory signal out of the PET data itself, usually based on low spatial resolution but high time resolution sinograms. Here, we propose an extension to extract the cardiac beating signal as well. We illustrate this using a recently proposed technique that uses PCA to extract multiple signals from the data and show that the cardiac signal can be obtained in a two stage process. The method does not need any image reconstruction and is therefore very fast. We show results on PET patient data using various tracers. We compare the signals extracted using the data-driven method with those obtained with external monitors (including ECG). Our results indicate that in many clinical situations, the data-driven method can achieve equivalent dual-gated images compared to methods that use ancillary equipment.