Correction of Respiratory Motion in Dual Gated Cardiac Imaging in PET/CT

Abstract

Cardiac imaging suffers from both respiratory and cardiac motion. One of the proposed solutions involves double gated acquisitions. Although such an approach may lead to both respiratory and cardiac motion compensation there are issues associated with (a). the combination of data from cardiac and respiratory motion bins, and (b). poor quality images as a result of using only part of the acquired data. The objectives of this work were to evaluate different fashions of combining binned data in order to maximise the acquired data used in the reconstruction of motion free cardiac images. A physical phantom study as well as a human study were used in this evaluation. PET data was acquired in list mode (LM) using a GE Discovery VCT PET/CT system. An RPM and an ECG device were used to provide the respiratory and cardiac motion triggers registered within the LM file. Subsequently data were binned considering four cardiac gates (gates 1 to 3 in the systole and gate 4 in the diastole) in combination with three respiratory gates in amplitude. Each of the respiratory gates contained either the 3 cardiac bins over systole (representing 50% of the cardiac cycle) or bin 4 over the diastole (representing the other 50% of the cardiac cycle). Reconstructed images from each of the respiratory bins considered above were subsequently used in combination with an affine and an elastic registration algorithm to derive transformation parameters allowing the combination of all acquired data in a particular position in the respiratory cycle. Images were assessed in terms of signal to noise ratio and contrast. Superior SNR and contrast were seen in the combined images using the affine respiratory motion model and the diastole cardiac bin in comparison to the use of the systole bins. No such dependence on the cardiac gates included in the respiratory bins was seen in the case of the corrected images derived using the elastic respiratory motion model. Three amplitude respiratory bins seem to be adequate for both motion models used. Respiration corrected cardiac images acquired over the diastole phase lead to reduced motion artefacts as well as improved SNR and contrast. When images are corrected for both cardiac and respiratory movements considering the 12 bins independently, the elastic model performs better than the affine model, leading to images with a higher SNR and a better repositioning of the heart walls.