Respiratory gating is a helpful technique to reduce motion blur in PET images of the lungs. Efforts still need to be made on reconstruction schemes adapted to this kind of data, since gated data are usually of low statistical quality. A 4D jointestimation image reconstruction method is tested and compared to classic OSEM algorithms on list-mode data acquired with a Philips Gemini GXL PET-CT scanner, using the NEMA IEC body phantom. The Varian RPM system was used to synchronize the 20-minute PET acquisition with the respiration signal. The list-mode events were sorted into eight gates. Four reconstruction schemes were compared: 4D, independent-frame OSEM, independent-frame OSEM followed by temporal filtering between the 8 gates and non-gated acquisitions reconstructed with OSEM. Results were evaluated in terms of signal-to-noise ratio (SNR) and contrast as a function of iteration number for every reconstruction algorithm. The SNR was evaluated from ten different 2-minute intervals of the original list-mode data. Contrast was calculated as the ratio of the background activity to the activity in the cold cylinder inside the phantom. The influence of the number of basis functions used in the 4D reconstruction algorithm was studied. Results show that in the gated images, the SNR is 2.3 times greater when using 4D reconstruction compared to independent reconstruction. 4D reconstruction helps recover the same SNR as non-gated acquisition. Temporal filtering of independent-frame reconstructed images yielded similar contrast -SNR trade-off as 4D reconstruction only in gated frames with little residual movement, but not in the gated frames still including substantial movement. It was also checked that using 4D reconstruction affects only slightly the motion deblurring effect of respiratory gating.
展开▼
