Automatic Center Correction for Crystal Efficiency Normalization using Line Source

Introduction: PET is a clinically valuable tool for cancer diagnosis and staging and crystal efficiency normalization (a.k.a, normalization) plays a critical role in quantitative PET imaging accuracy. Since normalization is periodically performed on clinical PET scanners, a normalization technique that requires straightforward preparations would be very helpful, even more so for longer axial FOV scanners. In our previous work, we proposed a normalization method using a line source. The line source normalization method is based on a well centered line source but in practice, it is difficult to perfectly center the line source. In this work, we propose an automatic center correction (ACC) method for the line source normalization. With this method, the line source normalization approach will be much more practical to implement in clinical scanners.

Methods: In the line source normalization, the overall efficiencies of the detectors are decoupled into two parts:  1. The relative efficiencies of detectors transaxially within a ring are determined by the singles counts recorded at each crystal 2. The relative axial efficiencies of detectors in different axial rings are determined by the paired counts recorded at each ring. If the line source is not well centered, the singles counts distribution along a ring will be biased. Crystals closer to the source will have more counts, and vice versa. The normalization result will also be biased due to the line source being off center, using the directly measured singles counts.  In order to remove the bias in the singles counts distribution, a transaxial moving average of each ring on the singles events is calculated. The final singles counts distribution is then the measured singles counts over the moving average results. In experiment, two datasets were acquired for line source normalization from a Canon Cartesion TOF PET/CT scanner: in the first dataset the line source was purposely placed in the center of the scanner, and in the second one it was off-center by 5mm. We used the two acquisitions to generate normalization results using the line source normalization method, both with and without ACC.

Results: In the experiment we can see that even in the acquisition where the line source was meant to be centered, it is not perfectly centered. From the line source normalization results without ACC, we can see the nonuniformities in the crystal efficiency maps. These nonuniformities were caused by the line source being off center, both in the centered and off centered acquisitions. The nonuniformity in the off centered source case is more obvious as expected. When ACC is used, both of the acquisitions can produce uniform crystal efficiency maps. The crystal efficiency maps with centering correction from the two acquisitions are close, and they both are similar to the crystal efficiency map generated by the cylinder phantom. The reconstructed images are similar from using both line source normalization and cylinder source normalization. There are no artifacts resulting from the use of line source normalization with ACC. The quantitative evaluations measured by contrast recoveries of the hot spheres in the phantom show that line source normalization with ACC can produce very accurate quantitation compared to cylinder phantom normalization.

Conclusions: In this work, in order to the make the line source normalization approach less sensitive to line source positioning, we proposed an automatic centering correction method. With this method, the line source normalization was able to generate uniform crystal efficiency maps even when the line source was purposely placed 5mm off center. Experimental results show that line source normalization with centering correction can generate accurate reconstructions. Therefore, with the help of centering correction, line source normalization is much more robust to positioning in experiments, and is therefore more practical for clinical implementation.

• Download figure
• Open in new tab
• Download powerpoint