Imaging Performance of a 1 mm³ Resolution Clinical PET System in the Presence of Out of FOV Activity from a Realistic Torso Phantom

Objectives High resolution PET offers an effective way to guide the clinical management of breast cancer alongside traditional imaging modalities. We are building a dual panel 1 mm³-resolution PET camera dedicated to breast imaging, which increases photon sensitivity by placing the PET panels close to the breast. This setup puts the imaging heads near regions with high uptake outside the FOV, such as the liver or heart. Out of FOV (OFOV) activity causes increased dead time and random coincidences, both of which decrease the contrast-to-noise ratio (CNR) in the image. We have designed a realistic torso phantom to quantify the effects of OFOV uptake on imaging performance during a breast PET scan.

Methods Two scans are performed with a µ-Derenzo phantom, one with the OFOV torso phantom present, and one without. The scans are performed with two functioning sections of opposing flat panel imaging heads, with a sensitive FOV of 22mm x 160 mm x 64 mm. In the first study, the phantom is injected with 200 µCi of FDG, and imaged at the center of the FOV without any OFOV activity. For the second study, the µ-Derenzo phantom is filled in the same manner, while the torso phantom is filled with a total of 1.7 mCi and placed outside of the FOV in an anatomically consistent position relative to the breast. Both scans are taken over 105 minutes with a time window of 20 ns. The minimum-resolvable rod size, randoms fraction, CNR, timing and energy resolutions are reported.

Results Rods of 1.2mm are resolvable in both scans. The results are quantified in the table below.

Conclusions The imaging performance of a µ-Derenzo scan is quantified with and without the presence of OFOV activity from a realistic torso phantom. The results demonstrate that our high-resolution clinical breast PET scanner is resistant to image degradation in the presence of realistic OFOV uptake in the torso.

Research Support This work is funded in part by NIH grant R01 CA119056. D.F.C. Hsu and D.L. Freese are supported by the Stanford Graduate Fellowship. D.L. Freese is also supported by an NSF Fellowship.