Evaluation of respiratory motion-corrected PET/CT in oncological patients

228 Objectives: Respiratory motions of organs may lead to blurring in PET/CT images, resulting in lower detectability of lesions and inaccurate SUV measurement. Motion correction methods, Q.Freeze (based on 4D phase-matched PET/CT) are expected to overcome these problems. The aim of this study was to determine the appropriate acquisition time in Q.Freeze, and to clarify the characteristics of Q.Freeze compared with the non-respiratory gated images in the clinical images. Methods: First, we investigated the appropriate acquisition time in Q.Freeze to obtain the equal image quality as non-respiratory gated images. FDG PET/CT scans of 20 patients (12 females and 8 males, median age 65, range 46-82 years old) were acquired on a Discovery IQ PET/CT system (GE Healthcare) for 1) 2.5 min under free breathing (FB), 2) 2.5 min using Q.Freeze (QF2.5), and 3) 5 min using Q.Freeze (QF5) per bed. Scans were started at 1 hour after administration of FDG. The images were reconstructed using a conventional OSEM algorism VUE Point. Background standard deviation (SD) as an index of image quality was measured in the mediastinal blood pool or right lobe of the liver, with 2.0 cm and 3.0 cm diameter spherical VOIs respectively. Next, we evaluated the impact of Q.Freeze on the quantitative values compared with non-respiratory gated images. FDG PET/CT scans of 40 patients with lung cancers (8 females and 32 males, median age 74, range 52-89 years old) were acquired on the same setting. The SUVmax, metabolic tumor volume (MTV), and total lesion glycolysis (TLG) of tumors were measured on images of FB and QF5, and statistically analyzed. Lung cancers were divided into 2 groups; the tumors located above and below the tracheal bifurcation were categorized into the upper and lower groups (n=20 and 20), and also categorized into 2 groups according to size, small (≤2 cm, n=16) and large (> 2cm, n=24). The percentage difference in SUVmax, MTV, and TLG between FB and QF5 were calculated and statistically analyzed in each group. Results: About the background analysis, SDs for FB, QF2.5, and QF5 were 0.2±0.1, 0.3±0.1, 0.2±0.1, respectively. No significant difference was observed among SDs for FB and QF5, QF2.5 and QF5 (p=0.12, 0.34). SDs for FB were significantly lower than those for QF2.5 (p<0.01). About the tumor analysis, QF5 showed significantly higher SUVmax than FB (10.3±6.0 vs 9.9±6.0, p=0.04). No significant difference was observed in MTV amongFB and QF5 (12.0±13.6 vs 11.4±12.7, p=0.31), and in TLG among FB and QF5 (90658±149903 vs 95054±151546, p=0.27). In the percentage difference in SUVmax and TLG, the lower group showed slightly larger value, but not significant (7.9±8.7 vs 9.0±35.6, p=0.90, 9.4±22.5 vs 16.5±23.1, p=0.39, respectively). The percentage difference in MTV, no significant difference was observed among the upper and the lower groups (3.8±21.9 vs 2.1±19.0 p=0.67). In the lower group, the percentage difference in MTV and TLG in small group were larger than those in large group (19.2±7.3 vs -4.0±4.3, p=0.01, 32.2±10.9 vs 10.9±5.6, p=0.03, respectively). Conclusions: With 2.5 min acquisition time, the motion correction methods might be affected by the surrounding noises, and 5 min might be needed to get better quantitative performance. QF tended to show higher SUVmax and lower MTV than FB, especially for lesions having larger respiratory movement. Smallertumors are more influenced than larger tumors by Q.Freeze.