*2.3. PET Imaging and Data Analysis*

Patients fasted for at least 6 hours before PET imaging, performed using a PET/CT scanner (Biograph 6 or 16, Siemens Healthineers K.K., Japan) with a 585 mm field of view. Three-dimensional data acquisition was initiated 60 minutes after injecting 3.7 MBq/kg of FDG. We acquired eight bed positions (2-minute acquisition per bed position) according to the range of imaging. Attenuation-corrected transverse images obtained with 18F-FDG were reconstructed with the ordered-subsets expectation-maximization algorithm, based on the point spread function into 168 × 168 matrices with a slice thickness of 2.00 mm.

For the semiquantitative analysis, functional images of SUV were produced using attenuation-corrected transaxial images, injected dosage of 18F-FDG, patient's body weight, and he cross-calibration factor between PET and the dose calibrator. SUV was defined as follows:

SUV = Radioactive concentration in the region of interest (ROI) (MBq/g)/Injected dose (MBq)/Patient's body weight (g).

A nuclear physician conducted the volume of interest (VOI) analysis using CT scans, eliminating the physiological uptake in the heart, urinary tracts, and gastrointestinal tracts. We used GI-PET software (Nihon Medi-physics Co. Ltd., Tokyo, Japan) on a Windows workstation to semi-automatically calculate the MTV and TLG (= SUVmean × MTV), of each lesion using SUV thresholds in the liver VOI (= SUVmean + [1.5 × SUVStandard\_Deviation]). These SUV thresholds were the optimum values to generate VOIs in which the whole tumor mass is completely enclosed in all cases, with the CT image as the reference. SUVmax and SUVmean within the generated VOI were also calculated automatically. VOIs over all measurable lesions on pretreatment PET/CT were automatically registered. In case of multiple lesions in the same organ, a maximum of 100 lesions were measured.
