Abstract
BACKGROUND: The net influx rate (K (i)), a quantitative metric that complements the standardized uptake value (SUV), entails certain limitations, including its long acquisition time in clinical practice. This study aimed to investigate the clinical practicability of (18)F fluorodeoxyglucose ((18)F-FDG) Patlak imaging in a high-sensitivity short-axial positron emission tomography-computed tomography (PET/CT) scanner by comparing various Patlak protocols alongside the application of a deep learning-based denoising algorithm. METHODS: This study included 14 patients who received a dual-time injection of (18)F-FDG. Four virtual scans were generated with two injections and the scan time ranging from 0 to 86 min. The second scan was conducted 80 min after the first scan. Protocols had varying scan durations-0-75, 0-40 and 41-75, 41-75 and 80-86, and 41-75 min-which were defined as protocols 1, 2, 3, and 4, respectively. Four protocols were generated to account for different arterial input functions (AIFs), and whole-body (WB) passes (3, 4, 5, 6, and 7 passes × 5 min/pass) obtained between 41 and 75 min after the injection were used for Patlak fitting after denoising. Bland-Altman analysis was performed to compare the four AIF protocols in K (i) values in FDG-avid lesions and to analyze the impact of the number of WB passes on K (i) values. Additionally, Pearson correlation of K (i) values between the abbreviated protocols (protocols 2, 3, and 4) and the standard protocol (protocol 1) was performed. RESULTS: Fourteen participants completed the standard Patlak protocol (0-75 min), while 12 completed the full dual-injection protocol (0-86 min) required for all AIF methods. Two participants did not complete the full protocol due to physical discomfort from prolonged lying. Compared to the image-derived input function (IDIF) of the standard protocol, the abbreviated protocols exhibited a relatively lower area under the curve (AUC). K (i) values demonstrated good agreement and high correlation between different protocols, with r values ranging from 0.9451 to 1.0000. In comparison to the estimation obtained from protocol 1, protocol 4, derived from the population-based input function (PBIF) with 20 min of PET (i.e., 55 to 75 min after injection), yielded <3% bias and <15% precision error for K (i) in tumor lesions. The K (i) images acquired with different protocols were visually equivalent. CONCLUSIONS: The findings suggest that abbreviated protocols can provide acceptable K (i) from short-axial PET/CT systems. The 20-min PBIF-based protocol, enhanced by a deep learning-based denoising algorithm, demonstrated the potential to be applied in K (i) analysis for both scientific and clinical purposes.