RT Journal Article SR Electronic T1 Evaluation of population-based input functions for kinetic modelling of 18F-FDG datasets from a long axial FOV PET scanner JF Journal of Nuclear Medicine JO J Nucl Med FD Society of Nuclear Medicine SP 3183 OP 3183 VO 63 IS supplement 2 A1 Sari, Hasan A1 Eriksson, Lars A1 Mingels, Clemens A1 Hong, Jimin A1 Casey, Michael A1 Cumming, Paul A1 Alberts, Ian A1 Shi, Kuangyu A1 Conti, Maurizio A1 Rominger, Axel YR 2022 UL http://jnm.snmjournals.org/content/63/supplement_2/3183.abstract AB 3183 Introduction: Accurate kinetic modelling of 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET) data entails accurate knowledge of the available tracer concentration in the plasma during the scan time, known as the arterial input function (AIF). The gold standard method to derive the AIF requires collection of serial arterial samples. The introduction of long axial field of view (LAFOV) PET systems enables extraction of non-invasive image derived input functions (IDIF) from large blood pools such as the aorta. However, these protocols require an extensive dynamic PET acquisition which can be cumbersome in a busy clinical setting. Population-based input functions (PBIF) have previously shown potential in accurate Patlak modelling of 18F-FDG datasets [1]. In this work, we exploit the high sensitivity and temporal resolution of LAFOV PET systems and explore use of PBIF with abbreviated protocols in 18F-FDG total body kinetic modelling. Methods: Dynamic PET data were acquired in 24 oncological subjects for 65 minutes following the administration of 18F-FDG. The PET data were reconstructed in 62 frames using the following frame durations: 2 × 10 s, 30 × 2 s, 4 × 10 s, 8 × 30 s, 4 × 60 s, 5 × 120 s, and 9 × 300 s. IDIFs were extracted using the descending thoracic aorta. The data were split into 16 training and 8 testing sets. The PBIFs were generated from the training datasets using the following steps: The IDIFs were normalized to their area under curves (AUC). The normalized curves were fitted using Feng’s model [2] and fitted curves were adjusted to population mean time delay. The resulting curves were averaged to generate a PBIF. During the evaluation of the PBIF, we generated 3 scaled PBIFs (sPBIF) by scaling the PBIF with AUC of IDIF curve tails using various periods (35-65, 45-65, and 55-65 min). The sPBIFs were compared with the IDIFs using the AUCs and Patlak Ki estimates in tumor lesions. Patlak plot start time (t*) was also varied to evaluate the performance of shorter acquisitions on accuracy of Patlak Ki estimates. Patlak Ki estimates with IDIF and t*=35 min was used as reference and mean bias and precision (standard deviation of bias) were calculated to assess relative performance of different sPBIFs. Results: There was no statistically significant difference between AUCs of the IDIF and sPBIF35-65, sPBIF45-65, sPBIF55-65 (Wilcoxon test: P=0.44, P=0.80 and P=0.96 respectively). Similarly, sPBIF35-65, sPBIF45-65, and sPBIF55-65 yielded Patlak Ki estimates with no statistically significant difference to IDIF (Wilcoxon test: P=0.74, P=0.84 and P=0.92 respectively). As shown in figure 1, sPBIF55-65 showed the best performance with 1.5% bias and %6.8 precision. Using sPBIF55-65 with Patlak model, 20 minutes of PET data (i.e. 45 to 65 post injection) is needed to achieve <15% precision error on Ki estimates compared to Ki estimates with IDIF (figure 2). Conclusions: Results of this study demonstrate feasibility to perform accurate 18F-FDG Patlak modelling using sPBIF with 20 minutes of PET data using a long axial FOV PET scanner.