BACKGROUND: Accurate kinetic modeling of 18F-fluorodeoxyglucose ([18F]-FDG) positron emission tomography (PET) data requires 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 blood samples, but the introduction of long axial field of view (LAFOV) PET systems enables the use of non-invasive image-derived input functions (IDIFs) from large blood pools such as the aorta without any need for bed movement. However, such protocols require a prolonged dynamic PET acquisition, which is impractical in a busy clinical setting. Population-based input functions (PBIFs) have previously shown potential in accurate Patlak analysis of [18F]-FDG datasets and can enable the use of shortened dynamic imaging protocols. Here, we exploit the high sensitivity and temporal resolution of a LAFOV PET system and explore the use of PBIF with abbreviated protocols in [18F]-FDG total body kinetic modeling. METHODS: Dynamic PET data were acquired in 24 oncological subjects for 65 min following the administration of [18F]-FDG. IDIFs were extracted from the descending thoracic aorta, and a PBIF was generated from 16 datasets. Five different scaled PBIFs (sPBIFs) were generated by scaling the PBIF with the AUC of IDIF curve tails using various portions of image data (35-65, 40-65, 45-65, 50-65, and 55-65 min post-injection). The sPBIFs were compared with the IDIFs using the AUCs and Patlak Ki estimates in tumor lesions and cerebral gray matter. Patlak plot start time (t*) was also varied to evaluate the performance of shorter acquisitions on the accuracy of Patlak Ki estimates. Patlak Ki estimates with IDIF and t* = 35 min were used as reference, and mean bias and precision (standard deviation of bias) were calculated to assess the relative performance of different sPBIFs. A comparison of parametric images generated using IDIF and sPBIFs was also performed. RESULTS: There was no statistically significant difference between AUCs of the IDIF and sPBIFs (Wilcoxon test: P > 0.05). Excellent agreement was shown between Patlak Ki estimates obtained using sPBIF and IDIF. Using the sPBIF55-65 with the Patlak model, 20 min of PET data (i.e., 45 to 65 min post-injection) achieved < 15% precision error in Ki estimates in tumor lesions compared to the estimates with the IDIF. Parametric images reconstructed using the IDIF and sPBIFs with and without an abbreviated protocol were visually comparable. Using Patlak Ki generated with an IDIF and 30 min of PET data as reference, Patlak Ki images generated using sPBIF55-65 with 20 min of PET data (t* = 45 min) provided excellent image quality with structural similarity index measure > 0.99 and peak signal-to-noise ratio > 55 dB. CONCLUSION: We demonstrate the feasibility of performing accurate [18F]-FDG Patlak analysis using sPBIFs with only 20 min of PET data from a LAFOV PET scanner.

PURPOSE: Attenuation correction is a critically important step in data correction in positron emission tomography (PET) image formation. The current standard method involves conversion of Hounsfield units from a computed tomography (CT) image to construct attenuation maps (µ-maps) at 511 keV. In this work, the increased sensitivity of long axial field-of-view (LAFOV) PET scanners was exploited to develop and evaluate a deep learning (DL) and joint reconstruction-based method to generate µ-maps utilizing background radiation from lutetium-based (LSO) scintillators. METHODS: Data from 18 subjects were used to train convolutional neural networks to enhance initial µ-maps generated using joint activity and attenuation reconstruction algorithm (MLACF) with transmission data from LSO background radiation acquired before and after the administration of 18F-fluorodeoxyglucose (18F-FDG) (µ-mapMLACF-PRE and µ-mapMLACF-POST respectively). The deep learning-enhanced µ-maps (µ-mapDL-MLACF-PRE and µ-mapDL-MLACF-POST) were compared against MLACF-derived and CT-based maps (µ-mapCT). The performance of the method was also evaluated by assessing PET images reconstructed using each µ-map and computing volume-of-interest based standard uptake value measurements and percentage relative mean error (rME) and relative mean absolute error (rMAE) relative to CT-based method. RESULTS: No statistically significant difference was observed in rME values for µ-mapDL-MLACF-PRE and µ-mapDL-MLACF-POST both in fat-based and water-based soft tissue as well as bones, suggesting that presence of the radiopharmaceutical activity in the body had negligible effects on the resulting µ-maps. The rMAE values µ-mapDL-MLACF-POST were reduced by a factor of 3.3 in average compared to the rMAE of µ-mapMLACF-POST. Similarly, the average rMAE values of PET images reconstructed using µ-mapDL-MLACF-POST (PETDL-MLACF-POST) were 2.6 times smaller than the average rMAE values of PET images reconstructed using µ-mapMLACF-POST. The mean absolute errors in SUV values of PETDL-MLACF-POST compared to PETCT were less than 5% in healthy organs, less than 7% in brain grey matter and 4.3% for all tumours combined. CONCLUSION: We describe a deep learning-based method to accurately generate µ-maps from PET emission data and LSO background radiation, enabling CT-free attenuation and scatter correction in LAFOV PET scanners.

PURPOSE: Kinetic parameters from dynamic 18F-fluorodeoxyglucose (FDG) imaging offer complementary insights to the study of disease compared to static clinical imaging. However, dynamic imaging protocols are cumbersome due to the long acquisition time. Long axial field-of-view (LAFOV) PET scanners (> 70 cm) have two advantages for dynamic imaging over clinical PET scanners with a standard axial field-of-view (SAFOV; 16-30 cm). The large axial coverage enables multi-organ dynamic imaging in a single bed position, and the high sensitivity may enable clinically routine abbreviated dynamic imaging protocols. METHODS: In this work, we studied two abbreviated protocols using data from a 65-min dynamic 18F-FDG scan: (A) dynamic imaging immediately post-injection (p.i.) for variable durations, and (B) dynamic imaging immediately p.i. for variable durations plus a 1-h p.i. (5-min-long) datapoint. Nine cancer patients were imaged on the Biograph Vision Quadra (Siemens Healthineers). Time-activity curves over the lesions (N = 39) were fitted using the Patlak graphical analysis and a 2-tissue-compartment (2C, k4 = 0) model for variable scan durations (5-60 min). Kinetic parameters from the complete dataset served as the reference. Lesions from all cancers were grouped into low, medium, and high flux groups, and bias and precision of Ki (Patlak) and Ki, K1, k2, and k3 (2C) were calculated for each group. RESULTS: Using only early dynamic data with the 2C (or Patlak) model, accurate quantification of Ki required at least 50 (or 55) min of dynamic data for low flux lesions, at least 30 (or 40) min for medium flux lesions, and at least 15 (or 20) min for high flux lesions to achieve both 10% bias and precision. The addition of the final (5-min) datapoint allowed for accurate quantification of Ki with a bias and precision of 10% using only 10-15 min of early dynamic data for either model. CONCLUSION: Dynamic imaging for 10-15 min immediately p.i. followed by a 5-min scan at 1-h p.i can accurately and precisely quantify 18F-FDG on a long axial FOV scanner, potentially allowing for more widespread use of dynamic 18F-FDG imaging.