Kinetic modelling of dynamic 18F-FDG datasets from long axial field-of-view PET scanner
2021
1405 Objectives: Quantitative information from dynamic PET studies can be extracted by using kinetic modelling and parametric imaging. These methods require an accurate knowledge of the available tracer concentration in the plasma as a function of time, known as the arterial input function (AIF). Arterial blood sampling is currently the gold standard method to measure AIF in brain PET studies, but is usually avoided due to its invasiveness. Estimation of an image derived input function (IDIF) is a non-invasive alternative but suffers from limited spatial resolution of PET scanners and partial volume effects when measured from carotid arteries in brain PET studies. The AIF peak can also be missed when PET sampling frequency is not sufficient. Recently introduced long axial FOV PET scanners allow short frame durations for better temporal resolution and also enable use of an IDIF from a larger blood pool (i.e. aorta) for whole body kinetic modelling. In this work, we explore use of an IDIF extracted from aorta for kinetic modelling of TACs extracted from multiple organs and brain structures. Methods: Dynamic 18F-FDG data (mean activity: 242.5 MBq, mean weight: 73.5 kg) were acquired from three oncologic patients using Biograph Vision Quadra (Siemens Healthineers) PET/CT scanner. Data were acquired for 65 minutes from injection and were sorted 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. The PET images were reconstructed using ToF PSF reconstruction with a 2-mm Gaussian filter applied. Attenuation correction was performed using information from a CT scan. The CT images were used to automatically define organs of interests: whole brain, lungs, liver, aorta, spleen, kidneys and aorta, using a deep-learning based software prototype for organ segmentation. Furthermore, carotid arteries, grey matter and white matter structures were segmented using PMOD (PMOD Technologies, Switzerland). TACs were extracted from these regions with no partial volume correction applied. 2-tissue compartmental irreversible model was used to model whole brain, gray matter and white matter TACs using IDIFs derived from aorta (IDIFAORTA) and carotid arteries (IDIFCAROTIDS). Results: Figure 1 shows PET image of a representative subject, averaged over first 2 minutes post injection, together with IDIFs from aorta, carotids and TACs from different organs. It is seen that IDIFCAROTIDS suffers from spill out effects and have a lower peak and tail compared to IDIFAORTA curve. Table 1 shows a summary of kinetic rate constants estimated using both input functions. Compared to influx constant Ki estimated using IDIFAORTA, Ki estimated using IDIFCAROTIDS was 23.2%, 27.0% and 24.0% higher for grey matter, white matter and whole brain respectively. These differences are in similar range to previously reported differences between Ki values computed using IDIFCAROTIDS and AIF derived from arterial samples. Estimates of individual micro-parameters K1 were higher by 50.6%, 52.9% and 53.1% in grey matter, white matter and whole brain when computed using IDIFCAROTIDS instead of IDIFAORTA. Conclusion: This preliminary study indicates significant differences between kinetic macro-parameters and micro-parameters estimated by IDIFs derived from aorta and carotid arteries. These results suggest that IDIFs derived from aorta can be used to estimate kinetic parameters in neuroimaging studies.
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
0
References
0
Citations
NaN
KQI