Improving image reconstruction to quantify dynamic whole-body PET/CT: Q.Clear versus OSEM
Abstract Background The introduction of PET systems featuring increased count rate sensitivity has resulted in the development of dynamic whole-body PET acquisition protocols to assess 18F-FDG uptake rate ( $${K}_{i}$$ K i ) using 18F-FDG PET/CT. However, in short-axis field-of-view (SAFOV) PET/CT s...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2025-03-01
|
| Series: | EJNMMI Physics |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s40658-025-00736-5 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Background The introduction of PET systems featuring increased count rate sensitivity has resulted in the development of dynamic whole-body PET acquisition protocols to assess 18F-FDG uptake rate ( $${K}_{i}$$ K i ) using 18F-FDG PET/CT. However, in short-axis field-of-view (SAFOV) PET/CT systems, multiple bed positions are required per time frame to achieve whole-body coverage. This results in high noise levels, requiring higher 18F-FDG activity administration and, consequently, increased patient radiation dose. Bayesian penalized-likelihood PET reconstruction (e.g. Q.Clear, GE Healthcare) has been shown to effectively suppress image noise compared to standard reconstruction techniques. This study investigated the impact of Bayesian penalized-likelihood reconstruction on dynamic whole-body 18F-FDG PET quantification. Methods Dynamic whole-body 18F-FDG PET/CT data (SAFOV PET Discovery MI 5R, GE Healthcare) of healthy volunteers and one lung cancer patient, consisting of a ten-minute dynamic scan of the thoracic region followed by six whole-body passes, were reconstructed with Q.Clear and Ordered Subset Expectation Maximization (OSEM) according to EARL 2 standards. Image noise in the measured time-activity-curves (TAC) was determined for the myocardium, hamstring, liver, subcutaneous adipose tissue and lung lesion for both reconstruction methods. $${K}_{i}$$ K i values were calculated using Patlak analysis. Finally, bootstrapping was used to investigate the effect of image noise levels on $${K}_{i}$$ K i values (bias and precision) as a function of magnitude of $${K}_{i}$$ K i and volume-of-interest (VOI) size for both computationally simulated TACs ( $${K}_{i}$$ K i = 1.0–50.0·10–3·ml·cm−3·min−1) and the measured TACs. Results Compared to OSEM, Q.Clear showed 40–55% lower noise levels for all tissue types (p < 0.05). For the measured TACs no systematic bias in $${K}_{i}$$ K i with either reconstruction method was observed. $${K}_{i}$$ K i precision decreased with decreasing VOI size, with that of Q.Clear being superior compared to OSEM for small VOIs of 0.56 cm3 in all tissues (p < 0.05), with the largest difference in relative precision for small values of $${K}_{i}$$ K i . The simulated TACs corroborated these results, with Q.Clear providing the best precision for small values of $${K}_{i}$$ K i and small VOIs in all tissues. Conclusion Q.Clear reconstruction of dynamic whole-body PET/CT data yields more precise $${K}_{i}$$ K i values, especially for small values of $${K}_{i}$$ K i and smaller VOIs, compared to standard OSEM. This precision improvement shows Q.Clear’s potential to better detect and characterize small lesion metabolic activity in oncology and allows for lower administered activity dosage. |
|---|---|
| ISSN: | 2197-7364 |