Ultra-low dose imaging in a standard axial field-of-view PET

Abstract

Though ultra-low dose (ULD) imaging offers notable benefits, its widespread clinical adoption faces challenges. Long-axial field-of-view (LAFOV) PET/CT systems are expensive and scarce, while artificial intelligence (AI) shows great potential but remains largely limited to specific systems and is not yet widely used in clinical practice. However, integrating AI techniques and technological advancements into ULD imaging is helping bridge the gap between standard axial field-of-view (SAFOV) and LAFOV PET/CT systems. This paper offers an initial evaluation of ULD capabilities using one of the latest SAFOV PET/CT device. A patient injected with 16.4 MBq ¹⁸F-FDG underwent a local protocol consisting of a dynamic acquisition (first 30 min) of the abdominal section and a static whole body 74 min post-injection on a GE Omni PET/CT. From the acquired images we computed the dosimetry and compared clinical output from kidney function and brain uptake to kidney model and normal databases, respectively. The effective PET dose for this patient was 0.27 ± 0.01 mSv and the absorbed doses were 0.56 mGy, 0.89 mGy and 0.20 mGy, respectively to the brain, heart, and kidneys. The recorded kidney concentration closely followed the kidney model, matching the increase and decrease in activity concentration over time. Normal values for the z-score were observed for the brain uptake, indicating typical brain function and activity patterns consistent with healthy individuals. The signal to noise ration obtained in this study (13.1) was comparable to the LAFOV reported values. This study shows promising capabilities of ultra-low-dose imaging in SAFOV PET devices, previously deemed unattainable with SAFOV PET imaging.

Supplementary Information

The online version contains supplementary material available at 10.1186/s41824-025-00261-9.

Introduction

Positron emission tomography (PET) combined with computed tomography (CT) is a powerful imaging modality (Ayesa and Murphy 2022). Recent technological advancements have enabled significant progress in low- and ultra-low dose imaging, enhancing radioprotection for young or pregnant patients (Snick et al. 2024; Dias et al. 2025; Ting Xia, Alessio and Kinahan 2009; Bebbington et al. 2023), improving outcomes in diseases with favorable prognoses, and promoting cost-efficient use of radiopharmaceuticals—all while maintaining diagnostic accuracy (Tan et al. 2021, 2023; Mostafapour et al. 2024).

Research towards lowering the dose up to ULD has predominantly focused on LAFOV PET/CT systems (Alberts et al. 2023; Gorenstein et al. 2023), which offer an extended FOV ranging from 106 to 194 cm. These systems enable whole-body imaging in a single or few bed positions with significantly higher sensitivity compared to SAFOV systems. However, recent advancements in SAFOV technology, characterized by smaller fields of view (15 to 32 cm) and requiring multiple bed positions to cover the entire body, challenge the notion that ULD capabilities are confined to LAFOV systems. Technological improvements, including enhanced scintillation materials and digitalization of electronic components, have significantly boosted the performance of both SAFOV and LAFOV PET/CT systems (Dadgar et al. 2024).

Despite the benefits of ULD imaging, widespread clinical implementation presents challenges. LAFOV PET/CT systems are costly and have limited availability, restricting access for many healthcare facilities. The sought-after ULD imaging has evolved to include AI (Dadgar et al. 2024), and technological advancements, narrowing the gap between SAFOV and LAFOV PET/CT systems. This paper provides a preliminary assessment of ULD capabilities using the latest generation of standard axial field-of-view PET/CT device.

Methods

Case description

In this study, a male subject (49 years old, 81 kg) from our low-dose imaging program, with no evidence of disease, was injected with 16.4 MBq ¹⁸F-FDG (0.2 MBq/kg).

Device description

The patient was scanned on an Omni PET/CT scanner with a SAFOV (32 cm) from GE Healthcare (Chicago, United States). This device’s generation comprises of a digital PET/CT with bismuth germanate oxide (BGO) detectors and no time-of-flight (TOF) capabilities, with a system sensitivity of 47 kcps/MBq.

Acquisition protocol

The protocol included a 30-minute dynamic acquisition of the upper abdomen of 46 single-field of view frames: 12 frames of 5 s each (0–1 min), followed by 6 frames of 10 s each (1–2 min), and finally 28 frames of 60 s each (2–30 min). This was followed by a whole-body acquisition 74 min post-injection including 5 bed positions, 5 min/bed for a total acquisition time of 25 min.

Image reconstruction

This scanner uses a proprietary Bayesian penalized likelihood reconstruction method (Q-clear) set to beta factor 1200 to compensate for the low statistics present in the image in addition to AI (precisionDL, set to medium strength).

Analysis

Dosimetry

Both organ and effective doses were computed for this examination based on the total injected activity using MIRDcalc software (Kesner et al. 2023; Carter et al. 2023) and compared to the local injected activity (2 MBq/kg) and by the national diagnostic reference level (3.5 MBq/kg) for oncological studies.

Renal modeling

To understand the robustness and limitations of the ULD dynamic acquisition we compared the renal function, from collecting kidneys activity concentration from placed regions of interest to a normal kidney function model (Qiao et al. 2008). The proposed kidney model consisted of two parts: the transportation of FDG from blood and the excretion process of FDG out of kidney. , the timing delay constant, was the excretion time of FDG out of kidney from injection (Qiao et al. 2008). is the concentration of FDG in kidney, which can be detected by PET which is based on is the concentration of FDG in blood and the concentration of FDG in kidneys (Qiao et al. 2008)

1

2

For our results we used the mean and standard deviation values of the parameters proposed in the referred publication and compared with our kidney measurements, . For the blood concentration , we used the aorta concentration (Qiao et al. 2008).

Brain normal database

To compare our acquired brain images with normal datasets, we utilized the MIMneuro^® software package, a comprehensive tool for quantitative analysis of PET and SPECT brain imaging. The MIMneuro package supports multi-tracer analysis and provides intuitive visualization and automated reporting. Our workflow involved registering the acquired images to a standardized template using the BrainAlign™ deformation algorithm, ensuring accurate alignment with the normal control database. Quantitative comparisons were performed using voxel-based analysis, which highlighted statistically significant differences in tracer uptake with color-coded overlays. Additionally, region-based analysis was conducted to calculate Z-scores and standard uptake value ratios (SUVR) for specific brain regions. This approach allowed us to identify and quantify deviations from normal patterns, facilitating a robust assessment of the acquired images.

Overall image quality

we evaluated background noise to compare the overall image quality. The background noise was measured by the placement of a 14 cm3 volume-of-interest (VOI) in healthy liver tissue in the right liver lobe (Boellaard et al. 2015). We then compared with reported values in the literature (Alberts et al. 2021). We also compared the SNR for the protocol used with medium intensity precisionDL versus high strength precisionDL and without precisionDL to validate the choice of the chosen reconstruction. Additionally, images were analyzed by two nuclear medicine physicians with more than 10 years of experience. They assessed subjective image quality with 5-point Likert scales (Tan et al. 2022). The scoring criteria were as follows: 1 for very poor image quality, 2 for poor image quality, 3 for average image quality, 4 for good image quality, and 5 for very good image quality. Images with scores under 3 did not meet the need for clinical diagnosis.

Results

Dosimetry

The total effective dose from an injected activity of 16.4 MBq of ¹⁸F-FDG PET was 0.27 ± 0.01 mSv, calculated using MIRDcalc. While for the organs, the absorbed dose estimated were of 0.56 mGy for the brain, 0.89 mGy for the heart, and 0.20 mGy for the kidneys. These achievements corresponded to 10% of the total absorbed and effective dose from our standard clinical protocol (set at 2 MBq/kg), as shown in the organ-pictograms in Fig. 1.

Fig. 1.

MIP with represented standard axial field-of-view (SAFOV) system and dosimetry (organ absorbed dose (mGy) and total effective dose (mSv)) calculated using MIRDcalc software for clinical protocol based on 18 F-FDG injected activity of 16.4 MBq. Presenting dosimetry for brain, heart, kidneys, and total body. Additionally, axial views of brain and abdominal region with brain (colored normal database overlay) and kidneys (dynamic) activity concentrations

Renal modeling

Figure 2 illustrates the kidney concentration profiles obtained from our dynamic acquisition compared to the predicted kidney model. The recorded kidney concentration closely followed the kidney model, matching the increase and decrease in activity concentration over time. However, some differences were observed, particularly at the later time points, where the recorded data showed slight deviations from the model predictions. These discrepancies may be attributed to individual physiological variations or experimental conditions, but low statistics associated with the ULD protocol cannot be excluded. Overall, the recorder data is within the range of the model prediction, however this result is limited to a single patient and a larger cohort would be required to validate the method.

Fig. 2.

Kidney concentration measurements, represented by dots, alongside the model outputs for blood and kidney concentrations, denoted as C1(t) and C2(t), respectively. Additionally, the model includes CT(t), which represents the kidney concentration seen by the PET The recorded data points and model outputs are plotted over time

Brain normal database

Figure 3 displays the brain activity map obtained from the neuro analysis using the MIMneuro^® software. As anticipated, the normal values for the z-score were observed for this subject, indicating typical brain function and activity patterns consistent with healthy individuals.

Fig. 3.

Brain activity map obtained from the neuro analysis using MIM software. The axial view on the left displays different slices throughout the brain, with highlighted voxels indicating regions where lower uptake was observed compared to the normal database. On the right, the accompanying table presents the specific values observed for the evaluated pre-segmented regions

Overall image quality

The SNR obtained for medium strength precisionDL in this study was 13.1. For the high strength precisionDL and no precisionDL SNR was 9.6 and 10.9 respectively. Figure 4 shows the axial view of the liver and the MIP of the tested reconstructions. The subjective assessment of the image quality scores was 4.0, meeting the needs of clinical diagnosis.

Fig. 4.

MIP and coronal view for the evaluated reconstructions. A and B for the medium strength precisionDL, C and D for high strength precisionDL and E and F for the reconstruction without precisionDL

Discussions

This study demonstrated the potential of a current standard field-of-view PET scanner in ultra-low dose imaging, including dynamic acquisitions. It significantly broadens the accessibility of these protocols: current literature on ultra-low dose imaging focuses exclusively on long axial field-of-view systems, which typically offer superior image quality due to their extended coverage and sensitivity (Snick et al. 2024; Alberts et al. 2021, 2023; Smith et al. 2024; Liu et al. 2025) but at a significant cost premium.

Regarding the dosimetry, achievements with 0.2 MBq/kg injection of ¹⁸F-FDG demonstrated the feasibility of this study for ultra-low dose imaging in conventional PET/CT field-of-view. Soret et al. (2022) highlighted that the same activity level maintained imaging quality without compromising the measurements (Soret et al. 2022). By using the LAFOV protocol, Dias et al. (2025) extended the discussion using paediatric patients (5 months and 7 years old) and a 16-year-old patient (Dias et al. 2025). They showed that by decreasing the amount of injected activity for these patients, the effective dose was reduced by a factor of 10, compared to the typical SAFOV PET/CT dose. Our study demonstrated the same level of reduction of total effective dose and absorbed dose to organs, compared to our current clinical protocol which is already conservative with a normal injected dose of 2 MBq/kg.

In respect to image quality, Alberts et al. evaluated the SNR at different simulated statistics with a LAFOV [1]. The activity level used in our study corresponds to the LAFOV 1 min reported by the authors (10% of the injected activity). They reported a mean SNR of 7 [2,12]. Using the same methodology, our calculated SNR was 13.1, which is superior to the value obtained by the LAFOV. What it is further noteworthy of our results is that in the study from Alberts et al., 10% of their injected activity varied between from 0.35 MBq/kg for [18 F]FDG to 25 MBq and 15 MBq respectively for 18 F PSMA and 68Ga-DOTA-TOC, compared to the 0.2 MBq/kg in this study. The obtained SNR values in our study are within (if not superior) the SNR values for all ranges of statistics presented by Alberts et al. using a LAFOV. Our SNR are also comparable to the study presented by Tan et al. (Tan et al. 2022), where they reported a SNR of 14.6 +/- 2.5 for a 5-minute acquisition in a cohort injected with 25.3 +/- 4.8 MBq using a 194 cm LAFOV PET/CT.

The clinical findings from this study indicate that ultra-low dose imaging is achievable with SAFOV PET/CT devices, from an overall image quality standpoint as well as for quantitation such as brain uptake compared to normal database and renal function. This aligns with previous research demonstrating the potential of ULD with LAFOV devices (Tan et al. 2023; Alberts et al. 2023; Liu et al. 2025). The observed normal z-scores and renal function support the feasibility of SAFOV devices in achieving diagnostic accuracy while minimizing radiation exposure. Notably, the benefits of reducing the injected activity can be allocated between lowering the injected dose (thus reducing patient radiation exposure) and shortening the acquisition time (thereby improving patient comfort), without compromising the final image quality. This allocation can be adjusted based on clinical need.

One of the most compelling aspects of ULD imaging on SAFOV PET/CT systems is its economic impact. While LAFOV systems require significantly higher financial investment and maintenance costs due to their cutting-edge design and extended detector coverage in the order of 3–4 times the price and running costs for a SAFOV PET/CT, the latest generation SAFOV PET/CT systems are already in place in many more healthcare institutions.

The shift towards ULD imaging on SAFOV PET/CT scanners offers several key benefits from a patient perspective, which extend beyond economic considerations.

Ultra-low dose PET/CT imaging significantly reduces absorbed and effective doses, lowering long-term radiation risks. Faster scan times improve patient experience, minimizing movement artefacts and the need for repeat scans. The approach broadens clinical applications beyond oncology, enabling its use in preventive medicine and routine imaging for specific patient groups, especially for children and pregnant individuals. By utilizing SAFOV systems, this method enhances accessibility, ensuring more equitable healthcare delivery across diverse populations. More extensive studies with larger patient cohorts are needed to confirm the ULD capabilities of SAFOV devices and to fully understand their clinical implications.

Conclusions

In conclusion, this study highlighted the potential of SAFOV PET/CT systems in achieving ultra-low dose imaging with satisfactory image quality and diagnostic accuracy. This advancement broadens the applicability of ultra-low-dose imaging, previously thought to be unattainable with standard field-of-view PET imaging.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

AI

Artificial intelligence

BGO

Bismuth germanate oxide

CT

Computed tomography

FDG

Fluorodeoxyglucose

FOV

Field-of-view

LAFOV

Long-axial field-of-view

PET

Positron emission tomography

SAFOV

Standard axial field-of-view

SUVR

Standard uptake value ratios

TOF

Time-of-flight

ULD

Ultra-low dose

Author contributions

TL and AL conceptualized the study. TL and CVG analyzed and interpreted the patient data. PF helped with the data acquisition and data collection. TL, CVG and AL were the main contributors in writing the manuscript. All authors read and approved the final manuscript.

Funding

No funding was used for this work.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval

This case study was performed in line with the principles of the Declaration of Helsinki.

Consent to participate

Written informed consent was obtained from the patient.

Consent to publishcation

The authors affirm that human research participants provided informed consent for publication of the images in figure.

Competing interests

The authors have no relevant financial or nonfinancial interests to disclose. The Swiss Medical Network and GE Healthcare have a partnership for building the future of health. GE Healthcare was not involved in the imaging process nor in the writing phase of this case.