Clinical evaluation of data-driven respiratory gating for PET/CT in an oncological cohort of 149 patients: impact on image quality and patient management

Abstract

Objectives:

To evaluate the impact of fully automatic motion correction by data-driven respiratory gating (DDG) on positron emission tomography (PET) image quality, lesion detection and patient management.

Materials and Methods:

A total of 149 patients undergoing PET/CT for cancer (re-)staging were retrospectively included. Patients underwent a PET/CT on a digital detector scanner and for every patient a PET data set where DDG was enabled (PET_DDG) and as well as where DDG was not enabled (PET_nonDDG) was reconstructed. All PET data sets were evaluated by two readers which rated the general image quality, motion effects and organ contours. Further, both readers reviewed all scans on a case-by-case basis and evaluated the impact of PET_DDG on additional apparent lesion, change of report, and change of management.

Results:

In 85% (n = 126) of the patients, at least one bed position was acquired using DDG, resulting in mean scan time increase of 4:37 min per patient in the whole study cohort (n = 149). General image quality was not rated differently for PET_nonDDG and PET_DDG images (p = 1.000) while motion effects (i.e. indicating general blurring) was rated significantly lower in PET_DDG images and organ contours, including liver and spleen, were rated significantly sharper using PET_DDG as compared to PET_nonDDG (all p < 0.001). In 27% of patients, PET_DDG resulted in a change of the report and in a total of 12 cases (8%), PET_DDG resulted in a change of further clinical management.

Conclusion:

Deviceless DDG provided reliable fully automatic motion correction in clinical routine and increased lesion detectability and changed management in a considerable number of patients.

Advances in knowledge:

DDG enables PET/CT with respiratory gating to be used routinely in clinical practice without external gating equipment needed.

Introduction

Positron emission tomography (PET) images radiotracer uptake in patients, thereby enabling the detection and quantification of physiologic and pathophysiological processes in vivo.¹ In oncological patients, ¹⁸F-fludeoxyglucose (¹⁸F-FDG) PET is used to assess increased glucose metabolism in primary tumors, lymph node or distant metastases,^2,3 whereas CT enables a detailed assessment of morphological disease extent with a high spatial resolution.^4,5 Therefore, hybrid imaging with ¹⁸F-FDG PET/CT has evolved as an important tool for staging of cancer patients in many different etiologies.³ Further substances tracing prostate-specific membrane antigen (PSMA) and DOTA-peptides (DOTATATE) play also an important role in PET imaging, namely for the assessment of prostate cancer and neuroendocrine tumors.^6–8

However, PET image quality may be significantly degraded by respiratory motion, among other causes.⁹ The radiotracer distribution appears blurred on images affected by motion, rendering exams suboptimal for diagnostic purposes and/or quantitative assessment, particularly in the chest and upper abdomen.¹⁰ To overcome this issue (i.e. to decrease the effect of motion), respiratory gating can be applied. This requires the acquisition of a gating signal, which is usually done by external devices, such as a pressure belt or a video camera tracing infrared markers, both of which may be inconvenient for the patient, and may also prolong the examination time owing to device mounting and unmounting.¹¹ Also, a single misregistration may spoil motion correction of the exam, in fact an erroneous misregistration of the underlying static attenuation maps may cause an over- or underestimation of activity of the region. A respiratory gating signal may also be extracted from the acquired PET raw data using different methods, generally referred to as data-driven gating (DDG).^12,13 Until recently, this method was mainly reserved to academic environments and could only be used retrospectively (i.e. based on list-mode raw data) and not under real-world conditions while scanning patients,¹⁴ although nowadays DDG has become available on the market and is becoming more widespread.

Recently, a new digital PET detector system with lutetium-based scintillator crystal arrays and silicon photomultipliers was introduced. A first study carried out in a mixed population of cancer patients showed an improved performance of digital PET/CT with regard to pathologic and physiologic structures.¹⁵ This may potentially be further enhanced by the use of DDG.

Accordingly, the purpose of our study was to evaluate the impact of DDG using the latest generation digital PET/CT scanner with regard to PET (a) image quality, (b) lesion detection and (c) patient management in an unselected oncological cohort.

Methods and materials

Patients

This retrospective study included consecutive patients (n = 149) who underwent a clinical PET/CT for (re-)staging of oncological diseases during a pilot evaluation phase of 6 weeks at the Department of Nuclear Medicine of the University Hospital of Zurich who gave general written informed consent for the use of their data for scientific purposes. Overall, 132 patients (89%) underwent an ¹⁸F-FDG PET/CT scan and 17 (11%) underwent a non-FDG PET/CT scan (n = 7 PSMA; n = 9 DOTATATE; n = 1 DOPA), Table 1. The local ethics committee approved the study protocol (Trial-No. BASEC 2019-00207).

Table 1.

Demographic data of study subjects (n = 149)

Female/male, n (%)	61 (41%) / 88 (59%)
Age, years	61 ± 15 (24 - 89)
Body weight, kg	76 ± 18 (40 - 142)
Body height, m	1.72 ± 0.1 (1.50–1.95)
BMI, kg/m2	25.6 ± 5.5 (14.9–43.4)
Injected dose, MBq	190 ± 74 (88 - 317)
Type of primary disease
Head and neck cancer	14 (9.4%)
Breast cancer	15 (10.1%)
Lung cancer	19 (12.8%)
Esophageal cancer	3 (2.0%)
Small bowel cancer	0 (0%)
Colon cancer	8 (5.4%)
Pancreatic cancer	7 (4.7%)
Cholangiocarcinoma	4 (2.7%)
Urogenital cancer	18 (12.1%)
Melanoma	27 (18.1%)
Lymphoma	12 (8.1%)
Cancer of unknown origin	3 (2.0%)
Neuroendocrine tumor	10 (6.7%)
Paraneoplastic syndrome	9 (6.0%)
Tracer used
FDG	132 (88.6%)
PSMA	7 (4.7%)
DOTATATE	9 (6.0%)
DOPA	1 (0.7%)

BMI = body mass index; MBq = Mega-Becquerel; PET = positron emission tomography.

Values are given as absolute numbers and percentages in parenthesis or mean ± standard deviation (range).

PET/CT imaging protocol

All patients underwent a PET/CT on a digital detector scanner (GE Discovery Molecular Insights - DMI PET/CT, GE Healthcare, Waukesha, WI) with a high intrinsic system sensitivity of 22 cps/kBq.

132 patients (89%) underwent an ¹⁸F-FDG PET with a body mass index (BMI)-adapted ¹⁸F-FDG dosage regimen, as previously described¹⁶ and presented in Table 2. Participants fasted for at least 4 h prior to the scan, and blood glucose levels were below 160 mg dl⁻¹ at the time of the 18F-FDG injection. A free-breathing CT scan was performed from the vertex of the skull to the mid-thighs (or feet), and used for attenuation correction purposes as well as for anatomic localization of ¹⁸F-FDG uptake. The CT scan was performed using automated dose modulation (range 15–100 mA, 120 kV). Immediately after the CT scan, PET images were acquired covering the identical anatomical region. The PET acquisition time was set to 2.5 min per bed position, with 6–11 bed positions per patient (depending on patient size), with an overlap of 23%. All PET images were reconstructed using a Bayesian penalized likelihood reconstruction method (Q.Clear, GE Healthcare, Waukesha, WI). Data sets were reconstructed with a 256 × 256 pixel matrix. Spatiel resolution were radial 1.62 mm, tangential 2.30 mm and axial 3.09 mm, as assessed by full width at half maximum (FWHM) measurements at 1 cm from center of field of view were.¹⁷

Table 2.

PET imaging protocols for the tracers used in the study group (n = 149)

Tracer	Scan time p.i.	Injected activity	Bed time	Reconstruction setting
FDG	60 min	BMI-adapteda	2.5 min	Q.Clear (β-value of 450)
PSMA	90 min	3 MBq/Kgb	2.5 min	Q.Clear (β-value of 1200)
DOTA	60 min	150 MBq	2.0 min	Q.Clear (β-value of 1000)
DOPA	60 min	150 MBq	2.5 min	Q.Clear (β-value of 1000)

BMI, body mass index; FDG = Fludeoxyglucose; MBq = Mega-Becquerel; PET = Positron emission tomography; PSMA = prostate-specific membrane antigen.

^aFDG-dose of 1.5 MBq/kg body weight was injected for patients with a BMI of <20 kg/m², 2 MBq/kg body weight for patients with a BMI of 20–24.5 kg/m², and 3.1 MBq/kg body weight for patients with a BMI >24.5 kg/m² (without exceeding a maximum injected FDG dose of 320 MBq).

^bminimum = 200 MBq, maximum = 300 MBq.

17 patients (11%) underwent a PET/CT scan with a tracer other than ¹⁸F-FDG. The imaging protocols used for PSMA, DOTATATE and DOPA-imaging were described previously,¹⁸ and are presented in Table 2.

Use of DDG for PET acquisition

In our study, we used a software-based prospective DDG algorithm (MotionFree, GE Healthcare, Waukesha, WI) that was recently introduced on DMI PET/CT, in combination with a motion correction algorithm (Q.Static, GE Healthcare, Waukesha, WI) that utilizes the quiescent phase of the respiratory cycle. Unlike externally driven respiratory gating, which usually relies on infrared camera tracking of chest motion, DDG methods solely use PET raw data in combination with dimensionality reduction techniques, in order to extract the respiratory signal. This technique utilizes a principal component analysis to compute the spatiotemporal variation of list mode data. The algorithm measures respiration-like frequencies within the PET data. At the end of base acquisition time, for each bed position for which motion screening was prescribed, an on-the-fly decision is made as whether motion has been detected. The whole body DDG reconstructed image is composed of a non-gated part for bed positions that were not triggered, and a quiescent gated part on the bed positions that were triggered; with the non-gated bed positions being 2.5 min in duration, and the gated positions being 5 min (i.e. doubling the per-bed acquisition time) in duration but with only 50% of coincidences retained, namely those occurring in the defined quiescent periods. The algorithm provides a signal-to-noise measure of respiration-like frequencies within the data, denoted as R-value, that is configurable (R-value threshold). The R-factor (i.e. signal-to-noise ratio of respiration-like frequencies within the data operating as threshold for triggering) was set to R > 10.0.^19,20

For further analysis, for every patient a PET data set where DDG was enabled (hereafter referred to as PET_DDG) and another data set where DDG was not enabled (hereafter referred to as PET_nonDDG) was reconstructed; PET_nonDDG was processed using a non-gated reconstruction using the first 2.5 min per bed position, with the additional 2.5 min of data that were collected on DDG-triggered bed positions discarded.

Subjective PET image analysis and assessment of clinical impact of DDG

All PET data sets were evaluated by two readers (M.M. and M.W.H., with 3 and 8 years of experience in PET/CT reading, respectively) blinded to the acquisition mode used. All scans were reviewed independently on a dedicated workstation (Advantage Workstation, v. 4.6; GE Healthcare, Waukesha, WI) and in random order. Readers were blinded to the clinical information. In case of discrepancy of image rating, a final decision was reached by consensus.

The readers first rated the general image quality, motion effects and organ contours. For this purpose, data sets were viewed using maximum intensity projection (MIP) views of PET and axial views with reformatted sections. The two readers evaluated the images according the scores described in Table 3. The frequency of attenuation correction artifacts in PET_nonDDG and PET_DDG was assessed by noting the presence of curvilinear cold artifacts on dome of liver/diaphragm or at lung base because of respiration mismatch on PET images with CT attenuation correction.

Table 3.

Subjective image quality scores

Overall image quality	1 = poor 2 = acceptable 3 = good 4 = very good
Motion effects	1 = no 2 = slight 3 = moderate 4 = severe
Organ contoursa	1 = ill-defined 2 = moderate blurring 3 = slight blurring 4 = sharp

^aOrgan contour assessment included liver, spleen, stomach, kidney, collecting system, bowel, heart.

In a second session performed 8 week after initial scanning, both readers reviewed all scans in consensus on a case-by-case basis. Decisions were made for every patient as (a) whether additional lesions were apparent using PET_DDG, (b) whether the use of PET_DDG led to a change of report, and (c) whether PET_DDG resulted in a change of management according to the institutional protocols that is based on international, interdisciplinary guidelines.

Statistical analyses

Categorical variables are expressed as proportions, and continuous variables are presented as mean ± standard deviation or median (range), depending on the distribution of values. Qualitative image ratings (i.e. overall image quality, motion artefacts, and organ contours) were compared of PET_nonDDG and PET_DDG reconstruction using the Wilcoxon matched pairs signed rank test. Analyses were carried out using SPSS release 25.0 (IBM Corporation, Armonk, NY) and MedCalc v. 15.8 (MedCalc Software, Ostend, Belgium). A two-tailed p-value of <0.05 was considered to indicate statistical significance.

Results

A total of 149 patients (61 female, 88 male, mean age 61 ± 15 years) referred for the initial staging (n = 41) or restaging (n = 108) of different oncological diseases undergoing PET/CT participated in our study. Detailed demographic information is presented in Table 1.

Technical aspects of DDG in study cohort

The results of the technical aspects using DDG in the study cohort are given in Table 4. Overall, in 85% of the patients at least one bed was acquired using DDG, leading to an average scan time increase of 4:37 min in the whole study cohort (n = 149), and 5:28 min in the subgroup of patients whose exams were triggered by DDG (n = 126).

Table 4.

Technical aspects of DDG in study group

Gating yes / gating no	126 (84.6%) / 23 (15.4%)
Gating yes FDG	109/132 (83%)
Gating yes PSMA	7/7 (100%)
Gating yes DOTATATE	9/9 (100%)
Gating yes DOPA	1/1 (100%)
Number of bed positions acquired	6 ± 1 (5–11)
Amount of bed positions triggered	1.69 ± 0.1 (1.48–1.94)
no gating	23 (15%)
one bed triggered	18 (12%)
two beds triggered	68 (46%)
three beds triggered	32 (21%)
four beds triggered	8 (5%)
Total scan time, min	15:16 ± 3:02 (10:00 - 27:30)
Total time increase per study in whole cohorta, min	4:37 ± 3:02 (0:00 - 10:00)
Total time increase per study in DDG cohortb, min	5:28 ± 1:52 (2:30 - 10:00)

DDG, Data-driven gating.

Values are given as absolute numbers and percentages in parenthesis or mean ± standard deviation (range).

^an = 149.

^bn = 126.

Impact of DDG on subjective image quality

The results of the subjective image assessment including all study subjects are given in Table 5. General image quality was not rated differently for PET_nonDDG and PET_DDG images (p = 1.000). Motion effects (i.e. indicating general blurring) were rated significantly lower in PET_DDG images and contours of different organs, including liver and spleen, were rated sharper using PET_DDG as compared to PET_nonDDG (all p < 0.001), Table 5. The improvement of organ contours was most pronounced for the liver, spleen, heart and collecting system, and less pronounced for the stomach, kidney and bowel, see Table 5. Overall, there were nine cases of attenuation correction artefacts in PET_DDG as compared to seven cases in PET_nonDDG, respectively.

Table 5.

Subjective images analysis comparing

	PETnonDDG	PETDDG	p-value
Overall image qualitya	3.8	3.8	1.000
Motion effectsb	1.2	1.0	<0.001
Organ contoursc
Liver	3.5	3.8	<0.001
Spleen	3.6	3.9	<0.001
Stomach	3.8	3.9	<0.001
Kidney	3.8	4.0	<0.001
Collecting system	3.6	4.0	<0.001
Bowel	3.6	3.8	<0.001
Heart	3.5	3.8	<0.001
Mean increase in organ contour rating, %
Liver	reference	+9.7%
Spleen	reference	+10.6%
Stomach	reference	+3.1%
Kidney	reference	+6.5%
Collecting system	reference	+12.8%
Bowel	reference	+6.3%
Heart	reference	+10.0%

DDG = Data-driven gating; PET = Positron emission tomography.

^aOverall image quality scale: 1, poor; 2, acceptable; 3, good; 4, very good.

^bMotion effects scale: 1, no; 2, slight; 3, moderate; 4, severe.

^cOrgan contours: 1, ill-defined; 2, moderate blurring; 3, slight blurring; 4, sharp.

Impact of DDG on lesion detectability and clinical management

Overall, in 40/149 patients (27%) PET_DDG resulted in a change of the report, mainly due to additional apparent lesions, Figure 1. When reviewing all cases, we observed that in a total of 12 cases (8%), PET_DDG resulted in a change of further clinical management, Table 6.

Figure 1.

Impact of DDG in the study cohort (n = 149) on lesion detectability and patient management. DDG, data-driven respiratory gating.

Table 6.

Cases (12/149) of patients where DDG led to change in management

Patient	Tracer	Indication for PET	Advantage of DDG	Change in management
#01	FDG	Paraneoplastic inflammatory syndrome	Detection of increased uptake in adrenal gland	Initiation of laboratory work-up and hormone substitution
#02	FDG	Follow-up of pancreatic carcinoma	Liver metastasis detected	Upstaging and change of therapy regimen
#03	DOTA	Staging of pancreatic neuroendocrine tumor	Correct identification of intrapancreatic accessory spleen	Biopsy obviated
#04	FDG	Staging of pancreatic carcinoma	Additional lymph node detected	Alteration of surgical approach
#05	DOPA	Search for pheochromocytoma	Detection of small pheochromocytoma	Surgical resection
#06	FDG	Follow-up of breast cancer	Focal lesion in breast detected	Initiation of breast ultrasound and biopsy
#07	FDG	Follow-up of chorioncarcinoma	Liver metastasis detected	Change of therapy regimen
#08	FDG	Follow-up of nasal carcinoma	Additional lymph node detected	Radiotherapy-field adapted
#09	FDG	Staging of cerebral lymphoma	Focal lesion in liver detected	MRI of liver performed
#10	FDG	Follow-up of thyroid cancer	Homogenous liver uptake (focal lesion in non-DDG images)	Refrain from additional work-up
#11	DOTA	Staging of gastric neuroendocrine tumor	Lymph node metastasis detected	Upstaging and change of therapy regimen
#12	DOTA	Staging of familial MEN1	Retroperitoneal NET detected	Upstaging

DDG = Data-driven gating, FDG = Fludeoxyglucose, PET = Positron emission tomography.

A representative case illustrating an additional apparent lesion with PET_DDG is presented in Figure 2. Representative cases with change of clinical management by using PET_DDG for ¹⁸F-FDG, DOTATATE, and ¹⁸F-DOPA, respectively, are presented in Figures 3–5.

Figure 2.

Representative case of ¹⁸F-FDG-PET/CT imaging of a 62-year-old female with breast cancer, using a digital PET/CT system without respiratory gating (a–d) as well as using DDG;(e–h) . An additional apparent lesion in the liver (arrow) was noted in PET images using DDG leading to a change of report but in this case not clinical management as there are multiple other liver metastasis present; (SUV_max in PET_nonDDG = 3.2; and SUV_max in PET_DDG = 6.1). DDG, data-driven respiratory gating; FDG, fludeoxyglucose; PET, positron emission tomography; SUV standardized uptake value.

Figure 3.

Representative case of ¹⁸F-FDG-PET/CT imaging of a 37-year-old female (corresponding to patient #7 of Table 5) with chorioncarcinoma undergoing PET/CT for follow-up. PET/CT images without respiratory gating (a–c) as well as using (d–f). A tiny focus of increased FDG-uptake in the liver is noted (arrow) only in PET images using DDG compatible with a liver metastasis (SUV_max in PET_nonDDG = 3.7; and SUV_max in PET_DDG = 4.6). DDG, data-driven respiratory gating; FDG, fludeoxyglucose; PET, positron emission tomography; SUV standardized uptake value.

Figure 4.

Representative case of 68Ga-DOTATATE-PET/CT imaging of a 60-year-old male (corresponding to patient #11 of Table 5) with neuroendocrine tumor of the stomach undergoing PET/CT for follow-up. PET/CT images without respiratory gating (a–c) as well as using (d–f). A new retroperitoneal lymph node metastasis was detected (arrow) only in PET images using DDG; with magnified images of the retroperitoneal region (asterisk), (SUV_max in PET_nonDDG = 5.7; and SUV_max in PET_DDG = 8.5). DDG, data-driven respiratory gating; FDG, fludeoxyglucose; PET, positron emission tomography; SUV standardized uptake value.

Figure 5.

Representative case of ¹⁸F-DOPA-PET/CT imaging of a 48-year-old female (corresponding to patient #5 of Table 5) with previous pheochromocytoma on the left side that was surgically resected now undergoing PET/CT search for pheochromocytoma on the right side. PET/CT images without respiratory gating (a–d) as well as using (e–h). A tiny focus of increased DOPA-uptake in the right adrenal gland is noted (arrow) and the patient underwent surgical resection of the adrenal where a small pheochromocytoma was proven, (SUV_max in PET_nonDDG = 5.5; and SUV_max in PET_DDG = 10.3). DDG, data-driven respiratory gating; FDG, fludeoxyglucose; PET, positron emission tomography; SUV standardized uptake value.

Discussion

This study sought to evaluate the impact of DDG (i.e. software-based respiratory PET gating) on image quality, lesion detection and patient management in an unselected oncological cohort using a latest generation silicon-based digital detector PET/CT scanner.

The major findings of our study are as follows: (1) the real-time use of DDG was feasible in clinical routine during a test period of 6 weeks; (2) DDG was activated in 85% of the patients, resulting in a mean time increase by 4:37 min (mean total scan time 15:16 min); (3) while general image quality was not affected by DDG, its use resulted in a significant reduction of motion artifacts and sharper delineation of organs (particularly liver and spleen); and (4) the use of DDG resulted in additional apparent lesions in 27% of patients and (5) the use of DDG translated into a change of management in 8% of patients. Following the results of our study, we have implemented DDG in our clinical routine and use it in our standard protocols.

PET/CT has evolved as an invaluable tool for staging of cancer patients in many different etiologies.³ Respiratory motion during image acquisition may, however, significantly degrade PET image quality and even render the detectability of small lung or liver lesions impossible.^9,10,21 To decrease the effect of motion, respiratory gating can be used. This requires the acquisition of a gating signal, usually by external devices, such as a pressure belt or a video camera, which may be inconvenient for the patient and technically challenging for technicians.¹¹

Contrarily, software-based DDG is able to detect respiratory motion within PET data, using the static phase for reconstruction, without the need for additional hardware or inreased radiation dose.²² Indeed, the DDG software used in our study was introduced to our clinical routine from the beginning without any specific patient preparation or instruction used, and after minimal training of staff. Using an R-factor of >10.0, the fully automated respiratory gating method was activated in the majority of patients (i.e. in 85%). This prolonged the scan time by 4:37 min in the whole cohort (up to 10 min in single subjects). One may argue that this is a relevant time increase which potentially impacts on clinical workflow. However, given the significant reduction of motion artifacts, improved visualization of various organ and particularly the amount of additional apparent lesions, we believe this approach may merit consideration.

Previous studies already reported an advantage of respiratory gating for PET/CT.^23–25 Guerra et al, e.g. described in a multicenter, retrospective study the benefits in reporting accuracy and quantification of lung lesions by increasing the overall accuracy in lung lesion detection and characterization.²⁵ In another multicenter, retrospective study Crivellaro et al reported respiratory-gated PET/CT technique as a valuable clinical tool in diagnosing liver lesions by reducing undetermined findings, improving diagnostic accuracy, and confidence in reporting.²⁴ Further, benefits from PET respiratory gating in a prospective study of 74 patients were described by Büther et al.²³; improvements in diagnostic image quality, increase in SUV, and decrease in metabolic volumes were reported with gated PET images as compared to non-gated PET images.

More recently, two different prospective studies aimed to directly compare the data-driven method to an external device-based system for the respiratory gating in PET. In a cohort of 56 patients undergoing whole-body ¹⁸F-FDG PET, Buther et al²⁶ found that DDG-based motion-correction delivered qualities comparable to hardware-based approaches, even if SUV measurements were significantly higher in both reconstruction compared to static images (p < 0.001). On a larger cohort of 144 whole-body ¹⁸F-FDG PET examinations, Walker et al²⁰ found that the use of DDG resulted in an increase in lesions’ SUV_max (i.e. 0.66 ± 0.1 g ml⁻¹, p < 0.0005) and a reduction in lesions’ PET volume compared to the external device-based system. Moreover, while the device-based gated reconstruction failed in 16% of exams, the ones using DDG always provided a clinically acceptable image. In another recent study by Walker et al, the authors concluded that DDG provided superior performances to that of an external device-based system.²⁷ We observed a slightly higher rate of attenuation correction artifacts in DDG compared to non-gated images (9 cases vs to 7 cases), although not statistically significant, this is an issue that should be investigated further.

We acknowledge that our study has some limitations. First, this current evaluation of DDG included an unselected oncological cohort of consecutive patients which resulted in a relatively heterogenous study population. The impact of DDG in different diseases and different tracers used may vary, especially as different radiotracers have different distribution properties and are thus expected to result in more or less successful gating outcomes. We therefore strongly ecourage future studies to systematicelly assess the impact and value of DDG in different diseases and radiotracers. Second, a clinical reader assessment as the one performed in our study, might carry an inherent bias since it is virtually impossible to completely blind readers to the image ‘appearance’ of different reconstruction algorithms. Third, results in our study are based on an R-factor of >10.0. The use of other R-factors leads to a different time consumption of DDG and may ultimately also differently impact lesion detection and patient management. Fourth, we did not compare DDG with an external gating technique in our study. Fifth, we did not compare the impact of prolonged non-gated PET acqusition on image quality nor compared a prolonged non-gated PET with DDG. Sixth, we only evaluated one respiratory gating method in our study using prospective on-console reconstruction, which roughly doubles the acquisition time of gated bed positions, and not includes remote offline reconstruction of 4D respiratory phase-matched PET data. While the DDG approach used in our study involves additional acquisition time in beds where motion was detected (in our cohort on average 1.8 bed positons per patient, if motion was detected), other techniques, being partly not clinically available yet, do not prolong the acquisition times and may thereby be more accepted in clinical environments. Such techniques rely on post-processing of gated PET image datasets of step-and-shoot acquisition, utilizing 100% of coincidence events, or on continuous-bed-motion acquisition²⁸ or other techniques based on a optical flow-based de-blurring of motion effects within image reconstruction.²⁶

In conclusion, the deviceless DDG based on a principal component analysis of PET data was evaluated. It provided reliable fully automatic motion correction in clinical routine, and increased lesion detectability and changed management in a considerable number of patients. Thanks to this new technique, PET/CT with respiratory gating can be used routinely in clinical practice without external gating equipment needed. Future studies should systematically assess the impact of DDG in different disease entities and radiotracers to further improve the clinical value of this technique.