This is a single center cohort study. The need for written informed consent was waived by the local ethics committee as the study was conducted retrospectively from data obtained for clinical purposes (reference number 2021-9835). A sequential cohort of 230 patients with squamous cell carcinoma of the oropharynx, hypopharynx or larynx that received definitive (chemo)radiotherapy between November 2020 and October 2021 was screened for eligibility. Acquisition of a [18F]FDG PET-CT scan in radiation treatment position with both EARL1 and EARL2 PET image reconstructions was mandatory. As from November 2020, acquisition of both EARL1 and EARL2 reconstructed images is standard practice at our institution. Patients with previous oncologic treatment (e.g. radiotherapy or tumor reductive surgery) of the head and neck area were excluded.

A PET scan, a low-dose CT scan for attenuation correction and an iodine contrast enhanced diagnostic CT scan for radiation treatment planning was acquired in one session on a Biograph mCT40 PET-CT scanner (Siemens Medical Solutions, Knoxville TN, USA). Imaging was acquired in radiation treatment position, using a customized neck support (AccuForm Custom Cushions, Accuform, MEDTEC, Orange City, IA) and a five-point fixation mask for immobilization of the head, neck and shoulders (HNS Mask-Nose Hole in Efficast 2.0 mm MAXI, Orfit Masks, Orfit Industries NV, Wijnegem, Belgium). Prior to [18F]FDG administration, patients fasted for at least 4 h and a serum glucose level of < 11 mmol/L was mandatory. [18F]FDG was intravenously administrated approximately 60 min prior to the scanning procedures (dose calculated using Eq. 1) [7].

$$\left[ }} \right]} \left( }} \right) = \frac}*\frac}}}}}}*\frac}}}} \right)*}\; } \left( }} \right)}}}\; }\; }\; }\; }\; }\; \left( }}}}}}} \right)}}.$$

(1)

All patients were scanned from the lower border of the clavicle to the cranium. The acquisition time was 3 min per bed position with an overlap of 43% between bed positions. The slice thickness of the CT-scan was 3 mm. EARL1 images were reconstructed with an ordered-subsets expectation–maximization (OSEM) algorithm including point spread function and time-of-flight, Gaussian Filter 7.5 mm full width at half maximum, image matrix 256*256 and voxel size 4.1*4.1*5.0 mm. EARL2 images were reconstructed with point spread function and time-of-flight OSEM, Gaussian Filter 4.3 mm, image matrix 400*400 and voxel size 2.0*2.0*5.0 mm.

For all patients, [18F]FDG PET-CT scans were imported into the radiation treatment planning system Pinnacle version 3.2.0.27 (Philips Medical Systems, Fitchburg, MA, USA). Primary tumors were delineated on the CT-scan based on information gathered from physical examination and diagnostic imaging. The volume and the maximum [18F]FDG uptake was determined in terms of SUVmax and maximum standardized uptake ratio (SURmax) on both EARL reconstructions for each delineated tumor. The tumor to cervical spinal cord standardized uptake ratio (SUR) has been shown to improve the reproducibility of quantitative [18F]FDG-PET data in a multicenter setting compared to SUV based approaches [13]. SUVmax and SURmax were calculated using Eqs. 2 and 3.

$$}_}}} \left( }}}^ }}} \right) = \frac}\; } \;} \left( }}}}^ }}} \right)}}}\; } \left( }} \right)*2^} \;} \;} \;}\; } \;}\; \left( } \right)}}}\; }\; } \;} \left( } \right)}}}} }}*}\; } \left( } \right)$$

(2)

$$}_}}} = \frac}_}}} \text}\; }\; \left( }}}^ }}} \right)}}_}}} \;}\; } \;} \;} \left( }}}^ }}} \right)}}$$

(3)

Three different segmentation methods were used to determine MTVs on both EARL1 and EARL2 reconstructed images. Delineation of MTVs was performed automatically with customized Pinnacle scripting. Thresholds used were (1) an adaptive threshold as a percentage of SURmax (threshold = 116.93 * SURmax−0.75) [13], (2) a percentage of maximum [18F]FDG uptake (MAX40% and MAX50%) and (3) a static threshold of SUV or SUR (2.5, 3.5 and 4.5). For the SUV based segmentation methods, TLGs are calculated using Eq. 4.

$$} \left( } \right) = } \; \left( }^ } \right)*}_}}} \;} \;} \left( }}}^ }}} \right).$$

(4)

The method of classification errors (CE) was used to evaluate spatial overlap of the MTVs based on EARL1 and EARL2 reconstructed images [14]. An important advantage of the CE method is that it does not only take volume into account, but also the spatial position and shape of the contours due to both false-negative and false-positive volumes. The CE can range from 0 to infinite, in which a lower CE implies better spatial overlap, and is calculated using Eq. 5.

$$}\; } = \frac}\; }\; }\; \left( }^ } \right) + } \;}\; } \;\left( }^ } \right)}}}\; }\; } \;}1\, }\; \left( }^ } \right)}}$$

(5)

The false-negative volume is defined as MTV that is delineated on EARL1 but not on EARL2, and vice versa for the false-positive volume.

All lymph nodes having a short-axis diameter of ≥ 5 mm in the axial plane were manually delineated on the CT-scan. This threshold was chosen because histopathological validation studies suggest that nodal metastases of this size can be detected by [18F]FDG PET-CT [15, 16]. Necrotic lymph nodes with irrefutably disturbed [18F]FDG distribution were not considered. Short-axis diameters, nodal volumes and quantitative [18F]FDG uptake parameters (i.e., SUVmax and SURmax) on both EARL1 and EARL2 reconstructed images were determined for each node.

For the qualitative analysis of nodal [18F]FDG uptake, both the EARL1 and EARL2 reconstructed image series were read independently by two experienced nuclear medicine physicians (AA and MvR), specialized in head and neck cancer. Each lymph node was scored separately on EARL1 and EARL2 reconstructed images using a 4-point scale (1—definitely benign, 2—probably benign, 3—probably malignant, 4—definitely malignant). To minimize observer recall bias, the time between scoring EARL1 and EARL2 reconstructed images was at least 4 weeks. Discrepancies between observers involving scores ‘3—probably malignant’ or ‘4—definitely malignant’ were resolved by consensus. Visual scores on EARL1 and EARL2 images were compared to identify consequences for radiation treatment and for N-classification (8th edition of UICC TNM classification) [17]. A change from score 1 or 2 to score 3 or 4 or vice versa was assumed to have consequence for staging and treatment.

All statistical analyses were performed using SPSS version 26 (IBM Corporation, New York, NY, USA). Statistical significance level was set to p < 0.05. Normal distribution of data was tested using the Shapiro–Wilk test. Data characterized by normal distribution were presented as mean with 95% confidence interval (95% CI) and parameters not normally distributed as median with the interquartile range (IQR). Scatter plots and intraclass correlation coefficients (with a two-way mixed model testing absolute agreement) were used to describe the relationship of maximum [18F]FDG uptake (i.e., SUVmax and SURmax) on EARL1 and EARL2 reconstructed imaging. To evaluate the magnitude of differences between EARL1 and EARL2, the relative differences in SUVmax or SURmax were plotted against the average SUVmax or SURmax on both EARL reconstructions, according to the Bland–Altman method. Mean/median differences in quantitative metrics between both EARL reconstructions were calculated based on the differences of paired data. Comparison of means between groups was done using the Student T test for paired data in case of a normal distribution and the Wilcoxon signed rank test for data not normally distributed. In the qualitative nodal evaluation, agreement between observers was calculated by the kappa statistic. The kappa score can range between 0 and 1, with a score of 0.00–0.20 indicating none to slight interobserver agreement, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 substantial, and 0.81–1.00 almost perfect agreement [18].