To implement this study, a Biograph VISION 450 camera was used (Siemens Healthineers, Knoxville, Tennessee). This camera consists of 6 rings of detectors with a total axial field of view coverage of 19.7 cm. Each detector block is composed of an array of 4 × 2 detection units. Each detection unit consists of an array of 5 × 5 lutetium oxyorthosilicate cerium-doped (LSO:Ce) crystal elements of 3.2 × 3.2 × 20 mm size coupled with a SiPM detector with a dimension covering the entire scintillation area [17]. The camera was cross-calibrated with the two-dose calibrators used for the preparation of radioactive solutions in this study: a Unidose dispenser (Trasis, Ans, Belgium) equipped with a VIK-202 dose calibrator (Comercer, Joure, Netherlands) for 18F and a shielding hot cell equipped with a CRC-55tR dose calibrator (Capintec, Florham Park, New Jersey) for 68Ga.

Twenty-minute emission scans in one bed position were performed with list mode data acquisition and TOF measurement. To assess the impact of imaging 4 rats simultaneously on image quality and quantification analysis using appropriate phantoms, the tests were repeated for two configurations, as follows: first the measurement phantoms were acquired in a centred position in the FOV without a scattering source (C-SC−) and then in the off-centre position at a radial and tangential offset of 5 cm, with 3 additional scattering sources to mimic the presence of 4 rats (OC-SC+), as illustrated in Fig. 1a. The scattering sources used in the experiments were fillable cylinders with a height of 90 mm and an internal diameter of 4 mm. All experiments were carried out for both 18F and 68Ga radionuclides.

Diagram showing the different phantom locations for acquisitions mimicking the presence of 4 animals in the FOV seen in transverse and longitudinal planes (a) and drawings of the NEMA NU 4-2008 image quality phantom (b), Micro-hollow sphere phantom (c) and 3D-printed rat phantom (d) with the corresponding picture (e)

CT acquisitions were performed for attenuation correction purposes with the following parameters: tube voltage of 80 kV, tube current of 23 mA, table moving speed set to a pitch of 1, total collimation aperture of 19.2 mm, slice thickness of 1.5 mm and a matrix size of 512 × 512.

Images were reconstructed for attenuation and scatter diffusion correction with a zoom factor of 2 and a matrix size of 880 × 880. The resulting pixel size of the reconstructed images was 0.4125 × 0.4125 mm. Depending on the test performed, images were reconstructed using 3 different algorithms: filtered backprojection (FBP), 3D-ordered subset expectation maximization algorithm (3D-OSEM) and vendor-proprietary iterative algorithm modelling PSF (TrueX) using 5 iterations (i) and varying the number of subsets (s) from 3 to 30.

The spatial resolution of the system was assessed using capillaries with a 1 mm internal diameter filled with an 18F solution of 0.5 MBq/mL. Two emission scans of 10 min duration were performed with capillaries placed first in the centre of the FOV and then in position 1/4 (Fig. 1a), corresponding to the off-centre position. Images were reconstructed with the FBP, 3D-OSEM and PSF algorithms as described previously in the PET/CT section. Profiles passing through the maximum pixel value of each source were drawn in the radial, tangential and axial directions using AMIDE software [18]. The full width at half maximum (FWHM) and the full width at tenth maximum (FWTM) in mm were determined in each direction using linear interpolation between adjacent pixels at half and one-tenth of profile’s maximum value, respectively. The maximum value was determining using a parabolic adjustment of the peak, as recommended by the NEMA NU 4-2008 standard.

Image quality was assessed using a NEMA NU 4-2008 PET image quality phantom (Fig. 1b). This phantom consists of a fillable chamber with a height of 63 mm and an internal diameter of 33.5 mm and is composed of three different sections. The first part, containing two cylindrical inserts of 14 mm height and 8 mm inner diameter filled with air and cold water, respectively, was used to evaluate the spill-over ratio (SOR). A second part, free of structures, was used to evaluate the reconstructed image uniformity. A third part, containing a solid PMMA insert drilled with 5 holes of 5, 4, 3, 2 and 1 mm inner diameter, was used to evaluate recovery coefficients (RC).

The phantom was filled with either 18F or 68Ga solution of approximately 3.7 MBq of activity at the start of acquisition. Two emission scans of 20 min duration were performed for C-SC− and OC-SC+ acquisitions. The activity prepared for the 3 scattering sources was approximately 4 MBq. The time duration for the second acquisition was adapted to consider the radiation decay.

To assess uniformity, a cylindrical VOI 22 mm in diameter and 10 mm in length was drawn in the uniform part of the phantom. The average, minimum, maximum and standard deviation (SD) were measured.

Images were reformatted to obtain 10-mm-thick slices centred on the hot rods. Circular ROIs encompassing each rod were drawn, with dimensions twice the physical size of the rods. The maximum pixel value of each ROI was measured, and the RC was calculated as follows:

Line profiles (10 mm) were drawn for each rod in the axial direction, passing through the coordinate of the maximum pixel value measured previously. The mean (Meanlineprofile) and SD (SDlineprofile) of the pixel values measured along the profile were measured, and the SD of the RC were calculated as follows:

$$_=RCx\sqrt_}_})}^+(_}_})}^}$$

With Meanbackground and SDbackgroud, the mean and SD values were measured in the uniform part.

VOIs of 4 mm diameter and 7.5 mm length were drawn on the air and water insert to calculate the spill-over ratio as follows:

All measurements were performed in accordance with the NEMA NU 4 standards.

To evaluate quantification accuracy, RCs were computed using a hot spot phantom equivalent to a small animal in size, filled with an activity concentration ratio between spheres and background. The phantom used for this test was a micro-hollow sphere (MHS) phantom that consists of a fillable cylinder with an internal diameter of 4 cm and a height of 8.2 cm, containing 4 hollow spheres with inner diameters of 7.86, 6.23, 4.95 and 3.95 mm (Fig. 1c). Spheres and background were filled with either 18F or 68Ga solutions. For the F18-filled phantom, two different sphere-to-background ratios were tested, filling the background with an activity of 4 MBq at the start of acquisition and the spheres with an activity concentration of 0.3 MBq/cc to obtain a contrast ratio of 1/8 or 0.14 MBq/cc for a ratio of 1/4. For the 68Ga experiment, only the 1/8 contrast ratio was tested. For the image quality phantom experiment, C-SC− and OC-SC+ acquisitions were performed to test the impact of imaging 4 rats simultaneously. Images were reconstructed using 3D-OSEM and PSF algorithms by varying the total number of iterations to determine the optimal PET reconstruction settings for quantification purposes.

To determine the RC values, the method proposed in the EANM procedure guidelines for FDG tumour PET imaging was used [19]. A cylindrical region of interest (ROI) was drawn in the uniform part of the phantom to determine the mean activity in the background. Four spherical volumes of interest (VOIs) encompassing the hot spheres were drawn to determine the maximum pixel value of each sphere. 3D isocontours at 50% of the maximum value adapted for background activity measured in the uniform part were delineated to determine the mean activity in the spheres (A50).

RC max and RC A50 were calculated for each sphere as follows:

$$RC max=\frac^)}^\right)}$$

$$RC A50=\frac^\right)}^\right)}$$

To test quantification in preclinical conditions, four 3D-printed rat phantoms represented in Fig. 1d, e were used. These phantoms were modelled by manually contouring the main organs visible on CT images of a real rat previously imaged on a preclinical CT device by our team as described in a previous study [20]. The resulting rat phantoms consisted of two air cavities simulating the lungs and four fillable cavities of 5.6, 0.2, 0.6 and 0.7 mL simulating the liver, bladder (only for two of the four rats), right and left kidneys, respectively. Four spherical cavities with two different sizes were added to the upper left, upper right, bottom left and bottom right of the rat phantom to mimic the presence of pertinent sized subcutaneous tumours. The two tumours located on the left side of the rat had an internal diameter of 10 mm corresponding to a volume of 0.53 mL, and the two tumours located on the right side had a diameter of 8 mm corresponding to a volume of 0.27 mL. The phantoms were printed with a Tevo Tornado 3D printer (Tevo 3D Electronic Technology Co. Ltd., China) using polyethylene terephthalate glycol plastic as the printed material.

The liver, kidneys and bladder were filled with 18F solutions of 0.3, 1.2 and 2.4 MBq/mL at the start of acquisition. Tumours of the rats were filled with different activity concentrations to obtain different contrast ratios between tumours and organs. The four tumours were filled with 0.3, 0.6, 1.2, and 2.4 MBq/mL of 18F. The tumour locations of the different radioactive concentrations were alternated for each rat. Twenty-minute acquisitions were repeated five times every 30 min to simulate a tumour activity reduction. The resulting total activity present in the four rats varied approximately from 5.7 MBq to 2.85 MBq from the first to the last acquisition. The same experiment was repeated by filling the rats with 68Ga in place of 18F. A total of 80 tumours with variations in size, location and activity concentration were thus quantified in this study for both radionuclides. Images were reconstructed using the PSF algorithm with 20 i and 5 s. For 18F acquisitions, a 2-mm FWHM Gaussian filter was applied in addition to complying with the best reconstruction settings determined with the MHS phantom. For each tumour visible on the different acquisitions performed, the quantification parameters Amax and A50 were determined as described in the recovery coefficient section and compared to the theoretical activity concentration (Acalc) calculated at the acquisition start.