Blood sample detector design
The proposed detector consists of two detection units, each one made with a 50 × 50 × 25 mm3 CsI(Tl) scintillation crystal with diffuse surfaces (Scionix Holland, Bunnik, The Netherlands) wrapped with a white diffuse plastic reflector. On the center of one of the 50 × 25 mm2 faces, an area of 4 × 4 mm2 is left unwrapped for optical readout. CsI(Tl) was selected for its high detection efficiency, high light yield, and absence of intrinsic radioactivity, which allows performing measurements in single and coincidence modes making it suitable for PET and SPECT radiotracers. Each crystal is coupled with optical grease to a SiPM (ASD-RGB4S-P, AdvanSiD, Trento, Italy) with an active area of 4 × 4 mm2 operated at 31 V. SiPMs are compact, MR compatible, can be operated at low voltage, and have a very high gain (106). SiPMs are connected to amplification boards based on inverting transimpedance amplifiers (ASD-EP-EB-N - SiPM Evaluation Board, AdvanSiD, Trento, Italy). Crystals are placed side-by-side along one of the 50 × 50 mm2 faces with an 11.5-mm gap between them (see Fig. 1a). The crystals, SiPMs, and readout electronics are placed within a 3D-printed enclosure (see Fig. 1b). The blood sampling catheter is placed in the center of the gap between crystals in a 3D-printed U-grooved holding cassette. This piece can be fixed to the detector enclosure, allowing for measurements in single and coincidence modes and ensuring a good reproducibility. The detector shielding is made of a double layer of 3-cm-thick lead bricks in order to minimize the detection of external radiation.
Data acquisition and signal processing
The signals from both detection units are digitized (15 MS/s) with an oscilloscope (Picoscope Series 2206A, Pico Technology Ltd, Cambridgeshire, UK) and sent to a PC for further processing using a custom-made application based on the oscilloscope’s C++ API. All pulses triggered by a falling edge discriminator with a threshold of −50 mV are integrated for 2 μs (no filters were applied) to obtain the energy deposited by the gamma photons and stored as single events for each detector. The signal is also recorded for 1 μs pre-trigger, and the average of those pre-trigger samples is used for baseline correction. Time stamps are generated by the leading edge discrimination in both detectors. Single events with a time difference below 300 ns are also stored as coincidence events. Random coincidences are estimated based on the single rates [19]. Environmental background events are subtracted from single and coincidence events. For that purpose, a 2-h measurement with no activity in the catheter was recorded.
Device characterization
Sensitivity, energy resolution, coincidence resolving time and count rate losses
In order to characterize the detector, an 0.8-mm internal diameter (ID) catheter (Tygon S3 E-3603, Saint-Gobain Performance Plastics Co., Akron, OH) was filled with 18FDG at an initial activity concentration of 900 kBq/mL and placed in the detector with the active length centered with the detection units. A 15-h acquisition was performed consisting of 150 frames of 1-min duration with 5-min gap intervals. The energy spectra were obtained and calibrated using the 511-keV peak for every acquisition in order to compensate for temperature-dependent gain variations. The energy resolution at the 511-keV photopeak was obtained by fitting to a Gaussian.
Sensitivity was computed as the slope of the linear fit of the events rate (singles or coincidences) measured within an energy window of 350–700 keV against their corresponding activity concentration. This fit was performed for the low count rate measurements. Count rate losses were estimated for single events assuming that no events are lost at low count rate. Coincidence resolving time (CRT) was derived from the time difference histogram for coincidence events within the same energy window of 350–700 keV by interpolation between the two bins on each side of the maximum that are immediately above and below the half maximum.
Minimum detectable activity
Minimum detectable activity (MDA) determines the smallest activity concentration that can be detected with a certain confidence level [20]. MDA depends on sensitivity of the detector (s), the duration of the measurement (T), and background events (NB) detected during T. In terms of kBq/mL, and at a 95% confidence level, MDA can be described as:
$$ \mathrm{MDA}=\frac{4.65\ \sqrt{N_B+2.71}}{fTs} $$
(1)
where f is the branching ratio for β+ decays (0.967 for 18F). NB was measured without and with the presence of an external source of activity. In the first case, only the environmental radiation contributes to background radiation, whereas in the second case, the radiation emitted from the patient itself can penetrate the shielding and contribute to background signal. Consequently, MDA was determined in both scenarios for both single and coincidence events using an energy window of 350–700 keV. Details about the setup employed for the background measurements with an external activity source are given in the following section. In each case, background measurements of 2–3 min were performed in three independent acquisitions. The results are presented as mean ± SD.
In vivo evaluation
The performance of the detector was evaluated in vivo in animals injected with 18FDG. Three healthy female large white pigs (mean weight = 45 ± 4 kg) were anesthetized by intramuscular injection of ketamine (20 mg/kg), xylazine (2 mg/kg), and midazolam (0.5 mg/kg), and maintained by continuous intravenous infusion of ketamine (2 mg/kg/h), xylazine (0.2 mg/kg/h), and midazolam (0.2 mg/kg/h). Oxygen saturation levels via pulse oximetry, and electrocardiogram signal, were monitored throughout the study. The coccygeal artery of the animal was cannulated, and arterial blood was withdrawn through an 0.8-mm ID catheter at 5 mL/min using a peristaltic pump. The animals received a bolus injection of heparin and the catheter was washed with heparinized saline to prevent clotting. The animal was placed in a PET/CT Gemini TF-64 scanner (Philips Healthcare, Best, The Netherlands) and the scanner table was moved to the position were cardiac PET acquisitions are routinely performed. The detector was placed at about 40 cm from the animal’s tail in order to minimize blood dispersion inside tubing. 18FDG (155 ± 12 MBq) was prepared in 6 mL and infused at a rate of 1.0 mL/s through a marginal ear vein, followed by a 6-mL saline flush at the same rate. The acquisition with the detector started with the radiotracer injection and lasted for 5 min. The AIF was obtained with the detector (AIFD) in consecutive 5-s frames using the single events recorded in the 350–700 keV energy window and converted to activity concentration using the sensitivity previously obtained. Decay correction was applied. Afterwards, measurements for MDA determination with an external source of activity were performed by placing an empty catheter in the device and leaving everything else in the same position.
In order to validate AIFD results, blood samples were collected into sample tubes after passing through the detector following this temporal scheme: 20 × 5 s, 8 × 10 s, 6 × 20 s. Then, the tubes were briefly centrifuged to provide a reproducible geometrical distribution of the blood and later analyzed using a well counter (Wallac 1470 Perkin Elmer, Waltham, Massachusetts, USA) applying dead time and decay corrections. The well counter was previously calibrated to convert measurements to activity values. The volume for each blood sample was determined as the weight difference between empty and filled tubes and applying a blood density of 1.03 g/mL [21]. Finally, the activity concentration of 18FDG was calculated for each blood sample obtaining the AIF derived from the well counter (AIFWC). The time delay existing between AIFD and AIFWC was corrected by maximizing the cross-correlation between both curves, which had been previously interpolated every 5 s.
Multi-tracer AIF detection by spectroscopic analysis
Pure β+ isotopes such as 18F only emit positrons that result in 511-keV annihilation photons, while non-pure β+ isotopes like 68Ga emit additional non-annihilation photons, although only those with 1.077 MeV are emitted in a significant fraction of the decays (3.22%) [22]. Thus, the ratio of events recorded at high-energy (> 750 keV) and low-energy windows (350–700 keV) (see Fig. 2) can be used to determine the amount of 18F and 68Ga in a sample containing any combination of both isotopes.
Two acquisitions were made in order to obtain the calibration data required to implement the proposed method. An 0.8-mm ID catheter was filled with 68Ga or 18F respectively with an initial activity concentration of 1200 kBq/mL (measured with an activimeter). Afterwards, the catheter was placed in the device and data was recorded for several hours acquiring frames of 1 min with 5-min gap intervals. Each frame was processed to obtain the total single rate at high-energy window (750–2000 keV, SHE) and the coincidences rate at low-energy window (350–700 keV, CLE). The ratio between SHE and CLE could be used, in theory, to calibrate the measurements for each isotope. However, we observed that, at higher count rates, the increase of SHE is not linear with CLE due to pile-up events. Therefore, in order to include this effect, SHE must be calibrated as a function of CLE. For that purpose, the variation of SHE against CLE was represented and fitted to a third-degree polynomial for each isotope (SHE,F(CLE) and SHE,G(CLE) for pure 18F and 68Ga respectively). When measuring SHE and CLE for a sample containing an unknown mixture of 18F and 68Ga (SHE,Mix), the relative activity of each isotope (aGa,D and aF,D where aF,D + aGa,D = 1) can be obtained using the following expression:
$$ {S}_{\mathrm{HE},\mathrm{Mix}}\left({C}_{\mathrm{LE}}\right)={a}_{\mathrm{Ga},\mathrm{D}}\times {S}_{\mathrm{HE},\mathrm{G}}\left({C}_{\mathrm{LE}}\right)+\left(1-{a}_{\mathrm{Ga},\mathrm{D}}\ \right)\times {S}_{\mathrm{HE},\mathrm{F}}\left({C}_{\mathrm{LE}}\right) $$
(2)
assuming that the contribution of pile-up events is equally distributed between both isotopes.
The ability of the developed detector to obtain separated AIFs of tracers labeled with different isotopes (68Ga and 18F) was tested in vitro. To do so, a catheter filled with a mixture of 73% 68Ga and 27% 18F with an initial total activity concentration of 1200 kBq/mL was placed in the detector. The activity was measured separately for each isotope with an activimeter and the mixture was prepared afterwards. Data was recorded for several hours acquiring frames of 1 min with 5-min gap intervals. The relative activity of both isotopes was constantly changing over time due to their different half-lives (109 min for 18F and 68 min for 68Ga).
The initial activity concentration (c(ti) = 1200 kBq/mL) was much higher than the typical values that can be measured in an AIF of an in vivo experiment on large animals. In the later studies, the activity concentration usually ranges between 10 and 200 kBq/mL, and the duration of the time frames of the dynamic PET acquisition is frequently set from 5 s (in those frames where the activity concentration is high and changes rapidly) up to 60–180 s (for those frames where the activity concentration is low and the changes over time are less significative). Hence, in order to evaluate our experiment in more realistic conditions, our 1-min acquisitions were trimmed depending on the activity concentration of each acquisition as follows:
$$ \left\{\begin{array}{c}\kern1em t=60\ \mathrm{s}\kern2.25em ,\kern0.5em \mathrm{when}\kern4.75em \max\ \left(c(t)\right)\le 10\ \mathrm{kBq}/\mathrm{mL}\kern8.25em \\ {}\kern1em t=10\ \mathrm{s}\kern2.25em ,\kern0.5em \mathrm{when}\kern4.5em 10\ \mathrm{kBq}/\mathrm{mL}<\max\ \left(c(t)\right)\le 50\ \mathrm{kBq}/\mathrm{mL}\ \kern1.5em \\ {}\kern0.75em t=5\ \mathrm{s}\kern3em ,\kern0.5em \mathrm{when}\kern4.5em 50\ \mathrm{kBq}/\mathrm{mL}<\max\ \left(c(t)\right)\kern8em \end{array}\right. $$
(3)
SHE,Mix and CLE were obtained for each frame and aGa,D was calculated using Eq. (2). These results were compared against theoretical values (aGa,A) obtained from their initial activity and taking into account their half-lives. The mean relative difference between aGa,D and aGa,A was obtained and reported as percentage.