Abstract
Objectives
Yttrium 90 (90Y) microspheres are utilized in the treatment of liver tumors. This study aimed to prepare 90Y suitable for experimental studies by separating it from within the microspheres and imaging it cartographically in positron emission tomography/computed tomography (PET/CT).
Methods
90Y microspheres were incubated for 2 hours in a solution containing 5.5 molar sodium chloride (NaCl) and 6 molar hydrochloric acid at pH 5. After this period, separation was achieved by passing through a 0.22 µm pore size polytetrafluoroethylene filter. A liver phantom, mimicking human liver equivalent dimensions and containing two simulated tumors, was filled with 90Y and imaged scintigraphically in PET/CT.
Results
The true 90Y ratios between the large simulated tumor and the small simulated tumor compared to normal liver tissue were approximately 10:1 and 2:1, respectively. PET/CT images yielded these ratios as 11.3 and 2.13 for the large and small tumors, respectively. Homogeneous distribution of 90Y within the simulated liver tissue was observed.
Conclusion
The separation technique enabled preparation of a homogeneous ionic 90Y solution suitable for quantitative PET/CT phantom imaging and may facilitate future phantom-based dosimetric studies.
Introduction
Selective internal radiation therapy using yttrium-90 (90Y)-loaded microspheres is an effective treatment modality for liver tumors. In this method, high radiation doses are delivered directly to the tumor via the hepatic artery using catheter-based techniques (1,2,3,4,5). Radioembolization is a targeted technique involving the administration of 90Y-containing microspheres into specific branches of the hepatic artery. After intrahepatic arterial infusion, 90Y microspheres become trapped within the microvascular bed. When 90Y decays, it emits beta (β-) radiation with an average energy of 0.93 MeV and a maximum energy of 2.26 MeV. Electrons at these energies can generate Bremsstrahlung (deceleration X-rays) when interacting with tissue. Gamma camera imaging utilizes these X-rays. In addition, 90Y emits low-yield annihilation photons at a rate of approximately 32 per million decays, enabling positron emission tomography (PET) imaging (6).
In liver tumor treatment, two commonly used dosimetry approaches exist: pre- and post-treatment dosimetry.
Pre-treatment dosimetry is performed using 99mTc-MAA-based scintigraphic imaging with a gamma camera. The first step in dosimetric calculations is the determination of the lung shunt fraction, which reflects the fraction of activity shunted from the liver to the lungs. In addition, the radiation tolerance of normal liver tissue and the intended tumor absorbed dose can be calculated for treatment planning (7,8,9,10). After calculations performed with 99mTc-MAA, the determined dose is administered to the patient using 90Y microspheres (11). It is well known that the intrahepatic distribution of 99mTc-MAA microspheres differs from that of 90Y microspheres. Therefore, post-treatment dosimetric studies using 90Y microspheres are also performed (12,13). For this purpose, some studies have used phantoms simulating the human liver and lungs labeled with 99Tc (14). However, due to technical limitations, no pre-dosimetric phantom studies using 90Y microspheres have been reported. In contrast, post-dosimetric studies using 90Y microspheres in patient imaging do exist. One such study was performed by D’Arienzo et al. (15) using a phantom with 90Y chloride. In that study, 90Y was used in chloride form without microspheres, allowing a homogeneous distribution of 90Y in PET/computed tomography (CT) imaging and enabling dosimetric analysis.
Partial volume effect (PVE) is a phenomenon that complicates scintigraphic differentiation in small liver tumors, as in all nuclear medicine imaging. PVE is a complex process influenced by several factors, including tumor size and shape, background activity in surrounding tissues, spatial resolution of the imaging system, voxel size, and imaging modality. Chiesa et al. (16) reported that lesions smaller than 3 cm are affected by PVE, and particularly tumors smaller than 2 cm cannot be reliably corrected for PVE and therefore should be excluded from evaluation. These investigators studied PVE both in phantoms and in patients using 99mTc. Further phantom studies are required for PET imaging of small liver tumors using pure 90Y, as tumor diameter cannot be adequately characterized.
Before separation, 90Y-labeled microspheres contain a partially cross-linked cation-exchange polymer matrix composed of polystyrene resin. During labeling, 90Y3+ replaces sodium ions (Na+) via ion exchange and binds to the resin matrix, thereby attaching to the microspheres. Subsequently, 90Y is immobilized on the microspheres by precipitation in the form of a phosphate salt. The 90Y-microsphere complex is maintained together by physical forces rather than chemical bonds, resulting in a relatively weak association that may be disrupted in highly ionic environments. SIR-Spheres consist of biocompatible resin microspheres radiolabeled with 90Y. During ion exchange, sodium ions (Na+) are replaced by 90Y3+, which binds to the polystyrene cation-exchange resin layer. Subsequently, 90Y is immobilized on the resin microspheres by phosphate precipitation (17).
Although previous phantom studies have used 90Y chloride solutions, studies investigating the extraction of 90Y from commercial resin microspheres and its subsequent use in phantom imaging are extremely limited. The ability to obtain a homogeneous ionic 90Y solution from microspheres may provide a significant advantage for cost-effective phantom-based dosimetric research and imaging studies. The aim of this study is to separate 90Y from microspheres into ionic form, optimize the preparation of 90Y solution for experimental studies, and validate its accuracy using PET/CT imaging and quantitative analysis.
Material and Methods
No live data were used in this study, and all data were produced from phantom studies. Due to the harmful effects of radiation on living organisms, experimental studies involving radiation are often conducted using organ-mimicking phantoms. In this study, we used a phantom resembling the human liver, containing two simulated tumors with known volumes. The study was planned in two stages: the first stage involved filtration of solutions with microspheres, and the second stage involved imaging and quantification of the phantom using PET/CT. The imaging was performed both when the phantom was labeled with microspheres and after the microspheres were separated from the 90Y.
90Y Microspheres Characteristics and Separation Technique
Commercially available resin-based 90Y microspheres were used as the radionuclide source in this study. The microspheres consist of a polystyrene resin matrix with an average diameter of 32±10 µm and a specific gravity of approximately 1.6 g/dL (15,16). The association between 90Y and the resin matrix can be disrupted under highly ionic conditions, allowing the radionuclide to be separated from the microspheres. To separate the 90Y isotope from this resin, the microspheres were treated with a solution composed of a mixture of 5.5 M NaCl and 6 M HCl at pH 5 [NaCl (pKa: 7.15), HCl (pKa: -6.3), and Y+3 (pKa: 10.64)]. After incubation at room temperature for 2 hours, the solution containing ionic 90Y was separated from the resin particles by passing it through a 0.22 µm PTFE filter. Since the initial trial was successful, further experimental replication was not performed. The microspheres within the 90Y complex have dimensions ranging from 20 to 60 microns (with an average of 32±10 microns), which prevented them from passing through sterilizing filters due to their size. Consequently, when these microspheres were transferred into the product vial in their free form, they became trapped within the vial. Meanwhile, the 90Y radionuclide remained in its free form within the product vial, completing the separation process. The separated ionic form of 90Y radionuclide was then utilized in the second phase of phantom imaging studies.
Liver Phantom and Water Tank
The liver model used in this study consists of three main sections: normal liver tissue, a large tumor, and a small tumor. Collaboration with the chemical engineering department was employed to determine the materials for the phantom. The liver phantom was fabricated from filament material called polylactic acid using fused deposition modeling on 3D printers. The model was designed to have a specific gravity similar to human tissue (d=1.24 g/cm3). The dimensions of the liver phantom are 112 mm x 191 mm with an empty mass of 350 grams, and it includes three caps for filling with radiopharmaceutical solution.
The leftmost cap is used to fill the 690 cm3 volume representing the hepatic parenchyma (normal liver) with liquid radioactive material. The other caps, designed as cylindrical structures with volumes of 9.5 cm3 and 202 cm3 respectively, are used to fill liquid radioactive material into two separate and independently designed tumors of different sizes (Figure 1). These tumor sizes represent commonly encountered small and relatively large volumes. Scintigraphic imaging of the liver phantom was conducted using a tank made of plexiglass that matches the dimensions of a normal human abdominal area and can be filled with water.
Measurement of 90Y Radioactivity, Imaging Protocol and Quantification
Radioactivity measurements were performed using a dose calibrator (Biodex Atom lab 300 dose calibrator). Accuracy tests of the dose calibrator were performed with a Cs-137 calibration source prior to measurement. The activities of 90Y in ionic form, to be injected into the simulated tumors and empty spaces representing normal liver tissue in the phantom, were drawn into syringes and measured in the dose calibrator. Subsequently, the 90Y radionuclide was diluted with water, and activities of 684 MBq, 2025 MBq, and 18.70 MBq were respectively filled into the locations representing normal liver, large tumor, and small tumor in the phantom. Empty syringes were measured again for background correction. This process allowed the determination of the net activity amounts deposited into the phantoms.
The PET/CT images of the phantom were acquired using a Siemens Horizon Biograph model scanner with the subject positioned on a single bed for a duration of 20 minutes. The imaging process was carried out in two stages: in the initial stage, imaging was performed using the radiopharmaceutical labeled with 90Y microspheres. In the second stage, following separation from the 90Y microspheres, the phantom was filled with the microspheres, and imaging was repeated under identical conditions using the same PET/CT scanner.
The image reconstruction was performed according to routine clinical protocol using the OSEM 3D algorithm with 16 subsets and 8 iterations. Quantification PET/CT images was carried out using the OsiriX Dicam Viewer program. In the first stage, after filling the phantom with a solution of 90Y microspheres into the tumor-simulating regions and empty spaces representing normal liver tissue, PET/CT imaging was performed. The PET images showed the settling of 90Y microspheres towards the bottom. Volumes of interest (VOIs) were drawn around the areas where microspheres settled and named as VOI1 and VOI2. in accordance with conventional use. Similarly, the entire liver (including tumors) was named VOI3 (Figure 2). Subsequently, the counts for VOI3–(VOI1+VOI2) were calculated to determine the ratio of settled 90Y microspheres. The ratio of settled 90Y microspheres [(VOI1+VOI2)/VOI3] was determined using the equation.
In the second stage, the actual values of the 90Y activity per unit volume (MBq/mL) for normal liver (N) and tumor (T) were calculated based on the amounts of 90Y activity filled into the phantom. PET images were used to draw 10 VOIs of the same size over the normal liver and separately over the tumors to determine the counts.
Statistical Analysis
The T/N ratios in the 90Y microsphere PET/CT images and the T/N ratios in the 90Y PET/CT images were evaluated using the Mann-Whitney U test. A significance level of p<0.05 was used to indicate a statistically significant difference.
Results
After filling the phantom with the solution of 90Y microspheres into the tumor-simulating regions and empty spaces representing normal liver tissue, PET/CT imaging was conducted. VOIs were drawn around the areas where microspheres settled. These regions were named VOI1 and VOI2, and their total counts were determined (Figure 2). The total count for VOI1 was 82484 and for VOI2 it was 62039. The total count for the entire liver (including tumors) was determined by drawing the VOI around the area of interest, resulting in a count of 151204 for VOI3. Using these counts, the ratio of settled microspheres [(VOI1+VOI2)/VOI3] was calculated as (82484+62039)/151204=0.95858 (95.58%).
The actual 90Y unit activity amounts Mega Becquerel/cm3 (MBq/cm3) placed into the phantom were 1 MBq/cm3 for normal liver (N), 9.9 MBq/cm3 for the large tumor (Ltm), and 4.73 MBq/cm3 for the small tumor (Stm), considering the volumes of the tumors. The T/N ratios for Ltm and Stm were determined to be 1/10 and 1/2, respectively. After separating the microspheres and filling pure 90Y into the phantom and tumor-simulating regions, PET/CT imaging was performed. The obtained images showed that the 90Y microspheres did not settle at the bottom but distributed relatively homogeneously. VOIs were drawn around the outer boundaries of the tumors on the images. The large and small tumors were labeled as Ltm and Stm, respectively. Additionally, 10 VOIs were drawn from different regions corresponding to normal liver tissue (N) on the cross-sectional image (Figure 3). The average VOI counts were 70120 for Ltm and 13848 for Stm. The T/N ratios were calculated by dividing the average VOI counts of Stm and Ltm by the VOI counts of N separately (Table 1).
Data distribution was assessed and non-parametric statistics were therefore applied. Statistical Comparison of Ltm and Stm T/N Values Using Mann-Whitney U Test Revealed a Highly Significant Difference at the p=0.00018 Level.
Discussion
The present study demonstrated that 90Y can be successfully separated from commercially available resin microspheres and converted into an ionic solution suitable for phantom-based PET/CT imaging studies. When the separated ionic 90Y solution was introduced into the phantom, it exhibited a homogeneous distribution, thereby enabling reliable quantitative PET analysis.
The differences between the two preparation methods before and after separation of the particulate resin microspheres were demonstrated both visually and quantitatively in Figures 2 and 3. Approximately 95.6% of the detected activity from the intact microspheres accumulated in the dependent (lower) regions of the phantom, confirming that gravitational sedimentation dominates the distribution of resin microspheres under static phantom conditions. Such sedimentation renders phantom studies unsuitable for quantitative imaging and dosimetric validation because the intended activity distribution cannot be maintained. In contrast, following separation of 90Y from the microspheres, the activity was distributed throughout the phantom without visible sedimentation, resulting in PET images that more accurately reflected the intended activity concentrations.
An important finding of this study was the close agreement between the prepared activity concentrations and the quantitative measurements obtained from PET imaging. The planned tumor-to-normal liver (T/N) activity concentration ratio was 10:1 for the large tumor and 2:1 for the small tumor. PET imaging yielded mean T/N ratios of 11.3±0.8 and 2.26±0.17, respectively. Although minor deviations were observed, these differences were within the expected range attributable to the PVE, the spatial resolution of PET, image reconstruction, VOI delineation, and counting statistics. Furthermore, the Mann-Whitney U test demonstrated a highly significant difference between the T/N ratios of the two tumor models (p=0.00018). These findings confirm that the prepared phantom successfully generated two distinct activity concentrations that were quantitatively distinguishable by PET/CT.
Most phantom studies reported in the literature have used commercially available 90Y chloride (90YCl3) solutions because of their inherently homogeneous distribution (15). Son et al. (18) used 90Y chloride in a phantom dosimetry study and demonstrated that its homogeneous activity distribution was suitable for quantitative PET imaging. Similarly, Mille et al. (19) performed phantom experiments using a homogeneous 90YCl3 solution. They emphasized the usefulness of this approach for image-based dosimetry while also recommending that future studies investigate more realistic heterogeneous activity distributions. In contrast, the present study demonstrates that by separating ionic 90Y from residual resin microspheres obtained from clinical applications, a homogeneous mixture exceeding 95% uniformity can be achieved for phantom studies. This approach therefore provides an alternative radionuclide source for phantom preparation without requiring commercially available 90Y chloride.
Clinical studies have shown that 90Y resin microspheres do not sediment under the influence of gravity but instead become distributed throughout the hepatic arterial microvasculature. Gulec et al. (11) reported consistent tumor-to-normal liver uptake ratios following radioembolization, demonstrating that the microspheres become physiologically embolized within the hepatic arterioles. Likewise, Son et al. (18) reported good agreement between dosimetric estimates obtained using 99mTc-MAA single photon emission computed tomography (SPECT)/CT and post-treatment 90Y PET/CT findings. Furthermore, Elschot et al. (20) demonstrated that PET/CT provides more accurate post-treatment activity quantification than Bremsstrahlung SPECT imaging. These clinical observations fundamentally differ from static phantom experiments because, under in vivo conditions, the microspheres are retained within the tumor microvasculature rather than freely settling under gravity. Consequently, the sedimentation observed in our initial phantom experiment should not be interpreted as a limitation of clinical radioembolization but rather as a phenomenon specific to the static phantom model. For phantom applications requiring homogeneous activity distribution, separation of 90Y from the microspheres therefore appears to be necessary.
Although the separated ionic solution produced a substantially more homogeneous activity distribution, minor count differences were still observed among the VOIs representing N tissue (Table 1). These small variations are more likely attributable to the intrinsic characteristics of PET imaging, including image noise, positron range, limited spatial resolution, and VOI placement, rather than incomplete mixing of the radionuclide. In addition, because image acquisition and reconstruction were performed using a routine clinical PET/CT protocol, no recovery coefficient or partial volume correction was applied, ensuring that the results reflected standard clinical imaging conditions.
Study Limitations
This study has several limitations. First, only a single separation protocol employing a 0.22 μm sterile membrane filter was evaluated, and alternative separation techniques were not investigated. Second, only PET/CT imaging was assessed, whereas other quantitative imaging modalities, such as SPECT/CT, were not examined. Finally, the experiments were performed using a static phantom, which cannot fully reproduce the complex vascular trapping mechanisms and heterogeneous microsphere distribution encountered during clinical radioembolization. Future studies should evaluate alternative separation methods, different phantom geometries, and multimodal imaging approaches to provide a more comprehensive validation of this technique.
Conclusion
In conclusion, when used under static phantom conditions, 90Y resin microspheres tend to sediment because of their particulate nature and higher density. To achieve a homogeneous distribution of 90Y in any medium, appears to be necessary for phantom studies requiring homogeneous activity distribution. In this study, it was demonstrated that ionic 90Y radionuclide can be successfully obtained from 90Y resin microspheres by filtration through a sterile 0.22 µm membrane filter (Merck Millipore).
