Abstract
This feasibility study demonstrates 90Y quantitative bremsstrahlung imaging of patients undergoing high-dose myeloablative 90Y-ibritumomab treatment. Methods: The study includes pretherapy 111In SPECT/CT and planar whole-body (WB) imaging at 7 d and therapy 90Y SPECT/CT at 6 d and 90Y WB imaging at 1 d. Time–activity curves and organ-absorbed doses derived from 90Y SPECT images were compared with pretherapy 111In estimates. Organ activities derived from 90Y WB images at the first day were compared with corresponding pretherapy estimates. Results: Pretherapy 111In images from 3 patients were similar to the 90Y images. Differences between absorbed-dose estimates from pretherapy 111In and 90Y therapy were within 25%, except for the lungs. Corresponding activity differences derived from WB images were within 25%. Differences were ascribed to incomplete compensation methods and real differences in pharmacokinetics between pretherapy and therapy. Conclusion: Quantitative bremsstrahlung imaging to estimate organ activities and absorbed doses is feasible.
Radioimmunotherapy is established for the treatment of relapsing follicular or transformed B-cell lymphomas. Two radioimmunoconjugates, 90Y-ibritumomab (Zevalin; Spectrum Pharmaceuticals) and 131I-tositumomab (Bexxar; GlaxoSmithKline), were approved. A dose–response relationship can be inferred from several observations (1,2), and the best clinical results published have made use of myeloablative radioimmunotherapy (3).
In standard 90Y-ibritumomab treatments, administered activity is based on patient weight. For high-dose ibritumomab studies involving bone-marrow stem-cell support, an accurate dosimetry is required. The organs at risk in these studies are the liver, kidneys, and lungs. To maximize the therapy effect, it is important to not exceed the maximum-tolerable dose (MTD). We have an ongoing clinical absorbed dose–escalation study to determine MTD for the liver based on a pretherapy dose planning. The pretherapy dosimetry is performed by imaging with 111In-labeled ibritumomab to predict the 90Y activity required for treatment. There is also a need to monitor the actual treatment for dose verification. A mixture of 111In- and 90Y-labeled ibritumomab could allow for imaging; however, this method has potential drawbacks. First, any labeling instability produces free-circulating 111In, which gives nonrepresentative image information. Second, 90Y bremsstrahlung may contaminate the 111In energy windows, leading to errors in the activity quantification. Third, an 111In-ibritumomab kit can be costly. If quantitative 90Y bremsstrahlung imaging is feasible, such a study could confirm both targeting and delivery of the prescribed absorbed dose.
Previously, we have experimentally investigated quantitative bremsstrahlung SPECT and planar whole-body (WB) imaging (4,5). To our knowledge, no studies have been performed on quantitative bremsstrahlung imaging of patients given radiolabeled monoclonal antibody. In this work, the feasibility of quantitative bremsstrahlung imaging to verify predicted absorbed doses was investigated for SPECT and WB imaging using data from our escalation study. That study includes a pretherapy study with 111In-labeled ibritumomab in which SPECT/CT and WB imaging allow for a comparison of quantitative 90Y with quantitative 111In imaging.
MATERIALS AND METHODS
Patients and Study Protocol
In this work, 3 patients were evaluated (2 men [ages, 74 and 57 y; weights, 74 and 79 kg] and 1 woman [age, 71 y; weight, 70 kg]). The organ at risk was the liver, because bone-marrow stem-cell support was given. All patients received 300 MBq of 111In-ibritumomab in the pretherapy study, followed by SPECT/CT and WB imaging on 7 occasions (at 1, 24, 48, 72, 144, 166, and 192 h after imaging). The SPECT/CT data determined the necessary therapy activity to give 12 Gy to the liver, and the WB studies served as an independent activity-quantification method to confirm the SPECT dosimetry. The 90Y activities were calculated to 2,915, 4,990, and 1,825 MBq. Before 90Y infusion, all patients received cold rituximab. The therapy study included 6 measurements at 1, 24, 48, 120, 144, and 166 h after injection. Both SPECT/CT and WB images were acquired on the first occasion, but only SPECT/CT measurements were performed on the other 5 occasions.
Imaging System
A SPECT/CT Discovery VH system (GE Healthcare), equipped with 2.54-cm NaI(TI) crystals and a Hawkeye CT, was used. 111In images were acquired from two 15% energy windows centered over the 245- and 172-keV peaks. A medium-energy general-purpose collimator was used for 111In imaging, and 90Y images were acquired using a high-energy general-purpose collimator and a 60% energy window centered at 150 keV. Anterior and posterior WB images were acquired in 384 × 1,024 matrices, with a pixel size of 2.21 mm. The scan speed was 20 cm/min for the first 3 measurements in the pretherapy study and 10 cm/min for all other measurements. SPECT data were acquired in a 64 × 64 matrix for 60 projections and 360°. For each time point, a CT study was conducted. The acquisition times per SPECT projection for 111In and 90Y were 45 and 60 s, respectively. All images were processed using the LundAdose software (6). The system sensitivities were measured from a known activity in air for 111In and behind a 10-mm clear acrylic sheet for 90Y and were 72, 12, 1, and 0.16 cps/MBq for 111In SPECT, 111In WB, 90Y SPECT, and 90Y WB imaging, respectively.
SPECT Activity Quantification
SPECT images were reconstructed with an iterative ordered-subset expectation maximization algorithm (7) using 6 angles per subset. Attenuation correction was made using a CT-based density map (8) scaled to proper photon energy, with either soft-tissue– or bone-equivalent mass-attenuation coefficients, depending on a threshold of 1.2 g·cm−3. Scatter was modeled using the effective source scatter estimation (9) including compensation for the collimator–detector response. Both effective source scatter estimation and collimator–detector response kernels were generated from Monte Carlo simulations (10). The 111In images were reconstructed using 6 iterations and the 90Y images using 3 iterations. The mean organ activity concentration was calculated from volumes of interest delineated with sufficient margins to avoid partial-volume effects.
Planar Activity Quantification
Organ activities were quantified using a pixel-based conjugate-view method (4,6). A patient-specific narrow-beam attenuation map was determined from an x-ray scout image (11) to correct for attenuation. To scale the map to the 90Y energy window, energy abundance–weighted linear attenuation coefficients were calculated from bremsstrahlung emission spectra obtained using MCNPX (12). Because of the 40-cm difference between the camera heads and x-ray unit, the patients were repositioned between the emission and scout scans. A software-based image registration, based on mutual information between the geometric mean–averaged WB image and the scout image (13,14), was applied for a correct pixel-based attenuation correction (15). The spatial change included the transformation of regions covering the head, torso, and left and right legs and was based on second-degree polynomials with translation, rotation, shearing, and second-degree curving included (15). Compensation for scatter, collimator response, and counts from backscattered photons was applied (4).
Organ activities were calculated from regions of interest (ROIs), and corrections for background activity and overlapping activities were made on a pixel basis in the activity images (16).
Absorbed-Dose Calculation and Evaluation
SPECT 90Y images were evaluated by comparing 111In and 90Y SPECT–based time–activity curves and absorbed doses for the liver, spleen, kidneys, and lungs. Differences in administered activities, half-lives, and times were considered by renormalizing the 111In activity, AIn:
Planar WB 90Y images were evaluated by comparing 111In and 90Y activities in the liver, spleen, kidneys, lungs, and total body, for which 111In images were normalized as:
RESULTS
SPECT
Figure 1 shows time–activity curves for patient 1. The 90Y time–activity curve for the liver is consistent with the 111In time–activity curve, apart from the first time point. The spleen time–activity curves differ somewhat between 111In and 90Y, showing a slightly lower initial uptake for 90Y. The 111In and 90Y time–activity curves for the kidneys correspond well, and the left and right kidneys exhibit almost identical kinetics. For the lungs, the 90Y time–activity curves are elevated, compared with the 111In time–activity curve, but the kinetics for the left and right lungs are similar.
Mean activity concentration in liver, spleen, kidneys, and lungs calculated from SPECT for patient 1. Act. conc. = activity concentration.
Table 1 shows calculated organ-absorbed doses. For the liver, organ-absorbed doses agreed well for patients 1 and 3, but for patient 2 the absorbed-dose calculated from 90Y images was 25% higher than that estimated from 111In images. The spleen values were within 20% for patients 1 and 2 and within 30% for patient 3, and the kidney values were within 14%. Considerable differences were seen in the absorbed doses to the lungs.
Absorbed Doses to Organs, Calculated from SPECT Images
Figure 2 shows SPECT images through the liver and spleen for 111In and 90Y for patient 1. It is evident that the 111In images have better spatial resolution and image contrast. However, the 90Y images compare well with 111In images, despite apparent background and nonuniform organ boundaries.
111In and 90Y images from patient 1 at first time point.
WB Imaging
Table 2 summarizes activity estimates based on the planar 111In images and 90Y images. Figure 3 demonstrates the improvement in image quality from the restoration filtering and the quantitative procedure. The ROIs used to quantify both the 111In and the 90Y images are also displayed. For all patients and organs, the differences in activities obtained from the 90Y images and from 111In imaging are within 25%. Estimated total-body activities, compared with the administered activities, showed deviations of −7%, −5%, and −7% for the 111In-based estimates and 9%, 5%, and 8% for the 90Y-based estimates for patients 1, 2, and 3, respectively.
Organ Activities at 1 Hour After injection, Calculated from WB Images
Unprocessed (A) and processed (B) 90Y WB images and corresponding unprocessed (C) and processed (D) 111In images. Curvature comes from registration process. ROIs used are shown in D.
DISCUSSION
The feasibility of clinical quantitative bremsstrahlung imaging has been investigated by comparing absorbed doses calculated from multiple SPECT/CT scans of 111In-labeled ibritumomab acquired before therapy with those acquired during therapy with 90Y-ibritumomab. The activities obtained from WB 111In images and planar WB 90Y images have also been compared.
The difference between the 111In and 90Y time–activity curves (Fig. 1), the corresponding absorbed-dose estimates, and the planar WB activities were mostly of the same magnitude for all patients. There were some differences that could be related to measurement uncertainties or actual differences in pharmacokinetics. Regarding the liver, the absorbed doses corresponded well for patients 1 and 3. For patient 1, a slightly higher activity was obtained for the 90Y-based estimation at the first SPECT measurement, and a similar difference was seen for the WB-derived value. For this patient, there could be a difference in the actual uptake because the pretherapy study was performed 4 mo before therapy. The 90Y liver time–activity curve, compared with the 111In time–activity curve, for patient 2 (not shown) was elevated, and the absorbed dose to the liver was estimated to be 14.7 Gy from the 90Y time–activity curve but only 11.8 Gy from the 111In extrapolation. The reason for the 24.6% difference is not yet understood. For patient 3, the liver time–activity curves and absorbed doses were in good agreement. The 11 Gy estimated from the 111In images was based on the actual amount of delivered 90Y activity and therefore deviates from the prescribed 12 Gy.
For all patients, the difference in the absorbed dose before therapy and during therapy was larger for the spleen than for the liver and kidneys. The spleen is relatively small, and size differs between the patients. Partial-volume effects could therefore influence the estimates differently between the patients. The kidneys are also relatively small, but 111In- and 90Y-derived kidney absorbed doses were similar for all patients. The differences for the spleen could therefore be interpreted as a real difference in uptake. The washout rate for patient 1 (Fig. 1) seems similar, but the initial uptake is slightly lower for the therapeutic infusion. Generally, the spleen can be regarded as a target organ, and the number of available binding sites of the antigen CD20 before therapy and during therapy can differ because of the infusion of cold monoclonal antibody.
For the lungs, the 90Y time–activity curves, compared with the 111In time–activity curves, were generally elevated, with corresponding absorbed-dose deviations up to 65%. These deviations may relate to insufficient compensation methods. A bremsstrahlung image has a more diffuse background with a higher intensity than an 111In image (Fig. 3). Most counts seen outside the patient boundary in Figure 3A stem from photons that have scattered in or penetrated the septa, passed the crystal, and then backscattered. This scatter background, combined with the fact that lungs are located between organs with elevated uptake (liver and blood in the heart), could explain some of the obtained differences. The effective source scatter estimation kernels are invariant regarding density, and the method is therefore not expected to be accurate in areas with heterogeneous attenuation.
Total-body activities, compared with administered activities, from WB 90Y images were approximately 10% higher than corresponding 111In estimates—on average 7%. There are 2 reasons for this. First, the mean energy of the bremsstrahlung image in the 105- to 195-keV energy interval is about 140 keV, which means that the overestimation that arises from the source thickness when applying the geometric mean is larger for 90Y. Second, the diffuse background in 90Y images is difficult to correct for. It has been found that organ activities based on WB studies can be underestimated by a few percentage points (4). Also for this study, the 90Y activity estimates were often lower than 111In estimates, contrary to results from SPECT. These lower estimates were especially evident for the lungs, for which 90Y activity concentration from SPECT estimates for the first time point were between 10% and 25% higher than the 111In-based estimates. The planar-based 90Y activity estimates were up to 20% lower than the 111In-based estimates. Generally, for the 111In and 90Y measurements at the first time point the WB estimates deviated by the same order of magnitude as the SPECT estimates. However, conjugate-view quantification may be more prone to variation because it is sensitive to subjective actions such as defining ROIs.
CONCLUSION
This work shows that adequate compensations for attenuation, scatter, and collimator response make 90Y bremsstrahlung imaging feasible, with a relatively good image quality and useful quantitative accuracy. These compensations may be of great importance for absorbed-dose planning of high-dose radioimmunotherapy and for future improved dosimetry protocols for present 90Y-based radionuclide therapies, such as standard 90Y-ibritumomab treatment.
Acknowledgments
This work was funded by the Swedish Research Council, Swedish Cancer Foundation, Gunnar Nilsson Foundation, Bertha Kamprad Foundation, and Lund University Medical Faculty Foundation and by Lund University Hospital Donation Funds.
- © 2010 by Society of Nuclear Medicine
REFERENCES
- Received for publication June 3, 2010.
- Accepted for publication September 8, 2010.
