SPECT/CT image‐based dosimetry for Yttrium‐90 radionuclide therapy: Application to treatment response

Abstract This work demonstrates the efficacy of voxel‐based 90Y microsphere dosimetry utilizing post‐therapy SPECT/CT imaging and applies it to the prediction of treatment response for the management of patients with hepatocellular carcinoma (HCC). A 90Y microsphere dosimetry navigator (RapidSphere) within a commercial platform (Velocity, Varian Medical Systems) was demonstrated for three microsphere cases that were imaged using optimized bremsstrahlung SPECT/CT. For each case, the 90Y SPECT/CT was registered to follow‐up diagnostic MR/CT using deformable image registration. The voxel‐based dose distribution was computed using the local deposition method with known injected activity. The system allowed the visualization of the isodose distributions on any of the registered image datasets and the calculation of dose‐volume histograms (DVHs). The dosimetric analysis illustrated high local doses that are characteristic of blood‐flow directed brachytherapy. In the first case, the HCC mass demonstrated a complete response to treatment indicated by a necrotic region in follow‐up MR imaging. This result was dosimetrically predicted since the gross tumor volume (GTV) was well covered by the prescription isodose volume (V150 Gy = 85%). The second case illustrated a partial response to treatment which was characterized by incomplete necrosis of an HCC mass and a remaining area of solid enhancement in follow‐up MR imaging. This result was predicted by dosimetric analysis because the GTV demonstrated incomplete coverage by the prescription isodose volume (V470 Gy = 18%). The third case demonstrated extrahepatic activity. The dosimetry indicated that the prescription (125 Gy) isodose region extended outside of the liver into the duodenum (178 Gy maximum dose). This was predictive of toxicity as the patient later developed a duodenal ulcer. The ability to predict outcomes and complications using deformable image registration, calculated isodose distributions, and DVHs, points to the clinical utility of patient‐specific dose calculations for 90Y radioembolization treatment planning.


| INTRODUCTION
Radionuclide therapy using Yttrium-90 ( 90 Y) microspheres has emerged as an effective treatment modality for the management of patients with primary and metastatic hepatocellular carcinoma (HCC). 1 SIR-Spheres is FDA-PMA (Food and Drug Administration-Premarket Approval) approved for metastatic colorectal cancer to the liver. 2 TheraSphere is FDA approved under HDE (humanitarian device exemption) for radiation treatment or as neo-adjuvant to surgery or transplantation in patients with HCC who can have appropriately placed hepatic arterial catheters. 3 This device is indicated for HCC patients with partial or branch portal vein thrombosis/occlusion when clinical evaluation warrants the treatment. 4 Current practice uses manufacturer-recommended prescription activity and dose calculations, derived from the Medical Internal Radiation Dose (MIRD) Committee of the Society of Nuclear Medicine, which are based on the size of the intended treatment volume. 5 Despite positive results from published studies, 6,7 the reported doses assume that the activity distribution within the treatment volume is uniform. Quantitative patient-specific 90 Y dosimetry has been studied using bremsstrahlung single photon emission computed tomography (SPECT)/computed tomography (CT) [8][9][10][11] and positron emission tomography (PET)/CT. [12][13][14][15][16] However, such dosimetric studies have utilized research software that is not readily available to the medical community and have not illustrated scenarios where voxel-based dosimetry could be used to predict treatment response.
In this study, we demonstrate how commercial voxel-based absorbed dose calculation software applied to post-therapy 90 Y bremsstrahlung SPECT imaging can facilitate dosimetric verification during the course of 90 Y therapy. This work is meant as a proof-ofconcept to highlight the potential clinical merits of using a patient-specific 90 Y dose calculation to quantify the treatment plan. Specifically, we demonstrate how voxel-based dosimetry can be used to predict treatment response and treatment complications. To our knowledge, this is the first time a study has illustrated the use of isodose distributions and dose-volume histograms in 90 Y therapy to predict not only tumor response but also normal tissue complications.

| ME TH ODS
Three cases were selected to demonstrate the utility of patient-specific dosimetry in 90 Y therapy. All patients underwent imaging and clinical evaluation before treatment. This included arterial mapping and abdominal Technetium-99m-macroaggregated albumin ( 99m Tc-MAA) SPECT/CT to assess the potential for extrahepatic uptake and determine the feasibility, safety, and number of injections required for treatment. Planar SPECT imaging was used for the liver-lung shunt calculation. For therapy, 90 Y glass microspheres (TheraSphere, BTG Inc., Ottawa, Canada) were administered with an activity calculated according to the manufacturer's specifications. 3 Within 90 min after the administration of the 90 Y, a bremsstrahlung SPECT/CT was acquired using a Siemens Symbia T6 (Siemens Medical Solutions Inc., Malvern, PA) with an energy window of 90-125 keV based on a highly optimized acquisition and reconstruction protocol published by Siman et al. 17 The SPECT projection images were acquired with a matrix size of 128 × 128, 4.8 mm pixel size, 28 s/view for 128 views over 360°. The SPECT images were reconstructed using the manufacturer's three-dimensional ordered subset expectation maximization algorithm (Flash3D, Siemens) using 8 iterations with 16 subsets, geometric collimator response modeling, CT-based attenuation correction using effective energy of the imaging window, dual-energy window-based scatter correction, and a 9.6 mm FWHM postreconstruction Gaussian filter. 17 Dosimetry was performed using the 90 Y microsphere dosimetry navigator (RapidSphere) within Velocity (Varian Medical Systems, Atlanta, GA) as illustrated in Fig. 1.
The first step within Velocity was to deformably register the 90 Y SPECT/CT images to the pre-and post-treatment diagnostic images (MRI and CT). The deformable registration algorithm applies multiresolution free-form deformations based on cubic spline interpolation between sparse, uniformly distributed control points as its transformation model. 18,19 The registration provided a one-to-one correlation between voxels on different images and time points, allowing for mapping of anatomical data and structure sets from diagnostic to follow-up imaging. Velocity automatically segmented the patient's external body contour on the SPECT/CT which was later used to cumulate the SPECT counts within the patient (excluding the lungs) for the dose calculation as described below.
The second step used the 90 Y dosimetry navigator to calculate the absorbed dose distribution from the 90 Y SPECT/CT images. The 90 Y dose calculation used the local deposition method (LDM), 20,21 which was previously demonstrated to be the most accurate when using SPECT imaging. 22 In this technique, 90 Y β-particles released by decay within a voxel deposit all energy locally with the same voxel. This is an accurate approximation considering that the mean range of β-particles in tissue is 3.8 mm which is within the typical SPECT voxel size (4.8 mm cubic voxels in our study). The absorbed dose within a voxel is where A = injected activity, LSF = lung shunt fraction, T 1/2 = 90 Y physical half-life (64.24 hours), E avg = average β-particle energy per disintegration (0.935 MeV), C voxel = SPECT counts within voxel, ΔV = voxel volume, ρ = tissue density, and C total = total SPECT counts within the patient (excluding the lungs). The derivation of Eq. (1) can be found in the Appendix. The only patient-specific parameters that had to be manually entered into the software prior to the dose calculation were the injected activity (GBq) and the lung shunt fraction (expressed as %). The default tissue density was set to that of soft tissue (1.04 g/cm 3 ). However, the software does provide the ability to enter tissue-specific densities for different contoured structures.
The patient was a 64-year-old male with hepatitis C induced cirrhosis and HCC. This patient was determined to be unresectable due to the degree of cirrhosis and was referred for radionuclide therapy

3.C | Case #3 -Extrahepatic activity
The patient was a 69-year-old male with HCC. The lung shunt fraction was determined to be 3.7% from the SPECT/CT study. A dose of 125 Gy was prescribed to the medial and lateral segments (1400 cc volume). The patient received 3.563 GBq of 90 Y using 3.2 million microspheres. Fig. 6 illustrates the deformably registered post-treatment SPECT/CT and the 11-week follow-up CT with the superimposed 90 Y isodose distribution. Extrahepatic activity was noted after treatment when new supraduodenal arteries had formed since the arterial mapping study and there was a catheter malfunction at the time of injection. Notice that the ≥125 Gy dose region extended outside the liver into the duodenum. The DVH of the duodenum, with a maximum dose of 178 Gy, is shown in Fig. 7. This patient developed a duodenal ulcer because of this high radiation dose.

| DISCUSSION
Radiation Oncology treatment modalities require an accurate estimation of the dose in Gray delivered to tumors and normal tissue. This fundamental principle also applies to 90 Y therapy which is a form of brachytherapy. However, 90 Y radioembolization has been practiced without a strong understanding of the actual quantity of radiation exposure in liver tumors, healthy liver tissue, and tissue surrounding the liver. It is our hope that this work will be a catalyst to opening a new era of patient-specific 90 Y dosimetry using commercially available software.
There is some debate on whether to use bremsstrahlung SPECT/

ACKNOWLEDG EMENTS
We wish to thank Eduardo Moros, PhD for his useful comments and guidance.
F I G . 7. Dose-volume histograms of the duodenum, and medial and lateral segments for Case #3.

APPEN DIX
The following describes the derivation of Eq. (1). The absorbed dose within a voxel is where E voxel = energy deposited within a voxel, m voxel = mass of voxel However, where ρ = tissue density, ΔV = volume of voxel Also, where E avg = average β-particle energy per disintegration, N = total number of disintegrations from a given amount of activity as a function of time within the voxel, integrated over all time. But, where A voxel = initial activity within the voxel, λ = decay constant And, where T 1/2 = physical half-life (physical decay only, that is, no biological clearance is assumed in the case of Y-90 microspheres). However, where C voxel = SPECT counts within voxel, C Total = total SPECT counts within patient (excluding lungs), A Total = total activity within the patient (excluding lungs). Finally, where A = injected activity, LSF = lung shunt fraction Therefore, using Eqs. (A2), (A3), (A4), (A5), (A6), and (A7) into Eq. (A1) we obtain: