Predictive SIRT dosimetry based on a territorial model
- Nadine Spahr1Email authorView ORCID ID profile,
- Philipp Schilling2,
- Smita Thoduka3,
- Nasreddin Abolmaali3 and
- Andrea Schenk4
Received: 11 May 2017
Accepted: 17 October 2017
Published: 31 October 2017
Abstract
Background
In the planning of selective internal radiation therapy (SIRT) for liver cancer treatment, one major aspect is to determine the prescribed activity and to estimate the resulting absorbed dose inside normal liver and tumor tissue. An optimized partition model for SIRT dosimetry based on arterial liver territories is proposed. This model is dedicated to characterize the variability of dose within the whole liver. For an arbitrary partition, the generalized absorbed dose is derived from the classical partition model. This enables to consider normal liver partitions for each arterial perfusion supply area and one partition for each tumor for activity and dose calculation. The proposed method excludes a margin of 11 mm emitting range around tumor volumes from normal liver to investigate the impact on activity calculation. Activity and dose calculation was performed for five patients using the body-surface-area (BSA) method, the classical and territorial partition model.
Results
The territorial model reaches smaller normal liver doses and significant higher tumor doses compared to the classical partition model. The exclusion of a small region around tumors has a significant impact on mean liver dose. Determined tumor activities for the proposed method are higher in all patients when limited by normal liver dose. Activity calculation based on BSA achieves in all cases the lowest amount.
Conclusions
The territorial model provides a more local and patient-individual dose distribution in normal liver taking into account arterial supply areas. This proposed arterial liver territory-based partition model may be used for SPECT-independent activity calculation and dose prediction under the condition of an artery-based simulation for particle distribution.
Keywords
Background
Selective internal radiation therapy (SIRT) is a kind of brachytherapy used in interventional radiology to treat cancer [1]. In particular, patients with unresectable cancers, such as hepatocellular carcinoma, resistant to conventional chemotherapy and poorly accessible in conventional radiotherapy [2], or liver metastases, are eligible for SIRT. This type of therapy exploits the fact that the vascular system of tumor tissue differs from that of normal liver. Nearly 70–80% of a normal liver is supplied by the portal vein; the remainder is supplied by the hepatic artery [3]. However, liver tumors are usually supplied by arterial vessels. SIRT takes advantage of these supply characteristics for selective tumor treatment by transarterial radioembolization. During an interventional treatment session, yttrium-90 (Y-90)-labelled microspheres are ideally administered through a catheter into the tumor-feeding arteries. In practice, these arteries supply the tumors but also normal liver parenchyma. The microspheres embolize the capillaries and emit beta radiation due to their Y-90 radionuclide, thereby irradiating the embolized tumor and potential liver parenchyma regions.
Prior to SIRT, an evaluation procedure is performed for treatment planning purposes. The evaluation is necessary, and diagnostic angiography is used to select the best catheter positions for tumoral targeting and to identify digestive arteries arising from the hepatic artery. The latter are coiled during the evaluation procedure to avoid digestive damage. When the optimal catheter positions are identified, a radiopharmaceutical, e.g., technetium-99m (Tc-99m) macroaggregated albumin (MAA), is delivered to the hepatic artery. This step is an approach to simulate the treatment intervention using Tc-99m MAA as a surrogate for Y-90 microspheres. Afterwards, the MAA particle distribution is visualized by planar bremsstrahlung imaging or by SPECT imaging depending on the center and radiologist. The acquired MAA SPECT is then routinely used to determine the lung shunt fraction. It is also used in research studies for Y-90 activity and dose planning purposes. Empirical activity planning by the body-surface-area (BSA) method is inferior to the more advanced partition model (PM) [4], which additionally allows to predict mean dose values inside predefined organ partitions [5].
The PM introduces three main tissue compartments for SIRT dosimetry: lung, normal liver, and tumor. Based on the MAA SPECT/CT, the amount of deposited activity is determined for each partition. The absorbed dose can then be calculated. This model enables straightforward dose calculation in short computation times. However, assumptions and simplifications are made affecting the determined absorbed dose. The PM gathers large volumes of interest in one partition and assumes a uniform distribution of activity throughout the partition. Especially for normal liver, this might be critical in case of dose calculation. The PM would gather the whole normal liver activity and distribute it evenly over the whole normal liver partition during dose calculation. Regional activity distribution is not considered and also the handling of multiple tumors is not feasible.
Kao et al. [6] suggested an artery-specific partition modeling for radioembolization. Arterial regional margins are delineated for each catheter position via invasive catheter-directed CT hepatic angiography during the MAA intervention. Activity and dose planning according to the PM are performed for each region individually. The method of Kao et al. [6] does not take into account all patient-individual arterial territories at the same time but relies on the MAA SPECT/CT distribution, which is a composition of the whole administered activity. Activity assignment to individual catheter positions might not be feasible.
We propose an optimized, extended partition model, called territorial model (TM), introducing more partitions to the normal liver based on arterial liver territories [7]. In addition, we permit several tumor partitions for non-connected tumors or metastasis.
Methods
In the following, we introduce the arterial liver territories that were used for the optimized partition model. Afterwards, the numerical details of SIRT dose calculation based on the partition model is described, the extended model is introduced, and activity calculation is presented briefly. Information about patient images and data analysis is given. Here, we focus on Y-90 resin SIR-Spheres®; (Sirtex Medical Limited, North Sydney, Australia), but the approach is not limited to them.
Determination of arterial liver territories
Visualization of 3D rendered liver territories from anterior (a) and posterior (b) view and translucent territories with opaque visualization of the arterial vessel system (c)
SIRT dose calculation
Partition model
Extended partition model based on arterial territories
The classical MIRD partition model gathers large volumes of interest in one partition and assumes a uniform distribution of activity throughout each partition. Our territorial model keeps this basic assumption but introduces more partitions to the normal liver based on arterial liver territories, see the “Determination of arterial liver territories” section. Additionally, we permit several tumor partitions for non-connected tumors or metastases. The lung partition will be handled as in the PM, where lung dose is determined from the lung shunt fraction.
where A i is the activity in partition i and A Σ∖i describes the activity in Σ without partition i.
SIRT activity calculation
Constrained partition activity calculation
This approach was directly transferred to the TM. To determine the activity to deliver, the mean normal liver dose over all liver partitions and the lung dose are restricted to thresholds given in Table 1.
BSA method
with tumor volume V T and total liver and tumor volume V total.
This empirical method assumes a relation between the tumor size in the liver and the patient’s size. Despite reasonable concerns on the use of BSA, e.g., in [9], it is also given here for comparison purposes as it is still used in clinical practice.
Patients, imaging, and image analysis
Voxel sizes of acquired contrast-enhanced T1-weighted MR images
Patient | Voxel size |
---|---|
Pat1 | 1.562 × 1.562 × 2 mm |
Pat2 | 0.7422 × 0.7422 × 5 mm |
Pat3 | 0.7813 × 0.7813 × 2.2 mm |
Pat4 | 0.7422 × 0.7422 × 2.2 mm |
Pat5 | 0.7422 × 0.7422 × 2.4 mm |
Volume and mass measurements were performed on the CT image using segmentation masks provided via coregistration. Lung segmentation was performed automatically on the CT images [11]. The liver territories of Pat1, Pat2, and Pat4 correspond to the eight Couinaud segments. For Pat5, the fourth segment was split into two separate segments. Pat3 has only four segments due to right hemihepatectomy. Segments five to eight were resected. Activity counts were determined on SPECT images.
Results
Activity and dose calculation were performed for five patients according to TM and the classical PM
Pat1 | Pat2 | Pat3 | Pat4 | Pat5 | ||
---|---|---|---|---|---|---|
Information | No. tumors | 1 | 1 | 5 | 1 | 2 |
Largest tumor [ml] | 274.5 | 582.8 | 54.8 | 454.8 | 13.09 | |
No. liver territories | 8 | 8 | 4 | 8 | 9 | |
BSA model | A 0 [GBq] | 1.64 | 1.92 | 2.01 | 1.85 | 1.67 |
PM | A 0 [GBq] | 3.97 | 3.88 | 4.05 | 2.71 | 3.03 |
D L [Gy] | 10.68 | 25.00 | 8.68 | 9.23 | 9.65 | |
D NL [Gy] | 70.00 | 43.41 | 70.00 | 70.00 | 70.00 | |
D T [Gy] | 287.95 | 190.55 | 167.48 | 88.75 | 144.69 | |
TM | A 0 [GBq] | 5.97 | 3.88 | 4.78 | 4.62 | 3.31 |
D L [Gy] | 16.12 | 24.78 | 10.62 | 15.73 | 10.75 | |
D NL [Gy] | 69.88 | 29.67 | 70.08 | 70.43 | 70.06 | |
D T [Gy] | 433.75 | 190.23 | 199.75 | 150.98 | 158.82 |
Activity and dose calculation were performed for five patients according to TM and the classical PM
Pat1 | Pat2 | Pat3 | Pat4 | Pat5 | ||
---|---|---|---|---|---|---|
PM | A 0 [GBq] | 2.27 | 3.58 | 2.31 | 1.55 | 1.73 |
D L [Gy] | 6.10 | 23.03 | 4.96 | 5.27 | 5.52 | |
D NL [Gy] | 40.00 | 40.00 | 40.00 | 40.00 | 40.00 | |
D T [Gy] | 164.54 | 175.56 | 95.7 | 50.71 | 82.68 | |
TM | A 0 [GBq] | 3.41 | 3.88 | 2.73 | 2.64 | 1.9 |
D L [Gy] | 9.35 | 24.78 | 6.04 | 8.96 | 6.29 | |
D NL [Gy] | 39.90 | 29.67 | 39.94 | 40.25 | 40.00 | |
D T [Gy] | 247.71 | 190.23 | 113.96 | 86.13 | 90.88 |
Discussion
Visualization of dose distribution for Pat1 using PM (left) and TM (right) and normal liver thresholds T 1 in the first row and T 2 in the second row. The dose color bar, relating color to dose values, is given below. Each quartet shows axial, sagittal, coronal, and anterior 3D view. a PM activity and dose calculation using T1. b TM activity and dose calculation using T1. c PM activity and dose calculation using T2. d TM activity and dose calculation using T2. e Dose color bar in Gy
The calculated activity to deliver obviously depends on the threshold’s limiting dose to normal liver. However, the results also show a strong dependence on the method used for activity calculation. Typically, mean normal liver dose is the limiting factor in activity calculation, except for cases like Pat2, where lung dose limits the activity to deliver due to high lung shunting. Therefore, all determined activities are smaller when using the small normal liver threshold of T 2 than using the threshold in T 1. For T 2, this results in insufficient tumor coverage for Pat3, Pat4, and Pat5 in case of PM. However, TM achieves the desired tumor dose for Pat3, and both methods were successful in Pat1 and Pat2. High tumor doses like 434 Gy for Pat1 with TM indicate a high tumor coverage in this case and suggest to extend the activity calculation by an additional constraint on the mean tumor dose. This will result in a smaller lung and normal liver dose exposure. A validation with post-interventional data should analyze the actually achieved doses with PM and TM and investigate whether locally higher liver doses can be accepted in return for sparing another major part of the liver.
The assumption that high normal liver counts in close neighborhood of tumor regions might originate from the activity at the tumor is demonstrated in Pat2. With T 1 the activity to deliver is limited by the maximum lung dose using PM as well as TM. Therefore, both methods determine the same activity to deliver of 3.88 GBq. For PM, this results in a normal liver dose of 43.41 Gy. The TM excludes activity counts in a neighborhood of maximum Y-90 emitting range of 11 mm for normal liver dose calculation. This results in a significant lower normal liver dose of 29.67 Gy. The relatively small volume in close neighborhood to the tumor seems to have a large impact on normal liver dose. This might be caused by the beta particle range, producing a spill-out effect on the tumor, or partial volume effects. Because the specified margin of 11 mm is only approximately twice the original SPECT image resolution, see the “Patients, imaging, and image analysis” section, the assumption of excluding this region from normal liver dose calculation seems to be reasonable. All other patients show a similar behavior with normal liver dose limited by the threshold but a higher prescribed activity and therefore slightly higher lung dose and significant higher tumor doses. For Pat1 and Pat4, tumor doses could be increased by a factor of approximately 1.5, whereas the increase in tumor doses is smaller for Pat3 and Pat5. This observation suggests that higher tumor doses in TM are associated with a higher tumor volume. A clear correlation between the number of tumors and an activity increase from PM to TM was not demonstrated in the results of five patients. Further studies on larger databases should investigate this. A systematic evaluation and a detailed investigation of different margins have to be analyzed in a next step. Potential critical side effects have to be investigated carefully.
One advantage of the presented model is that it can be used in cases with several tumors present in liver tissue. The restriction of the PM to clearly differentiable tumors still remains true due to the dependence from segmentations. Therefore, tumors have to be differentiable and delineated on CT images or coregistered MR images rather than on SPECT images. The margin around the tumor was determined here by the maximum Y-90 emitting range of 11 mm. Other possibilities, e.g., margin selection depending on SPECT image resolution or mean Y-90 emitting range, might be considered and its influence should be analyzed. The impact of image noise as well as segmentation and coregistration accuracy should be investigated.
The presented approach, as well as the partition model in general, assumes that MAA is a suitable surrogate for Y-90 microspheres and that the MAA particle distribution is similar to the Y-90 particle distribution. This is controversially discussed: A high correlation of MAA and calibrated beta-probe was shown in [12], and [3] confirms that the assumption of microsphere and MAA particle distribution similarity introduces less error into dose calculations than the assumption of uniform activity throughout a volume of interest. The recommendation of Dezarn et al. [3] to use PM, which is relying on the MAA SPECT/CT, is contradicted by results of Wondergem et al. [13]. Poor correlation between MAA and Y-90 is also reported in [14] caused by systematic errors, like differences in catheter position, injection techniques, or differences in particle sizes, flow hemodynamics, or disease progression. A good correlation of predictive MAA SPECT-based dosimetry with post-radioembolization Y-90 PET dosimetry was demonstrated in [15]. Also, [16] showed a good correlation of MAA SPECT and Y-90 PET tumor-to-normal uptake ratios. Gnesin et al. [17] showed that the MAA SPECT provides a good estimate of absorbed doses compared to post-treatment PET for tumor and non-tumor tissues in HCC radioembolization. Despite this, MAA SPECT imaging is currently the only possibility for predictive three-dimensional dose assessment.
The reliance upon MAA SPECT is one of the main limitations for a partition-based approach for SIRT activity and dose calculation and also applies to the presented approach. A comparison of predictive TM dosimetry based on MAA SPECT and post-interventional TM dosimetry based on Y-90 PET/CT imaging is planed to be investigated in a next step. As a prospect for the future, the proposed liver territory-based partition model is designed to enable SPECT-independent activity calculation and dose prediction under the condition of an artery-based simulation for particle distribution. MAA SPECT imaging would be used then only to respect safety-oriented tasks, e.g., estimation of extra-hepatic shunting.
Recently, more advanced methods for voxel-based dose calculation, such as Monte Carlo [18], dose point kernel [19], and local deposition methods [20], have been developed. However, they still rely on the MAA SPECT/CT and the recommendation of the microsphere manufacturers [8] remains on activity and dose calculation based on the PM which is widely used in clinical practice. Reason for this is, it can be easily performed and offers a practical option for individual activity planning [21]. Provided that there is a reliable artery-based simulation for particle distribution, the proposed liver territory-based partition model enables SPECT-independent activity calculation and dose prediction based on the territories. This is the decisive advantage over three-dimensional voxel-based dosimetry methods, which cannot achieve that.
Conclusions
An extended partition model based on arterial liver territories is proposed for SIRT activity and dose calculation. This method is able to better account for non-uniformly distributed activity in normal liver tissue. By this, the proposed method is also able to provide a more local dose distribution. Compared to the classical partition model, both methods predict the same normal liver dose, whereas the predicted activity and dose for lung and tumor tissue is lower in the classical model than in the territorial model using a 11-mm margin. This leads to the conclusion that tumors increase normal liver dose and that excluding a defined, small region around tumors from normal liver in case of normal liver dose calculation based on a partition model can estimated liver partition dose more precise. Studies on a larger database and post-interventional images should investigate this further.
Declarations
Acknowledgements
The authors would like to thank Christiane Engel and Andrea Koller (Fraunhofer Institute for Medical Image Computing MEVIS, Bremen, Germany) for segmenting liver, tumors, and arteries in the image data and reconstructing arterial hepatic territories. We thank David Black (Fraunhofer Institute for Medical Image Computing MEVIS, Bremen, Germany) for proofreading.
Funding
This work was funded by the Fraunhofer-Gesellschaft.
Authors’ contributions
NS and AS developed the proposed method. NS implemented the algorithm, performed the data analysis, contributed to interpretation and discussion of results and drafted the manuscript. ST and NA collected the patient data and analyzed corresponding results. PS, NA, and AS helped to interpret and discuss the results. AS coordinated the project SIRTOP, in which the research presented was achieved, and AS revised the manuscript. All authors read and approved the final version of the manuscript.
Ethics approval and consent to participate
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. For this type of study, formal consent is not required.
Competing interests
The authors declare that they have no competing interests.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Authors’ Affiliations
References
- Kennedy A, Salem R. Radioembolization (yttrium-90 microspheres) for primary and metastatic hepatic malignancies. Cancer J. 2010; 16:163–75.View ArticlePubMedGoogle Scholar
- Avila MA, Berasain C, Sangro B, Prieto J. New therapies for hepatocellular carcinoma. Oncogene. 2006; 25:3866–84.View ArticlePubMedGoogle Scholar
- Dezarn WA, Cessna JT, Dewerd LA, Feng W, Gates VL, Halama J, Kennedy AS, Nag S, Sarfaraz M, Sehgal V, Selwyn R, Sabin MG, Thomadsen BR, Williams LE, Salem R. Recommendations of the American Association of Physicists in Medicine on dosimetry, imaging, and quality assurance procedures for 90Y microsphere brachytherapy in the treatment of hepatic malignancies. Med Phys. 2011; 38(8):4824–45.View ArticleGoogle Scholar
- Ho S, Lau WY, Leung TWT, Chan M, Ngar YK, Johnson PJ, Li AKC. Partition model for estimating radiation doses from yttrium-90 microspheres in treating hepatic tumors. Eur J Nucl Med Mol Imaging. 1996; 23:974–52.View ArticleGoogle Scholar
- Kao YH, Tan EH, Ng CE, Goh SW. Clinical implications of the body surface area method versus partition model dosimetry for yttrium-90 radioembolization using resin microspheres: a technical review. Ann Nucl Med. 2011; 25(7):455–61.View ArticlePubMedGoogle Scholar
- Kao YH, Tan AEH, Burgmans MC, Irani FG, Khoo LS, Lo RHG, Tay KH, Tan BS, Chow PKH, Ng DCE, Goh ASW. Image-guided personalized predictive dosimetry by artery-specific SPECT/CT partition modeling for safe and effective 90Y radioembolization. J Nucl Med. 2011; 53(4):559–66.View ArticleGoogle Scholar
- Selle D, Preim B, Schenk A, Peitgen HO. Analysis of vasculature for liver surgical planning. IEEE Trans Med Imag. 2002; 21(11):1344–57.View ArticleGoogle Scholar
- SIR-Spheres®; Microspheres Training Program Physicians and Institutions. (TRN-RW-04) http://foxfireglobal.sirtex.com/sites/foxfireglobal.sirtex.com/files/user/trn-rw-04_for_eu_au_nz_and_asia.pdf.
- Kennedy A, Coldwell D, Sangro B, Wasan H, Salem R. Radioembolization for the treatment of liver tumors general priciples. Am J Clin Oncol. 2012; 35:91–9.View ArticlePubMedGoogle Scholar
- SOMATOM Sensation 10/16 Application Guide. 2005.Google Scholar
- Lassen B, Kuhnigk JM, Schmidt M, Krass S, Peitgen HO. Lung and lung lobe segmentation methods at Fraunhofer MEVIS; 2011. pp. 185–199.Google Scholar
- Lau WY, Leung TW, Ho S, Chan M, Leung NW, Lin J, Metreweli C, Li AK. Diagnostic pharmaco-scintigraphy with hepatic intra-arterial technetium-99m macroaggregated albumin in the determination of tumour to non-tumour uptake ratio in hepatocellular carcinoma. Br J Radiol. 1994; 67:136–9.View ArticlePubMedGoogle Scholar
- Wondergem M, Smits MLJ, Elschot M, de Jong HWAM, Verkooijen HM, van denBosch MAAJ, Nijsen JFW, Lam MGEH. 99mTc-macroagrregated albumin poorly predicts the intrahepatic distribution of 90Y resin microspheres in hepatic radioembolization. J Nucl Med. 2013; 54:1294–301.View ArticlePubMedGoogle Scholar
- Knesausrek K, Machac J, Muzinic M, DaCosta M, Zhang Z, Heiba S. Quantitative comparison of Yttrium-90 (90Y)-microspheres and technetium-99m (99mTc)-macroaggregated albumin SPECT images for planning 90Y therapy of liver cancer. Technol Cancer Res Treat. 2010; 9:253–62.View ArticleGoogle Scholar
- Kao YH, Steinberg JD, Tay YS, Lim GKY, Yan J, Townsend DW, Budgeon CA, Boucek JA, Francis RJ, Cheo TST, Burgmans MC, Irani FG, Lo RHG, Tay KH, Tan BS, Chow PKH, Satchithanantham S, Tan AEH, Ng DCE, Goh ASW. Post-radioembolization yttrium-90 PET/CT—part 2: dose-response and tumor predictive dosimetry for resin microspheres. EJNMMI Res. 2013; 3:57.View ArticlePubMedPubMed CentralGoogle Scholar
- Rhee S, Kim S, Cho J, Park J, Eo JS, Park S, Lee E, Kim YH, Choe JG. Semi-quantitative analysis of post-transarterial radioembolization (90)Y microsphere positron emission tomography combined with computed tomography (PET/CT) images in advanced liver malignancy: comparison with (99m)Tc macroaggregated albumin (MAA) single photon emission computed tomography (SPECT). Nucl Med Mol Imaging. 2016; 50(1):63–9.View ArticlePubMedGoogle Scholar
- Gnesin S, Canetti L, Adib S, Cherbuin N, Monteiro MS, Bize P, Denys A, Prior JO, Baechler S, Boubaker A. Partition model based 99mTc-MAA SPECT/CT predictive dosimetry compared to 90Y TOF PET/CT post-treatment dosimetry in radioembolization of hepatocellular carcinoma: A quantitative agreement comparison. J Nucl Med. 2016; 57:1672–8.View ArticlePubMedGoogle Scholar
- Ljungberg M, Sjögreen K, Liu X, Frey E, Dewaraja Y, Strand SE. A 3-dimensional absorbed dose calculation method based on quantitative SPECT for radionuclide therapy: evaluation for (131)I using Monte Carlo simulation. J Nucl Med. 2002; 43:1101–9.PubMedPubMed CentralGoogle Scholar
- Bolch WE, Bouchet LG, Robertson JS, Wessels BW, Siegel JA, Howell RW, Erdi AK, Aydogan B, Costes S, Watson EE, Brill AB, Charkes ND, Fisher DR, Hays MT, Thomas SR. MIRD pamphlet No. 17: the dosimetry of nonuniform activity distributions–radionuclide S values at the voxel level. Medical Internal Radiation Dose Committee. J Nucl Med. 1999; 40:11–36.Google Scholar
- Pasciak AS, Bougeois AC, Bradley YC. A comparison of techniques for 90Y PET/CT image-based dosimetry following radioembolization with resin microspheres. Front Oncol. 2014; 4:121.PubMedPubMed CentralGoogle Scholar
- Song YS, J C Paeng H-CK, Chung JW, Cheon GJ, Chung JK, Lee DS, Kang KW. PET/CT-based dosimetry in 90Y-microsphere selective internal radiation therapy: single cohort comparison with pretreatment planning on 99mTc-MAA imaging and correlation with treatment efficacy. Medicine. 2015; 94:945.View ArticleGoogle Scholar