Strategies for CT tissue segmentation for Monte Carlo calculations in nuclear medicine dosimetry P. E. N. Braad a) and T. Andersen Department of Nuclear Medicine, Odense University Hospital, Sdr. Boulevard 29, Odense C DK-5000, Denmark S. B. Hansen Department of Nuclear Medicine and PET-Centre, Aarhus University Hospital, Nørrebrogade 44, Aarhus C DK-8000, Denmark P. F. Høilund-Carlsen Department of Nuclear Medicine, Odense University Hospital, Sdr. Boulevard 29, Odense C DK-5000, Denmark (Received 4 January 2016; revised 22 September 2016; accepted for publication 22 October 2016; published 15 November 2016) Purpose: CT images are used for patient specific Monte Carlo treatment planning in radionuclide therapy. The authors investigated the impact of tissue classification, CT image segmentation, and CT errors on Monte Carlo calculated absorbed dose estimates in nuclear medicine. Methods: CT errors as a function of patient size, CT reconstruction, and tube current modulation methods were assessed in a phantom experiment on a clinical CT system. The impact of tissue segmentation methods and CT number variations on EGSnrc Monte Carlo calculated absorbed dose distributions was assessed for 99m Tc and 131 I in the ICRP/ICRU male phantom and in a patient PET/CT-scanned with 124 I prior to radioiodine therapy. Results: CT number variations <20 HU were obtained for whole-body CT examinations at effective CT doses ∼2 mSv. Monte Carlo calculated absorbed doses depended on both the number of media types and accurate calibration of the CT number-to-density conversion ramp. Tissue segmentation by a 13-tissue CT conversion ramp, calibrated by a stoichiometric method, resulted in low (<4%) dose errors in selected organs for both isotopes. Conclusions: A calibrated CT scanner specific conversion ramp is required for accurate patient specific dosimetry in nuclear medicine. Accurate dosimetry was obtained with a 13-tissue ramp that included five different bone types. C 2016 American Association of Physicists in Medicine. [http://dx.doi.org/10.1118/1.4967267] Key words: radionuclide therapy, nuclear medicine, CT, dosimetry, Monte Carlo, PET/CT, SPECT/CT 1. INTRODUCTION In recent years, numerous research groups have explored the potential and feasibility of improving clinical dosimetry in nuclear medicine (NM) by using Monte Carlo (MC) methods for dose calculations. 1–6 Most NM clinical dosimetry studies are based on the simple S-value formalism, 7 where absorbed doses are calculated from patient specific activity distribu- tions and tabulated S-values. The errors of internal dosimetry calculations for diagnostic or therapeutic studies with current conventional methods may be as large as 30%–100% or even higher. 8 Large dose errors limit the value of dosimetry and may indirectly be responsible for an inefficient and non-optimal use of targeted radionuclide therapies (TRTs) in clinical cancer treatments. Radioiodine therapy (RAIT) of thyroid cancer 9 is an example of a therapeutic approach with growing interest in patient specific treatment planning. The increased availability of 124 I has raised the interest in PET-based pre-RAIT dosim- etry and several clinical studies are currently investigating the potential of this approach. 10–12 Likewise, it has been argued in a recent study 13 that radiation doses to patients from diagnostic radiopharmaceuticals in NM may better be evaluated in real patients by MC methods rather than from tabulated S-values calculated in reference phantoms. MC dose calculations in NM depend on the availability of combined PET and CT or SPECT and CT information. The NM scans provide information on the biodistribution of radionuclides in the patient and CT numbers provide detailed information on the radiological properties of the tissues. 3,14 Before MC methods can be used for routine calculations of patient specific dose estimates, it is pivotal that all technical aspects in the calculations are evaluated and optimized care- fully in realistic anthropomorphic phantom experiments. One aspect that has only received limited attention is the use of CT for tissue segmentation. CT information is required to calculate patient specific physical cross-sections for the MC simulation, and the accuracy of CT numbers therefore has a direct impact on the accuracy of dose calculations. 15,16 Clinical dosimetry studies often consider density variations but neglect tissue composition variations by only using water cross-sections. 3,17–20 For megavoltage external photon beam radiotherapy (EBRT), the differences in atomic compositions 6507 Med. Phys. 43 (12), December 2016 0094-2405/2016/43(12)/6507/10/$30.00 © 2016 Am. Assoc. Phys. Med. 6507