Corrections for improved quantitative accuracy in SPECT and planar scintigraphic imaging ANNE LARSSON Department of Radiation Sciences, Radiation Physics Umeå University, Sweden 2005 Cover illustration: The logarithm of a 123I point spread function. The point source was simulated at 14.0 cm distance from the collimator in air, with the SIMIND version described in Paper V, using parameters corresponding to a GE Millennium MPR with 2.3 × 2.3 m2 field of view. The grey scale is processed. • © Anne Larsson 2005 ISBN: 91-7305-938-2 Printed by Print & Media, 2001342, Umeå, Sweden 2 To Björn, family and friends If you don't care where you are, you ain't lost. Rune’s Rule 3 ABSTRACT A quantitative evaluation of single photon emission computed tomography (SPECT) and planar scintigraphic imaging may be valuable for both diagnostic and therapeutic purposes. For an accurate quantification it is usually necessary to correct for attenuation and scatter and in some cases also for septal penetration. For planar imaging a background correction for the contribution from over- and underlying tissues is needed. In this work a few correction methods have been evaluated and further developed. Much of the work relies on the Monte Carlo method as a tool for evaluation and optimisation. A method for quantifying the activity of 125I-labelled antibodies in a tumour inoculated in the flank of a mouse, based on planar scintigraphic imaging with a pin-hole collimator, has been developed and two different methods for background subtraction have been compared. The activity estimates of the tumours were compared with measurements in vitro. The major part of this work is attributed to SPECT. A method for attenuation and scatter correction of brain SPECT based on computed tomography (CT) images of the same patient has been developed, using an attenuation map calculated from the CT image volume. The attenuation map is utilised not only for attenuation correction, but also for scatter correction with transmission dependent convolution subtraction (TDCS). A registration method based on fiducial markers, placed on three chosen points during the • SPECT examination, was evaluated. The scatter correction method, TDCS, was then optimised for regional cerebral blood flow (rCBF) SPECT with 99mTc, and was also compared with a related method, convolution scatter subtraction (CSS). TDCS has been claimed to be an iterative technique. This requires however some modifications of the method, which have been demonstrated and evaluated for a simulation with a point source. When the Monte Carlo method is used for evaluation of corrections for septal penetration, it is important that interactions in the collimator are taken into account. A new version of the Monte Carlo program SIMIND with this capability has been evaluated by comparing measured and simulated images and energy spectra. This code was later used for the evaluation of a few different methods for correction of scatter and septal penetration of 123I brain SPECT. The methods were CSS, TDCS and a method where the corrections for scatter and septal penetration are included in the iterative reconstruction. This study shows that quantitative accuracy in 123I brain SPECT benefits from separate modelling of scatter and septal penetration. 4 LIST OF PAPERS This thesis is based on the following papers which are referred to by their Roman numerals in the text. I. Anne Larsson, Lennart Johansson, Rauni Rossi Norrlund, Katrine Riklund Åhlström. Methods for estimating uptake and absorbed dose in tumours from 125I labelled monoclonal antibodies, based on scintigraphic imaging of mice. Acta Oncologica 1999 38 361-365 II. Anne Larsson, Lennart Johansson, Torbjörn Sundström, Katrine Riklund Åhlström. A method for attenuation and scatter correction of brain SPECT based on CT-images. Nucl Med Commun 2003 24 411-420 III. Anne Larsson, Lennart Johansson. Transmission dependent convolution subtraction of 99mTc-HMPAO rCBF SPECT – A Monte Carlo study. IEEE Trans Nucl Sci 2005 52 231-237 IV. Anne Larsson, Lennart Johansson. Scatter-to-primary based scatter fractions for transmission dependent convolution subtraction of SPECT images. Phys Med Biol 2003 48 N323-N328 V. Michael Ljungberg, Anne Larsson, Lennart Johansson. A new collimator simulation in SIMIND based on the Delta-Scattering technique. Accepted for publication IEEE Trans Nucl Sci VI. Anne Larsson, Michael Ljungberg, Susanna Jakobson Mo, Katrine Riklund, Lennart Johansson. Correction for scatter and septal penetration in 123I brain SPECT imaging – A Monte Carlo study. Submitted Paper I-V have been reproduced with permission from the publishers. 5 TABLE OF CONTENTS 1 INTRODUCTION 9 2 SCINTIGRAPHIC IMAGING 11 2.1 The gamma camera 11 2.1.1 Gamma cameras used in this work 12 2.2 Single photon emission computed tomography (SPECT) 15 2.2.1 Reconstruction of SPECT images 15 2.3 Radiopharmaceuticals 17 2.4 Photon interactions with matter 19 2.4.1 Attenuation 20 2.5 Scintigraphic studies in this work 21 3 QUALITY OF SCINTIGRAPHIC IMAGES 22 3.1 Contrast 22 3.2 Noise 22 3.3 Quality parameters of a gamma camera 24 3.3.1 Spatial resolution 24 3.3.2 Sensitivity 25 3.3.3 Uniformity 26 3.3.4 Energy resolution 27 • 3.3.5 Count rate performance 27 4 QUANTITATIVE IMAGING 28 4.1 Absolute quantification 28 4.1.1 Absolute quantification in this work 29 4.2 Relative quantification 30 4.2.1 Relative quantification in this work 30 5 THE MONTE CARLO METHOD 32 5.1 Monte Carlo simulations in this work 33 5.1.1 Monte Carlo simulations of collimator interactions 33 6 CORRECTIONS FOR QUANTITATIVE SPECT 35 6.1 Attenuation correction 35 6.1.1 Transmission measurements with a radionuclide 37 6.1.2 Transmission measurements with a CT scanner 39 6.1.3 Attenuation correction in this work 40 6.2 Scatter correction 42 6.2.1 Energy window techniques 44 6.2.2 Convolution techniques 46 6.2.3 Scatter correction in an iterative reconstruction procedure 47 6.2.4 Scatter correction in this work 48 6.3 Compensation for collimator-detector response (CDR) 51 6.3.1 CDR compensation in this work 54 6 6.4 Correction for septal penetration 56 6.4.1 Energy window techniques 58 6.4.2 Convolution techniques 59 6.4.3 Septal penetration in an iterative reconstruction procedure 59 6.4.4 Correction for septal penetration in this work 60 7 CORRECTIONS FOR PLANAR SCINTIGRAPHIC IMAGING 62 7.1 Attenuation correction 62 7.2 Background correction 65 7.3 Attenuation and background correction in this work 66 8 SUMMARY AND CONCLUSIONS 69 8.1 Paper I 69 8.2 Paper II 69 8.3 Paper III 69 8.4 Paper IV 70 8.5 Paper V 70 8.6 Paper VI 71 9 FUTURE PERSPECTIVES 72 10 ACKNOWLEDGEMENTS 73 11 REFERENCES 75 7 ABBREVIATIONS 123I-IMP N-isopropyl-p-123I-amphetamine 1D One-dimensional 2D Two-dimensional 3D Three-dimensional CDR Collimator-detector response cpm Counts per minute cps Counts per second CSS Convolution scatter subtraction CT Computed tomography DAT Dopamine transporter ESSE Effective source scatter estimation FBP Filtered backprojection FWHM Full width at half maximum FWTM Full width at tenth maximum GP General purpose HEGP High energy general purpose HMPAO Hexamethyl propyleneamine oxime HR High resolution LE Low energy LEGP Low energy general purpose • ME Medium energy MEGP Medium energy general purpose MIBG Meta-iodobenzylguanidine MIRD Medical Internal Radiation Dose MLEM Maximum likelihood expectation maximisation MRT Magnetic resonance tomography NEMA National Electrical Manufacturers Association NMSE Normalised mean square error OSEM Ordered subsets expectation maximisation PET Positron emission tomography PM Photomultiplier rCBF Regional cerebral blood flow RIT Radioimmunotherapy ROI Region of interest SDSE Slab-derived scatter estimation SLSF Scatter line spread function SPECT Single photon emission computed tomography TDCS Transmission dependent convolution subtraction TEW Triple energy window VOI Volume of interest 8 1 INTRODUCTION In single photon emission computed tomography (SPECT) and planar scintigraphic imaging, the distribution of a radiopharmaceutical in vivo is studied. In absence of some physical degrading factors, the SPECT images would reflect the 3D distribution of the radiopharmaceutical, while the planar images would reflect the corresponding 2D projections through the imaged object. In reality the images are seriously affected by attenuation, scatter and detector response, and for some radionuclides also septal penetration. It is however possible to correct for these effects, more or less accurately. For a visual interpretation, planar images are seldom corrected. The effect of attenuation in such cases can even be an advantage since it can provide depth information of the radiopharmaceutical distribution, if images from opposing views are acquired. If SPECT images are corrected for the purpose of visual interpretation it is in most cases only for attenuation. A visual interpretation should however benefit from corrections also for scatter and detector response in many cases since the resulting increase in image quality can simplify lesion detection. Corrections are more important and widely applied for quantitative evaluations of the images, especially for absolute quantification where the activity is calculated. To know the activity in a specific organ or a tumour is essential for calculating the absorbed dose to that part of the body, which for example can be of interest for dose planning in radionuclide therapy. Absolute quantification from planar images requires usually also a background correction for the contribution of registrations from activity in over- and underlying tissues. For a relative quantification, the calibration to activity is not necessary, but the images should preferably reflect the distribution of the radiopharmaceutical, which requires corrections. In this case the counts in different regions in the images can be compared and this is often of interest for diagnostic purposes. The aims of this work were to evaluate and further develop corrections that improve the quantitative accuracy of SPECT (in particular) and planar scintigraphic imaging. Most of the work concerns brain imaging, since most of the SPECT research at Norrlands University Hospital, Umeå University has been concentrated on regional cerebral blood flow (rCBF) or the dopamine transporter and receptor system in the brain. In many cases, however, the methods are also applicable to other parts of the body. This work is focused on corrections for background activity, attenuation, scatter, septal penetration and detector response. The Monte Carlo method is used as a tool for evaluation and optimisation in many parts of the work. More specific the aims of the present work were to: • Develop and evaluate a quantitative method for determining the uptake of 125I- labelled compounds in a tumour inoculated in the flank of a mouse, based on planar scintigraphic imaging with a pin-hole collimator. The evaluation includes the comparison of two different methods to correct for the background 9 activity. This work on planar scintigraphic imaging can be seen as an introduction to the more complex world of SPECT. The method is described in Paper I. • Develop and evaluate a method for attenuation and scatter correction of brain SPECT based on computed tomography (CT) images of the same patient, where the registration of the CT and SPECT image volumes should be performed using fiducial markers that only need to be present during the SPECT examination. The method is evaluated for 99mTc Hexamethyl propyleneamine oxime (HMPAO) rCBF SPECT and is described in Paper II. • Further developing of the scatter correction method used in the previous study, transmission dependent convolution subtraction (TDCS), by determining the optimal geometry for deriving scatter fractions as a function of attenuation path length, and the optimal scatter kernel, for 99mTc HMPAO rCBF SPECT. TDCS is also compared with convolution scatter subtraction (CSS) and this is described in Paper III. The aim was also to introduce a few adjustments in the correction method to make it suitable for many iterations, which is described in Paper IV. • To contribute to the evaluation of a new version of the Monte Carlo program SIMIND that can take interactions in the collimator into account. This is • important when the effect of septal penetration is studied using Monte Carlo simulations. This is described in Paper V. • To use this Monte Carlo program to evaluate a few correction methods for scatter and septal penetration; CSS, two TDCS versions and one method included in the iterative reconstruction which uses the effective source scatter estimation (ESSE) to model scatter and collimator-detector response (CDR) including septal penetration. The aim was also to implement a distance- independent compensation for CDR for geometrically mean valued projection images. This study is described in Paper VI. 10