ct imaging based digitally reconstructed radiographs and ... · ct imaging based digitally...

CT imaging based digitally reconstructed radiographs and its application in brachytherapy

N. Milickovic

Department of Medical Physics & Engineering, Strahlenklinik, Kliniken Offenbach, 63069 Offenbach, Germany.

D. Baltas

Department of Medical Physics & Engineering, Strahlenklinik, Kliniken Offenbach, 63069 Offenbach, Germany,

and

Institute of Communication & Computer Systems, National Technical University of Athens, 15773 Zografou, Athens, Greece. S. Giannouli, M. Lahanas

Department of Medical Physics & Engineering, Strahlenklinik, Kliniken Offenbach, 63069 Offenbach, Germany. N. Zamboglou

Department of Medical Physics & Engineering, Strahlenklinik, Kliniken Offenbach, 63069 Offenbach, Germany,

and

Institute of Communication & Computer Systems, National Technical University of Athens, 15773 Zografou, Athens, Greece.

Corresponding author:

Natasa Milickovic, Ph.D. Dept. of Medical Physics & Engineering

Strahlenklinik Kliniken Offenbach Starkenburgring 66 63069 Offenbach am Main, Germany Tel.: + 49 - 69 - 8405 - 4480 or - 4522 Fax: + 49 - 69 - 8405 - 4481 E-mail: [email protected]

Natasa Milickovic et al. CT imaging based digitally reconstructed radiographs...

2

ABSTRACT

The aim of our study was to develop an algorithm to simulate the Digitally Reconstructed Radiograph (DRR) calculation process for different beam qualities (photon energies) in the range of 50 keV to 12 MeV. This was achieved using volumetric anatomical data of the patient obtained from three-dimensional diagnostic CT images. These DRR images can be used in three-dimensional treatment planning for external beam radiotherapy as well as for brachytherapy in the same way as conventional radiographic films. The advantages of using such DRRs in modern 3D brachytherapy treatment planning are shown. A number of tools tools are described, illustrating that the application of the DRRs in brachytherapy is very useful.


3

1. INTRODUCTION

Modern brachytherapy treatment planning is image based with CT playing an increasingly important role. This is because CT scans can provide accurate three-dimensional information on the size and position of the target volume, and the position of any critical organs or structures of interest.

The CT data set enables construction of the three-dimensional volumetric data set (VDS) where each voxel is characterized by a characteristic CT number given in Hounsfield units (HU). From this VDS, DRRs can be calculated, having first been defined by a number of parameters selected by the user.

Several papers have described the process of DRR calculation (Bahner et al 1999, Chaney et al 1995, Schlegel 1991, Sherouse et al 1990). For example, Sherouse et al 1990, present a ray tracing algorithm which simulates the calculation of the DRR for the energy at which the CT images were acquired. Their algorithm enables an approximation for the images that would result from purely photoelectric or Compton interactions.

Our aim was to develop an algorithm for simulation of the DRR calculation process for different energies including those for radiotherapy simulators and accelerators. Such calculated (high energy) DRRs can be used in the process of the patient position check: matching with portal image. We have also developed a number of novel tools which make such calculated DRRs very useful in brachytherapy treatment planning. These tools include those for the following four applications. (1) Implementation of geometrical, anatomical and catheter filters. (2) Catheter reconstruction from two DRRs. (3) Navigation between DRRs and the three-dimensional view window. (4) Navigation between two DRRs.

The 3D volume data can be used directly for catheter reconstruction. The imaging based 3D treatment planning for brachytherapy combines in ideal way the catheter reconstruction with an anatomy definition: PTV and critical structures, ICRU 50 and 62, and anatomy based evaluation of dose distribution. Our group has published a method for automatic reconstruction of catheters from CT 3D data set (Milickovic N et al 2000). We believe in the future of this method but we also realize the need of an intermediate method like the DRR based catheter reconstruction: for the reasons of generality and to enable user to choose the method which he prefers.

There are many clinics where the traditional way of catheter reconstruction based on projectional methods (radiological films) is combined with the imaging based anatomy definition. Furthermore, for the case of permanent prostate implants, radiographs still play an important role for the seeds identification and reconstruction. It follows that there was a concrete need for an approach to produce digitally the information which is usually available from classical film and in this way to enable user having the anatomical and catheter data presented on the same set of CT slices. For this reason we developed our DRR based catheter reconstruction algorithm which can completely change classical films and in the same time there is established relationship between the reconstructed catheters and patient anatomy. It is very easy to choose the two best “views” (gantry angles) to get the best catheter presentation on DRR films. The next improvement is the ability for easy reconstruction of crossing-implants that could otherwise cause confusion or implants whose projections cover each other. This paper was written to give an insight on our achievement in DRR reconstruction and a number of tools which have not been published until now.

Significant improvements in computers with regard to speed, graphical ability, memory, and also the availability of CT scanners in radiotherapy clinics have enabled the development of algorithms to support the treatment planning process.


4

2. METHODS AND MATERIALS

2.1. Introduction

Algorithms have been developed for simulation of DRR calculations for user selected energies from low energy diagnostic machines to high energy linear accelerators: 50 keV to 12 MeV. Using as an example, high dose rate (HDR) interstitial brachytherapy, the process is based on post-implantation acquired CT images with the catheters in situ in their final positions. This includes the relevant patient anatomy, target volume(s), organs at risk, and also the catheters.

The three-dimensional VDS is calculated by trilinear interpolation from CT images in a pre-processing step. Our algorithm does not require the interslice distance or slice thickness to be constant. As the radius of catheters can be less than 2 mm we must take care to avoid loss of catheter area contrast in the interpolation process. We achieve this by selecting a high resolution of VDS and our algorithm takes into account each intersected voxel. Two voxel dimensions of the VDS are defined by the CT pixel dimensions and the third one is selected to be less than 2mm, although user can select any other values.

The VDS results in a three dimensional matrix of which each element, i.e. voxel, is uniquely defined by its position in this three dimensional volume and by its CT number. This specific CT number for each voxel is related to the electron density of the material in the corresponding matrix element. We use this property of CT scanning to establish a relationship between the CT numbers and their corresponding linear attenuation coefficients and relative electron densities. This data can further be used either for dosimetric purposes or in the process of DRR calculation.

2.2. Experimental estimation of the HU ↔ ρwe relationship

A CT scan, i.e. a Computed Axial Transmission Tomography Image, is based on taking a large number of one-dimensional attenuation profiles of a body slice from many different directions and reconstructing the anatomical detail within the slice. Digitisation is applied during the reconstruction process and results in a matrix where each element, i.e. pixel, is defined by the CT number expressed in Hounsfield units (HU) (Dendy et al 1999). The specific value for each pixel is related to the electron density of the tissue in the corresponding matrix element.

As the chemical compounds of different tissues are well known from the literature (Hubbel 1994, ICRP 23 1995, ICRU 46 1989), the next step is to calibrate the CT scanner and obtain the curve describing the dependence of relative electron density on Hounsfield number. This calibration was made using the 33 cm diameter phantom, RMI Electron Density Phantom model 465∗. This has 20 holes of 2.8 cm diameter into which 6 rods made of different commercially available plastics and 11 rods simulating different tissues with known chemical composition and electron densities, are inserted (Constantinou 1974, Constantinou et al 1992). The remaining three rods are made of solid water which is the same material as the bulk of the phantom.

The relationship between relative electron densities and Hounsfield numbers is described by a density conversion function, HU to ρw

e. By scanning the phantom the HU values of each rod with known ρw

e can be obtained. These values are then used to establish a scanner-specific density conversion function. In the case of our Somatom Plus 4 CT scanner# these equations are:

1. Known relative electron density of material with its CT number unknown.

∗Radiation Measurements Inc., Middleton, WI, USA # Siemens, Medizinische Technik, Erlangen, Germany


5

2051.391?1953.923HU 1.083)if(?

527.635?541.2442HU1.083)?if(1.001

1012.1429?1018.608HU 1.001)if(?

iweii

we

iweii

we

iweii

we

−⋅=⇒>

−⋅=⇒≤<

−⋅=⇒≤

2. Known CT number of material with its relative density coefficient unknown.

1.05HU0.00051?58)if(HU

0.975HU0.00185? 58)HUif(7

0.994HU0.00098? 7)if(HU

iiwei

iiwei

iiwei

+⋅=⇒>

+⋅=⇒≤<

+⋅=⇒≤

Figure 1. shows the calibration curve we have obtained for our Somatom Plus 4 CT scanner#.

The relative electron density conversion function varies between CT scanners and the scanner-specific density function must be derived for each scanner. The scanner-specific parameters used, such as kVp, beam quality, beam hardening, filter and reconstruction algorithm, can affect the CT number of each volume element because the CT number is related to the attenuation coefficient of the tissue in the volume element. Periodical checks of the scanner-specific function ( )HUf?w

e = are recommended.

2.3. Experimental estimation of the HU↔µwe relationship

The entire range of CT numbers available from our treatment planning system PLATO BPS v13.7♣ and CT scanner Somatom Plus 4 CT scanner#: [-1024 ÷ 3071] HU is subdivided into 23 sub-ranges defined by materials of known relative electron density and chemical composition. Using the XCOM (Berger et al 1987) program we next calculate the linear attenuation coefficients of those 23 materials for the selected photon energies (monoenergetic beams). For energy E the linear attenuation coefficient µ of material x characterized by the CT number HUx∈[HUa, HUb], where [HUa, HUb] is one of 23 sub-ranges, is calculated by linear interpolation between the µa and µb:

bab

axa

ab

xbx µ

HUHUHUHU

µHUHUHUHU

µ ⋅−−

+⋅−−

= .

In this way the inhomogeneity correction factors can also be included in dose calculations and thus improve the accuracy of treatment planning.

The linear attenuation coefficient µx for the CT voltage applied during the CT slice acquisition is given by ( Battista et al 1980,Sherouse et al 1990)

µx=[(HUx / 1000) + 1]⋅µw,

where µw is the linear attenuation coefficient of water for the same CT voltage.

# Siemens, Medizinische Technik, Erlangen, Germany ♣ Nucletron B.V., Veenendaal, The Netherlands


6

The importance of scattered radiation in dosimetry must not be underestimated, but in the DRR calculation we obtain a much better image quality by the calculation of only primary energy attenuation. To perform this, we establish the relationship HUx↔µx, x = 1,..,4096, for all HU∈[-1024÷3071], and for all accelerator voltages of interest. This relationship is used to prepare the look-up tables which are then stored and used in the calculation process for DRRs.

2.4. Selection of DRR parameters

Before the process of calculation starts, the user needs to make a number of choices, including the following seven.

(1) Method for the DRR calculation of which there are two possibilities: the simulation method described in this paper or the splatting method described elsewhere (Crawfis et al 1993, Mao et al 1995, Max et al 1990, Westover 1986, Westover 1989, Williams 1992). The reasons for use of each of methods is described in Results.

(2) Area of calculation which we term the geometrical filter. We can choose between projecting the entire VDS information, only the information that is within the body contour, or projecting the user defined region which is selected through the Graphical User Interface (GUI).

(3) Resolution of VDS which is the number of pixels in the x and y directions (the CT image plane) and voxel size in the z direction.

(4) Accelerator voltage to be used for the simulation. This choice will be made from 11 in the range [0.050 ÷ 12.0] MeV together with the voltage used to acquire the CT scans.

(5) Isocenter coordinates are by default set at the centre of the target.

(6) Gantry angle, focus-axis-distance and focus-film-distance are by default set respectively to 0.00, 1000 mm and 1400 mm.

(7) DRR range filters which are anatomical (bone, fat, soft tissue) and catheter (plastic or metallic). Analysis was made of the HU characteristics for three kinds of catheters, see Appendix. These data were used for developing of the catheter filter.

2.5. Coordinate systems

The reference coordinate system we use is the CT coordinate system as defined in DICOM definition (NEMA 1993). As well as this reference coordinate system, each DRR has its own local coordinate system, figure 2. From figure 2 a relationship is established between the two coordinate systems. This enables easy navigation between all objects. The position of each VDS voxel is well defined in the reference coordinate system. The positions of each DRRs’ pixel is defined according to both the reference and the local DRR coordinate systems.

2.6. DRR calculation

Volume visualization of scalar data, or volume rendering, is the process of generating a two-dimensional image from a scalar function defined over a given region of three-dimensional space. This volume visualization is used to produce DRRs from the three-dimensional data set formed by patient CT images.

We present an algorithm for simulation of individual x-rays which are generated from the x-ray source and travel through human tissue onto the DRR image (Milickovic 1999a,


7

Milickovic et al 1999b). As input data for our algorithm we use VDS, user selected parameters and previously described look-up tables.

When the CT voltage is selected for which the CT images are to be acquired, the process of calculation is similar to that described by Sherouse et al 1990. Alternatively, look-up tables are used.

The basic DRR calculation algorithm is presented in figure 3. As can be seen, the attenuation data for an individual x-ray is accumulated while this ray travels voxel-by-voxel through the VDS. We do not use a constant step in the DRR calculation process for the following two reasons. (1) If we chose a small step this would not significantly improve our algorithm in terms of speed. (2) If we increase this step we could loose details such as catheter information which are very important when we use DRR in the field of brachytherapy.

When an individual ray has passed through the VDS, its value is:

∑=

−⋅=n

1i

dµo

iieII ,

where Io is the original ray value, i is the voxel through which the ray passes, µI is the linear attenuation coefficient of the material in voxel i and di is the segment between the entrance and output point of the ray in voxel i.

Images for six rep representative energies are presented in figure 4. Decrease of contrast in DRR images can be noticed as the accelerator voltage increases. This is caused by the decrease of differences between the linear attenuation coefficients of different human tissues with the increase of voltage (Cunningham et al 1980, Johns et al 1983, Milickovic 1999a. Milickovic et al 1999b). Figure 5 presents data measured for our Somatom Plus 4 CT scanner#. This effect occurs for the following reasons. The photoelectric coefficient per electron varies with atomic number as Z3 and per atom as Z4 and photoelectric cross-section varies with energy approximately as E-3. In this way the photoelectric effect plays an important role up to an energy of 150 keV. Between 150 keV and 10 MeV, the Compton effect which does not depends on the atomic number Z, is dominant and that causes that images acquired in this energy region do not have very good contrast between the different human tissues (Johns and Cunningham 1983).

2.7. Additional tools for work with DRRs

In 2.4 we have mentioned that the user can choose to include different kinds of filters. Geometrical filters enables a choice of the region we wish to project on the DRR image, figure 6. Anatomical filters enable us to include or exclude any kind of tissue from the DRR calculation process as illustrated in figure 7. This can be very useful, for example, if the region of interest is covered by bony structures: for example, catheters. We than can exclude bones from the calculation process and replace them by soft tissue or air.

The reconstruction of brachytherapy implants has usually been performed using two or more projected radiographs of the patient that were taken after the implantation. The catheter points are then digitized separately on films or alternatively, the films are scanned using a high resolution film scanner. A number of algorithms have been developed for catheter reconstruction from such projected radiographs (Bao 1991, Bullit et al 1997, Kassaee et al 1994, Li et al 1994, Li et al 1996, Metz et al 1989, Siddon and Chin 1984, Tabushi et al 1992). When making catheter reconstruction from the conventional radiological films, problems can occur if catheters are entirely or partly obscured by bones or even by soft tissue and fat in the case of an obese person. This is especially a problem if plastic catheters are implanted. Our idea was to apply catheter reconstruction algorithms using two isocentric DRRs (Milickovic 1999a, Milickovic et al 1999b). In this case, by using plastic and metallic

# Siemens, Medizinische Technik, Erlangen, Germany


8

catheter filters this problem can be solved, figure 8. The process is very user friendly, as there are navigation tools developed which indicate corresponding catheter points on two DRR images and guide the user through the reconstruction process, figure 9.

Making catheter reconstruction from two DRRs also enables us to easily skip from the usually ill-defined cases. One example is the case when two catheters are laying in the same projection plane and are seen on the DRR as one. This is solved by changing the table angle which in our case means rotation around the z-axis of the reference coordinate system.

Anatomical contours can be projected onto the DRR and seen as contours or surfaces with changeable transparencies. Also, markers and catheters, if they are already reconstructed, can be projected onto DRR images. Furthermore, we can see on the DRR the region of influence of each CT slice on the final DRR image.


9

3. RESULTS AND DISCUSSION

The time required for DRR simulation is dependent on the size of the three-dimensional volumetric data set and the DRR image resolution. We performed all tests using a Silicon Graphics O2 work-station with 180 MHz processor. The 512 x 512 DRR image is computed in about 55 s from a 512 x 512 x 128 data matrix, and in less than 22s from 256 x 256 x 128 data matrix. As the software is hardware independent it can easily be transferred onto a PC with a much higher processor speed. Corresponding times for the splatting method, which uses the same look-up tables as the simulation algorithm, are certainly much faster (12 s and 3 s). If anatomical or catheter filters are included in DRR calculation the calculation time is significantly reduced, as this algorithm makes a calculation only for the voxels that are within of HU region of interest. In this way the calculation of DRRs where catheter filters are included becomes a real time process. This enables the user to work interactively and to produce in real time number of DRR images from different views to select the two best views that are used of catheter reconstruction process. DRRs calculated by the splatting algorithm are of less quality than those calculated with simulation algorithm.

The catheter reconstruction from two DRRs developed using our algorithms was applied and tested in clinical practice. Our conclusion is that DRRs developed by our method can be used instead of conventional radiographs for brachytherapy treatment planning.

Some of advantages of the use of DRR for brachytherapy treatment planning instead of the use conventional radiological films when the patient CT images are available are as follows.

(a) Use of anatomical and catheter filters enables reconstruction of the catheters that would otherwise be completely or partly obscured by bony and soft tissue.

(b) There is no requirement for the radiological film acquisition and digitisation which significantly speeds up the process of catheter reconstruction.

(c) Possible patient motion during the CT slice acquisition using the new generation of spiral CT machines is insignificant compared to the motion present between the two radiological films acquisition. This is because the set of CT slices is acquired within the 10 s even for the big fields, and the patient can stop breathing for this time. Also, the slice thicnesses in use for brachytherapy treatment planning do not exceed 3 mm, which means that the reconstruction error due to this parameter is not greater than 1.5 mm. This discussion concerns the geometrical error in the catheter reconstruction process and a necessity to acquire at least two isocentric radiological films in the case of catheter reconstruction from two films. Opinion opposite to ours was presented as a poster by Hensley et al. 1999. but we are not certain which CT scanning machine and slice parameters were used in this study.

(d) There is no need for matching between two DRRs, as in the case of radiological films since DRRs are produced from the same volumetric data set and the same isocenter.

(e) Simple solutions for the (usually) ill-defined cases: explained in Methods.

As we can produce DRR images for selected voltages, this can serve for calculation of DRR images of very similar characteristics to those of available portal imaging. We can then use our DRR image and a portal image for an image correlation procedure for the correction of patient position in the same way as if working with a conventional mega voltage radiograph.


10

4. CONCLUSIONS

Presented tools and user friendly graphical user interface enable significant simplification of brachytherapy treatment planning process comparing to the classical based on projection techniques using the radiological films. Furthermore, the anatomical and catheter data are presented on the same set of CT slices.


11

APPENDIX

Hounsfield number properties of the catheters

The HU profile of the catheters on CT slice depends on the HU properties of the neighbouring tissues or materials, on the slice thickness and on the angle at which the catheter enters the CT slice. This is because the CT images are smoothed during the reconstruction process of the CT slice acquisition.

We have analysed the HU profiles of the flexible plastic catheters which have an outer diameter of 2.0 mm, wall thickness of 0.25 mm and effective wall density of 1.019 g/cm3. We have also analysed profiles of brain implant flexible needles with an outer diameter of 2.0 mm, wall thickness of 0.3 mm and effective wall density of 1.42 g/cm3. Finally, we analysed profiles for stainless steel trocar point needles with an outer diameter of 1.9 mm, wall thickness of 0.2 mm and wall density 8.02 g/cm3 . Example of our results is shown in figure 10. All have been obtained using a Somatom Plus 4 CT scanner#.

The HU profile observed for catheters depend on slice thickness, HU properties of the surrounding material and angle ϕ between the catheter central axis at the catheter entrance position to the CT slice and the vertical axis through the CT slice. When the catheter is not orthogonal to the CT slice (ϕ 0≠ ) the catheter area on the CT slice has an ellipsoid shape and its HU profile along the ellipse’s major axis is shown in figure 10 for ϕ=70o. The default HU values we use are in the following ranges, [-600, -200] HU for typical plastic catheter material and [2800, 3071] HU for typical metallic catheter material.

# Siemens, Medizinische Technik, Erlangen, Germany


12

REFERENCES

- Bahner ML, Debus J, Zabel A, Levegrun S and Van Kaick G 1999 Digitally reconstructed radiographs from abdominal CT scans as a new tool for radiotherapy planning Invest. Radiolog. 34 643-7

- Bao J 1991 On Three-Dimensional Reconstruction and Modeling of Coronary Arteries from Biplane X-Ray Angiograms Ph.D. Thesis Technishe Universität Berlin

- Battista JJ, Rider WD, van Dyke J 1980 Computed tomography for radiotherapy planning Int. J. Radiat. Oncol. Biol. Phys. 6 99-107

- Berger MJ and Hubbel JH 1987 XCOM: Photon Cross Sections on a Personal Computer NBSIR 87-3597

- Bullit E, Liu A and Pizer SM 1997 Three-dimensional reconstruction of curves from pairs of projection views in the presence of error. I. Algorithms Med. Phys. 24 1671-1678

- Chaney EL, Thorn JS, Tracton G, Cullip T, Rosenman JG and Trepper JE 1995 A portable software tool for computing digitally reconstructed radiographs Int. J. Radiat. Oncol. Biol. Phys. 32 491-497

- Constantinou C 1974 Tissue substitutes for particulate radiations and their use in radiation dosimetry and radiotherapy PhD Thesis London University

- Constantinou C and Harrington J 1992 An electron density calibration phantom for CT based treatment planning computers Med. Phys. 19 325-7

- Crawfis R and Max N 1993 Texture Splats for 3D Scalar and Vector Field Visualization Proceedings, Visualization ’93, IEEE Computer Society Press, Los Alamitos, CA 261-266

- Cunningham JR and Johns HE 1980 Calculation of the average energy absorbed in photon interactions Med. Phys. 7 51-54

- Dendy PP and Heaton B 1999 Physics for Diagnostic Radiology Medical Science Series, 2nd Edition, IOP Publishing Ltd.

- Hensley FW, Harms W, Krempien R, Fritz P, Berns Ch and Wannenmacher M 1999 Analysis of the geometrical accuracy of CT-based interstitial brachytherapy reconstruction Radiotherapy & Oncology 51 (supp. 2) S31

- Hubbel JH 1994 Bibliography of Photon Total Cross Section (Attenuation Coefficient) Measurements 10eV to 13.5GeV, 1907-1993 NISTIR 5437

- ICRP 1975 Report of the Task Group on Reference Man ICRP Publication 23

- ICRU 1989 Tissue Substitutes in Radiation Dosimetry and Measurement ICRU Report 46

- ICRU 1993 Prescribing, Recording, and Reporting Photon Beam Therapy ICRU Report 50

- ICRU 1999 Prescribing, Recording, and Reporting Photon Beam Therapy (Supplement to ICRU Report 50) ICRU Report 62

- Johns HE and Cunningham DJ 1983 The Physics of Radiology Thomas Springfield III 4th Edition

- Kassaee A and Altschuler MD 1994 Semiautomated matching and seed position location for implanted ribbons Med.. Phys. 21 643-650

- Li S, Chen GTY, Pelizzari CA, Reft C, Roeske JC and Lu Y 1996 A new source localization algorithm with no requirement of one-to-one source correspondence between biplane radiographs Med. Phys. 23 921-927

- Li S, Pelizzari CA and Chen GTY 1994 Unfolding patient motion with biplane radiographs Med. Phys. 21 1427-1433


13

- Mackie TR, El-Khatib E, Battista J, Scringer J, vanDyk J and Cunningham JR 1985 Lung dose corrections for 6 and 15 MV x-rays Med. Phys. 12 327

- Mao X, Hong L and Kaufman A 1995 Splatting of curvilinear volumes Proceedings of Visualization Atlanta 61-68

- Max N, Hanrahan P and Crawfis R 1990 Area and volume coherence for efficient visualization of 3D scalar functions Computer Graphics San Diego Workshop on Volume Visualization 24 27-33

- Metz CE and Fencil LE 1989 Determination of three-dimensional structure in biplane radiography without prior knowledge of the relationship between the two views: Theory Med. Phys. 16 45-51

- Milickovic N 1999a Three Dimensional CT Based Reconstruction Techniques in Modern Brachytherapy Treatment Planning Ph.D. Thesis National Technical University of Athens

- Milickovic N, Baltas D, Giannouli S, Uzunoglu N and Zamboglou N 1999b Catheter autorecognition and DRR based catheter reconstruction in brachytherapy Physica Medica XV-3 158, 178.

- Milickovic N, Giannouli S, Baltas D, Lahanas M, Kolotas C, Zamboglou N and Uzunoglu N 2000 Catheter autoreconstruction in computed tomography based brachytherapy treatment planning Med. Phys. 27 1047-1057

- NEMA National Electrical Manufacturers Association 1993 Digital imaging and communications in medicine (DICOM) NEMA Standards Publication PS 3.6 - 1993 NEMA: Washington.

- Schlegel W 1991 3D reconstruction in radiotherapy planning Radiologie 31 457-66

- Sherouse GW, Novins K and Chaney EL 1990 Computation of digitally reconstructed radiographs for use in radiotherapy treatment design Int. J. Radiat. Oncol. Biol. Phys. 18 651-658

- Siddon RL and Chin LM 1984 Two-film brachytherapy reconstruction algorithm Med. Phys. 12 77-83

- Tabushi K, Itoh S, Sakura M, Kutsutani-Nakamura Y, Iinuma TA, Arai T, and Irifune T 1992 Two-radiograph reconstruction using six geometrical solution sets and least-squares method Med. Phys. 19 1307-1310

-Westover L 1986 Footprint evaluation for volume rendering Computer Graphics SIGGRAPH ’86 Proceedings 20 367-376

-Westover L 1989 Interactive volume rendering Computer Graphics Chapel Hill Workshop on Volume Visualization Chapel Hill 9-16

-Williams P 1992 Interactive splatting of nonrectilinear volumes Poceedings of Visualization 37-44


14

FIGURE LEGENDS

Figure 7. DRR images resulting from calculation with different anatomical filters included.


15

Figure 1. Calibration curve of a Somatom Plus 4 CT scanner.


16

Figure 2. Relationship between reference CT coordinate system and local DRRs’ coordinate systems.


17

Figure 3. Presentation of the DRR simulation algorithm. If Io is the relative unit, then I is the new value after it pass through VDS.


18

Figure 4. Resulting DRR images for different accelerator voltages. The decrease of contrast can be noticed as the accelerator voltage increases.


19

Figure 5. Graph showing how the linear accelerator coefficients of different material depend on the accelerator voltage.


20

Figure 6. (a) CT image with an user selected region (blue) that was used for the DRR calculation process. (b) Resulting DRR image.


21

Figure 7. DRR images resulting from calculation with different anatomical filters included.


22


23

Figure 8. (a) DRR calculated for the case of prostate cancer with four metallic catheters implanted. (b) DRR calculated using the metallic catheter filter. (c) Resulting DRR image for the case of breast cancer with 10 plastic catheters implanted. (d) DRR image calculated using the plastic catheter filter. Notice that the plastic catheters are completely obscured in the case when the whole VDS (c) was used for DRR calculation.


24

Figure 9. GUI window for the case of catheter reconstruction from two DRRs calculated with metallic catheter filter included. The first DRR is calculated for the gantry angle 60o and the second one for 200o. We can choose the two best views, as there is no additional patient irradiation during the DRR calculation, as in the case of conventional film acquisition. For the selected point (red) on the first DRR, navigation tool points on the corresponding point (blue arrow and red point) on the second DRR.


25


26

Figure 10. Representative CT slices and HU profiles, respectively, for a plastic flexible

brain needle on the slices of 1 mm, 3 mm, 5 mm and 10 mm thicknesses: (a), (b) in air and

(c), (d) in water. The angle ϕ is defined as the angle between the catheter axis and the

orthogonal on the CT plane. For ϕ = 70° the profiles are calculated along the ellipse’s major

axis of the catheter area.

ct imaging based digitally reconstructed radiographs and ... · ct imaging based digitally...

Documents