A NEW APPROACH IN THE DETERMINATION OF PATIENT S REFERENCE POINT IN CONFORMAL EXTERNAL RADIOTHERAPY Kordatzakis Antonios, Pragmatefteli Maria, Baltas Dimos, Koutsouris Dimitrios 4 Biomedical Engineering Laboratory, National Technical University of Athens, 577 Zografou, Greece, akorda@biomed.ntua.gr Biomedical Engineering Laboratory, National Technical University of Athens, 577 Zografou, Greece, mpragma@biomed.ntua.gr Department of Medical Physics and Engineering, Strahlenklinik, Klinikum Offenbach, 6069 Offenbach, Germany, dimos.baltas@klinikum-offenbach.de 4 Biomedical Engineering Laboratory, National Technical University of Athens, 577 Zografou, Greece, dkoutsou@biomed.ntua.gr ABSTRACT Our study involves mainly patient s immobilization and imaging in conformal radiotherapy process. The goal is the accurate automatic determination of the patient s reference point in regard to table coordinates, which is very important for the successful execution of the whole treatment, as patient s set up must be literally maintained between imaging and treatment phase. The significance of the determination of the reference point, emerge in the treatment phase for the accurate transfer of the planned irradiation technique to the patient. The determination of RP to table coordinates is based mainly on detection of markers on patient s skin and in table in a DICOM image, generated during the imaging process. The results are compared with the same results acquired manually by a tape measure and by the use of commercial software for D simulation of external beam radiation therapy. KEY WORDS Radiotherapy, image processing, patient s set-up patient set-up in external beam radiotherapy is based on ink marking on the surface of the skin. This set up is usually verified by comparing portal images acquired during treatment delivery with reference images acquired during treatment planning. The reference images show the planned position of PTV and Organs in Risk in relation to the radiation field, which is compared with their actual position during treatment []. When used with a formal verification protocol, portal imaging is very efficient to eliminate large systematic errors []. In some cases, set up verification based on portal imaging of the treatment field is difficult. Recently, the use of Mega-Voltage Cone Beam CT (MVCT) acquired at the treatment time for patient positioning has been proposed. The MVCT may be compared and matched with the planning CT for accurate determination of patient offset [4]. Our approach presents the first step of a navigation technique which takes into account RP coordinates and patient D model, produced combining a D camera and imaging results, and control the positioning of patient at the beginning of each fraction.. Introduction Conformal radiotherapy is, after surgery the most frequently used and successful treatment for cancer. It is applied in more than 50% of all cancer patients []. The course of radiotherapy can be described by the following steps: Patient immobilization, imaging, tumour localization, treatment planning, treatment, quality assurance and verification. In the first step, patient is placed on the table for imaging. This position should be exactly the same during treatment. This demands the adjustment of the treatment table with the patient until the isocenter position matches the pre-calculated coordinates. The ultimate goal of our work is to describe a reliable solution to patient positioning problem. Typically the. Description of the problem As mentioned above the first goal is the automatic determination of patients reference point to table coordinates. The exact procedure is the following: Patient is placed on the table and ink marking is placed on the surface of his/her skin. This marking should match to the room-mounted orthogonal lasers, which define the isocenter. Patient s reference point is the point in patient s body, which matches exactly to the isocenter. If this point is accurately calculated, then the irradiation field can be successfully estimated in regard to this point. In order to find the exact position of the RP in relation to the table, a specific table has been constructed by Prof. Dr. Baltas in Offenbach hospital. This table is fitted with metallic 458-4 54
sticks, which are visible in CT image. The table contains three metallic sticks along its length. Each stick has 0mm distance from the other. All technical characteristics of the table are shown in Figure. In this way someone can accurately calculate the exact position of a slice on the table by the number and structure of visible sticks. The determination of RP uses the CT slice in DICOM image format [5,6], produced during imaging, which includes all three markers on patient body, as in Figure. The white arrows show the three markers on patient s skin, which define the RP point of the patient. The yellow circles present the three upper table markers. Under these markers there should be a number of lower markers regarding the position on the table. Figure : Example of a DICOM image after successful automatic detection of the table and CT markers.. Implementation The calculation of the RP to Table coordinates is the most important process in this study. There are currently two different ways supported, to calculate the RP to table point (Figure ). Manually by user determination of the CT table markers and RP point and automatically, by user bounding and automatic determination of CT markers and RP point. The exact coordinates (x, y, z) of RP location on table s coordinate system are calculated as follows: X = X RP - X M Z = abs(y RP Average (Y L, Y M, Y R )) offset Y = - (N-) * 400 d, where Figure : Technical characteristics of CT table 55
(X L,Y L ) Coordinates of Left Marker point (X M,Y M ) Coordinates of Middle Marker point (X R,Y R ) Coordinates of Right Marker point Coordinates of RP to Table point (set (X RP,Y RP ) by the user) Coordinates (in IEC) of RP to Table (X,Y,Z) point (to be calculated) The number of Position points set by N the user ( N 4) The distance (in Z axis) between the Offset center of Marker points and the surface of the table The distance (in X axis) between the leftmost and the rightmost Marker d point. d = X R - X L Figure 4: Coordinate system of the CT table The automatic way of RP location in table s coordinate system calculation utilizes a CT markers detection algorithm, which follows (Figure 5): Set a region containing the markers Find blobs in image that have HU in the specified range The Y coordinate is the same for all points since all points are defined in the same CT image (slice). Import DICOM image file Patient's Table's or patient's CT markers? Table's Find the right most blob Find the three upper blobs Manual Manual or automatic calculation? Automatic Find the left most blob Find the first lower blob under the upper middle blob Set left upper marker of CT table Set a region containing upper and lower markers of the CT table Find the top most blob Search for other blobs with the same y coordinate Set middle upper marker of CT table Set a region containing the markers of the reference point Set number of lower markers of CT table Set right upper marker of CT table Set number of lower markers of CT table Return the coordinates of the center of masses of all detected markers Set reference point Figure 5: Flowchart describing the algorithm for markers detections Calculation of reference point's coordinates x,y,z in regard to the CT table Figure : Flowchart describing the calculation of reference point in regard to the tables coordinates The coordinate system of the table used in all calculations is the following (Figure 4). Table CT markers: The table has two rows of markers capable of accurately determining CT slice s position on the table. Upper markers are always three: left, middle, and right. Lower markers can differ in number from to 4 in respect to the position on the table. Patient CT markers: CT markers on patient s skin can easily determine the reference point. 56
This algorithm can detect all CT markers contained in a user defined region. It is based on low level image characteristics (Hounsfield s) and location information. The detection of CT table markers, except a window of Hounsfield s, calculates also other specific characteristics as their number, which can be between 4 and 7, and their relative position. All Hounsfield s of CT images were retrieved directly from the DICOM files. The detection of the RP point takes also into account that these CT markers should be placed on patient s body, so they should be the top-, left- and right-most points. The markers detection algorithm takes as input an HU and its half range. The automatic calculation of table CT markers and RP point is based on markers HU. Because these s can differ from image to image, user can set a window of HU s, which better detect the markers on each image. User can configure separately HU and half range for table markers and RP point marker. All above algorithms were implemented in C++ and were integrated into the Anticollision, Registration and Information System for Radiation Therapy (ARIS) software [7]. 4. Evaluation In the framework of the presented study, measurements about the position of 8 patients were taken in different ways, in order to compare their differences and reliability. The measured is the location of RP in table s coordinate system in Cartesian coordinates. At first the RP location was measured manually on the CT table by a tape measure. Then it was calculated manually on Exomio [8] using appropriate measurement tools provided by the software and finally using ARIS both in manual and automatic mode. The differences of the above results are shown in the following graphics: Difference (mm) 0 Differences between measurments in ARIS manual and automatic mode 4 5 6 7 8 90456789045678 Cases x y z Difference (mm),00,50,00,5 0,0 0 0,50 0,00.5.5.5 0.5 0 Differences between measurements in Exomio and in ARIS automatic mode 4 56 Manual- Exomio 7 89 0 4 5 6 7 8 9 0 4 5 6 7 8 Cases s of differences Aris manual - Aris automatic Aris automatic- Exomio x y z x y z It is quite obvious that the mean s of the differences in RP location in the table coordinate system between ARIS and Exomio are below mm, which is considered a very accurate estimation. Manual calculation with tape measure is the most inaccurate as expected, because of the various practical limitations e.g. the patient s set-up on the table. It is also very important that the mean of the difference of the three coordinates of RP between the two modes of ARIS is below mm. This means, that the automatic calculation of RP to table coordinates is not only easier and more user friendly, but also very accurate and reliable. The differences between manual tape measurements and Exomio are quite great as expected, with averages of.70mm,.44mm and.9mm for axes X, Y, Z respectively. The average of difference between Exomio and ARIS automatic mode is 0.86mm, 0.79mm and 0.9mm for axes X, Y, Z respectively. 4. Conclusion This approach presents a different way of estimating the RP location in table s coordinate system. The advantage of this method is that it offers an automatic and easy interface; it can be easily integrated in radiotherapy 57
process and can provide reliability not only in imaging, but also in radiation treatment phase. Our future work includes the combination of this study with a patient D model produced by a D camera during imaging and treatment phases for patient s set up verification (computer-aided set-up). The D camera will be mounted on the ceiling on both rooms of CT and accelerator and will produce patient D model. Markers visible from the camera will be placed on the table and patient s skin. The exact position of D model on the CT table will be specified utilizing the above method and all movements regarding the initial position will be easily calculated. 5. Acknowledgements This study is performed in close cooperation with the Medical Physics Department of the Public Hospital of Offenbach, Germany. This research was supported through a European Community Marie Curie Fellowship (HPMT-CT-000-000). References: [] Wolfgang Schlegel, Andreas Mahr, D Conformal Radiation therapy: A multimedia introduction to methods and techniques (Springer, Heidelberg, 00) [] Gilhuijs KG, van Herk M. Automatic on-line inspection of patient setup in radiation therapy using digital portal images. Med Phys, 99 0:667-677 [] A. Bel, R. Keus, R.E. Vijlbrief, and J.V. Lebesque, "Setup deviations in wedged pair irradiation of parotid gland and tonsillar tumors, measured with an electronic portal imaging device," Radiother. Oncol. 7, 995, 5 59 [4] D tomographic reconstruction from portal imaging for patient positioning, Matthias Mitschke, Ali Bani- Hashemi, Farhad Ghelmansarai, Ali Khamene and Jean Pouliot, Proc. 7 th CARS Conf on Computer Assisted Radiology and Surgery, International Congress Series Volume 56, June 00, Pages 07- [5] American College of Radiology, National Electrical Manufactures Association: ACR-NEMA Digital Imaging and Communications in Medicine (DICOM): Version.0, ACR-NEMA Committee, Working Group VI, Washington DC, 99 [6] NEMA official DICOM web site: http://medical.nema.org/ and current version available online http://medical.nema.org/dicom/00.html [7] Miltiadis F. Tsiakalos, Spiros Vathis, Dimos Baltas, ARIS: A D treatment simulation and collision detection computer tool for radiotherapy treatment planning, Med. Phys. 7(7), 00, 59-6 [8]Exomio software by medintec, available online http://www.medintec.de/mse/indexmse.html 58