Evaluation of accuracy in determination of stereotactical isocenter: from image acquisition to treatment.

Transcription

1 Evaluation of accuracy in determination of stereotactical isocenter: from image acquisition to treatment. ANNA SARDO 1, EUGENIA MADON 1, EDOARDO TREVISIOL 1, DANILO SANTUARI 1, VERONICA RICHETTO 1, ROBERTO PELLEGRINI 2, ALESSANDRO URGESI 1, M.D. 1 Department of Radiotherapy, OIRM S.Anna Hospital, Turin, Italy 2 3Dline Medical Systems s.r.l, Milan, Italy. Purpose: To analyse sources of spatial uncertanty on target localisation due to image acquisition, patients immobilization and linac movements accuracy. Methods and materials: Longitudinal error in target localization due to image acquisition has been evaluated using a home made phantom on 3Dline re-locatable immobilization system. To analyse random and sistematic errors in target localization due to patient immobilization, 15 patients submitted to stereotactic radiotherapy have been analysed througt fusion on DRR on TC images, simulation and linac computed radiography. Accuracy on linac isocenter positionig has been evaluated checking laser and optical tracking system, available on radiotherapy department. Results: Mean error on TC localization is strongly dependent to image acquisition parameters: for pitch less than 1,5 mm and slice thickness not superior than 3 mm that value is 1,28 ± 0,2 mm for elicoidal scans and 1,1 ± 0,28 mm for assial scans. About patient immobilization accuracy, from fusion images analysis a mean error of 1,75 ±1,6 mm have been evaluated. Mechanical movements on linac give precision of 0,4 ± 0,5 mm in isocenter localization based on laser, a precision of 0,07 ± 0,17 mm when optical fibres have been used. Conclusion: According to different clinical intent, different treatment must be chosen: total error can be variable from 1,28 ± 0,26 mm in better case of radiosurgery with invasive frame and optical fibers localisation, to 2,21 ± 1,68 mm in worst case of stereotactical radiotherapy with non invasive re-locatable frame and laser localization. INTRODUCTION In recent years, advances in imaging, treatment planning and treatment delivery have changed the perspective of radiotherapy of intracranial lesions: focused radiation with different types of tertiary collimation systems has enabled to deliver high doses to small targets. Stereotactic techniques are now available at many institutions and allow precise localization of small targets; imobilization with invasive systems is generally used for stereotactic radiosurgery (SRS) which delivers high doses in a

2 single fraction. However, there are a great number of tumors which can better be treated with dose fractionation due to their size or encasement of critical structures. Fractionated stereotactic radiotherapy (fsrt) is an attractive option to treat such tumors; fsrt utilizes the same technology as SRS for tumor localization, treatment planning and irradiation delivery; patient positioning however must use non-invasive relocatable systems. Fixation with thermoplastic masks has been used for many years in radiotherapy for the treatment of lesions in the head and neck area; widespread knowledge of the technology, quick and simple use and low cost make it an attractive mean of fixation when a high number of stereotactic treatments is anticipated. Acceptance testing of a SRS system has been described in AAPM report 54 (2) : inaccuracies in isocenter positioning may arise from the imaging system, the localization software, isocenter positioning in the treatment room and the delivery system (gantry and couch rotation). In fsrt the additional errors are due to inter and intra-fraction head movements within the imobilization device and daily repeated localization in the treatment room. Gantry and couch rotation testing and tolerances have been accurately described (2) and don t need to be further discussed. Errors concerning isocenter localization and positioning are evaluated and discussed in this report. Digital images resolution Target volume definition and treatment planning computation are more frequently performed on CT images. Further MR or PET/SPECT images give more precision on tumor boundary or make evidence on differences into target volume or between it and surronding necrotic areas. CT images are a discrete spatial system depending on pixel dimension, slice thickness on longitudinal axis and spacing between two

3 consecutive slices. Correlation between accuracy in point definition on stereotactic coordinates and CT images spatial resoluction has been studied (4,5, 6). Stereotactical frames In stereotactic system an imobilization obtained with invasive frame is able to guarantee precision of 0.1 mm, because of solidity between frame and skull. When thermoplastic masks are used for fsrt, accuracy is influenced by mask conformation and thermoplastic material rigidity: inaccuracies that may arise from head translation and rotation movements, due to different head position on support, or from errors in patient re-position, can be evaluated. Linac localization systems The final step in the stereotactic chain is alignement between patient isocenter and delivery system isocenter in the treatment room: in fsrt this procedure is repeated several times and a random component is added to the systematic error caused by misalignement between the two isocenters. Traditionally alignement between patient and Linac isocenter is accomplished with lasers: optical tracking is another option that may increase precision, especially in fractionated treatments. In our Department fsrt for intracranial patologies was started in 1997: imobilization/localization was based on a relocatable frame allowing head fixation with a thermoplastic mask and/or a bite block: the purpose of this paper is the evaluation of all the components of the final localization and positioning of the isocenter in fsrt when the mask-based localization/immobilization system is used.

4 MATERIAL AND METHODS Imobilization system and stereotactic equipment Imobilization in fsrtwere performed with a stereotactic relocatable system based on a face mask attached to a ring and couch stand; for stereotactic radiosurgery an invasive frame, mounted as the relocatable one, is employed (3D Line International s.r.l.). Stereotactic localization has been performed using adapted supports with radiopaque fiducials for CT acquisition and millimetric scale for geometric coordinates individuation on linac. All equipment are by 3D Line International s.r.l.. The stereotactic treatment have been performed with different diameters circular collimators and with micro/minimultileaf collimator developed by 3D line International s.r.l.. Both systems were commissioned with the 6 MV beam produced by an Elekta SL 75-5 linear accelerator. Phantoms To verify localization geometric accuracy an apposite phantom has been used (figure 1). It is based on fifteen identical plexiglass plates (15 x 15 cm 2 ) with mean thickness of 9,92 mm; all plates are in position by two plexiglass pullings posed on phantom opposite corners. In the thirteen inner plates three small steel spheres (0,8 mm of diameter) have been posed on rectum triangle apexes (30, 39, 49,2 mm of side respectively). Each triangle baricentrum is on each plate centrum. On central plate baricentrum an other steel sphere has been posed. It represents phantom isocenter. Each single sphere positionig error is not higher than ± 0,15 mm and spheres distances vary between 9,92 mm and 138,1 mm. Phantom has been posed into localization equipment as mentioned above and for each steel sphere stereotactical coordinated have been evaluated with fluoroscopy. Four central plate reperi and six

5 spheres on previous and successive plates have been considered. Stereotactic scale and fluoroscopy have the same precision of 1 mm: sistematic error on each measure 2 2 can be so evaluated as = 0, 7 mm Image acquisition Different CT images on phantom have been performed. CT scanner is a CT Twin by Elscint with two detectors array of 1 mm of dimension on trasversal axes and 10 mm on longitudinal axes. Slice thickness is given by beam collimation, centered in the middle of detectors array. To increase spatial resolution on cross direction ultra high resolution scanner modality has been used: detection surface results 0,5 x 10 mm. CT scanner has been used in axial and spiral modality: different axial scans have been performed varying slice thickness; all available combination of pitch and slice thickness have been used for helicoidal scans. In table 1 all CT acquisitions performed are summerized. Acquisition matrix dimension is 512 x 512 pixel, with a pixel dimension of 0,57 x 0,57 mm 2 : it has been the same for all acquisitions. Localization All performed images have been localized (SCT Module on ERGO by 3Dline International Systems) and mean calibration error on each images set has been evaluated. Difference between stereotactical coordinates by fluoroscopy and by localization software has been evaluated. Sequential verification films To evaluate head patient shift into mask, 15 patients have been studied. On every mask surface five radioopaque markers have been positionated, having so reference points indipendent from head patient position. During treatment two RX

6 images (AP and LL) have been performed once a week. Through DicomView software by Tecnologie Avanzate (TA) a projections point registration has been performed, using three of the five markers on patients masks. Registration error has been calculated by difference of markers not used for registration. Anatomical markers shift from reference image has been evaluated. In figure 2 and 3 anatomical markers chosen for analysis are shown in AP and LL projection respectly. Real patient shift measure into mask has been evaluated subtracting registration error to position shift. Patients and treatment parameters For patient localization during stereotactic treatment linac routine equipments have been used: laser and optical tracking. The three laser are parallel and intersect themself at isocenter. They are frequently verified and deviation accepted is less than 1 mm. Optical tracking is based on three cameras with infrared LED and a serie of positioning reflecting spheres posed on circular basis. These spheres are positioned directly on patients or on immobilization systems. Spheres image has been acquired by the three cameras that shield completely visible radiation. After calibration tracking system stability has been verified using a five spheres array centred on isocenter. With this equipment also laser stability can be verified. RESULTS Spatial resolution on digital image Localization software can give calibration error for each image set and for each single slice. Mean calibration error measured is 0,7 mm ± 0,1 mm. Standard deviation is very small: error measured isn t due to acquision modality but to localization reperi dimension that isn t punctual. CT scanner point spread function

7 can be calculated by gray intensity FWHM: there is an inverse proportionality between resoluction and this function. In figure 4 grey level distribution of one localization repere, chosen by chance from a central acquisition slice, has been shown. From Gaussian fit a mean intensity value of 3595,1 HU with a standard deviation of 159, 8 HU has been extracted. That value corresponds to a FWHM of 0,8 mm and is comparable to mean calibration error. Differences between stereotactic coordinates by fluoroscopy and by software have been evaluated. Sistematic errors in measures must be summed: total sistematic error is 1,4 mm. Both shift proceeding on longitudinal direction than total distance have been considered. In figure 5 mean total difference with interpolation equation and relative correlation coefficient has been shown. This parameter can be obtained by threedimensional Ω 2. In figure 6 bidimensional matrix of total shift varying slice thickness and pitch has been displayed. Difference has a polinomial course with a mean correlation coefficient of 0,94. For axial scans different behaviour has been found: there is a higher gradient also for small slice thickness. Quadratic term in interpolation equation is negative; differences calculated are also higher because casual error in point localization on trasversal slices must be considered. That parameter has been calculated making different localization of a same point with a standard deviation of 0,15 mm. Mean difference is always less than 1,5 mm for scans with pitch 1 and for slice thickness 3 mm as shown in the last figures. For helicoidal scans it is measured as 1,28 ± 0,2 mm, while for axial scans mean error is calculated 1,1 ± 0,28 mm. Imobilization system First analysis concerns casual error in point localization: making anatomical repere coordinate determination for 50 times a standard deviation of 0,6 has been found.

8 That parameter differs from error deriving from radiopaque repere individuation becouse of different view and consequent localization: anatomical reperi mark is more operating dependent than radiopaque one. Fusion error related to each projection regard reference image has been evaluated. In figure 7 normal distribution of such error is illustrated with mean and standard deviation. Different projection registration has been performed for each single patient and for every various image: mean total different as quadratic sum of three direction shift has been calculated, as shown in figure 8. In table 2 single direction shift as been summerized.shift value from AP direction is higher than others: rotational movements into mask causes a more evident movement of anatomical reperi considered in vertical axis rather than orizzontal one. As example in figure 9 a patient fusion with this phenomenon has been shown. On the right side fusion between two LL projections has been shown; particular on anatomical reperi shift is on the left one. Head rotation gives a shift more evident on AP direction. Mean total error on all 15 patients results 1,75 ± 1,6 mm. Patient and treatment parameters All laser and optical tracking systems are checking every stereotactic radiosurgery or two times a week for fractionated stereotactic radiotherapy. To correct a casual shift from tollerance it can be possible to adjust laser system with micrometric screws; for optical tracking equipment it always possible to recalibrate cameras. In table 3 and 4 casual deviation for laser and optical tracking has been quantified making day by day registration for a whole working week in all three spatial directions. Mean laser localization error is 0,55 ± 0,26 mm, mean optical tracking deviation is 0,15 ± 0,06 mm.

9 CONCLUSIONS Choice of immobilization systems and localization equipments have been made through clinical and physical criteria, as regard all error sources analysed in that work. Such criteria are based on target dimensions and on surrounding organ at risk. In table 5 chiose criteria are summerized and related localization error is shown. REFERENCES 1 Leksell DG. Stereotactic radiosurgery: Present status and future trends. Neurol Res 1987; 9:60. 2 Schell MC, Bova FJ, Larson DA, et al. AAPM Report No. 54. Stereotactic radiosurgery. Report of AAPM Task Group Siddon RL, Barth WH. Stereotaxic localization of intracranial targets. Int. J. Radiat. Oncol. Biol. Phys. 1987; vol.13, 1241: ChengY, Apuzzo MLJ, Chi-Shing and Petrovich Z. A phantom study of the geometric accuracy of computed tomographic and magnetic resonance imaging stereotactic localization with the Leksell stereotactic system. Neurosurgery. 2001; vol.48, 1092: Henri CJ, Collins DL and Peters TM. Multimodality image integration for stereotactic surgical planning. Med. Phys. 1991; vol.18, 167:177.

10 6 Murphy MJ. The importance of computed tomography slice thickness in radiographic patient positioning for radiosurgery. Med. Phys. 1999;vol.26, 171:175 7 Yeung D, Palta J, Fontanesi J, et al. Systematic analysis of errors in target localization and treatment delivery in stereotactic radiosurgery. Int. J. Radiat. Oncol. Biol. Phys 1993; vol.28, 493: Mack A, Czempiel H, Durr G, et al. Quality assurance in stereotactic space. A system test for verifying the accuracy of aim in radiosurgery. Med. Phys. 2002; vol.29, 561: Lemieux L, Kitchen ND, Hughes SW, Thomas DG. Voxel-based localization in frame-base and frameless stereotaxy and its accuracy. Med. Phys. 1994; vol.21, 1301: Kortmann RD, Becker G, Perelmouter J, et al. Geometric accuracy of field alignment in fractionated stereotactic radiotherapy of brain tumors. Int. J. Radiat. Oncol. Biol. Phys. 1999; vol.43, 921: Dong L, Boyer AL. An image correlation procedure for digitally reconstructed radiographs and electronic portal images. Int. J. Radiat. Oncol. Biol. Phys. 1995; vol.33, 1053: Willner J, Flentje M, Bratengeier K. CT simulation in stereotactic brain radiotherapy analysis of isocenter reproducibility with mask fization. Radiother. Oncol.1997; vol.45,83: 88.

11 FIGURES AND TABLES captions Figure 1: Phantom used for localization geometric accuracy analisys. Table 1: CT acquisitions performed. Figure 2: AP projection : on red pen radiopaque reperi on mask, on blue pen anatomical reperi, used for shift calculation, are visualized. Figure 3: LL projection: on red pen radioopaque reperi on mask, on blue pen anatomical reperi, used for shift calculation, are visualized. Figure 4: Histogram on grey levels distribution of a stereotactic repere. Figure 5: Mean total shift with interpolation equation and relative correlation coefficient. Figure 6: Mean total shift as a function of slice thickness and pitch. Figure 7: Mean normal distribution of fusion error. Figure 8: Mean total fusion shift on patient. Table 2: Single direction mean fusion shift. Figure 9: Fusion between two LL projections on the right; anatomical reperi visualization on the left side. Table 3: Optical tracking daily deviations on three spatial directions. Table.4: Laser daily deviations on three spatial directions. Table.5:Choise criteria for stereotactic treatment and relative localization error.

12 FIGURES AND TABLES Figure 1 Pitch Slice thickness (mm) 0,5 0,6 1,3 3,2 6,5 10,4 13 0,7 0,5 1,1 2,7 5,5 8, ,6 1,3 3,2 6,5 10,4 13 1,5 0,5 1,1 2,7 5,5 8, ,6 1,3 3,2 6,5 10,4 13 Table 1 Supero-lateral corner of right orbit Supero-lateral corner of left orbit Anterior nasal spine Figure 2

13 Superior frontal sinus Anterior clinoid process External auditory meatus Figure 3 Figure 4 Figure 5

14 Figure 6 Figure 7 Total shift 15 shift (mm) patients Figure 8

15 Mean (mm) Dev.St (mm) AP Direction 0,55 0,91 LL Direction 1,09 1,98 CC Direction 0,44 1,88 Table 2 ΔAP ΔCC Figure 9 Table 3 Shift (mm) lateral vertical longitudinal ,2 0,3 0,1 3 0,1 0,1 0,2 4 0,2 0,2 0,2 5 0,3 0,2 0,1 mean 0,16 0,16 0,12 dev.st 0,11 0,11 0,08 Shift (mm) lateral vertical longitudinal ,9 0,9 0,8 3 0,6 1, ,2 0,6 0,5 5 0,7 0,8 0,2 mean 0,48 0,68 0,5 dev.st 0,37 0,42 0,41 Table 4

16 Lesion characteristic Treatment type Immobilization System Localization System Imaging Total Localization error Critical structure inside lesion (i.e cavernous sinus) fsrt Mask error 1,75 ± 1,6 mm Optical Tracking error 0,15± 0,06 mm TC error 1,28± 0,2 mm 2,17 ± 1,61 mm Lesion distance from critical structures 5 mm; Lesion diameter 30 mm SRS Invasive frame Optical Tracking error 0,15± 0,06 mm TC error 1,28± 0,2 mm 1,29 ± 0,2 mm Lesion distance from critical structures 5 mm; Lesion diameter > 30 mm fsrt Mask error 1,75 ± 1,6 mm OpticaTracking error 0,15± 0,06 mm TC error 1,28± 0,2mm 2,17 ± 1,61 mm Lesion distance from critical structures > 5 mm; Lesion diameter 30 mm SRS Invasive frame Laser error 0,55 ± 0,26 mm TC error 1,28± 0,2 mm 1,39 ± 0,32 mm Lesion distance from critical structures > 5 mm; Lesion diameter >30 mm fsrt Mask error 1,75 ± 1,6 mm Laser error 0,55 ± 0,26 mm TC error 1,28± 0,2 mm 2,24 ± 1,63 mm Table 5