- Email: [email protected]
A clinical application of an automated phantom-film QA procedure for validation of IMRT treatment planning and delivery
Medical Dosimetry, Vol. 29, No. 4, pp. 279-284, 2004 Copyright © 2004 American Association of Medical Dosimetrists Printed in the USA. All rights reserved 0958-3947/04/$–see front matter
doi:10.1016/j.meddos.2004.04.008
A CLINICAL APPLICATION OF AN AUTOMATED PHANTOM-FILM QA PROCEDURE FOR VALIDATION OF IMRT TREATMENT PLANNING AND DELIVERY ALEXANDER KAPULSKY, PH.D., GLEN GEJERMAN, M.D., and JOSEPH HANLEY, PH.D. Department of Radiation Oncology, Hackensack University Medical Center, Hackensack, NJ (Received 18 November 2003; accepted 5 April 2004)
Abstract—To quantify the correlation between planned and delivered intensity-modulated radiation therapy (IMRT) dose distributions, IMRT plans for 37 prostate carcinoma patients were analyzed. IMRT treatment plans were converted into hybrid phantom plans using a commercially available treatment planning system and delivered to a specialized film phantom via a static-tomotherapy technique. The films were analyzed using a commercial film dosimetry system. Hybrid phantom axial dose maps and film images were normalized, registered to one another, and subtracted to calculate the overall relative dose difference throughout the entire film area on a pixel-by-pixel basis. The average percentage of pixels with dose-difference values greater than ⴞ 3% among analyzed hybrid patient plans was 8.6% ⴞ 3%. The average percentage of pixels with dose differences greater than ⴞ 5% was 1.7% ⴞ 1.0%. The number of pixels with more than ⴞ 10% dose differences was negligible. An initial subset of hybrid plans was used to develop a quantitative criterion to verify for positional accuracy based on dosimetric verification of intensity map of IMRT plans for prostate patients in our institution. Plans with less than 5% of the pixels outside the ⴞ 5% dose-difference range were accepted. This method could be implemented for other anatomical sites or treatment planning and delivery systems. © 2003 American Association of Medical Dosimetrists. © 2004 American Association of Medical Dosimetrists. Key Words: IMRT treatment plan verification, Film dosimetery, QA phantom.
consuming. Diodes have limited accuracy and can only be used in low-dose gradient areas. Polymer gel can produce 3D dose distributions but is limited by expense, availability of magnetic resonance imaging (MRI) equipment, and relatively high thermal sensitivity. EPIDs could be used for profile verification and absolute dose measurements but their application in IMRT verification is still under development. High resolution, 2D dose distributions can be obtained using radiographic films10 and analyzed using commercially available film analysis systems. With the advance of automated dosimetry systems and the improved characteristics of extended dose range film, relative film dosimetry has become the method of choice for routine IMRT plan verification. The standard method of comparison for 2D dose distributions consists of overlaying hardcopy plots of measured and calculated doses in the form of cross-beam profiles, depth doses, or isodose distributions.11 To improve the efficiency and accuracy of IMRT quality assurance (QA), automated dose distribution verification methods have been implemented.12,13 We performed a plan/film analysis, using a computer-assisted registration technique, and determined the relative difference between the planned and delivered dose distributions on a pixel-by-pixel basis throughout the entire area of the radiographic film. To develop a criterion of acceptability, statistical analysis for 37 hybrid plans was performed on the dose differences to provide an objective score for the
INTRODUCTION Dosimetric verification criteria for conformal treatment plans is based upon either the analysis of a limited number of points in low-dose gradient areas or the measurement of distances between isodose lines in high-dose gradient areas. Quality assurance for intensity-modulated radiation therapy (IMRT) requires a more detailed protocol to quantify the steep dose-gradient regions associated with IMRT dose distributions. Because very high dose gradients can be created by individually modulated beamlets, and dose discrepancies may be small in one part of the image while large elsewhere, it is imperative that a global, rather than local, criterion for plan acceptance be developed. Ion chambers, thermoluminescent dosimeters (TLDs), diodes, radiographic and radiochromic films, polymer gels, and electronic portal imaging devices (EPIDs) have been used to measure IMRT dose distributions.1–9 While ion chambers are accurate instruments for absolute dose verification, their large size makes them impractical for analysis of the high-dose gradient areas commonly found in IMRT dose distributions. Although TLDs are relatively small, their analysis is time Reprint requests to: Dr. Alexander Kapulsky, Department of Radiation Oncology, Hackensack University Medical Center, 40-05 Kuiken Terrace, Hackensack, NJ 17601. E-mail: akapulsky@ humed.com Presented at the ASTRO 44th Annual Meeting, October 6 –10, 2002, New Orleans, LA 279
280
Medical Dosimetry
Volume 29, Number 4, 2004
fit. The objective scores were analyzed and an average number of pixels exceeding the ⫾ 3%, ⫾ 5%, and ⫾ 10% dose difference were calculated. Based on this initial subset of hybrid plans, acceptability limits were determined and a 2-parameter criterion was derived and applied to prostate patients treated with the same technique at our institution. The proposed criterion of acceptability reflects what we consider to be practically achievable in this clinical setting. The same methodology can be applied to other treatment techniques or treatment sites. METHODS AND MATERIALS IMRT plan specification The CORVUS v.4.0 treatment planning system was used to generate IMRT plans for the treatment of the prostate gland and seminal vesicles using 6-MV photons and a MIMiC multileaf collimator (NOMOS Corp., Sewickley, PA). Patients underwent treatment planning computed tomography (CT) simulation in the supine position with VacLok immobilization (Med-Tec, Orange City, IO) from the knees to the lower back. The clinical tumor volume (CTV) included the prostate and seminal vesicles, with an additional 3-mm margin (except at the rectal interface) around the prostate. The rectal and bladder volumes were contoured from the prostate apex to the superior extension of the seminal vesicles. The Corvus planning system automatically performed a true 3D expansion of the CTV to create the planning target volume (PTV). The margins for the expansion were determined by combining the immobilization and localization uncertainties in quadrature. The total uncertainty resulted in the following treatment margins: anterior - 10.2 mm, posterior - 10.2 mm, right - 5.8 mm, left - 5.8 mm, superior - 10.2 mm, and inferior - 10.2 mm. Inverse treatment planning enabled dose optimization so that the prescription goal to the PTV could be met while constraining the normal tissue dose. To enable effective sparing of the rectum and bladder, high-dose gradients between the target and critical organs are an integral part of the inverse treatment planning procedure. Hybrid plan delivery and analysis The intended/delivered plan verification procedure includes hybrid phantom plan calculation, plan dose map file transfer to the RIT film dosimetry system (Radiological Imaging Technology, Colorado Springs, CO), film phantom irradiation, and RIT plan/film registration and analysis (Fig. 1). This method utilizes an analysis of the entire axially-oriented film area of 15.2 ⫻ 12.7 cm (6 ⫻ 4 in). The details of this process were described elsewhere.13 Thirty-seven prostate treatment plans were converted into hybrid phantom plans using the Corvus treatment planning system. The hybrid phantom plan is created using the patient’s treatment plan beam’s configuration applied to the phantom geometry. This
Fig. 1. Plan/film IMRT QA procedure flowchart.
option is now available in most IMRT treatment planning systems.14 To simplify plan/film analysis, we have developed a specialized film phantom Bluebox (TARGETQA, Fair Lawn, NJ), which can be loaded with radiographic films and an ionization chamber. To account for any out-of-plane rotations, the phantom was aligned on a treatment table using lasers and fiducial marks on its surface. Hybrid plans were delivered using a 6-MV photon beam (VARIAN 6EX, Varian Medical Systems, Inc., Palo Alto, CA), and the exposed EDR-2 radiographic films (KODAK, Rochester, NY) were then processed on a KODAK X-Omat RP film processor and digitized using a VXR-12 film densitometer (Vidar Systems Corp., Herndon, VA). Image registration and analysis of the planned and delivered dose distributions were performed using the RIT film dosimetry system v.3.13. To support plan/film registration, the Bluebox film phantom has a simple geometrical relationship between 3 special grooves on its surface, the position of the film inside the phantom, and the corners of the region of interest (ROI), which includes the film image (6 ⫻ 5 in) with two 1-in margins (black rectangular areas on Fig. 2). The ROI width of 8 in and height of 5 in are equal to the distances between the milled grooves on the phantom. The hybrid plan axial dose map and the film image can therefore be easily registered using 3 corners of the ROI and 3 grooves on a phantom surface (Fig. 2). Because the dimensions of the ROI can be accurately specified in the RIT dosimetry system during film digitization, the registration process does not require any specific marks on the film surface. Since the position of the film inside the phantom does not change from one hybrid plan to another, the same registration file can be used for all hybrid plans, thus significantly reducing the overall time of registration. In relative film dosimetry, it is customary to nor-
Phantom-film QA for IMRT planning and delivery ● A. KAPULSKY et al.
Fig. 2. Plan and film images superimposed after registration: Registration points, shown with arrows, placed at the corners of rectangular region of interest (ROI) and the phantom grooves.
malize the calculated and measured dose distributions to a common point, usually the isocenter. We have previously described a novel method of plan/film normalization that uses a relatively small area in a high-dose low-gradient region13 of dose distribution. To reduce the uncertainties associated with dose measurements in highgradient areas of the dose distribution, it is important to include in this area only the low-gradient part of the dose distribution. The dimensions of the normalization area may change from one plan to another. For the treatment targets associated with prostate treatment plans included in our analysis, we used a rectangular normalization area
281
(approximately 2 cm2) near the isocenter consisting of 13,653 pixels (pixel size 0.169 mm2). If the planned and measured dose distributions were normalized correctly, we would expect a mean dose-difference value close to zero, similar to the histogram shown in Fig. 3 To achieve proper normalization of the planned and measured dose distributions, the dose-difference histogram and its peak were moved toward the 0% dosedifference position via an interactive adjustment of the plan dose scale factor (PDSF). PDSF is a single dosescale factor used in the RIT dosimetry system to convert the values in the plan file to actual dose units. To calculate the overall relative dose difference, normalized plan and film dose matrices were subtracted throughout the entire film area of approximately 500,000 pixels, including the high-gradient and the low-gradient dose regions (Fig. 3). Variables studied in the statistical analysis of the dose-difference distributions included the total number of pixels in the image, the number and percentage of pixels exceeding the threshold value for acceptability, and the maximum, minimum, and mean dose differences. The dose difference in a high-dose low-gradient area was also independently verified by point dose measurements using a PTW N30001 (0.6 cc) ionization chamber (CNMC Company, Inc., Nashville, TN). While absolute dose verification is an important part of IMRT QA and is a part of our ongoing QA, the description of this technique is beyond the scope of this paper. The RIT dosimetry system requires the user to
Fig. 3. Normalized plan/film dose difference histogram after PDSF adjustment. Maximum pixels count corresponds to the zero plan/film dose difference.
282
Medical Dosimetry
choose a threshold value for acceptability (TVA), or tolerance level, for analysis of the images. Each subtracted image contains a certain number of pixels with the plan/film dose differences within and above the TVA. The percentage of pixels above the TVA, called the exceedence, averaged over all analyzed plans, determines the confidence limits in the statistical analysis of the dose-difference distributions. Therefore, the acceptance criterion is a 2-parameter function, consisting of the tolerance level and the corresponding confidence level, which is related to the number of pixels above the TVA. For any chosen TVA level, we use the following formula for calculation of the confidence level (CL): CL ⫽ MEAN EXCEEDENCE ⫹ 3 STDV
(1)
where mean exceedence and standard deviations (STDV) are calculated from the initial subset of hybrid phantom plans. Venselaar et al.15 and, more recently, Palta et al.16 have introduced similar concepts. They recommended defining confidence limits to quantify the accuracy of the delivered dose as a mean dose-difference deviation plus several standard deviations. Palta chose a multiplication factor of 1.96 for standard deviation, which implies that 5% of the individual points remains in excess of the tolerance for that particular situation. The use of the multiplication factor 3 indicates that 99.7% of the pixels in the analyzed hybrid plans were within the confidence level. The TVA and the corresponding confidence level therefore reflect the achievable precision of an IMRT hybrid plan delivery in our institution. TVA values of 3%, 5%, and 10%, which corresponded to ⫾ 3%, ⫾ 5%, and ⫾ 10% plan/film dose differences, were tested during the statistical analysis of the subtracted images. RESULTS Hybrid phantom plans and film images were registered, normalized, and subtracted as described above and in more detail previously.13 The percentage of the pixels in high-dose lowgradient areas of the dose distribution for all analyzed hybrid plans was within the 3% tolerance level. The percentage of the pixels throughout the entire area of the dose distribution for all analyzed hybrid plans was within the 10% tolerance level. Therefore, all plans have met the recently proposed criterion16 for low- and high-gradient regions of an IMRT dose distribution. The percentage of pixels in the entire film area (including high-gradient regions) with dose-differences above the TVA for 37 hybrid patient plans is shown in Fig. 4. The percentage of pixels outside the 3% TVA, averaged over the initial subset of hybrid plans, was 8.6% ⫾ 3.0%. The greatest plan/film difference, or maximum exceedence, was 14.9% and the minimum was
Volume 29, Number 4, 2004
Fig. 4. Tolerance statistics for 37 prostate IMRT hybrid phantom plans using 3% and 5% TVA levels.
2.7%. The average percentage of pixels outside the 5% TVA was 1.7% ⫾ 1.0%. The maximum exceedence was 5.5% and the minimum was 0%. For the 37 hybrid plans analyzed in this study, the confidence levels calculated from Eq. (1) were 4.7% for a TVA of 5% and 17.6% for a TVA of 3%. DISCUSSION QA phantoms and plan/film registration Dose verification of IMRT is typically performed with phantoms. Tsai et al.1 used a humanoid phantom for QA of dynamic IMRT. Anthropomorphic phantoms are useful for initial acceptance testing but are not convenient for individual patient plan verification. Film phantoms capable of daily patient dose verification have been developed. Nomos developed the first commercially available IMRT phantom17 using films stacked between polystyrene plates. The MED-TEC IMRT/3D QA phantom uses ion chambers, TLD chips, MOSFET detectors, diodes, verification films, or radiochromic films. Standard Imaging (Standard Imaging, Middleton, WI) introduced an IMRT dose verification phantom for both IMRT commissioning and routine dose verification. We have chosen the Bluebox film phantom because it supports fast and precise plan/film registration. Using the Bluebox film phantom, we have achieved a very high correlation between the registration points without additional labor-intensive measurements. Registration point errors calculated by the RIT dosimetry system were negligible. The registration precision was therefore defined only by the size of the precut radiographic films, which was within 1 mm from its expected value, and their tight fit inside the phantom. The registration process, as well as the results of plan/film analysis, is very sensitive to small deviations from the correct registration points position. To illustrate the impact of the registration on a plan/film analysis, we shifted one of the registration points in a model calculation away from its correct position by 1 mm, 2 mm, and
Phantom-film QA for IMRT planning and delivery ● A. KAPULSKY et al.
283
Fig. 5. Subtracted dose images. (A) Different position of registration points. (1) optimal position, (2) 1-mm shift from the optimal position, (3) 2-mm shift, (4) 6-mm shift, (5) 8.5-mm diagonal shift. (B) Different PDSF factors. (A1)⫺optimal value of PDSF, increased (upper row) and decreased (lower row) factors from the optimal PDSF: by 2.5% (1), by 5% (2), and by 7.5% (3).
6 mm laterally, and by 8.5 mm diagonally. The corresponding percentage of pixels exceeding 5% TVA increased from 0.83% to 1.5%, 2.9%, 10.2%, and 23.4%, respectively. Dose difference maps (Fig. 5A) enable a visual assessment of the agreement between plan and film. The white area represents dose differences within the TVA level. To demonstrate the importance of a proper normalization, plan dose scale factors were also changed in the model calculation from the correct value, and the dose difference maps were recalculated (Fig. 5B). The percent of pixels with plan/film dose differences exceeding 5% TVA increased from 0.83% to 2.5% (1.0%), 8.7% (7.6%), and 29% (33%) when the PDSF was increased (decreased) by 2.5%, 5.0%, and 7.5% from its optimal value. Criterion of acceptability Radiation Therapy Committee Task Group 53 acknowledges that “the dosimetric accuracy required or achievable for treatment planning purposes has been the subject of much discussion. Each planning system, institution, and dosimetric situation will have its own requirements, capabilities, and limitations.”11 As treatment planning and delivery have become more and more complex, and verification methods have become more and more sophisticated, the criterion of acceptability has also undergone a continuing evolution. Cunningham et al.19 have indicated that an overall accuracy of 5% in dose delivery may be an acceptable goal on radiobiological grounds. The criteria of dose verification acceptability published by Van Dyk et al.19 and by the AAPM (TG53)11 are reflective of the complexity involved in dose distribution comparisons for conformal radiotherapy. The TG40 report on comprehen-
sive QA for radiation oncology recommends a tolerance of 2% for dose calculation accuracy of single fields and 2 mm in regions of high-dose gradients.20 Van Dyk’s criteria of dose calculation acceptability in conventional radiation therapy (2D and 3D) is 3% in relative dose accuracy in low-gradient regions and 4 mm spatial accuracy in regions with high-dose gradients. The dosimetric accuracy of IMRT treatment planning systems for single fields has been reported to be better than 3% or 3mm.21 Zhe Chen et al.,22 using the MIMiC multileaf collimator, has shown that direct dose measurements at the isocenter agreed with the calculations within ⫾ 3%. Recent studies have attempted to address the accuracy of calculated dose for the total plan transferred to a phantom compared with measurements for a series of patients. Tsai et al.1 accepted discrepancies within 5% as the criterion for treatment verification approval in the QA of dynamic IMRT. Localization accuracy of the planned distribution was obtained by manually comparing locations of planned and measured isodose lines relative to the gantry axis. Olch23 performed plan/film analysis in coronal planes by manual superposition of isodoses. He reported an agreement in the high-dose low gradient region within 1.7% ⫾ 2%. Xia and Change14 noted that while in a high-dose, relatively homogeneous region, ion chamber measurements are usually within 4% of the calculations; deviations in low-dose regions of more than 10% could be registered in some hybrid phantom plans. Kapulsky et al.13 have registered axial images using the RIT dosimetry system registration tools and have compared areas of high-dose low-gradient dose distribution on a pixel-by-pixel basis. They concluded that the plan/film differences in those areas were less than 5%. Most recently, Palta et al.16
284
Medical Dosimetry
have proposed a confidence level of 3% in high-dose, low-gradient areas, and 10% in high-dose and high-gradient areas of IMRT dose distribution as an acceptance-criterion. These recommendations are based on the results of a survey of facilities actively involved in IMRT. Because for IMRT a simple subdivision in regions with high- and low-dose gradients is no longer possible24 or may not be easily achievable, we apply a computerassisted registration and analysis technique to the entire area of a radiographic film. To assess an achievable accuracy in IMRT treatment planning and delivery in our institution, we have analyzed 3 TVA levels for each hybrid plan. A threshold value of 10% was not a meaningful level because observed dose differences were all less than 10%. A threshold value of 3%, used commonly for conformal treatment plans, tends to underestimate differences in the areas of high-dose gradient because a large number of pixels have plan/film dose differences beyond this TVA. In our clinical practice, we have chosen the TVA level of ⫾ 5%, with the confidence level of 5% (4.7% calculated from Eq. [1]), and applied this criterion to all prostate IMRT plans in our department. Although we have used only 37 plans for this statistical analysis, the values are homogenous with a very narrow confidence interval, indicating very little variability around the mean value. CONCLUSIONS An automated plan/film analysis technique provides comprehensive and precise QA of IMRT plans. It takes less then an hour to scan, register, and compare radiographic film and the corresponding planned dose distribution using the RIT dosimetry system. A Bluebox film phantom is a convenient tool for plan/film registration, which is a crucial part of the plan/film analysis. We have developed a criterion of acceptability by analyzing the results for 37 prostate cancer patients planned on the CORVUS treatment planning system and treated with a MIMiC multileaf collimator. This 2 parameter criterion, consisting of the tolerance level and the corresponding confidence level, is not universal but rather specific for the treatment site and treatment planning and delivery systems described in this paper. It is also dependent upon the type of the phantom and dosimetry system available. We believe that the approach we have described and the precision we have achieved using our method of plan/film analysis may provide a useful baseline in developing similar criteria for other treatment sites and in other institutions. A proposed criterion for acceptance quantifies practically achievable accuracy between planned and delivered IMRT dose distributions.
Acknowledgments—The authors thank John Napoli for useful discussion during manuscript preparation.
Volume 29, Number 4, 2004
REFERENCES 1. Tsai, J.-S.; Wazer, D.E.; Ling, M.N.; et al. Dosimetric verification of the dynamic intensity-modulated radiation therapy of 92 patients. Int. J. Radiat. Oncol.Biol. Phys. 40:1213–30; 1998. 2. Low, D.A.; Gerber, R.L.; Mutic, S.; Purdy, J.A. Phantoms for IMRT dose distribution measurements and treatment verification. Int. J. Radiat. Oncol. Biol. Phys. 40:1231–5; 1998. 3. Collaborative Working Group. Intensity-modulated radiotherapy: Current status and issues of interest. Int. J. Radiat. Oncol. Biol. Phys. 51:880 –914; 2001. 4. Oldham, M.; Wedd, S. Intensity-modulated radiotherapy by means of static tomotherapy: A planning and verification study. Med. Phys. 24:827–36; 1997. 5. Low, D.A.; Dempsey, J.F.; Venkatesan, R.; et al. Evaluation of polymer gels and MRI as a 3-D dosimeter for intensity-modulated radiation therapy. Med. Phys. 26:1542–51; 1999. 6. Low, D.A.; Mutic, S.; Dempsey, J.F.; et al. Quantitative dosimetric verification of an IMRT planning and delivery system. Radiother. Oncol. 49:305–16; 1998. 7. Xing, L.; Curran, B.; Holmes, R.; et al. Dosimetric verification of commercial inverse planning system. Phys. Med. Biol. 44:463–78; 1999. 8. Williamson, J.F.; Khan, F.M.; Sharma, S.C. Film dosimetry of megavoltage photon beams: A practical method of isodensity-toisodose curve conversion. Med. Phys. 8:94 –9; 1981. 9. Chang, J.; Mageras, G.S.; Chui, C.S.; et al. Relative profile and dose verification of intensity-modulated radiation therapy. Int. J. Radiat. Oncol. Biol. Phys. 47:231–240; 2000. 10. Robar, J.L.; Clark, B.G. The use of radiographic film for linear accelerator stereotactic radiosurgical dosimetry. Med. Phys. 26: 2144 –50; 1999. 11. American Association of Physicists in Medicine. Radiation Therapy Committee Task Group 53: Quality assurance for clinical radiotherapy treatment planning. Med. Phys. 25:1773–829; 1998. 12. Low, D.A.; Dempsey, J.F.; Markman, J.; et al. Toward automated quality assurance for intensity modulated radiation therapy. Int. J. Radiat. Oncol. Biol. Phys. 53:443–52; 2002. 13. Kapulsky, A.; Mullokandov, E.; Gejerman, G. An automated phantom-film QA procedure for intensity-modulated radiation therapy. Med. Dosim. 27:201–207; 2002(3). 14. Xia, P.; Change, C. Patients specific quality assurance in IMRT. In: Palta, J.R., Mackie, T.R., editors. Intensity-Modulated Radiation Therapy. The State of the Art. Medical Physics Monograph No. 29. Madison, WI: Medical Physics Publishing; 2003:495–514. 15. Venselaar, J.; Welleweerd, G.H.; Mijnheer, B. Tolerances for the accuracy of photon beam dose calculations of treatment planning systems. Radiother. Oncol. 60:191–201; 2001. 16. Palta, J.R.; Kim, S.; Li, J.G.; Liu, C. Tolerances limits and action levels for planning and delivery of IMRT. In: Palta, J.R., Mackie, T.R., editors. Intensity-Modulated Radiation Therapy. The State of the Art. Medical Physics Monograph No. 29. Madison, WI: Medical Physics Publishing; 2003:593–612. 17. Low, D.A.; Gerber, R.L.; Mutic, S.; Purdy, J.A. Phantoms for IMRT dose distribution measurements and treatment verification. Int. J. Radiat. Oncol. Biol. Phys. 40:1231–5; 1998. 18. Cunningham, J. Quality assurance in dosimetry and treatment planning. Int. J. Radiat.Oncol. Biol. Phys. 10:105–9; 1984. 19. Van Dyk, J.; Barnett, R.B.; Cygler, J.E.; Shragge, P.C. Commissioning and quality assurance of treatment planning computers. Int. J. Radiat. Oncol. Biol. Phys. 26:261–73; 1993. 20. American Association of Physicists in Medicine. Radiation Therapy Committee TaskGroup 40: Comprehensive QA for Radiation. Med. Phys. 21:37; 1994. 21. Papatheodorou, S.; Rosenwald, J.C.; Zefkili, S. Dose calculation and verification of intensity modulation generated by multilieaf collimators. Med. Phys. 27:960 –71; 2000. 22. Chen, Z.; Xing, L.; Nath, R. Independent monitor unit calculation for IMRT using MIMiC multileaf collimator. Med. Phys. 29:2041– 51; 2002. 23. Olch, A.J. “Dosimetric accuracy of the ITP™ inverse treatment planning system. Med.Phys. 29:2484 –88; 2002. 24. Georg, D.; Kroupa, B.; Winkler, P.; Potter, R. Normalized sensitometric curves for the verification of hybrid IMRT treatment plans with multiple energies. Med. Phys. 30:1142–50; 2003.
doi:10.1016/j.meddos.2004.04.008
A CLINICAL APPLICATION OF AN AUTOMATED PHANTOM-FILM QA PROCEDURE FOR VALIDATION OF IMRT TREATMENT PLANNING AND DELIVERY ALEXANDER KAPULSKY, PH.D., GLEN GEJERMAN, M.D., and JOSEPH HANLEY, PH.D. Department of Radiation Oncology, Hackensack University Medical Center, Hackensack, NJ (Received 18 November 2003; accepted 5 April 2004)
Abstract—To quantify the correlation between planned and delivered intensity-modulated radiation therapy (IMRT) dose distributions, IMRT plans for 37 prostate carcinoma patients were analyzed. IMRT treatment plans were converted into hybrid phantom plans using a commercially available treatment planning system and delivered to a specialized film phantom via a static-tomotherapy technique. The films were analyzed using a commercial film dosimetry system. Hybrid phantom axial dose maps and film images were normalized, registered to one another, and subtracted to calculate the overall relative dose difference throughout the entire film area on a pixel-by-pixel basis. The average percentage of pixels with dose-difference values greater than ⴞ 3% among analyzed hybrid patient plans was 8.6% ⴞ 3%. The average percentage of pixels with dose differences greater than ⴞ 5% was 1.7% ⴞ 1.0%. The number of pixels with more than ⴞ 10% dose differences was negligible. An initial subset of hybrid plans was used to develop a quantitative criterion to verify for positional accuracy based on dosimetric verification of intensity map of IMRT plans for prostate patients in our institution. Plans with less than 5% of the pixels outside the ⴞ 5% dose-difference range were accepted. This method could be implemented for other anatomical sites or treatment planning and delivery systems. © 2003 American Association of Medical Dosimetrists. © 2004 American Association of Medical Dosimetrists. Key Words: IMRT treatment plan verification, Film dosimetery, QA phantom.
consuming. Diodes have limited accuracy and can only be used in low-dose gradient areas. Polymer gel can produce 3D dose distributions but is limited by expense, availability of magnetic resonance imaging (MRI) equipment, and relatively high thermal sensitivity. EPIDs could be used for profile verification and absolute dose measurements but their application in IMRT verification is still under development. High resolution, 2D dose distributions can be obtained using radiographic films10 and analyzed using commercially available film analysis systems. With the advance of automated dosimetry systems and the improved characteristics of extended dose range film, relative film dosimetry has become the method of choice for routine IMRT plan verification. The standard method of comparison for 2D dose distributions consists of overlaying hardcopy plots of measured and calculated doses in the form of cross-beam profiles, depth doses, or isodose distributions.11 To improve the efficiency and accuracy of IMRT quality assurance (QA), automated dose distribution verification methods have been implemented.12,13 We performed a plan/film analysis, using a computer-assisted registration technique, and determined the relative difference between the planned and delivered dose distributions on a pixel-by-pixel basis throughout the entire area of the radiographic film. To develop a criterion of acceptability, statistical analysis for 37 hybrid plans was performed on the dose differences to provide an objective score for the
INTRODUCTION Dosimetric verification criteria for conformal treatment plans is based upon either the analysis of a limited number of points in low-dose gradient areas or the measurement of distances between isodose lines in high-dose gradient areas. Quality assurance for intensity-modulated radiation therapy (IMRT) requires a more detailed protocol to quantify the steep dose-gradient regions associated with IMRT dose distributions. Because very high dose gradients can be created by individually modulated beamlets, and dose discrepancies may be small in one part of the image while large elsewhere, it is imperative that a global, rather than local, criterion for plan acceptance be developed. Ion chambers, thermoluminescent dosimeters (TLDs), diodes, radiographic and radiochromic films, polymer gels, and electronic portal imaging devices (EPIDs) have been used to measure IMRT dose distributions.1–9 While ion chambers are accurate instruments for absolute dose verification, their large size makes them impractical for analysis of the high-dose gradient areas commonly found in IMRT dose distributions. Although TLDs are relatively small, their analysis is time Reprint requests to: Dr. Alexander Kapulsky, Department of Radiation Oncology, Hackensack University Medical Center, 40-05 Kuiken Terrace, Hackensack, NJ 17601. E-mail: akapulsky@ humed.com Presented at the ASTRO 44th Annual Meeting, October 6 –10, 2002, New Orleans, LA 279
280
Medical Dosimetry
Volume 29, Number 4, 2004
fit. The objective scores were analyzed and an average number of pixels exceeding the ⫾ 3%, ⫾ 5%, and ⫾ 10% dose difference were calculated. Based on this initial subset of hybrid plans, acceptability limits were determined and a 2-parameter criterion was derived and applied to prostate patients treated with the same technique at our institution. The proposed criterion of acceptability reflects what we consider to be practically achievable in this clinical setting. The same methodology can be applied to other treatment techniques or treatment sites. METHODS AND MATERIALS IMRT plan specification The CORVUS v.4.0 treatment planning system was used to generate IMRT plans for the treatment of the prostate gland and seminal vesicles using 6-MV photons and a MIMiC multileaf collimator (NOMOS Corp., Sewickley, PA). Patients underwent treatment planning computed tomography (CT) simulation in the supine position with VacLok immobilization (Med-Tec, Orange City, IO) from the knees to the lower back. The clinical tumor volume (CTV) included the prostate and seminal vesicles, with an additional 3-mm margin (except at the rectal interface) around the prostate. The rectal and bladder volumes were contoured from the prostate apex to the superior extension of the seminal vesicles. The Corvus planning system automatically performed a true 3D expansion of the CTV to create the planning target volume (PTV). The margins for the expansion were determined by combining the immobilization and localization uncertainties in quadrature. The total uncertainty resulted in the following treatment margins: anterior - 10.2 mm, posterior - 10.2 mm, right - 5.8 mm, left - 5.8 mm, superior - 10.2 mm, and inferior - 10.2 mm. Inverse treatment planning enabled dose optimization so that the prescription goal to the PTV could be met while constraining the normal tissue dose. To enable effective sparing of the rectum and bladder, high-dose gradients between the target and critical organs are an integral part of the inverse treatment planning procedure. Hybrid plan delivery and analysis The intended/delivered plan verification procedure includes hybrid phantom plan calculation, plan dose map file transfer to the RIT film dosimetry system (Radiological Imaging Technology, Colorado Springs, CO), film phantom irradiation, and RIT plan/film registration and analysis (Fig. 1). This method utilizes an analysis of the entire axially-oriented film area of 15.2 ⫻ 12.7 cm (6 ⫻ 4 in). The details of this process were described elsewhere.13 Thirty-seven prostate treatment plans were converted into hybrid phantom plans using the Corvus treatment planning system. The hybrid phantom plan is created using the patient’s treatment plan beam’s configuration applied to the phantom geometry. This
Fig. 1. Plan/film IMRT QA procedure flowchart.
option is now available in most IMRT treatment planning systems.14 To simplify plan/film analysis, we have developed a specialized film phantom Bluebox (TARGETQA, Fair Lawn, NJ), which can be loaded with radiographic films and an ionization chamber. To account for any out-of-plane rotations, the phantom was aligned on a treatment table using lasers and fiducial marks on its surface. Hybrid plans were delivered using a 6-MV photon beam (VARIAN 6EX, Varian Medical Systems, Inc., Palo Alto, CA), and the exposed EDR-2 radiographic films (KODAK, Rochester, NY) were then processed on a KODAK X-Omat RP film processor and digitized using a VXR-12 film densitometer (Vidar Systems Corp., Herndon, VA). Image registration and analysis of the planned and delivered dose distributions were performed using the RIT film dosimetry system v.3.13. To support plan/film registration, the Bluebox film phantom has a simple geometrical relationship between 3 special grooves on its surface, the position of the film inside the phantom, and the corners of the region of interest (ROI), which includes the film image (6 ⫻ 5 in) with two 1-in margins (black rectangular areas on Fig. 2). The ROI width of 8 in and height of 5 in are equal to the distances between the milled grooves on the phantom. The hybrid plan axial dose map and the film image can therefore be easily registered using 3 corners of the ROI and 3 grooves on a phantom surface (Fig. 2). Because the dimensions of the ROI can be accurately specified in the RIT dosimetry system during film digitization, the registration process does not require any specific marks on the film surface. Since the position of the film inside the phantom does not change from one hybrid plan to another, the same registration file can be used for all hybrid plans, thus significantly reducing the overall time of registration. In relative film dosimetry, it is customary to nor-
Phantom-film QA for IMRT planning and delivery ● A. KAPULSKY et al.
Fig. 2. Plan and film images superimposed after registration: Registration points, shown with arrows, placed at the corners of rectangular region of interest (ROI) and the phantom grooves.
malize the calculated and measured dose distributions to a common point, usually the isocenter. We have previously described a novel method of plan/film normalization that uses a relatively small area in a high-dose low-gradient region13 of dose distribution. To reduce the uncertainties associated with dose measurements in highgradient areas of the dose distribution, it is important to include in this area only the low-gradient part of the dose distribution. The dimensions of the normalization area may change from one plan to another. For the treatment targets associated with prostate treatment plans included in our analysis, we used a rectangular normalization area
281
(approximately 2 cm2) near the isocenter consisting of 13,653 pixels (pixel size 0.169 mm2). If the planned and measured dose distributions were normalized correctly, we would expect a mean dose-difference value close to zero, similar to the histogram shown in Fig. 3 To achieve proper normalization of the planned and measured dose distributions, the dose-difference histogram and its peak were moved toward the 0% dosedifference position via an interactive adjustment of the plan dose scale factor (PDSF). PDSF is a single dosescale factor used in the RIT dosimetry system to convert the values in the plan file to actual dose units. To calculate the overall relative dose difference, normalized plan and film dose matrices were subtracted throughout the entire film area of approximately 500,000 pixels, including the high-gradient and the low-gradient dose regions (Fig. 3). Variables studied in the statistical analysis of the dose-difference distributions included the total number of pixels in the image, the number and percentage of pixels exceeding the threshold value for acceptability, and the maximum, minimum, and mean dose differences. The dose difference in a high-dose low-gradient area was also independently verified by point dose measurements using a PTW N30001 (0.6 cc) ionization chamber (CNMC Company, Inc., Nashville, TN). While absolute dose verification is an important part of IMRT QA and is a part of our ongoing QA, the description of this technique is beyond the scope of this paper. The RIT dosimetry system requires the user to
Fig. 3. Normalized plan/film dose difference histogram after PDSF adjustment. Maximum pixels count corresponds to the zero plan/film dose difference.
282
Medical Dosimetry
choose a threshold value for acceptability (TVA), or tolerance level, for analysis of the images. Each subtracted image contains a certain number of pixels with the plan/film dose differences within and above the TVA. The percentage of pixels above the TVA, called the exceedence, averaged over all analyzed plans, determines the confidence limits in the statistical analysis of the dose-difference distributions. Therefore, the acceptance criterion is a 2-parameter function, consisting of the tolerance level and the corresponding confidence level, which is related to the number of pixels above the TVA. For any chosen TVA level, we use the following formula for calculation of the confidence level (CL): CL ⫽ MEAN EXCEEDENCE ⫹ 3 STDV
(1)
where mean exceedence and standard deviations (STDV) are calculated from the initial subset of hybrid phantom plans. Venselaar et al.15 and, more recently, Palta et al.16 have introduced similar concepts. They recommended defining confidence limits to quantify the accuracy of the delivered dose as a mean dose-difference deviation plus several standard deviations. Palta chose a multiplication factor of 1.96 for standard deviation, which implies that 5% of the individual points remains in excess of the tolerance for that particular situation. The use of the multiplication factor 3 indicates that 99.7% of the pixels in the analyzed hybrid plans were within the confidence level. The TVA and the corresponding confidence level therefore reflect the achievable precision of an IMRT hybrid plan delivery in our institution. TVA values of 3%, 5%, and 10%, which corresponded to ⫾ 3%, ⫾ 5%, and ⫾ 10% plan/film dose differences, were tested during the statistical analysis of the subtracted images. RESULTS Hybrid phantom plans and film images were registered, normalized, and subtracted as described above and in more detail previously.13 The percentage of the pixels in high-dose lowgradient areas of the dose distribution for all analyzed hybrid plans was within the 3% tolerance level. The percentage of the pixels throughout the entire area of the dose distribution for all analyzed hybrid plans was within the 10% tolerance level. Therefore, all plans have met the recently proposed criterion16 for low- and high-gradient regions of an IMRT dose distribution. The percentage of pixels in the entire film area (including high-gradient regions) with dose-differences above the TVA for 37 hybrid patient plans is shown in Fig. 4. The percentage of pixels outside the 3% TVA, averaged over the initial subset of hybrid plans, was 8.6% ⫾ 3.0%. The greatest plan/film difference, or maximum exceedence, was 14.9% and the minimum was
Volume 29, Number 4, 2004
Fig. 4. Tolerance statistics for 37 prostate IMRT hybrid phantom plans using 3% and 5% TVA levels.
2.7%. The average percentage of pixels outside the 5% TVA was 1.7% ⫾ 1.0%. The maximum exceedence was 5.5% and the minimum was 0%. For the 37 hybrid plans analyzed in this study, the confidence levels calculated from Eq. (1) were 4.7% for a TVA of 5% and 17.6% for a TVA of 3%. DISCUSSION QA phantoms and plan/film registration Dose verification of IMRT is typically performed with phantoms. Tsai et al.1 used a humanoid phantom for QA of dynamic IMRT. Anthropomorphic phantoms are useful for initial acceptance testing but are not convenient for individual patient plan verification. Film phantoms capable of daily patient dose verification have been developed. Nomos developed the first commercially available IMRT phantom17 using films stacked between polystyrene plates. The MED-TEC IMRT/3D QA phantom uses ion chambers, TLD chips, MOSFET detectors, diodes, verification films, or radiochromic films. Standard Imaging (Standard Imaging, Middleton, WI) introduced an IMRT dose verification phantom for both IMRT commissioning and routine dose verification. We have chosen the Bluebox film phantom because it supports fast and precise plan/film registration. Using the Bluebox film phantom, we have achieved a very high correlation between the registration points without additional labor-intensive measurements. Registration point errors calculated by the RIT dosimetry system were negligible. The registration precision was therefore defined only by the size of the precut radiographic films, which was within 1 mm from its expected value, and their tight fit inside the phantom. The registration process, as well as the results of plan/film analysis, is very sensitive to small deviations from the correct registration points position. To illustrate the impact of the registration on a plan/film analysis, we shifted one of the registration points in a model calculation away from its correct position by 1 mm, 2 mm, and
Phantom-film QA for IMRT planning and delivery ● A. KAPULSKY et al.
283
Fig. 5. Subtracted dose images. (A) Different position of registration points. (1) optimal position, (2) 1-mm shift from the optimal position, (3) 2-mm shift, (4) 6-mm shift, (5) 8.5-mm diagonal shift. (B) Different PDSF factors. (A1)⫺optimal value of PDSF, increased (upper row) and decreased (lower row) factors from the optimal PDSF: by 2.5% (1), by 5% (2), and by 7.5% (3).
6 mm laterally, and by 8.5 mm diagonally. The corresponding percentage of pixels exceeding 5% TVA increased from 0.83% to 1.5%, 2.9%, 10.2%, and 23.4%, respectively. Dose difference maps (Fig. 5A) enable a visual assessment of the agreement between plan and film. The white area represents dose differences within the TVA level. To demonstrate the importance of a proper normalization, plan dose scale factors were also changed in the model calculation from the correct value, and the dose difference maps were recalculated (Fig. 5B). The percent of pixels with plan/film dose differences exceeding 5% TVA increased from 0.83% to 2.5% (1.0%), 8.7% (7.6%), and 29% (33%) when the PDSF was increased (decreased) by 2.5%, 5.0%, and 7.5% from its optimal value. Criterion of acceptability Radiation Therapy Committee Task Group 53 acknowledges that “the dosimetric accuracy required or achievable for treatment planning purposes has been the subject of much discussion. Each planning system, institution, and dosimetric situation will have its own requirements, capabilities, and limitations.”11 As treatment planning and delivery have become more and more complex, and verification methods have become more and more sophisticated, the criterion of acceptability has also undergone a continuing evolution. Cunningham et al.19 have indicated that an overall accuracy of 5% in dose delivery may be an acceptable goal on radiobiological grounds. The criteria of dose verification acceptability published by Van Dyk et al.19 and by the AAPM (TG53)11 are reflective of the complexity involved in dose distribution comparisons for conformal radiotherapy. The TG40 report on comprehen-
sive QA for radiation oncology recommends a tolerance of 2% for dose calculation accuracy of single fields and 2 mm in regions of high-dose gradients.20 Van Dyk’s criteria of dose calculation acceptability in conventional radiation therapy (2D and 3D) is 3% in relative dose accuracy in low-gradient regions and 4 mm spatial accuracy in regions with high-dose gradients. The dosimetric accuracy of IMRT treatment planning systems for single fields has been reported to be better than 3% or 3mm.21 Zhe Chen et al.,22 using the MIMiC multileaf collimator, has shown that direct dose measurements at the isocenter agreed with the calculations within ⫾ 3%. Recent studies have attempted to address the accuracy of calculated dose for the total plan transferred to a phantom compared with measurements for a series of patients. Tsai et al.1 accepted discrepancies within 5% as the criterion for treatment verification approval in the QA of dynamic IMRT. Localization accuracy of the planned distribution was obtained by manually comparing locations of planned and measured isodose lines relative to the gantry axis. Olch23 performed plan/film analysis in coronal planes by manual superposition of isodoses. He reported an agreement in the high-dose low gradient region within 1.7% ⫾ 2%. Xia and Change14 noted that while in a high-dose, relatively homogeneous region, ion chamber measurements are usually within 4% of the calculations; deviations in low-dose regions of more than 10% could be registered in some hybrid phantom plans. Kapulsky et al.13 have registered axial images using the RIT dosimetry system registration tools and have compared areas of high-dose low-gradient dose distribution on a pixel-by-pixel basis. They concluded that the plan/film differences in those areas were less than 5%. Most recently, Palta et al.16
284
Medical Dosimetry
have proposed a confidence level of 3% in high-dose, low-gradient areas, and 10% in high-dose and high-gradient areas of IMRT dose distribution as an acceptance-criterion. These recommendations are based on the results of a survey of facilities actively involved in IMRT. Because for IMRT a simple subdivision in regions with high- and low-dose gradients is no longer possible24 or may not be easily achievable, we apply a computerassisted registration and analysis technique to the entire area of a radiographic film. To assess an achievable accuracy in IMRT treatment planning and delivery in our institution, we have analyzed 3 TVA levels for each hybrid plan. A threshold value of 10% was not a meaningful level because observed dose differences were all less than 10%. A threshold value of 3%, used commonly for conformal treatment plans, tends to underestimate differences in the areas of high-dose gradient because a large number of pixels have plan/film dose differences beyond this TVA. In our clinical practice, we have chosen the TVA level of ⫾ 5%, with the confidence level of 5% (4.7% calculated from Eq. [1]), and applied this criterion to all prostate IMRT plans in our department. Although we have used only 37 plans for this statistical analysis, the values are homogenous with a very narrow confidence interval, indicating very little variability around the mean value. CONCLUSIONS An automated plan/film analysis technique provides comprehensive and precise QA of IMRT plans. It takes less then an hour to scan, register, and compare radiographic film and the corresponding planned dose distribution using the RIT dosimetry system. A Bluebox film phantom is a convenient tool for plan/film registration, which is a crucial part of the plan/film analysis. We have developed a criterion of acceptability by analyzing the results for 37 prostate cancer patients planned on the CORVUS treatment planning system and treated with a MIMiC multileaf collimator. This 2 parameter criterion, consisting of the tolerance level and the corresponding confidence level, is not universal but rather specific for the treatment site and treatment planning and delivery systems described in this paper. It is also dependent upon the type of the phantom and dosimetry system available. We believe that the approach we have described and the precision we have achieved using our method of plan/film analysis may provide a useful baseline in developing similar criteria for other treatment sites and in other institutions. A proposed criterion for acceptance quantifies practically achievable accuracy between planned and delivered IMRT dose distributions.
Acknowledgments—The authors thank John Napoli for useful discussion during manuscript preparation.
Volume 29, Number 4, 2004
REFERENCES 1. Tsai, J.-S.; Wazer, D.E.; Ling, M.N.; et al. Dosimetric verification of the dynamic intensity-modulated radiation therapy of 92 patients. Int. J. Radiat. Oncol.Biol. Phys. 40:1213–30; 1998. 2. Low, D.A.; Gerber, R.L.; Mutic, S.; Purdy, J.A. Phantoms for IMRT dose distribution measurements and treatment verification. Int. J. Radiat. Oncol. Biol. Phys. 40:1231–5; 1998. 3. Collaborative Working Group. Intensity-modulated radiotherapy: Current status and issues of interest. Int. J. Radiat. Oncol. Biol. Phys. 51:880 –914; 2001. 4. Oldham, M.; Wedd, S. Intensity-modulated radiotherapy by means of static tomotherapy: A planning and verification study. Med. Phys. 24:827–36; 1997. 5. Low, D.A.; Dempsey, J.F.; Venkatesan, R.; et al. Evaluation of polymer gels and MRI as a 3-D dosimeter for intensity-modulated radiation therapy. Med. Phys. 26:1542–51; 1999. 6. Low, D.A.; Mutic, S.; Dempsey, J.F.; et al. Quantitative dosimetric verification of an IMRT planning and delivery system. Radiother. Oncol. 49:305–16; 1998. 7. Xing, L.; Curran, B.; Holmes, R.; et al. Dosimetric verification of commercial inverse planning system. Phys. Med. Biol. 44:463–78; 1999. 8. Williamson, J.F.; Khan, F.M.; Sharma, S.C. Film dosimetry of megavoltage photon beams: A practical method of isodensity-toisodose curve conversion. Med. Phys. 8:94 –9; 1981. 9. Chang, J.; Mageras, G.S.; Chui, C.S.; et al. Relative profile and dose verification of intensity-modulated radiation therapy. Int. J. Radiat. Oncol. Biol. Phys. 47:231–240; 2000. 10. Robar, J.L.; Clark, B.G. The use of radiographic film for linear accelerator stereotactic radiosurgical dosimetry. Med. Phys. 26: 2144 –50; 1999. 11. American Association of Physicists in Medicine. Radiation Therapy Committee Task Group 53: Quality assurance for clinical radiotherapy treatment planning. Med. Phys. 25:1773–829; 1998. 12. Low, D.A.; Dempsey, J.F.; Markman, J.; et al. Toward automated quality assurance for intensity modulated radiation therapy. Int. J. Radiat. Oncol. Biol. Phys. 53:443–52; 2002. 13. Kapulsky, A.; Mullokandov, E.; Gejerman, G. An automated phantom-film QA procedure for intensity-modulated radiation therapy. Med. Dosim. 27:201–207; 2002(3). 14. Xia, P.; Change, C. Patients specific quality assurance in IMRT. In: Palta, J.R., Mackie, T.R., editors. Intensity-Modulated Radiation Therapy. The State of the Art. Medical Physics Monograph No. 29. Madison, WI: Medical Physics Publishing; 2003:495–514. 15. Venselaar, J.; Welleweerd, G.H.; Mijnheer, B. Tolerances for the accuracy of photon beam dose calculations of treatment planning systems. Radiother. Oncol. 60:191–201; 2001. 16. Palta, J.R.; Kim, S.; Li, J.G.; Liu, C. Tolerances limits and action levels for planning and delivery of IMRT. In: Palta, J.R., Mackie, T.R., editors. Intensity-Modulated Radiation Therapy. The State of the Art. Medical Physics Monograph No. 29. Madison, WI: Medical Physics Publishing; 2003:593–612. 17. Low, D.A.; Gerber, R.L.; Mutic, S.; Purdy, J.A. Phantoms for IMRT dose distribution measurements and treatment verification. Int. J. Radiat. Oncol. Biol. Phys. 40:1231–5; 1998. 18. Cunningham, J. Quality assurance in dosimetry and treatment planning. Int. J. Radiat.Oncol. Biol. Phys. 10:105–9; 1984. 19. Van Dyk, J.; Barnett, R.B.; Cygler, J.E.; Shragge, P.C. Commissioning and quality assurance of treatment planning computers. Int. J. Radiat. Oncol. Biol. Phys. 26:261–73; 1993. 20. American Association of Physicists in Medicine. Radiation Therapy Committee TaskGroup 40: Comprehensive QA for Radiation. Med. Phys. 21:37; 1994. 21. Papatheodorou, S.; Rosenwald, J.C.; Zefkili, S. Dose calculation and verification of intensity modulation generated by multilieaf collimators. Med. Phys. 27:960 –71; 2000. 22. Chen, Z.; Xing, L.; Nath, R. Independent monitor unit calculation for IMRT using MIMiC multileaf collimator. Med. Phys. 29:2041– 51; 2002. 23. Olch, A.J. “Dosimetric accuracy of the ITP™ inverse treatment planning system. Med.Phys. 29:2484 –88; 2002. 24. Georg, D.; Kroupa, B.; Winkler, P.; Potter, R. Normalized sensitometric curves for the verification of hybrid IMRT treatment plans with multiple energies. Med. Phys. 30:1142–50; 2003.