- Email: [email protected]
The evolution of quality assurance for intensity- modulated radiation therapy (IMRT): sequential tomotherapy
Int. J. Radiation Oncology Biol. Phys., Vol. 56, No. 1, pp. 274 –286, 2003 Copyright © 2003 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/03/$–see front matter
doi:10.1016/S0630-3016(03)00097-X
3D-CRT
THE EVOLUTION OF QUALITY ASSURANCE FOR INTENSITYMODULATED RADIATION THERAPY (IMRT): SEQUENTIAL TOMOTHERAPY SHIAO Y. WOO, M.D., WALTER GRANT III, PH.D., JOHN E. MCGARY, PH.D., BIN S. TEH, M.D., AND E. BRIAN BUTLER, M.D. Baylor College of Medicine, The Methodist Hospital, Houston, TX Purpose: To identify the pertinent issues to be addressed in successfully implementing IMRT using sequential tomotherapy into clinical reality and presenting the maturation of quality assurance (QA) programs for both the delivery system and patient treatments that allow routine clinical use of the system. Methods and Materials: Initially, a cubic phantom containing silver halide film was exposed to the entire treatment before patient treatment. The processed films were digitized with a laser densitometer and the dose distributions were compared with that generated by the planning system. Later, software that calculates the dose delivered to any phantom employing the intensity patterns developed in the inverse planning system for an individual patient was implemented for point checks of dose. A measurement phantom for use with this software was developed and evaluated on a large number of patients. Invasive fixation was used for all cranial patients initially. To use sequential tomotherapy for other sites and larger targets, noninvasive immobilization systems using two types of thermoplastic masks for cranial targets and reusable, evacuated body cradles were evaluated for positional accuracy and suitability for use with port films for patient QA. Results: The program for equipment validation is divided into daily, weekly, and monthly programs that add only small amounts of time to routine QA programs. For the first 15 patients treated with this modality, the maximum dose measured on the film was within 5% of that predicted by the planning computer. The prescription isodose line was measured in the anteroposterior and lateral dimensions and the average discrepancy between measured and predicted was less than 2 mm. For an isodose line between 50% and 70% of the prescribed dose, the agreement was better than 3 mm. Success with the volume QA program was followed by a point check QA program that reduced the time required for individual patient QA from days to hours. Phantom measurements compared with computer predictions for 588 data points resulted in only 8% being outside a ⴞ5% criterion. These cases were identified and allow a further reduction in the frequency of tests. Thermoplastic mask materials have adequate restraint characteristics for use with the system and port films on 21 patients resulted in one standard deviation ⴝ 1.3 mm. Body cradles are less accurate and require more frequent port films. A QA system that reduces the frequency of port films was developed. Conclusion: The evolution of sequential tomotherapy in our department has been from a maximum of 3 cranial patients per day with invasive fixation to 60 patients per day for treatment of cranial, head-and-neck, and prostate tumors using different immobilization techniques. With proper preparation and refinement of tools used in commissioning and validation, sequential tomotherapy IMRT can become a routine clinical treatment modality. © 2003 Elsevier Inc. Intensity-modulated radiation therapy, Quality assurance.
INTRODUCTION
al radiotherapy system (1). Before the clinical implementation of this technology, we reviewed what was considered to be similar technologies to determine what potential problems might be encountered and how they might be resolved. These technologies are linear accelerator-based SRS and high-dose-rate (HDR) brachytherapy. The former involves small-field dosimetry with precision arc delivery requirements. The latter involves inverse treatment planning and
Sequential tomotherapy (The Peacock System, NOMOS Corp., Sewickley, PA) consists of an inverse planning system and a special multileaf collimator that operates dynamically during an arc therapy delivery. The system initially was introduced as a possible stereotactic radiosurgery tool (SRS), but was implemented as a three-dimensional conformReprint requests to: Shiao Y. Woo, M.D., The Methodist Hospital Annex, Radiotherapy Department, 6565 Fannin St., MS 121B, Houston, TX 77030. Tel: (713) 790-2637; Fax: (713) 7931300; E-mail: [email protected] Presented at the 3rd S. Takahashi Memorial International Work-
shop on 3-Dimensional Conformal Radiotherapy, December 8 –10, 2001, Nagoya, Japan. Received Feb 22, 2002, and in revised form Oct 11, 2002. Accepted for publication Oct 16, 2002. 274
Evolving clinical QA for IMRT
computer-controlled delivery systems. From these we were able to identify and resolve issues regarding possible modifications to our linear accelerator, the treatment room, patient planning, dose validation, and patient-specific quality assurance (QA). This article identifies these areas and solutions that were implemented, allowing us to introduce intensity-modulated radiation therapy (IMRT) into our clinic safely. In addition, further refinements that allow the expansion of the IMRT program into routine clinical use are discussed. METHODS AND MATERIALS The sequential tomotherapy system Sequential tomotherapy using the Peacock System is described in detail (2– 4), but it is important to highlight the unique features of the system. The system delivers IMRT using arc therapy and a beam that has its intensity modulated by a special collimator called the MIMiC (multivane intensity-modulating collimator). The MIMiC consists of 40 vanes, each of which is an 8-cm-thick tungsten block that projects to a nominal 1 ⫻ 1 cm2 field at isocenter. In 1994, SRS employed divergent cones that were symmetric about the central axis of the open radiation field to collimate the accelerator beam. The MIMiC collimator differs in that the central ray of each vane is divergent from the machine isocenter and no vane contains the central axis of the open radiation field. This allows the treatment of targets that do not include the isocenter of the accelerator. The system is a precursor of a tomotherapy system first described by Mackie (5). The treatment is delivered in an arc, and the vanes are dynamically moved to either allow passage of the beam or block the beam (a binary system), thereby creating the intensity-modulated pattern. The vanes are operated pneumatically and normally move from open to closed in approximately 50 ms. There are two rows of 20 vanes each, so an arc treats only a nominal 2-cm length, and, therefore, the system requires precision indexing of the couch in the longitudinal direction to treat anything but small tumors (6). This sequence of table indexing leads to the definition of the process as sequential tomotherapy (7). It is possible to remove mechanical stops and allow the vanes to open a nominal 2 cm in the longitudinal direction, thereby treating a nominal 4-cm length per arc. The MIMiC fits in the accessory tray holder of the linear accelerator and is connected to linear accelerator electronics only via the “door interlock” circuit. An attached computer called the controller box, operated via a touch screen, dictates the MIMiC’s operation. This computer receives the information for vane position as a function of gantry angle via a 3.5-inch floppy disk created by the planning computer and monitors gantry speed, gantry position, and vane position as a function of gantry angle. It does not monitor dose rate or beam flatness. A second on-board computer measures changes in gantry speed, gantry angle, and time for vanes to open or close. If either computer detects an error, the vanes are closed and a signal is sent to the linear
● S. Y. WOO et al.
275
accelerator that triggers the “door open” interlock and stops the operation of the linear accelerator. The MIMiC is designed for easy removal and replacement via locking pins created at installation by the company. Equipment QA Program The main sources of error with the mechanics of the system were evaluated. In addition to the correct indexing of the patient, the other major part of treatment accuracy is the proper alignment of the MIMiC. Simple film tests were developed for validating the MIMiC alignment for the superior-inferior direction, the left-right direction, and for collimator rotation. The test for the cross-plane (relative to the path of electrons in the accelerator waveguide) alignment requires that the gantry be at a lateral position and a film placed at isocenter, perpendicular to the beam. With central vanes open on opposite rows of vanes, a 180° arc is delivered. The test for the in-plane involves a double exposure in which all MIMiC vanes are open, the gantry placed at a lateral position, and the film placed close to the source. After the first exposure, the gantry is rotated to the opposite lateral position and a second exposure made. Because the MIMiC radiation field is rectangular rather than cylindrical as are cones for SRS, another required check is collimator rotational position. This test is set up as for the cross-plane test, except that all vanes are open. An exposure is made at the initial gantry angle and a second is made at the opposite lateral gantry angle. Precision indexing of the table is accomplished with a device known as the Crane (NOMOS Corp.) that is attached to the couch top hand rails. The Crane has a digital micrometer readout that is based on a laser illuminating a reflective strip. The readout can switch between millimeters and inches, although all treatment plans dictate indexing in millimeters. The Crane comes in two models: the Crane I, which is floor mounted, and the Crane II, which mounts to the side of the couch. Pretreatment patient quality assurance The major component of the first QA system was the use of cut film in a cubic phantom. The phantom, made of black polystyrene, is 17 cm ⫻ 17 cm ⫻ 18 cm high, and contains 19 plates that are 14.0 cm ⫻ 12.8 cm ⫻ 0.6 cm thick. Films normally are oriented in the transverse plane of the patient. For the transverse exposures, the lateral thickness of the phantom to the center of the film is 8.5 cm. When using the nominal 1-cm slice width mode, three films are placed in the phantom for each arc and a film is placed at the beginning and another at the end to ensure coverage of the entire treatment. These films are processed with a set of calibration film exposed at a 5-cm depth in phantom to a known dose and digitized on a laser densitometer. The chemical processing and densitometer combination permits a maximum of 3 Gy to be delivered to the film before there is saturation. Because the Hurter and Driffield curve for silver halide films is nonlinear, the pixel intensity values of the film are converted to dose using a fifth-order polynomial and are
276
I. J. Radiation Oncology
● Biology ● Physics
Volume 56, Number 1, 2003
Fig. 1. Patient immobilized on treatment table by the invasive fixation system.
plotted using commercial software (EasyPlot, Spiral Software, Lyme, NH, or Neosys Transform, Research Systems, Inc., Boulder, CO). Both of these software packages can read the unsigned binary output data of the laser densitometer, perform the curve fitting mentioned previously, and then apply the conversion of the pixel value to dose. The initial system incorporated both quantitative and qualitative techniques to verify that the plan generated by the computer was delivered. These were: 1. The exposure of a commercial prepackaged (“Ready Pack”) film (XV-2, 10 ⫻ 12 inch, Eastman Kodak Co., Rochester, NY) in a transverse orientation to check that the dose pattern is correct relative to the isocenter of the linear accelerator. Pin marks corresponding to the axes of the room lasers are made on the film. The distance of the dose pattern on the processed film is compared to the planner axes. 2. The exposure of a sagittal or coronal film in a Ready Pack to the entire series of arcs to check for overlaps or cold spots between arcs and verify that the longitudinal indexing of the couch is correct. 3. Exposure in the phantom of a series of films separated by 0.6 cm to the entire series of arcs. These films are qualitatively compared for shape and intensity with the series of TIFF (i.e., tagged-image file format) images generated by the computer at the same separation. The films are cut to fit the phantom described previously. 4. The films from the phantom are digitized with a laser densitometer and converted to dose. The lateral and anteroposterior dimensions of the prescription isodose line (range 86%–90% of the maximum dose) and another
Fig. 2. (a) Thermoplastic mask with added strapping for rigidity used for noninvasive immobilization. (b) New thermoplastic mask used for noninvasive immobilization.
isodose line, 50% of maximum for the first 11 patients and 70% of maximum for later patients, are compared quantitatively with those generated in the computer. To assist with the comparison, patients were scanned on an axial computed tomography (CT) scanner with a 3-mm slice width and a 3-mm table index. This made each film in the phantom, separated by 6 mm, related to every other axial scan and allowed for comparison between film and planner. A later QA system compared the absolute dose measured in a phantom with the dose predicted by computer planning software that calculated the dose delivered to a phantom by an intensity pattern associated with an actual patient delivery. For use with this software, we designed a plastic phantom to measure point doses with p-type diodes or thermoluminescent dosimeters (TLD) chips. One only needs to identify common points in the patient plan and the phantom image set, and then the software rescales the distance differences and recalculates the tissue-maximum ratios to display the resultant dose distribution in the phantom.
Evolving clinical QA for IMRT
● S. Y. WOO et al.
277
Fig. 3. The results of the film test for the cross-plane (transverse) alignment of the multivane intensity-modulating collimator.
Patient immobilization An invasive fixation system was used initially for all patient immobilization. Two titanium screws with threaded inserts were inserted into the vertex of the patient’s skull and remained in place for the duration of the treatment. For planning CT scans and daily treatment, an aluminum bar was attached to the skull screws and the patient was locked to the couch by coupling the bar to a docking receptacle that had been integrated into the couch. Figure 1 is a patient locked on the treatment couch. A pointer rod supplied by the manufacturer was attached to the top of the docking receptacle during CT to define the origin of the coordinate system for planning and treatment. Two noninvasive immobilization systems for cranial patients using thermoplastic masks were tested. The first, Type A (Med-Tec Inc., Orange City, IA), was mounted to a U-frame and had additional strapping to add rigidity (Fig. 2a); the second, Type B (Orfit Industries, Wijnegem, Belgium), is mounted to a T-frame, and, because it can be prestretched, produces a more uniform restraint and was tested without added strips (Fig. 2b). Before clinical implementation, we made masks of both types for a physicist, and, using pressure-sensitive Dial Indicators (The L.S. Starrett Co., Athol, MA), determined how much movement was possible and how well the system returned the physicist to his or her original position. The Dial Indicators were placed on both zygomatic arches, the frontal bone (center of the forehead), and the mental protuberance of the mandible. The lasers on the CT provide the setup points for treatment, and thin solder wire is shaped as an “X” crossing at the laser line to provide visualization of the setup point on the CT data set. In addition, the scout CT image has a vertical line
Fig. 4. The results of the film tests for collimator rotation (left) and in-plane multivane intensity-modulating collimator alignment (right).
placed at the setup point as well as lines 20-mm superior and inferior to the setup line to aid in visualization of the port film. These can also be used as references for port films that are taken using high-contrast film (EC-L Localization Film, Eastman Kodak Co., Rochester, NY). For treatment of prostate, the patients are immobilized in a special cradle that contains an evacuated, shaped insert. A rectal balloon is inserted to minimize organ motion (8, 9). The cradle contains removable fiducials that allow localization of the symphysis pubis on both the planning CT and portal films for verification of organ position relative to the fiducials. RESULTS Accelerator and MIMiC QA Figure 3 shows the results of the cross-plane film test for
278
I. J. Radiation Oncology
● Biology ● Physics
Volume 56, Number 1, 2003
Fig. 5. The film and associated treatment plan for the qualitative quality assurance test for documentation of the dose distribution relative to the accelerator isocenter.
both incorrect and proper alignment. The diagram in the middle shows the vanes that are open for the test. The test, including processing of the film, takes approximately 5 min. Figure 4 shows both a collimator rotation film that is deliberately misaligned and the in-plane film test results, labeled “In/Out.” Again, each of these is a 5-min test. After a review of approximately 3 years of clinical use of the Crane I, on six occasions the readout was in inches and not millimeters. On one occasion, the reflective strip loosened and the readings would jump between correct and a 5.67-mm error.
Patient immobilization Most of the evaluation of the success of invasive fixation was determined by posttreatment studies. Figure 7 is a 6-month follow-up of a patient showing that the tumor necrosis is centered in the target. Figure 8 shows the scout image from the planning CT with three additional vertical lines to visualize the predicted port film. The arrows have been added to indicate the bone Table 1. Dosimetric evaluation of first 15 patients, all cranial Planning-Film (in mm)*
Pretreatment patient quality assurance Figure 5 is Ready Pack film taken in an axial orientation and the corresponding axial slice from the planning system. The pin holes on the film are connected by a line and the density of the film and the dose of the planning system relative to machine isocenter can be compared visually. Table 1 shows the results of the use of film dosimetry with the film phantom. The prescription isodose line and a second isodose line were measured in the anteroposterior and lateral dimensions for three films or targets and for the comparative axial slice from the planning system. The average difference for the prescription isodose line was 1.1 mm or better except for two cases, and better than 2.0 mm for all cases. Table 1 also lists the ratio of the maximum doses from the film and the planning system. In all cases the ratio is within 5% of unity. Figure 6 is a plot of the quantitative check results used in the point measurement QA system. All points are within the prescription isodose line in a low gradient region of the plan. Eight percent of the 588 data points exceed a ⫾5% criterion.
Prescription isodose
Other isodose†
Patient
Average
Maximum
Average
Maximum
1 2 2a 3 4 5 6 7 8 9 10 11 12 13 14 15
1.7 0.6 0.6 1.8 1.1 1.0 0.8 0.8 1.2 1.3 1.0 1.0 0.8 1.0 0.8 0.8
5.0 2.0 2.0 5.0 2.0 2.0 1.0 2.0 2.0 4.0 2.0 4.0 4.0 2.0 2.0 2.0
5.3 3.0 4.0 2.5 2.0 1.4 2.0 2.0 2.5 1.8 2.5 2.0 2.0 2.0 2.5 1.0
6.0 5.0 11.0 4.0 4.0 3.0 4.0 5.0 5.0 4.0 6.0 4.0 5.0 4.0 4.0 2.0
Dose ratio (%) 3 5 5 1 3 5 2 4 4 3 4 3 4 5 5
* The absolute value of the difference between computed and measured isodose lines. † 50% of the prescribed isodose line for the first 11 patients and 70% of the prescribed isodose line for the remainder.
Evolving clinical QA for IMRT
● S. Y. WOO et al.
279
latter, the average setup variation is ⫺0.04 mm (SD ⫽ 1.86 mm). If the two outlying points, indicated on Fig. 9 by the circle, are excluded, the average setup variation is ⫺0.05 mm (SD ⫽ 1.00 mm). DISCUSSION
Fig. 6. Frequency distribution of dose measured with a diode vs. predicted dose by the computer.
in the orbit and the air in the auditory process that is used for comparison of setup with the portal film (shown in the right-hand image of Fig. 8). Port films were taken for the first two fractions and twice weekly thereafter. A review of the port films for 21 patients produced a mean error in the longitudinal direction of 0.28 mm with a standard deviation (SD) of 1.3 mm. Table 2 lists the restraint properties of the two different mask systems that have been used. The results indicate that a 1-mm error can be introduced even under ideal conditions, but that both masks have the ability to return the patient to the initial position within 0.25 mm. Figure 9 (top) shows the monitoring results of the first full-course prostate patient whose setup variation has a mean error of ⫺0.23 mm with a SD of 2.2 mm, whereas the bottom image in Fig. 9 shows similar results for the tenth patient treated. If all the data points are included in the
Because delivery of this type of IMRT had never been used clinically, there were no established techniques of validation or criteria of acceptability. As with most new technologies, one looks for similar technologies as guides and begins with an extensive QA program. Over time, such programs are modified, if possible, to reduce delays in initiating therapy and to reduce costs associated with the manpower required for QA. This has been our experience with the implementation of sequential tomotherapy. First, there are recognizable similarities among sequential tomotherapy, HDR, and SRS, and we looked carefully at the latter two technologies to assist in the implementation of IMRT. Table 3 lists the common features that must be addressed to implement each of the three technologies. We used identical or modified tests for SRS and HDR for application to IMRT with sequential tomotherapy. The American Association of Physicists in Medicine has published two reports that can serve as excellent guidelines (10, 11). Modifications required to the accelerator, couch, and room are provided by the manufacturer and are not discussed in this work. Accelerator and MIMiC QA Because sequential tomotherapy relies on computer-controlled delivery using an arc, commissioning and validation are arduous and require a consideration to detail that was not necessary for systems that used static delivery methods, such as wedges or tissue compensators. The machine-re-
Fig. 7. Pre- and posttreatment scans of a patient immobilized with invasive fixation.
280
I. J. Radiation Oncology
● Biology ● Physics
Volume 56, Number 1, 2003
Fig. 8. Reference scan from the planning computed tomography scan with vertical lines to simulate the multivane intensity-modulating collimator (MIMiC) field (left) and a port film with all MIMiC vanes open (right).
lated QA procedures devised during commissioning of the mechanical system became part of a routine QA program. Table 4 lists the tests employed for a basic QA program for sequential tomotherapy. This program is divided into daily, weekly, and monthly components based on frequency and magnitude of possible error. Estimates of time are included to demonstrate that this is not an excessively demanding program. The film tests discussed for MIMiC alignment require little time and yet provide documental evidence that the system is precisely aligned. In addition, these tests are Table 2. Restraint of thermoplastic masks Maximum motion (mm) Action
Type A
Type B
At rest Chin up Chin down Return at rest At rest Rotate right Rotate left Return to at rest
0.00 0.97 0.64 0.08 0.00 0.76 1.14 0.25
0.00 0.30 0.43 0.20 0.00 0.33 0.64 0.15
extremely sensitive. To physically correct the lateral MIMiC setup error in Fig. 3, the unit was adjusted a distance of 0.125 mm (0.005 inch). This means that visual confirmation of symmetry in any of the tests results in corrections that are not dosimetrically evaluable because of the physical limitations of systems. Likewise, for the inplane test, the distance of the film from the source is not critical, but the larger the distance between the double exposures, the more evident misalignment becomes. Errors in table indexing were detected by simple mechanical checks that require little additional equipment or time. These can be either a mechanical ruler attached to the couch or marks on patients or immobilization devices. Both work and eliminate the fear that index errors have occurred. Pretreatment patient quality assurance The complexity achievable with the planning and delivery systems is demanding. The planning system allows for each of the 20 vanes in a slice to create an intensity pattern ranging from 0 to 100% transmission in steps of 10% for every 5° of arc, thereby resulting in 5.9 ⫻ 1022 possible intensity patterns over a 270° arc. In addition, the initial planning system dose algorithm had approximations that
Evolving clinical QA for IMRT
● S. Y. WOO et al.
281
Fig. 9. Results of daily review of port film setups for the first prostate patient treated (top) and the tenth prostate patient treated (bottom).
required a large correction factor that was a constant for all treatments (for 10 MV X-rays, we assigned a factor of 0.79). We believed that is was necessary to review the dosimetry of every patient in a volumetric manner. A review of the summation of these patterns seemed the only prudent approach, so we devised a combination of both qualitative and quantitative tests to convince ourselves that the computer plan was being delivered. We selected film (Kodak-XV, Eastman Kodak Co., Rochester, NY) because it had been used successfully for other small-field dosimetry systems,
such as stereotactic radiosurgery (12–15) and intensity modulation (16, 17). Visualization of the radiation pattern in space with the MIMiC is difficult because all 40 vanes are divergent and only some are part of the delivery process. For a qualitative position check, a specially designed film phantom (Fig. 10) was constructed to determine the shape and location of the dose distribution in space relative to the isocenter of the machine. It holds two pieces of film that are separated by 1 cm and has scribe lines to facilitate alignment to room lasers
282
I. J. Radiation Oncology
● Biology ● Physics
Volume 56, Number 1, 2003
Table 3. Similarities of SRS, HDR, and IMRT Process
HDR
SRS
IMRT
Commissioning time, including training Special dosimetry equipment Room modification Precision laser system Air/electric for MiMiC Linac modification Gantry isocenter verified Couch isocenter verified Counter weight for gantry balance Collimator isocenter verified Longitudinal couch motion Treatment couch head adapter Accelerator tuned for large MU/degree Collimator rotation lock Patient immobilization CT couch head adapter Special treatment planning Inverse planning optimization Accurate and secure transfer of “dwell times” from a planning computer to a delivery computer Treatment interruption documentation and restart Documentation of computerized delivery Precision couch indexing Pretreatment quality assurance Physicist present at treatment
4–6 weeks Y Y
6–8 weeks Y Y Y
8–12 weeks Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y
Y Y Y Y Y Y Y Y
Y Y
Y Y Y
Y Y Y Y Y
Y Y
Y Y Y Y Y Y
Abbreviations: HDR ⫽ high-dose rate; SRS ⫽ stereotactic radiosurgery tool; IMRT ⫽ intensity-modulated radiation therapy; Y ⫽ yes; MIMiC ⫽ multivane intensity-modulating collimator; MU ⫽ monitor units; CT ⫽ computed tomography.
and imbedded pins to mark the film. Delivering an arc treatment to the phantom records the output of each set of vanes and one can compare both the location of the dose in space and the general shape of the pattern with that intended in the planning system. For example, it will demonstrate that the dose is delivered to the left anterior quadrant of the space relative to isocenter rather than the right anterior or even right posterior quadrants. The ability of the qualitative film test to detect inaccuracy in dose placement saved two
potential errors involving planning of a patient with the incorrect immobilization system, thereby introducing a 1.5-cm error in the isocenter placement. This problem was corrected by the vendor allowing the user to have only those immobilization systems in use at each institution in an active mode on the planning system.
Table 4. QA program for accelerator and MIMiC Frequency Daily
Weekly Monthly Room MIMiC off MIMiC on
Test
Time
Machine output (dose/MU) Laser and couch top Collimator rotation check (laser) Patient arc with interrupt Verify MU/degree for arc Crane check Compressor Room lasers (marks on wall)
5 min 5 min 2 min
Mechanical laser check Constancy check Energy check Flatness/symmetry check Radiation isocenter alignment (3 film tests)
5 min
Abbreviation: MU ⫽ monitor units.
5 1 5 5 2
min min min min min
Total ⫽ 30 min 15 min Fig. 10. The phantom used to expose film for determination of the dose distribution relative to the room isocenter as defined by the lasers.
Evolving clinical QA for IMRT
Table 5. Physical dimensions of first 15 patients Patient
Anterior
Separation, cm
1 2 2a 3 4 5 6 7 8 9 10 11 12 13 14 15
7.0 7.2 9.0 10.0 10.6 8.0 7.2 4.5 4.0 5.5 4.1 7.1 6.9 11.4 6.6 7.3
13.0 15.3 8.5 13.5 12.0 16.2 15.0 15.0 11.7 15.0 12.5 14.8 15.5 16.1 15.8 13.6
We also selected film for the quantitative tests. We employed to the phantom designed for commissioning for patient QA because arcs reduce the importance of phantom topography; our earliest patients had cranial targets, and the phantom was near the size of a head. The data in Table 5 indicate that the phantom dimensions of 17-cm wide and 8-cm deep are representative of the majority of cases; therefore, dose differences related to anatomy differences were expected to be less than 2%. We compared film isodose patterns, as shown in Fig. 11, with the planner isodose patterns. The prescription isodose line, as a percentage of the maximum dose in the plan, and the second isodose line were measured in the anteroposterior and lateral dimensions for a minimum of three films per target. Initially, the second isodose line was selected arbitrarily as 50% of the prescription isodose line. The results in Table 1 show that, for a variety of patient shapes, the average difference in the measured dose distribution for the prescription isodose line was 1.1 mm or better, except for 2 cases, and better than 2.0 mm for all cases. A given pixel intensity value has an uncertainty of 2% in dose. The difference in dimensions of the isodose lines with 2% difference in dose is about 2 mm. For this reason, a 2-mm displacement is considered within the uncertainty of the film dosimetry system. The major discrepancies occurred for the 50% isodose line in Patients 1 and 2, in whom the lesion was in the neck rather than the brain. In both cases, the dose pattern became large and reached near the edges of the phantom. We then changed the selection of the second isodose line to 70% of the prescription isodose line for all cases. We had set a limit of discrepancy at 4 mm for the prescribed isodose, but that did not occur in validating a plan. Table 1 also lists the ratio of the maximum doses from the film and the planner and in all cases is within 5% of unity. This parameter is used to confirm that the film dosimetry system has no major errors. If the ratio differs from unit by more than 5%, the test is repeated. This occurred once and was caused by the use of an incorrect film batch. We were
● S. Y. WOO et al.
283
satisfied that, for the cranial patients treated, the phantom used was adequate. The ability to use a simple phantom facilitated the process of validation, but the effort was not trivial. To run all the film exposures and perform the quantitative checks took 4 h for a three-arc treatment. The qualitative evaluation took another 4 to 6 h. We allowed 2 days between the approval of a plan and the scheduling of the first treatment to ensure that this documentation was completed and reviewed. Later users of sequential tomotherapy also employed film successfully (18 –20). The valuable experience gained by looking at the volumetric dosimetry gave us the confidence to begin reducing the amount of QA required. The desire to treat other disease sites led us to the concept of forward-planning software to calculate the dose at points in a generic phantom. This would allow the development of QA procedures that are not so labor-intensive that they prohibit the treatment of large numbers of patients, a desirable progression of QA for IMRT. It is equally important that necessary safeguards are not excluded just because they are too cumbersome. For this next generation of QA, we used a plastic phantom to measure point doses using this software. It contains removable plugs that are placed on a 25.4-mm grid. The plugs are milled to hold two TLD chips, or an insert that holds a diode can be used (Fig. 12). One needs only to identity common points in the patient plan and the phantom image set, and the software would scale the distance differences and recalculate the tissue-maximum ratios to display the resultant dose distribution in the phantom. The software was first tested for multiple points per patient in comparison to the film phantom for 5 patients. With agreement within 2% for multiple points, the QA procedure was reduced to measure a single point in a low-gradient region within the prescription isodose line. Review of the data in Fig. 6 determined the 8% of outlying measurement points were associated with plans that had either one table index or at least five table indexes. The former were very small cranial targets and the latter were head-and-neck treatments that also had at least one arc with significant modulation as indicated by an accelerator setting of approximately three monitor units per degree setting for gantry speed. All prostate patients had checks within 5%. The outlying points occur because the dose algorithm did not model head leakage and the correction factor required to model the dose was chosen for three table indexes and modulation with a gantry speed of two monitor units per degree. These experiences have led to our current QA program, which is to perform the qualitative and quantitative checks on two prostate cases per week, chosen at random, and any cases that meet the parameters that lead to exceeding the 5% tolerance. If there are no presentations that should exceed 5%, then two cranial or head-and-neck cases are chosen at random. The monitor units are adjusted for any plan that exceeds the 5% limit.
284
I. J. Radiation Oncology
● Biology ● Physics
Volume 56, Number 1, 2003
Fig. 11. (a) The treatment plan slice to be validated with film dosimetry as used in the initial quantitative quality assurance system. (b) The film from the phantom. (c) The resultant isodose distribution from the film.
Evolving clinical QA for IMRT
● S. Y. WOO et al.
285
Table 6. Properties of thermoplastic masks
Fig. 12. The plastic phantom designed for measurements using either a diode or TLD.
Patient immobilization Invasive fixation of cranial targets afforded us the opportunity to focus on setup error without consideration for target and structure motion. We elected to use conventional fractionation to treat all patients to minimize the effect of a possible major error, and the review of posttreatment scans was an effective way to gain confidence in positioning at a time when there were no high-contrast imaging systems available. It is important that any immobilization system be verified
Property
Type-A
Type-B
Time for shrinkage Adherence to skin/hair Frame mount Cost/mask
24 h Requires release materials U-Shape ⬃$61.00
10 min Nonstick T-shape ⬃$38.00
by the user, regardless of its success elsewhere. This requires the establishment of tests that may be cumbersome and timeconsuming. The introduction of noninvasive fixation to sequential tomotherapy occurred at the time that high-contrast film became available, and its use has proven valuable. With the use of fiducials and bony anatomy, one can estimate positioning errors associated with immobilization techniques. It is important that reproducibility of positioning and knowledge of target and normal tissue locations is most important as dose distributions become more conformal. McGary et al. (21, 22) have discussed the prostate immobilization system and analysis of setup error in detail and it will not be repeated here. It should be noted, however, that the system required monitoring daily to obtain the required precision. In addition to showing that we could improve the setup precision of patients with time and training, this type of information led us to the current policy for using the system as described in the flow chart in Fig. 13. If one can minimize the setup variations early in the treatment by frequent monitoring and adjustment if necessary, setup variation of less than 5 mm is achievable. It is evident that new and possibly better immobilization systems are arriving. We have moved from one type of mask to another and, although either mask has adequate restraint properties, the new material has improved properties, as shown in Table 6, that permit patients to have the mask and planning CT performed within 1 h and a cheaper cost per mask. However, new techniques always require close observation and usually represent a temporary reduction in clinical efficiency. CONCLUSION
Fig. 13. Flow chart for port films for prostate patients.
Validation of IMRT systems provides functional tools that can be used to establish simplified QA programs to treat large numbers of cases per day. We have presented the evolution of such a clinical program for sequential tomotherapy that allows us to treat 60 patients per day prudently, approximately half of the total patient load. Perhaps only some of the tools presented here are applicable to other systems, but certainly the philosophies are universal and the solutions not unique.
REFERENCES 1. Smith AR, Ling CC. Implementation of three-dimensional conformal radiotherapy [Special Issue]. Int J Radiat Oncol Biol Phys 1995;33:979 –1343. 2. Carol MP. Beam modulation conformal radiotherapy. In:
Purdy JA, Emami B, editors. 3D radiation treatment planning and conformal therapy. Madison, WI: Advanced Medical Publishing; 1993. p. 435–445. 3. Carol MP, Woo SY, Butler EB, et al. Intensity modulated
286
4.
5.
6.
7.
8.
9.
10. 11. 12.
I. J. Radiation Oncology
● Biology ● Physics
radiation therapy treatment. In: Tobias JS, Thomas PRM, editors. Current radiation oncology, Vol. 3. London: Arnold; 1998. p. 376–395. Grant WH, III. Experience with intensity modulated beam delivery. In: Mackie TR, Palta JR, editors. Teletherapy: present and future. Madison, WI: Advanced Medical Publishing; 1996. p. 793–804. Mackie TR, Holmes TW, Swedloff S, et al. Tomotherapy: A new concept for the delivery of conformal radiotherapy using dynamic compensation. Med Phys 1993;20:1709–1719. Carol MP, Grant WH, Bleier AR, et al. The field-matching problem as it applies to the Peacock 3-D conformal system for intensity modulation. Int J Radiat Oncol Biol Phys 1996;34: 183–187. Intensity Modulated Radiation Therapy Collaborative Working Group. Intensity-modulated radiotherapy: Current status and issues of interest. Int J Radiat Oncol Biol Phys 2001;51: 880–914. D’Amico AV, Manola J, Loffredo M, et al. A practical method to achieve prostate gland immobilization and target verification for daily treatment. Int J Radiat Oncol Biol Phys 2001;51:1431–1436. Teh BS, Mai WY, Uhl BM, et al. Intensity-modulated radiation therapy (IMRT) for prostate cancer with the use of a rectal balloon for prostate immobilization: Acute toxicity and dosevolume analysis. Int J Radiat Oncol Biol Phys 2001;49:705–712. AAPM Report 41. Remote afterloading technology. Woodbury, NY: American Institute of Physics; 1993. AAPM Report 54. Stereotactic radiosurgery. Woodbury, NY: American Institute of Physics; 1995. Houdek PV, VanBuren JM, Fayos JV. Dosimetry of small
Volume 56, Number 1, 2003
13. 14. 15.
16. 17.
18. 19. 20. 21. 22.
radiation fields for 10-MV x rays. Med Phys 1983;10:333– 336. Rice RK, Hansen JL, Svensson GK, et al. Measurements or dose distributions in small beams of 6MV x-rays. Phys Med Biol 1987;32:1087–1099. Wu A, Lindne G, Maitz AH, et al. Physics of gamma knife approach on convergent beams in stereotactic radiosurgery. Int J Radiat Oncol Biol Phys 1990;18:941–949. Serago CF, Houdek PV, Hartmann GH, et al. Tissue maximum ratios (and other parameters) of small circular 4, 6, 10, 15 and 24 MV x-ray beams for radiosurgery. Phys Med Biol 1992;37:1943–1956. Wang X, Spirou S, LoSasso T, et al. Dosimetric verification of intensity-modulation fields. Med Phys 1996;23:317–327. Bortfeld TR, Boyer AL, Schlegel W, et al. Realization and verification of three-dimensional conformal radiotherapy with modulated fields. Int J Radiat Oncol Biol Phys 1994;30:899– 908. Low DA, Mutic S, Dempsey JF, et al. Quantitative dosimetric verification of an IMRT planning and delivery system. Radiother Oncol 1998;49:305–316. Tsai J, Wazer DE, Ling M, et al. Dosimetric verification of the dynamic intensity-modulated radiotherapy of 92 patients. Int J Radiat Oncol Biol Phys 1998;40:1213–1230. Low DA, Dempsey JF, Venkatesan R, et al. Evaluation of polymer gels and MRI as a 3-D dosimeter for intensitymodulated radiation therapy. Med Phys 1999;26:1542–1551. McGary JE, Grant W III. A clinical evaluation of setup errors for a prostate immobilization system. J Appl Clin Med Phys 2000;1:138–147. McGary JE, Teh BS, Butler EB, et al. Prostate immobilization with a rectal balloon. J Appl Clin Med Phys 2002;3:6–11.
doi:10.1016/S0630-3016(03)00097-X
3D-CRT
THE EVOLUTION OF QUALITY ASSURANCE FOR INTENSITYMODULATED RADIATION THERAPY (IMRT): SEQUENTIAL TOMOTHERAPY SHIAO Y. WOO, M.D., WALTER GRANT III, PH.D., JOHN E. MCGARY, PH.D., BIN S. TEH, M.D., AND E. BRIAN BUTLER, M.D. Baylor College of Medicine, The Methodist Hospital, Houston, TX Purpose: To identify the pertinent issues to be addressed in successfully implementing IMRT using sequential tomotherapy into clinical reality and presenting the maturation of quality assurance (QA) programs for both the delivery system and patient treatments that allow routine clinical use of the system. Methods and Materials: Initially, a cubic phantom containing silver halide film was exposed to the entire treatment before patient treatment. The processed films were digitized with a laser densitometer and the dose distributions were compared with that generated by the planning system. Later, software that calculates the dose delivered to any phantom employing the intensity patterns developed in the inverse planning system for an individual patient was implemented for point checks of dose. A measurement phantom for use with this software was developed and evaluated on a large number of patients. Invasive fixation was used for all cranial patients initially. To use sequential tomotherapy for other sites and larger targets, noninvasive immobilization systems using two types of thermoplastic masks for cranial targets and reusable, evacuated body cradles were evaluated for positional accuracy and suitability for use with port films for patient QA. Results: The program for equipment validation is divided into daily, weekly, and monthly programs that add only small amounts of time to routine QA programs. For the first 15 patients treated with this modality, the maximum dose measured on the film was within 5% of that predicted by the planning computer. The prescription isodose line was measured in the anteroposterior and lateral dimensions and the average discrepancy between measured and predicted was less than 2 mm. For an isodose line between 50% and 70% of the prescribed dose, the agreement was better than 3 mm. Success with the volume QA program was followed by a point check QA program that reduced the time required for individual patient QA from days to hours. Phantom measurements compared with computer predictions for 588 data points resulted in only 8% being outside a ⴞ5% criterion. These cases were identified and allow a further reduction in the frequency of tests. Thermoplastic mask materials have adequate restraint characteristics for use with the system and port films on 21 patients resulted in one standard deviation ⴝ 1.3 mm. Body cradles are less accurate and require more frequent port films. A QA system that reduces the frequency of port films was developed. Conclusion: The evolution of sequential tomotherapy in our department has been from a maximum of 3 cranial patients per day with invasive fixation to 60 patients per day for treatment of cranial, head-and-neck, and prostate tumors using different immobilization techniques. With proper preparation and refinement of tools used in commissioning and validation, sequential tomotherapy IMRT can become a routine clinical treatment modality. © 2003 Elsevier Inc. Intensity-modulated radiation therapy, Quality assurance.
INTRODUCTION
al radiotherapy system (1). Before the clinical implementation of this technology, we reviewed what was considered to be similar technologies to determine what potential problems might be encountered and how they might be resolved. These technologies are linear accelerator-based SRS and high-dose-rate (HDR) brachytherapy. The former involves small-field dosimetry with precision arc delivery requirements. The latter involves inverse treatment planning and
Sequential tomotherapy (The Peacock System, NOMOS Corp., Sewickley, PA) consists of an inverse planning system and a special multileaf collimator that operates dynamically during an arc therapy delivery. The system initially was introduced as a possible stereotactic radiosurgery tool (SRS), but was implemented as a three-dimensional conformReprint requests to: Shiao Y. Woo, M.D., The Methodist Hospital Annex, Radiotherapy Department, 6565 Fannin St., MS 121B, Houston, TX 77030. Tel: (713) 790-2637; Fax: (713) 7931300; E-mail: [email protected] Presented at the 3rd S. Takahashi Memorial International Work-
shop on 3-Dimensional Conformal Radiotherapy, December 8 –10, 2001, Nagoya, Japan. Received Feb 22, 2002, and in revised form Oct 11, 2002. Accepted for publication Oct 16, 2002. 274
Evolving clinical QA for IMRT
computer-controlled delivery systems. From these we were able to identify and resolve issues regarding possible modifications to our linear accelerator, the treatment room, patient planning, dose validation, and patient-specific quality assurance (QA). This article identifies these areas and solutions that were implemented, allowing us to introduce intensity-modulated radiation therapy (IMRT) into our clinic safely. In addition, further refinements that allow the expansion of the IMRT program into routine clinical use are discussed. METHODS AND MATERIALS The sequential tomotherapy system Sequential tomotherapy using the Peacock System is described in detail (2– 4), but it is important to highlight the unique features of the system. The system delivers IMRT using arc therapy and a beam that has its intensity modulated by a special collimator called the MIMiC (multivane intensity-modulating collimator). The MIMiC consists of 40 vanes, each of which is an 8-cm-thick tungsten block that projects to a nominal 1 ⫻ 1 cm2 field at isocenter. In 1994, SRS employed divergent cones that were symmetric about the central axis of the open radiation field to collimate the accelerator beam. The MIMiC collimator differs in that the central ray of each vane is divergent from the machine isocenter and no vane contains the central axis of the open radiation field. This allows the treatment of targets that do not include the isocenter of the accelerator. The system is a precursor of a tomotherapy system first described by Mackie (5). The treatment is delivered in an arc, and the vanes are dynamically moved to either allow passage of the beam or block the beam (a binary system), thereby creating the intensity-modulated pattern. The vanes are operated pneumatically and normally move from open to closed in approximately 50 ms. There are two rows of 20 vanes each, so an arc treats only a nominal 2-cm length, and, therefore, the system requires precision indexing of the couch in the longitudinal direction to treat anything but small tumors (6). This sequence of table indexing leads to the definition of the process as sequential tomotherapy (7). It is possible to remove mechanical stops and allow the vanes to open a nominal 2 cm in the longitudinal direction, thereby treating a nominal 4-cm length per arc. The MIMiC fits in the accessory tray holder of the linear accelerator and is connected to linear accelerator electronics only via the “door interlock” circuit. An attached computer called the controller box, operated via a touch screen, dictates the MIMiC’s operation. This computer receives the information for vane position as a function of gantry angle via a 3.5-inch floppy disk created by the planning computer and monitors gantry speed, gantry position, and vane position as a function of gantry angle. It does not monitor dose rate or beam flatness. A second on-board computer measures changes in gantry speed, gantry angle, and time for vanes to open or close. If either computer detects an error, the vanes are closed and a signal is sent to the linear
● S. Y. WOO et al.
275
accelerator that triggers the “door open” interlock and stops the operation of the linear accelerator. The MIMiC is designed for easy removal and replacement via locking pins created at installation by the company. Equipment QA Program The main sources of error with the mechanics of the system were evaluated. In addition to the correct indexing of the patient, the other major part of treatment accuracy is the proper alignment of the MIMiC. Simple film tests were developed for validating the MIMiC alignment for the superior-inferior direction, the left-right direction, and for collimator rotation. The test for the cross-plane (relative to the path of electrons in the accelerator waveguide) alignment requires that the gantry be at a lateral position and a film placed at isocenter, perpendicular to the beam. With central vanes open on opposite rows of vanes, a 180° arc is delivered. The test for the in-plane involves a double exposure in which all MIMiC vanes are open, the gantry placed at a lateral position, and the film placed close to the source. After the first exposure, the gantry is rotated to the opposite lateral position and a second exposure made. Because the MIMiC radiation field is rectangular rather than cylindrical as are cones for SRS, another required check is collimator rotational position. This test is set up as for the cross-plane test, except that all vanes are open. An exposure is made at the initial gantry angle and a second is made at the opposite lateral gantry angle. Precision indexing of the table is accomplished with a device known as the Crane (NOMOS Corp.) that is attached to the couch top hand rails. The Crane has a digital micrometer readout that is based on a laser illuminating a reflective strip. The readout can switch between millimeters and inches, although all treatment plans dictate indexing in millimeters. The Crane comes in two models: the Crane I, which is floor mounted, and the Crane II, which mounts to the side of the couch. Pretreatment patient quality assurance The major component of the first QA system was the use of cut film in a cubic phantom. The phantom, made of black polystyrene, is 17 cm ⫻ 17 cm ⫻ 18 cm high, and contains 19 plates that are 14.0 cm ⫻ 12.8 cm ⫻ 0.6 cm thick. Films normally are oriented in the transverse plane of the patient. For the transverse exposures, the lateral thickness of the phantom to the center of the film is 8.5 cm. When using the nominal 1-cm slice width mode, three films are placed in the phantom for each arc and a film is placed at the beginning and another at the end to ensure coverage of the entire treatment. These films are processed with a set of calibration film exposed at a 5-cm depth in phantom to a known dose and digitized on a laser densitometer. The chemical processing and densitometer combination permits a maximum of 3 Gy to be delivered to the film before there is saturation. Because the Hurter and Driffield curve for silver halide films is nonlinear, the pixel intensity values of the film are converted to dose using a fifth-order polynomial and are
276
I. J. Radiation Oncology
● Biology ● Physics
Volume 56, Number 1, 2003
Fig. 1. Patient immobilized on treatment table by the invasive fixation system.
plotted using commercial software (EasyPlot, Spiral Software, Lyme, NH, or Neosys Transform, Research Systems, Inc., Boulder, CO). Both of these software packages can read the unsigned binary output data of the laser densitometer, perform the curve fitting mentioned previously, and then apply the conversion of the pixel value to dose. The initial system incorporated both quantitative and qualitative techniques to verify that the plan generated by the computer was delivered. These were: 1. The exposure of a commercial prepackaged (“Ready Pack”) film (XV-2, 10 ⫻ 12 inch, Eastman Kodak Co., Rochester, NY) in a transverse orientation to check that the dose pattern is correct relative to the isocenter of the linear accelerator. Pin marks corresponding to the axes of the room lasers are made on the film. The distance of the dose pattern on the processed film is compared to the planner axes. 2. The exposure of a sagittal or coronal film in a Ready Pack to the entire series of arcs to check for overlaps or cold spots between arcs and verify that the longitudinal indexing of the couch is correct. 3. Exposure in the phantom of a series of films separated by 0.6 cm to the entire series of arcs. These films are qualitatively compared for shape and intensity with the series of TIFF (i.e., tagged-image file format) images generated by the computer at the same separation. The films are cut to fit the phantom described previously. 4. The films from the phantom are digitized with a laser densitometer and converted to dose. The lateral and anteroposterior dimensions of the prescription isodose line (range 86%–90% of the maximum dose) and another
Fig. 2. (a) Thermoplastic mask with added strapping for rigidity used for noninvasive immobilization. (b) New thermoplastic mask used for noninvasive immobilization.
isodose line, 50% of maximum for the first 11 patients and 70% of maximum for later patients, are compared quantitatively with those generated in the computer. To assist with the comparison, patients were scanned on an axial computed tomography (CT) scanner with a 3-mm slice width and a 3-mm table index. This made each film in the phantom, separated by 6 mm, related to every other axial scan and allowed for comparison between film and planner. A later QA system compared the absolute dose measured in a phantom with the dose predicted by computer planning software that calculated the dose delivered to a phantom by an intensity pattern associated with an actual patient delivery. For use with this software, we designed a plastic phantom to measure point doses with p-type diodes or thermoluminescent dosimeters (TLD) chips. One only needs to identify common points in the patient plan and the phantom image set, and then the software rescales the distance differences and recalculates the tissue-maximum ratios to display the resultant dose distribution in the phantom.
Evolving clinical QA for IMRT
● S. Y. WOO et al.
277
Fig. 3. The results of the film test for the cross-plane (transverse) alignment of the multivane intensity-modulating collimator.
Patient immobilization An invasive fixation system was used initially for all patient immobilization. Two titanium screws with threaded inserts were inserted into the vertex of the patient’s skull and remained in place for the duration of the treatment. For planning CT scans and daily treatment, an aluminum bar was attached to the skull screws and the patient was locked to the couch by coupling the bar to a docking receptacle that had been integrated into the couch. Figure 1 is a patient locked on the treatment couch. A pointer rod supplied by the manufacturer was attached to the top of the docking receptacle during CT to define the origin of the coordinate system for planning and treatment. Two noninvasive immobilization systems for cranial patients using thermoplastic masks were tested. The first, Type A (Med-Tec Inc., Orange City, IA), was mounted to a U-frame and had additional strapping to add rigidity (Fig. 2a); the second, Type B (Orfit Industries, Wijnegem, Belgium), is mounted to a T-frame, and, because it can be prestretched, produces a more uniform restraint and was tested without added strips (Fig. 2b). Before clinical implementation, we made masks of both types for a physicist, and, using pressure-sensitive Dial Indicators (The L.S. Starrett Co., Athol, MA), determined how much movement was possible and how well the system returned the physicist to his or her original position. The Dial Indicators were placed on both zygomatic arches, the frontal bone (center of the forehead), and the mental protuberance of the mandible. The lasers on the CT provide the setup points for treatment, and thin solder wire is shaped as an “X” crossing at the laser line to provide visualization of the setup point on the CT data set. In addition, the scout CT image has a vertical line
Fig. 4. The results of the film tests for collimator rotation (left) and in-plane multivane intensity-modulating collimator alignment (right).
placed at the setup point as well as lines 20-mm superior and inferior to the setup line to aid in visualization of the port film. These can also be used as references for port films that are taken using high-contrast film (EC-L Localization Film, Eastman Kodak Co., Rochester, NY). For treatment of prostate, the patients are immobilized in a special cradle that contains an evacuated, shaped insert. A rectal balloon is inserted to minimize organ motion (8, 9). The cradle contains removable fiducials that allow localization of the symphysis pubis on both the planning CT and portal films for verification of organ position relative to the fiducials. RESULTS Accelerator and MIMiC QA Figure 3 shows the results of the cross-plane film test for
278
I. J. Radiation Oncology
● Biology ● Physics
Volume 56, Number 1, 2003
Fig. 5. The film and associated treatment plan for the qualitative quality assurance test for documentation of the dose distribution relative to the accelerator isocenter.
both incorrect and proper alignment. The diagram in the middle shows the vanes that are open for the test. The test, including processing of the film, takes approximately 5 min. Figure 4 shows both a collimator rotation film that is deliberately misaligned and the in-plane film test results, labeled “In/Out.” Again, each of these is a 5-min test. After a review of approximately 3 years of clinical use of the Crane I, on six occasions the readout was in inches and not millimeters. On one occasion, the reflective strip loosened and the readings would jump between correct and a 5.67-mm error.
Patient immobilization Most of the evaluation of the success of invasive fixation was determined by posttreatment studies. Figure 7 is a 6-month follow-up of a patient showing that the tumor necrosis is centered in the target. Figure 8 shows the scout image from the planning CT with three additional vertical lines to visualize the predicted port film. The arrows have been added to indicate the bone Table 1. Dosimetric evaluation of first 15 patients, all cranial Planning-Film (in mm)*
Pretreatment patient quality assurance Figure 5 is Ready Pack film taken in an axial orientation and the corresponding axial slice from the planning system. The pin holes on the film are connected by a line and the density of the film and the dose of the planning system relative to machine isocenter can be compared visually. Table 1 shows the results of the use of film dosimetry with the film phantom. The prescription isodose line and a second isodose line were measured in the anteroposterior and lateral dimensions for three films or targets and for the comparative axial slice from the planning system. The average difference for the prescription isodose line was 1.1 mm or better except for two cases, and better than 2.0 mm for all cases. Table 1 also lists the ratio of the maximum doses from the film and the planning system. In all cases the ratio is within 5% of unity. Figure 6 is a plot of the quantitative check results used in the point measurement QA system. All points are within the prescription isodose line in a low gradient region of the plan. Eight percent of the 588 data points exceed a ⫾5% criterion.
Prescription isodose
Other isodose†
Patient
Average
Maximum
Average
Maximum
1 2 2a 3 4 5 6 7 8 9 10 11 12 13 14 15
1.7 0.6 0.6 1.8 1.1 1.0 0.8 0.8 1.2 1.3 1.0 1.0 0.8 1.0 0.8 0.8
5.0 2.0 2.0 5.0 2.0 2.0 1.0 2.0 2.0 4.0 2.0 4.0 4.0 2.0 2.0 2.0
5.3 3.0 4.0 2.5 2.0 1.4 2.0 2.0 2.5 1.8 2.5 2.0 2.0 2.0 2.5 1.0
6.0 5.0 11.0 4.0 4.0 3.0 4.0 5.0 5.0 4.0 6.0 4.0 5.0 4.0 4.0 2.0
Dose ratio (%) 3 5 5 1 3 5 2 4 4 3 4 3 4 5 5
* The absolute value of the difference between computed and measured isodose lines. † 50% of the prescribed isodose line for the first 11 patients and 70% of the prescribed isodose line for the remainder.
Evolving clinical QA for IMRT
● S. Y. WOO et al.
279
latter, the average setup variation is ⫺0.04 mm (SD ⫽ 1.86 mm). If the two outlying points, indicated on Fig. 9 by the circle, are excluded, the average setup variation is ⫺0.05 mm (SD ⫽ 1.00 mm). DISCUSSION
Fig. 6. Frequency distribution of dose measured with a diode vs. predicted dose by the computer.
in the orbit and the air in the auditory process that is used for comparison of setup with the portal film (shown in the right-hand image of Fig. 8). Port films were taken for the first two fractions and twice weekly thereafter. A review of the port films for 21 patients produced a mean error in the longitudinal direction of 0.28 mm with a standard deviation (SD) of 1.3 mm. Table 2 lists the restraint properties of the two different mask systems that have been used. The results indicate that a 1-mm error can be introduced even under ideal conditions, but that both masks have the ability to return the patient to the initial position within 0.25 mm. Figure 9 (top) shows the monitoring results of the first full-course prostate patient whose setup variation has a mean error of ⫺0.23 mm with a SD of 2.2 mm, whereas the bottom image in Fig. 9 shows similar results for the tenth patient treated. If all the data points are included in the
Because delivery of this type of IMRT had never been used clinically, there were no established techniques of validation or criteria of acceptability. As with most new technologies, one looks for similar technologies as guides and begins with an extensive QA program. Over time, such programs are modified, if possible, to reduce delays in initiating therapy and to reduce costs associated with the manpower required for QA. This has been our experience with the implementation of sequential tomotherapy. First, there are recognizable similarities among sequential tomotherapy, HDR, and SRS, and we looked carefully at the latter two technologies to assist in the implementation of IMRT. Table 3 lists the common features that must be addressed to implement each of the three technologies. We used identical or modified tests for SRS and HDR for application to IMRT with sequential tomotherapy. The American Association of Physicists in Medicine has published two reports that can serve as excellent guidelines (10, 11). Modifications required to the accelerator, couch, and room are provided by the manufacturer and are not discussed in this work. Accelerator and MIMiC QA Because sequential tomotherapy relies on computer-controlled delivery using an arc, commissioning and validation are arduous and require a consideration to detail that was not necessary for systems that used static delivery methods, such as wedges or tissue compensators. The machine-re-
Fig. 7. Pre- and posttreatment scans of a patient immobilized with invasive fixation.
280
I. J. Radiation Oncology
● Biology ● Physics
Volume 56, Number 1, 2003
Fig. 8. Reference scan from the planning computed tomography scan with vertical lines to simulate the multivane intensity-modulating collimator (MIMiC) field (left) and a port film with all MIMiC vanes open (right).
lated QA procedures devised during commissioning of the mechanical system became part of a routine QA program. Table 4 lists the tests employed for a basic QA program for sequential tomotherapy. This program is divided into daily, weekly, and monthly components based on frequency and magnitude of possible error. Estimates of time are included to demonstrate that this is not an excessively demanding program. The film tests discussed for MIMiC alignment require little time and yet provide documental evidence that the system is precisely aligned. In addition, these tests are Table 2. Restraint of thermoplastic masks Maximum motion (mm) Action
Type A
Type B
At rest Chin up Chin down Return at rest At rest Rotate right Rotate left Return to at rest
0.00 0.97 0.64 0.08 0.00 0.76 1.14 0.25
0.00 0.30 0.43 0.20 0.00 0.33 0.64 0.15
extremely sensitive. To physically correct the lateral MIMiC setup error in Fig. 3, the unit was adjusted a distance of 0.125 mm (0.005 inch). This means that visual confirmation of symmetry in any of the tests results in corrections that are not dosimetrically evaluable because of the physical limitations of systems. Likewise, for the inplane test, the distance of the film from the source is not critical, but the larger the distance between the double exposures, the more evident misalignment becomes. Errors in table indexing were detected by simple mechanical checks that require little additional equipment or time. These can be either a mechanical ruler attached to the couch or marks on patients or immobilization devices. Both work and eliminate the fear that index errors have occurred. Pretreatment patient quality assurance The complexity achievable with the planning and delivery systems is demanding. The planning system allows for each of the 20 vanes in a slice to create an intensity pattern ranging from 0 to 100% transmission in steps of 10% for every 5° of arc, thereby resulting in 5.9 ⫻ 1022 possible intensity patterns over a 270° arc. In addition, the initial planning system dose algorithm had approximations that
Evolving clinical QA for IMRT
● S. Y. WOO et al.
281
Fig. 9. Results of daily review of port film setups for the first prostate patient treated (top) and the tenth prostate patient treated (bottom).
required a large correction factor that was a constant for all treatments (for 10 MV X-rays, we assigned a factor of 0.79). We believed that is was necessary to review the dosimetry of every patient in a volumetric manner. A review of the summation of these patterns seemed the only prudent approach, so we devised a combination of both qualitative and quantitative tests to convince ourselves that the computer plan was being delivered. We selected film (Kodak-XV, Eastman Kodak Co., Rochester, NY) because it had been used successfully for other small-field dosimetry systems,
such as stereotactic radiosurgery (12–15) and intensity modulation (16, 17). Visualization of the radiation pattern in space with the MIMiC is difficult because all 40 vanes are divergent and only some are part of the delivery process. For a qualitative position check, a specially designed film phantom (Fig. 10) was constructed to determine the shape and location of the dose distribution in space relative to the isocenter of the machine. It holds two pieces of film that are separated by 1 cm and has scribe lines to facilitate alignment to room lasers
282
I. J. Radiation Oncology
● Biology ● Physics
Volume 56, Number 1, 2003
Table 3. Similarities of SRS, HDR, and IMRT Process
HDR
SRS
IMRT
Commissioning time, including training Special dosimetry equipment Room modification Precision laser system Air/electric for MiMiC Linac modification Gantry isocenter verified Couch isocenter verified Counter weight for gantry balance Collimator isocenter verified Longitudinal couch motion Treatment couch head adapter Accelerator tuned for large MU/degree Collimator rotation lock Patient immobilization CT couch head adapter Special treatment planning Inverse planning optimization Accurate and secure transfer of “dwell times” from a planning computer to a delivery computer Treatment interruption documentation and restart Documentation of computerized delivery Precision couch indexing Pretreatment quality assurance Physicist present at treatment
4–6 weeks Y Y
6–8 weeks Y Y Y
8–12 weeks Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y
Y Y Y Y Y Y Y Y
Y Y
Y Y Y
Y Y Y Y Y
Y Y
Y Y Y Y Y Y
Abbreviations: HDR ⫽ high-dose rate; SRS ⫽ stereotactic radiosurgery tool; IMRT ⫽ intensity-modulated radiation therapy; Y ⫽ yes; MIMiC ⫽ multivane intensity-modulating collimator; MU ⫽ monitor units; CT ⫽ computed tomography.
and imbedded pins to mark the film. Delivering an arc treatment to the phantom records the output of each set of vanes and one can compare both the location of the dose in space and the general shape of the pattern with that intended in the planning system. For example, it will demonstrate that the dose is delivered to the left anterior quadrant of the space relative to isocenter rather than the right anterior or even right posterior quadrants. The ability of the qualitative film test to detect inaccuracy in dose placement saved two
potential errors involving planning of a patient with the incorrect immobilization system, thereby introducing a 1.5-cm error in the isocenter placement. This problem was corrected by the vendor allowing the user to have only those immobilization systems in use at each institution in an active mode on the planning system.
Table 4. QA program for accelerator and MIMiC Frequency Daily
Weekly Monthly Room MIMiC off MIMiC on
Test
Time
Machine output (dose/MU) Laser and couch top Collimator rotation check (laser) Patient arc with interrupt Verify MU/degree for arc Crane check Compressor Room lasers (marks on wall)
5 min 5 min 2 min
Mechanical laser check Constancy check Energy check Flatness/symmetry check Radiation isocenter alignment (3 film tests)
5 min
Abbreviation: MU ⫽ monitor units.
5 1 5 5 2
min min min min min
Total ⫽ 30 min 15 min Fig. 10. The phantom used to expose film for determination of the dose distribution relative to the room isocenter as defined by the lasers.
Evolving clinical QA for IMRT
Table 5. Physical dimensions of first 15 patients Patient
Anterior
Separation, cm
1 2 2a 3 4 5 6 7 8 9 10 11 12 13 14 15
7.0 7.2 9.0 10.0 10.6 8.0 7.2 4.5 4.0 5.5 4.1 7.1 6.9 11.4 6.6 7.3
13.0 15.3 8.5 13.5 12.0 16.2 15.0 15.0 11.7 15.0 12.5 14.8 15.5 16.1 15.8 13.6
We also selected film for the quantitative tests. We employed to the phantom designed for commissioning for patient QA because arcs reduce the importance of phantom topography; our earliest patients had cranial targets, and the phantom was near the size of a head. The data in Table 5 indicate that the phantom dimensions of 17-cm wide and 8-cm deep are representative of the majority of cases; therefore, dose differences related to anatomy differences were expected to be less than 2%. We compared film isodose patterns, as shown in Fig. 11, with the planner isodose patterns. The prescription isodose line, as a percentage of the maximum dose in the plan, and the second isodose line were measured in the anteroposterior and lateral dimensions for a minimum of three films per target. Initially, the second isodose line was selected arbitrarily as 50% of the prescription isodose line. The results in Table 1 show that, for a variety of patient shapes, the average difference in the measured dose distribution for the prescription isodose line was 1.1 mm or better, except for 2 cases, and better than 2.0 mm for all cases. A given pixel intensity value has an uncertainty of 2% in dose. The difference in dimensions of the isodose lines with 2% difference in dose is about 2 mm. For this reason, a 2-mm displacement is considered within the uncertainty of the film dosimetry system. The major discrepancies occurred for the 50% isodose line in Patients 1 and 2, in whom the lesion was in the neck rather than the brain. In both cases, the dose pattern became large and reached near the edges of the phantom. We then changed the selection of the second isodose line to 70% of the prescription isodose line for all cases. We had set a limit of discrepancy at 4 mm for the prescribed isodose, but that did not occur in validating a plan. Table 1 also lists the ratio of the maximum doses from the film and the planner and in all cases is within 5% of unity. This parameter is used to confirm that the film dosimetry system has no major errors. If the ratio differs from unit by more than 5%, the test is repeated. This occurred once and was caused by the use of an incorrect film batch. We were
● S. Y. WOO et al.
283
satisfied that, for the cranial patients treated, the phantom used was adequate. The ability to use a simple phantom facilitated the process of validation, but the effort was not trivial. To run all the film exposures and perform the quantitative checks took 4 h for a three-arc treatment. The qualitative evaluation took another 4 to 6 h. We allowed 2 days between the approval of a plan and the scheduling of the first treatment to ensure that this documentation was completed and reviewed. Later users of sequential tomotherapy also employed film successfully (18 –20). The valuable experience gained by looking at the volumetric dosimetry gave us the confidence to begin reducing the amount of QA required. The desire to treat other disease sites led us to the concept of forward-planning software to calculate the dose at points in a generic phantom. This would allow the development of QA procedures that are not so labor-intensive that they prohibit the treatment of large numbers of patients, a desirable progression of QA for IMRT. It is equally important that necessary safeguards are not excluded just because they are too cumbersome. For this next generation of QA, we used a plastic phantom to measure point doses using this software. It contains removable plugs that are placed on a 25.4-mm grid. The plugs are milled to hold two TLD chips, or an insert that holds a diode can be used (Fig. 12). One needs only to identity common points in the patient plan and the phantom image set, and the software would scale the distance differences and recalculate the tissue-maximum ratios to display the resultant dose distribution in the phantom. The software was first tested for multiple points per patient in comparison to the film phantom for 5 patients. With agreement within 2% for multiple points, the QA procedure was reduced to measure a single point in a low-gradient region within the prescription isodose line. Review of the data in Fig. 6 determined the 8% of outlying measurement points were associated with plans that had either one table index or at least five table indexes. The former were very small cranial targets and the latter were head-and-neck treatments that also had at least one arc with significant modulation as indicated by an accelerator setting of approximately three monitor units per degree setting for gantry speed. All prostate patients had checks within 5%. The outlying points occur because the dose algorithm did not model head leakage and the correction factor required to model the dose was chosen for three table indexes and modulation with a gantry speed of two monitor units per degree. These experiences have led to our current QA program, which is to perform the qualitative and quantitative checks on two prostate cases per week, chosen at random, and any cases that meet the parameters that lead to exceeding the 5% tolerance. If there are no presentations that should exceed 5%, then two cranial or head-and-neck cases are chosen at random. The monitor units are adjusted for any plan that exceeds the 5% limit.
284
I. J. Radiation Oncology
● Biology ● Physics
Volume 56, Number 1, 2003
Fig. 11. (a) The treatment plan slice to be validated with film dosimetry as used in the initial quantitative quality assurance system. (b) The film from the phantom. (c) The resultant isodose distribution from the film.
Evolving clinical QA for IMRT
● S. Y. WOO et al.
285
Table 6. Properties of thermoplastic masks
Fig. 12. The plastic phantom designed for measurements using either a diode or TLD.
Patient immobilization Invasive fixation of cranial targets afforded us the opportunity to focus on setup error without consideration for target and structure motion. We elected to use conventional fractionation to treat all patients to minimize the effect of a possible major error, and the review of posttreatment scans was an effective way to gain confidence in positioning at a time when there were no high-contrast imaging systems available. It is important that any immobilization system be verified
Property
Type-A
Type-B
Time for shrinkage Adherence to skin/hair Frame mount Cost/mask
24 h Requires release materials U-Shape ⬃$61.00
10 min Nonstick T-shape ⬃$38.00
by the user, regardless of its success elsewhere. This requires the establishment of tests that may be cumbersome and timeconsuming. The introduction of noninvasive fixation to sequential tomotherapy occurred at the time that high-contrast film became available, and its use has proven valuable. With the use of fiducials and bony anatomy, one can estimate positioning errors associated with immobilization techniques. It is important that reproducibility of positioning and knowledge of target and normal tissue locations is most important as dose distributions become more conformal. McGary et al. (21, 22) have discussed the prostate immobilization system and analysis of setup error in detail and it will not be repeated here. It should be noted, however, that the system required monitoring daily to obtain the required precision. In addition to showing that we could improve the setup precision of patients with time and training, this type of information led us to the current policy for using the system as described in the flow chart in Fig. 13. If one can minimize the setup variations early in the treatment by frequent monitoring and adjustment if necessary, setup variation of less than 5 mm is achievable. It is evident that new and possibly better immobilization systems are arriving. We have moved from one type of mask to another and, although either mask has adequate restraint properties, the new material has improved properties, as shown in Table 6, that permit patients to have the mask and planning CT performed within 1 h and a cheaper cost per mask. However, new techniques always require close observation and usually represent a temporary reduction in clinical efficiency. CONCLUSION
Fig. 13. Flow chart for port films for prostate patients.
Validation of IMRT systems provides functional tools that can be used to establish simplified QA programs to treat large numbers of cases per day. We have presented the evolution of such a clinical program for sequential tomotherapy that allows us to treat 60 patients per day prudently, approximately half of the total patient load. Perhaps only some of the tools presented here are applicable to other systems, but certainly the philosophies are universal and the solutions not unique.
REFERENCES 1. Smith AR, Ling CC. Implementation of three-dimensional conformal radiotherapy [Special Issue]. Int J Radiat Oncol Biol Phys 1995;33:979 –1343. 2. Carol MP. Beam modulation conformal radiotherapy. In:
Purdy JA, Emami B, editors. 3D radiation treatment planning and conformal therapy. Madison, WI: Advanced Medical Publishing; 1993. p. 435–445. 3. Carol MP, Woo SY, Butler EB, et al. Intensity modulated
286
4.
5.
6.
7.
8.
9.
10. 11. 12.
I. J. Radiation Oncology
● Biology ● Physics
radiation therapy treatment. In: Tobias JS, Thomas PRM, editors. Current radiation oncology, Vol. 3. London: Arnold; 1998. p. 376–395. Grant WH, III. Experience with intensity modulated beam delivery. In: Mackie TR, Palta JR, editors. Teletherapy: present and future. Madison, WI: Advanced Medical Publishing; 1996. p. 793–804. Mackie TR, Holmes TW, Swedloff S, et al. Tomotherapy: A new concept for the delivery of conformal radiotherapy using dynamic compensation. Med Phys 1993;20:1709–1719. Carol MP, Grant WH, Bleier AR, et al. The field-matching problem as it applies to the Peacock 3-D conformal system for intensity modulation. Int J Radiat Oncol Biol Phys 1996;34: 183–187. Intensity Modulated Radiation Therapy Collaborative Working Group. Intensity-modulated radiotherapy: Current status and issues of interest. Int J Radiat Oncol Biol Phys 2001;51: 880–914. D’Amico AV, Manola J, Loffredo M, et al. A practical method to achieve prostate gland immobilization and target verification for daily treatment. Int J Radiat Oncol Biol Phys 2001;51:1431–1436. Teh BS, Mai WY, Uhl BM, et al. Intensity-modulated radiation therapy (IMRT) for prostate cancer with the use of a rectal balloon for prostate immobilization: Acute toxicity and dosevolume analysis. Int J Radiat Oncol Biol Phys 2001;49:705–712. AAPM Report 41. Remote afterloading technology. Woodbury, NY: American Institute of Physics; 1993. AAPM Report 54. Stereotactic radiosurgery. Woodbury, NY: American Institute of Physics; 1995. Houdek PV, VanBuren JM, Fayos JV. Dosimetry of small
Volume 56, Number 1, 2003
13. 14. 15.
16. 17.
18. 19. 20. 21. 22.
radiation fields for 10-MV x rays. Med Phys 1983;10:333– 336. Rice RK, Hansen JL, Svensson GK, et al. Measurements or dose distributions in small beams of 6MV x-rays. Phys Med Biol 1987;32:1087–1099. Wu A, Lindne G, Maitz AH, et al. Physics of gamma knife approach on convergent beams in stereotactic radiosurgery. Int J Radiat Oncol Biol Phys 1990;18:941–949. Serago CF, Houdek PV, Hartmann GH, et al. Tissue maximum ratios (and other parameters) of small circular 4, 6, 10, 15 and 24 MV x-ray beams for radiosurgery. Phys Med Biol 1992;37:1943–1956. Wang X, Spirou S, LoSasso T, et al. Dosimetric verification of intensity-modulation fields. Med Phys 1996;23:317–327. Bortfeld TR, Boyer AL, Schlegel W, et al. Realization and verification of three-dimensional conformal radiotherapy with modulated fields. Int J Radiat Oncol Biol Phys 1994;30:899– 908. Low DA, Mutic S, Dempsey JF, et al. Quantitative dosimetric verification of an IMRT planning and delivery system. Radiother Oncol 1998;49:305–316. Tsai J, Wazer DE, Ling M, et al. Dosimetric verification of the dynamic intensity-modulated radiotherapy of 92 patients. Int J Radiat Oncol Biol Phys 1998;40:1213–1230. Low DA, Dempsey JF, Venkatesan R, et al. Evaluation of polymer gels and MRI as a 3-D dosimeter for intensitymodulated radiation therapy. Med Phys 1999;26:1542–1551. McGary JE, Grant W III. A clinical evaluation of setup errors for a prostate immobilization system. J Appl Clin Med Phys 2000;1:138–147. McGary JE, Teh BS, Butler EB, et al. Prostate immobilization with a rectal balloon. J Appl Clin Med Phys 2002;3:6–11.