Preclinical evaluation of the reliability of a 50 MeV racetrack microtron

Int. _I. Radiation

Oncology

Pergamon

Biol. Phys., Vol. 28. NO. 5, pp. 1219-1227, 1994 Copyright 0 1994 Elsevier Science Ltd Printed in the USA. All rights reserved 0360-3016/94 $6.00 + .OO

0360-3016(93)E0062-B

??Phvsics Original Contribution

PRECLINICAL EVALUATION OF THE RELIABILITY OF A 50 MEV RACETRACK MICROTRON M. E. MASTERSON, L. G. LARSSON,

M.S.,

S. A. LEIBEL, Dept. of Medical

G.

B. TSIRAKIS, M.D.,

S. MAGERAS, B. T.,

PH.D.,

R. FEBO,

Z. FUKS,

M.D.

T. LOSASSO,

R. MOHAN, AND

PH.D.,

PH.D.,

E. JORESKOG,

C. C. LING,

G. J. KUTCHER,

PH.D.,

PH.D.

Physics, Dept. of Radiation Oncology, Memorial Sloan-Kettering Cancer Center, NY; and Radiotherapy Division, Scanditronix AB, Uppsala, Sweden

Purpose: A 50 MeV racetrack microtron has been installed and tested at Memorial Sloan-Kettering Cancer Center. -isdesigned to execute multi-segment conformal therapy automatically under computer control using scanned X ray and electron beams from 10 to 50 MeV. Prior to acceptance of the machine from the manufacturer, formal reliability testing was carried out. Only in this way could confidence be gained in its usefulness for routine 3D computer-controlled conformal therapy. Materials and Methods: To assess reliability, a set of 25 multi-segment test cases, each consisting of 10 to 17 fixed segments, was developed. The field arrangements and modalities for some of the test cases were identical to 3D conformal treatments that were being delivered with multiple static fields on conventional linear accelerators at our institution. Other cases were designed to explore reliability under more complex sets of conditions. These cases were “treated” repeatedly during a total period of 45 hours, over 5 days. During the treatments, ion chambers attached to the head of the machine provided dosimetric data for each field. Data from sensors connected to every set-up parameter (for example, couch positions, gantry angle, collimator leaf positions, etc.) were recorded and verified by an external computer. Results: While preliminary tests indicated an interlock rate of 5%, final reliability test results demonstrated an interlock fault rate of approximately 0.5%. The reproducibility of dosimetric data and geometric setup parameters was within specifications. As an example, leaf position reproducibility in the patient plane was within 0.5 mm for 97% of the setups. The times required to carry out treatments were recorded and compared with the times to carry out identical treatments on a conventional linear accelerator with cerrobend blocks. Areas where additional time savings can be achieved were identified. Conclusion: As an integral part of acceptance testing, the Scanditronix MM50 was rigorously tested for reliability. The machine successfully passed these tests, providing increased confidence in its usefulness for routine 3D conformal therapy. Radiotherapy

machine, Conformal radiotherapy, Racetrack microtron, Computer control.

The Scanditronix’ 50 MeV Medical Microtron (MM50) is a fully computer-controlled racetrack microtron ( 1, 9) equipped with a multileaf collimator (MLC). A new generation of MMSOs, described in part elsewhere (7), can be used in an automated multi-segment mode. Each treatment segment (a specified combination of monitor units, field shape, gantry angle, collimator angle, and couch position) is set up under computer control. Radia-

tion is delivered only after the intended mechanical positions have been reached. Because of the inherent complexity of the accelerator, the computer control systems, and extensive hardware and software interlocks, evaluation of the machine’s reliability prior to acceptance was crucial. The rationale and methodology for these reliability tests are described in this paper, and illustrated with selected examples from the large volume of data obtained. The beam characteristics of this new generation of MM50 are discussed in a separate publication (7).

Presented at a Meeting of the American Society for Therapeutic Radiology and Oncology, San Diego, 10 November 1992. Reprint requests to: M. E. Masterson, Radiation Oncology Center, Holmes Regional Medical Center, 1350 South Hickory St., Melbourne, FL 32901.

Acknowledgment-Supported in part by grant CA54749 from the National Cancer Institute, Department of Health and Human Resources, Bethesda, MD. Accepted for publication 12 November 1993. ’Scanditronix, AB, Uppsala, Sweden.

INTRODUCTION

1219

1220

I. J. Radiation Oncology 0 Biology 0 Physics

Volume 28,

Number

Equipment Room

5, 1994

Control Room

rll

I

Desk

I

I-I

rf waveguide

‘4 Therapist’s Console

Accelerator Room

Fig. 1. Layout of the MM50 installation

Description of the MM50

Collimator

Treatment Room

at Memorial

Sloan-Kettering

Cancer Center.

ment gantry, although dual gantry systems are available. More detailed descriptions of the microtron, its beam transport system, its computer control systems, and its interface to an external computer are published elsewhere

The first of the new generation of computer-controlled MM50s2 with the capability of segmental therapy has recently been installed.2 As illustrated schematically in Figure 1, it occupies space on two floors. The treatment room and the accelerator room are located on the lower floor. The klystron, modulator, and magnet power supplies are in a room above the accelerator, while the accelerator control desk, the main computer and microprocessors are located above the treatment room. The total area occupied is approximately 1200 square feet. The MM50 is capable of producing electron beams from 10 to 50 MeV in 5 MeV increments. For the purposes of this study, the energies and modalities considered were 10, 15, 25, 40, and 50 MeV electrons and 10, 25, and 50 MV X rays. As shown in Figure 2, the microtron consists of a linear accelerator and two 180 degree bending magnets which steer the electrons in multiple passes through the linear accelerator. On each pass through the linear accelerator, the electrons gain an additional 5 MeV. The electrons move in wider orbits as they gain energy, producing patterns that resemble racetracks of different widths. When the desired energy is achieved, deflection magnets steer the electrons to exit the racetrack and enter the transport system where they are magnetically guided into the treatment gantry. Our system has a single treat-

(6, 7). When the electrons reach the treatment head, they either strike one of three targets to produce X rays or they are spatially modulated for electron beam therapy. In either case, the electrons are magnetically scanned in patterns that are specified by the user. In principle, the patterns can be adjusted arbitrarily to produce any intensity distribution. In practice, the X ray beam widths are large: the full width half maximum (FWHM) ranges from 10 cm at 100 cm from the target for 50 MV to 30 cm for 10 MV. The large size allows only for coarse intensity modulation (for example, wedge type distributions) over clinically useful field sizes. However, the electron beam widths range from about 1 to 6 cm at 100 cm from the target, such that finer intensity modulation may be feasible. The entire head is filled with helium to reduce the scattering in air and the associated broadening of the electron beams. To produce uniform X ray beam profiles, the electrons are magnetically scanned on the target to produce one or more concentric circular arrays of beam pulses in the patient plane. This differs somewhat from the patterns in use on the earlier MM50 (4). For electron beams, fields

’ Patient treatment with the MM50 at Memorial Sloan-Kettering Cancer Center (MSKCC) began in July, 1992. All further

discussion indicated)

of the machine and facilities refer (unless otherwise to the unit at MSKCC.

Reliability of a racetrack microtron 0 M. E. MASTERSON el al.

Left Main Magnet

Magnetic Shield 71

Evacuated Chamber

Extraction Magnets

Einear Accelerator (Linac) Fig. 2. Sketch of a 50 MeV racetrack

microtron

of uniform intensity are obtained by scanning the electrons in equispaced hexagonal grid patterns (3, 7). To assure flatness and symmetry, it is necessary for each treatment segment to contain an integral number of complete scans. This is accomplished with a unique dose rate regulation system. When irradiation begins, the MM50 sets the pulse length to achieve the prescribed dose rate. It monitors the dose rate and the remaining number of monitor units throughout the irradiation, and adjusts the pulse length for the final scan so that the treatment segment terminates at the completion of a scan and at the prescribed number of monitor units. As shown in Figure 3, the head of the MM50 contains two motor driven wedges. The two wedges (nominally 15 and 45 degrees) can be moved into the beam under computer-control, either singly, or together to produce a composite wedge (nominally 60 degrees). The MM50 contains a doubly focused multileaf collimator (MLC)3 consisting of 32 pairs of leaves. There is only one pair of block collimators upstream of the MLC. The MLC can be used to shape electron fields as well as X ray fields. The projected leaf width at 100 cm from the target is 1.25 cm. Each leaf can move from 5 cm over the centerline to 15 cm away from the centerline, with a speed of 1 cm/s at 100 cm from the target. The maximum available field size is 30 X 40 cm at isocenter. The average transmission of the collimator is 1% or less for all energies, and the maximum transmission between adjacent leaves is 3% (8).

3 Patented

by Scanditronix,

Uppsala,

Sweden.

1221

n

Znjection Magnets

showing the linear accelerator

Electron Gun

and associated

Right Main Magnet

/ Extraction Valve magnets.

The MM50 can be programmed to deliver a series of up to 50 static segments per treatment with the setup of each segment being automatic and under computer control. The setup conditions for each segment are downloaded from an external computer system known as the Conformal Therapy Verification Delivery and Recording System (CTVDR). Before treatment, the patient is appropriately set up on the couch. A “virtual treatment” is then performed with the radiation therapist engaging a “motion enable switch” in the treatment room. During a virtual treatment, each set of mechanical motions is executed under computer control in the same sequence that will be used during treatment, but without radiation. These motions include rotation of the gantry and collimator, and translation and rotation of the couch. If a collision appears possible, the “motion enable” switch is released, cutting power to all motors. If the “virtual treatment” proceeds without incident, the therapist then exits the treatment room, closes the shielding doors and proceeds with the treatment. During treatment delivery, the mechanical settings, responding to commands from the MM50 main computer, are adjusted to meet the specifications of each of the segments. When all parameters have reached their prescribed settings for a particular segment, a consistency check request is sent from the MM50 main computer to the CTVDR system, transmitting the sensed values of each setting. The CTVDR system compares the sensed values with the prescribed values, and if they are in agreement

I. J. Radiation

1222

Oncology

0 Biology 0 Physics

Volume 28. Number 5, 1994

Target Selector

Ch

Fig. 4. Diagram showing the three ion chambers (NACP parallel plate) in a solid plastic phantom attached to the treatment head during reliability testing. The rigid attachment device maintains a constant SSD of 100 cm, while the depth (d) of the ion chambers can be varied. One chamber is located on the central axis, while the other two chambers are located 8 cm (w) on either side of the central axis.

Collimator LC%XItTS

Fig. 3. Subassemblies ofthe MM50 treatment head. (Only three of the sixty-four multileaves are shown for simplicity.)

to within a tolerance, a “permit beam” signal is returned from CTVDR allowing irradiation to begin. After each segment is irradiated, the number of monitor units delivered and the reason for beam termination are transmitted from the MM50 to the CTVDR system for recording. The process is then repeated for the next segment. A detailed description of the CTVDR system and its interface to the MM50 have been described by Mageras et al. (6).

METHODS

AND MATERIALS

In order to assess the performance of the MM50 prior to clinical use, a stringent test was carried out to determine system reliability under conditions which simulate a full week of clinical treatments. Twenty-five multi-segment treatments were repeated five or more times during a 45 h period over 5 days. During this process, information was gathered on interlock frequency and the time required to complete treatments. Ion chambers in a full scatter phantom were attached to the head of the machine. They provided dosimetric data on the beam characteristics of each of the fields treated. As shown in Figure 4, an ionization chamber4 was located along the central axis, and two others were located 8 cm off axis on opposite sides. The depths of the ion chambers were 10 cm for X ray beams and the depth of maximum ionization (J,& for each electron beam. Lastly, position sensors on the MM50 provided data on the reproducibility of mechanical settings for each segment.

4 Model NACP-02,

Scanditronix,

AB, Uppsala,

Sweden.

The tests were divided in two classes: clinical and physics cases. The clinical cases were taken from actual beam configurations for 3D conformal plans being carried out at our institution on conventional linear accelerators with cerrobend blocks. They included four treatment sites: prostate, nasopharynx, brain, and lung. As shown in Table 1, each clinical case consisted of three to seven treatment segments. Four segments were added to each of the cases to represent two doubly exposed beam films. All segments were treated with the same X ray energy. A computercontrolled virtual treatment was performed before each actual treatment. The same clinical cases were also executed on a conventional linear accelerator using cerrobend blocks, and treatment times were also recorded. Each of the 2 1 physics cases contained 1 l- 17 segments. The beam configurations were designed to exercise the MM50 comprehensively, changing every available parameter over its entire range in a variety of sequences. The 2 1 cases included three X ray (10, 25, 50 MV) and five electron energies (10, 15, 25, 40, 50 MeV). A virtual treatment was not performed before the physics cases. For brevity. a detailed description of the exact conditions for the physics cases is not given here. The variable settings for each segment included modality, energy, scan matrix (describing the position and pulse length for each point in the scan pattern), position of the block collimators, positions of the leaves, wedge position, dose rate, gantry angle, collimator angle, couch angle, couch vertical, lateral, and longitudinal positions. Obviously, it was not possible to test all combinations of parameter settings; however, prudent selection of settings provided useful dosimetric data. Care was taken to study output, flatness, and symmetry for many gantry and collimator positions, and for the small numbers of monitor units characteristic of highly segmented treatments in 3-D conformal therapy. Wedge factors were repeatedly measured for various gan-

Reliability of a racetrack microtron 0 M. E. Table 1. Field arrangements Case number

Type of treatment

101

Prostate

102

Nasopharynx

103

Brain

104

Lung

for clinical cases

Description 2 doubly exposed beam films 2 laterals 2 anterior obliques 2 posterior obliques 2 doubly exposed beam films I posterior 6 wedged posterior obliques 2 doubly exposed beam films 2 wedged laterals 1 wedged apical 2 doubly exposed beam films 1 wedged AP 1 wedged PA 2 wedged anterior obliques

try and collimator

angles to assess the reproducibility of the positions of the motor-driven wedges. Dose rate dependence of the monitor chamber was evaluated over the clinically available range of dose rates for the X ray and electron energies studied. For this assessment, the collection efficiencies of the external ion chambers attached to the head were determined by varying their bias voltages to determine the number of cGy/mu for each modality and dose rate. Changes in the beam energy from one segment to another were also included in order to verify the adequacy of the ramping sequences of all of the magnets when changing beam energy. Improper control of the hysteresis effect could cause a loss of dose rate or more subtle changes in beam characteristics.

MASTERSON

et al.

1223

During this time, nine interlocks occurred. This interlock rate (0.56% of the total number of segments) was within the pre-established criterion of acceptability which was that 99% of the segments must proceed without any interlocks. There was no significant difference between the interlock rate for the clinical cases and the physics cases. There was no dominating type of interlock, and in each case the interlock was either immediately resettable or required less than 10 minutes to resolve.

Treatment times During the MM50 testing, a single physicist carried out the virtual treatment and the actual treatment sessions. On the conventional linear accelerator, two physicists engaged in multi-tasking, as is common for therapists in routine radiotherapy. Because every segment in the clinical cases required a change of blocks and often the wedges, the treatment room door had to be opened and closed between all segments on the linear accelerator. The times to complete a virtual treatment and an actual treatment for the clinical cases on the MM50 are shown in Figure 5 in comparison with the treatment times on a conventional linear accelerator. The virtual treatment time is defined as the period between the downloading of the prescription and the completion of the final virtual treatment segment. The actual treatment time is the period between the initiation and completion of the actual treatment. Figure 5 also includes the virtual and actual treatment times on the MM50 which are anticipated after modifications (discussed below) are implemented. Because these values do not include the time required to set up the patient on the couch nor the time required to set up and expose beam films or real time imagers, the indicated

RESULTS MM50 treatment (Z92)

Interlock frequency Because the MSKCC MM50 is the first of its generation and had not been used clinically at the time of the test, it was recognized that further development and refinement of software and hardware systems may be necessary. A round of “shakedown” reliability testing was carried out using the reliability treatment cases in order to identify problems in implementation. During the “shakedown” round of reliability testing, 1285 segments were treated and 67 interlocks occurred (a 5% interlock rate). These tests served to identify numerous areas requiring hardware or software re-design, repair or adjustment for both the MM50 and the CTVDR system. A detailed discussion of such problems is beyond the scope of this paper. The problems identified during this phase of reliability testing were resolved over the course of a month by the manufacturer and the hospital staff. After resolution of the problems identified during shakedown, final reliability testing was performed. One thousand two hundred thirty seven treatment segments and 356 virtual treatment segments were completed.

MM50 tmt + w tmt (2192) Linac treatment MM50 tmt + vir lmt (anticipated)

-z .-

z, E .-

10

0

prostate 6 segs

nasopharynx 7 segs

bram 3 segs

lung 4 segs

Fig. 5. Histogram showing the treatment and simulation times on the MM50 in comparison with a conventional linear accelerator using cerrobend blocks. The treatment time is the time from initiation of “beam on” for treatment until the end of treatment. The virtual treatment time is clocked from the download of the treatment setup parameters, through the execution of the dry run (from the first to the last treatment segment), and returning to the setup configuration for the first treatment segment. The anticipated times are the projected values after timesaving modifications (discussed in the text) are implemented on the MM50.

1224

1. J. Radiation Oncology 0 Biology 0 Physics Table 2. MM50 treatment

Treatment

type

Prostate Nasopharynx Brain Lung

No. of treatment segments

Volume

time per segment

28, Number

for the four clinical cases Couch rotation (requires “dummy” segment)

No. of wedged treatment segments

6 7 3 4

5, 1994

Time (min) per treatment segment

No No Yes No

0 6 3 4

0.6 1.1 2.0 1.0

With the anticipated modifications described above, the times for treatment and virtual treatment are shown in Figure 5 as the “anticipated” treatment times for the cases studied. As can be seen, the anticipated MM50 times are 46-56% of the times required for treatment on a linear accelerator with cerrobend blocks.

times should not be taken as reflective of total treatment times, but rather as providing a comparison of the times required to carry out similar technical operations on the two types of machines. The average virtual treatment time was 4.1 minutes for the four cases. Treatment times on the MM50 were 3.8, 8.0, 6.0, and 4.0 minutes for the prostate, nasopharynx, brain, and lung cases respectively. The times to execute the same treatments on a conventional linear accelerator were 12.3, 17.6, 10.0, and 9.5 minutes. No account has been taken of other personnel intensive activities such as fabrication of cerrobend blocks, or entering and checking the treatment files on the CTVDR system. Table 2 shows the time-per-treatment segment on the MM50. At the present time, segments involving wedges consume significantly more time than those without wedges.6 Of the clinical cases studied, the prostate case does not use wedges (see Table l), and the average time per segment is 0.6 min. For the lung and nasopharynx cases, where nearly all of the fields have wedges, the time per segment is 1.0-l. 1 min. The brain case includes a couch rotation for the vertex field as well as wedges. For safety reasons, a “dummy” segment (a segment for which there are no monitor units delivered) is added to this treatment so that the couch rotation is executed separately from the gantry rotation (6). The addition of a “dummy” segment, as well as the use of wedges, increases the time per treatment segment to 2.0 min for the brain case. Subsequent to these tests, the manufacturer reevaluated the design of the interlock system as it relates to wedges. After planned modifications are implemented, it is anticipated that the treatment time for segments involving wedges will be comparable to the treatment time for segments without wedges. The initial version delay of therapy software incorporated a lo- 15 s time delay before proceeding from one segment to the next. Reduction of this waiting time by 5 s per segment is anticipated following software upgrade. Also, for those cases which do not require rotation of the collimator or movement of the couch during treatment, a time saving can be achieved during the virtual treatment by starting at the final gantry position and ending at the gantry position for the initial treatment segment.

Fig. 6. Histogram of the reproducibility of MLC setup during reliability testing. The x-axis shows the difference in mm between the read (or sensed) value and the set (or intended) value. Bin size is 0.2 mm.

5 This is attributable to the design of the interlock system. When wedges are in motion, the trigger oscillator is turned off. When the wedge has reached its intended position, the trigger

oscillator is restarted. The beam is not passed through to the gantry until a servo system adjusts the high voltage. This process adds an additional 20-25 seconds each time a wedge moves.

Reproducibility ofgeometric set-up parameters The agreement between the actual physical location of the gantry, collimators, leaves, etc., and their sensed positions was confirmed as part of acceptance testing. During reliability testing, the digital readouts were stored on the CTVDR system for comparison with the intended values for each parameter. Shown in Figure 6 are the data for position of the leaves. As can be seen, leaf position reproducibility at isocenter was within +0.5 mm for 97% of the setups. The data for reproducibility of the other geometric parameters (within a 95% confidence limit) are given in Table 3. Dosimetric data During the 1237 treatment segments, a large volume of data was acquired from the three ion chambers which were attached to the head of the MM50. The clinical cases 30 -

1237

segments

8 0

o-2

-1

Read

0

Value

1

- Set Value

2

(mm)

Reliability of a racetrack microtron 0 M. E. Table 3. Computer-controlled set-up reproducibility (95% confidence level) Set-up reproducibility

Parameter Gantry angle Collimator angle

+0.4 degree kO.4 degree

Couch rotation

20.4 degree

Couch Couch Couch Block

+1.5 +1.5 +-1.5 +I.0

height lateral longitudinal collimator position

mm mm mm mm

provided data on the reproducibility of output for each segment. Both short term (separated by an hour) and long term (over the 5 days of reliability testing) reproducibility were assessed. Figure 7 shows the results of these tests. For each case, the dose measured on the central axis for each segment has been normalized to the dose measured for that segment the first time the case was run. The first four segments simulated two doubly exposed beam films.

1.15 -I

MASTERSON et al.

1225

Each segment delivered only 2 or 3 monitor units (mu). As expected, the reproducibility of the dose per mu for these segments is poorer than for the treatment segments. For the treatment segments, the reproducibility is within k1.5%. An example of the type of data obtained from the physics cases is given in Figure 8 which shows the variation of symmetry with gantry and collimator angles for a large field size using 10 MV X rays. The field size for segments 3,6,9, and 11 of Physics Case 1 was 30 X 30 cm, whereas the gantry and collimator angles varied for each segment. For each of these segments, only 10 mu were delivered. As described previously, two ion chambers were located on opposite sides of the central axis, 8 cm from the central axis. The ratios of the readings of the two chambers for each segment in this case and for each of the five repetitions are plotted in Figure 8. As can be seen 19 of the 20 measurements were within +- 1%. The remaining segment showed a variation of 1.5%. This degree of accuracy for segments having small numbers of monitor units is important for multi-segment conformal therapy.

1.15 ,

I

I Nasopharynx

Prostate

l.lO-

1.05: . .. . . .

._ t .OO : . . .

.

. . .

0.95 -

0.90 -

0.85

treatment

I

beam films ’ I

I

0

2

4

’

I 6

segment

treatment

segments .

I

I

0

10

segments

12

number

segment

a

number

b

Lung

0.90 beam films

beam films 0.85 0

segment C

number

2

4

6

segment

6

10

number

d

Fig. 7. Output constancy for the four clinical cases. The cGy/mu for each segment was determined from the reading of the ion chamber on the central axis at isocenter. This reading has been divided by the cGy/mu measured for that segment during the initial run of the case. The readings of the ion chamber in phantom have been corrected for ambient temperature and pressure at the time of the measurement.

1. J. Radiation

1226

Oncology

0 Biology 0 Physics

1.04

cu

1.03-

30 x 30 cm SSD

field

100 cm

0.980.97 0.96 2

4

6

segment

a

10

12

number

Fig. 8. Symmetry invariance versus gantry and collimator angle for 10 MV X rays on the MMSO. During case 1, segments 3, 6, 9, and 11 corresponded to a 30 X 30 cm field, irradiated at various gantry and collimator angles, for 10 monitor units. The ratio of the readings of the two ion chambers 8 cm off axis in phantom at isocenter have been plotted. The chambers were at 10 cm depth. The case was repeated five times during reliability testing.

DISCUSSION

Volume 28, Number

5, 1994

which the monitor chamber was electrically shorted due to contact with the mirror assembly. Acceptance tests did not include measurements at this collimator angle, and therefore the problem was not identified until the rehability tests were performed. The physics cases provide a good indication of the interlock rate under conditions which are more stressful to the system than the clinical cases. In this way, they might identify problems with future developments in treatment techniques, for example, change of modality and/or energy between treatment segments. Useful information on the stability of dosimetric characteristics has also been obtained. Data have been gathered on the reproducibility of the monitor chamber calibration under different temperature and pressure conditions, Confidence has been gained in the stability of beam profiles, their invariance with mechanical setup conditions. and in the reproducibility of motorized wedges. Lastly, both the clinical and physics cases provide data on the day-to-day reproducibility of computer-controlled mechanical setup parameters such as gantry angle, collimator angle, position of the multileaves, block collimators, and couch.

Development of a methodology

Interlock rate

The performance of rigorous reliability testing on any radiotherapy unit is an important component of pre-acceptance testing. The new generation Scanditronix MM50 incorporates the greatest level of computer control currently available on a medical accelerator. In addition to computer control, the technology of the hardware in the racetrack microtron, the beam transport system, and the gantry are new to the hospital environment. Therefore, in addition to well designed acceptance tests which demonstrate safe operation and assess compliance of individual parameters within specifications, the reliability of the entire system must be demonstrated. Only in this way, can sufficient confidence be gained in the reliable performance of a high technology machine like the MM50 before integrating it into a busy radiotherapy department. A methodology for performing reliability testing on a state-of-the-art radiotherapy machine was developed and implemented on the MM50. As described previously, 4 of the 25 multi-segment treatment cases were derived directly from actual conformal treatments which were underway at our institution for four disease sites. These clinical cases provide a good indication of the interlock rate to be expected using current conformal techniques. They also provide a useful indication of the time required to complete an actual treatment and a virtual treatment. Clinical cases are extremely helpful in uncovering potential problems associated with treatment implementation and with the interface between the MM50 computer and hospital computer systems. Such problems might go unnoticed during more standard acceptance testing. The remaining 2 1 “physics” cases were designed to test the robustness of the system by varying parameters often and over their full range. As an example, on the MM50 these tests identified a small range of collimator angles at

When compared with conventional radiotherapy machines, higher technology machines like the MM50 require hardware and software interlock systems of much greater complexity to operate safely. The potential for a greater number of fault conditions requires a higher degree of dependability of all subsystems in order to avoid excessive unplanned downtime. By mimicking the anticipated treatment load, reliability testing provides a useful indicator of overall system dependability. A radiotherapy unit which is to be used for conformal therapy with many static segments per fraction must have a very low interlock rate. During the “shakedown” tests, an interlock rate of 5% was observed. This round of testing served to identify many areas which needed redesign, repair, or adjustment. After modifications were made, reliability testing was carried out and the measured interlock rate was reduced by nearly an order of magnitude to 0.56%. Because problems were identified and rectified during the shakedown phase of reliability testing, the subsequent implementation of the MM50 for patient treatment proceeded very smoothly. An interlock rate of 0.56% leads to approximately one to two interlocks per week for 10 patients treated with a 6 field plan. Implementation of treatments at this rate for computer controlled plans with a moderate number of segments is acceptable. However, as the number of patients per day and the number of segments per treatment increases, a lower interlock rate is required. While progress has already been made in reducing the interlock rate further during the months since reliability testing was completed, it is important to realize that the need for fast, highly reproducible, and safe treatments with interlock rates approaching zero sets stringent requirements on computer-controlled systems.

Reliability of a racetrack microtron 0 M. E. MASTERSON ef al.

Treutment time Reducing patient treatment time was not the primary factor in the decision to install an MM50 at our institution. The unique capabilities of the MM50 (for example, high energy intensity modulated electron beams (5)) and the associated potential for developing improved treatment techniques outweighed concerns about throughput. However, treatment time affects the number of patient segments that can be treated in a reasonable time, and therefore is of concern. Reliability testing identified areas in which time savings can be effected. As discussed previously, these include a modification to the interlock system related to the use of wedges, reduction of watchdog time-outs during virtual treatment, and performance of a “reverse” virtual treatment when neither the collimator angle nor the couch position change during treatment. As shown in Figure 5, following implementation of these relatively straightforward changes, the time to execute virtual and actual treatment on the MM50 is expected to be approximately half the time required to carry out the same treatment on a linear accelerator with cerrobend blocks. Of course, additional timesaving modifications may be identified and implemented in the future which will further reduce the MM50 treatment times. ReproducihilitJl qf geometric set-up parameters The reproducibility of various geometric setup parameters is shown in Figure 6 and Table 3. The specifications for these parameters were 1 mm for block and leaf positions, 0.5 degree for gantry, collimator, and couch angles, and 2 mm for the couch lateral, longitudinal and vertical positions. As can be seen, these tolerances were met during reliability testing. The demonstrated ability to perform accurate geometric setups under computer control is essential on a unit that is to be dedicated to 3-D multisegment conformal therapy. Dosimetric data As shown in Figures 7 and 8, valuable dosimetric data can be obtained during reliability testing by attaching ionization chambers rigidly to the head of the machine

1221

and collecting data from them after each segment. The data obtained during execution of the clinical cases provided information on the reproducibility of dose delivery under clinical conditions. By carefully designing the physics cases, data were obtained which provided information on the long and short term reproducibility of dosimetric characteristics for a wide variety of conditions. Prudent design of test cases not only generates large volumes of dosimetric data, but also provides increased confidence that the system will perform reliably under conditions which are relevant to multisegment conformal therapy, for example, segments at a wide variety of gantry and collimator angles and with small numbers of monitor units. CONCLUSION A reliability test was performed on the Scanditronix MM50. During the course of a week, data were acquired on interlock rates, treatment times, and the reproducibility of dosimetric data and geometric setup conditions. Perhaps most importantly, systematic software and hardware problems were identified on the MM50. Problems were also identified which related to the external computer which downloads prescriptions, verifies settings, and records patient treatment parameters. These problems were remedied in a timely fashion and, as a result, the subsequent implementation of the machine for patient treatment proceeded very smoothly. The goal of providing multisegment conformal therapy safely, accurately, and in reasonable treatment times can only be achieved with a highly reliable radiotherapy unit. As radiotherapy machines become technologically more complex, assurance is needed that they will function reproducibly and without frequent interruption due to interlocks. By carrying out formal reliability testing prior to machine acceptance, confidence has been gained in the reliable performance of the MM50 racetrack microtron at our institution. It is recommended that a formal assessment of machine reliability be included as an essential component of acceptance testing on other computer-controlled radiotherapy machines.

REFERENCES 1. Brahme, A. Investigations

on the application ofa microtron accelerator for radiation therapy. Doctoral Thesis, University of Stockholm, 1975. 2. Brahme, A.; Kraepelien, T.; Svensson, H. Electron and photon beams from a 50 MeV racetrack Radiol. Oncol. 19:305-319; 1980.

microtron.

Acta

3. Karlsson, M.; Nystrom, H.; Svensson, H. Electron characteristics of the 50 MeV racetrack microtron. Phys. 19:307-3 15; 1992.

beam Med.

4. Karlsson, M.; Nystrom, H.; Svensson, H. Photon beam characteristics on the MM50 racetrack micro&on and a new approach for beam quality determination. Med. Phys. 20: 143-161: 1993. 5. Lo, Y. C.; Kutcher, G. J.; Masterson, M. E.; Leibel, S.; Fuks, Z. Conjecture on the use of high energy electrons for con-

formal radiotherapy (Abstract). Int. J. Radiat. Oncol. Biol. Phys. 24 (Suppl. 1):161; 1992. Mageras, G. S.; Podmaniczky, K. C.; Mohan, R. A model for computer-controlled delivery of 3-D conformal treatments. Med. Phys. 19:945-953; 1992. Masterson, M. E.; Kutcher, G. J.; Bjork, S.; Chui, C. S.; LoSasso, T. J.; Hung, D.; Febo, R.; Enstrom J. Beam characteristics of a 50 MeV racetrack microtron. Submitted to Med. Phys. Masterson, M. E.; LoSasso, T.; Larsson, A.; Mageras, G. S.; Hung, D.: Enstrom, J.; Kutcher, G. J. Design and performance of the multileaf collimator on the scanditronix 50 MeV racetrack microtron. Med. Phys. 19(3):8 16; 1992. Rosander. S.; Sedlacek, M.; Wernholm, 0. The 50 MeV racetrack microtron at the Royal Institute of Technology, Stockholm. Nucl. Instr. Meth. 204:1-20; 1982.