Long‐term dosimetric stability of multiple TomoTherapy delivery systems

Abstract The dosimetric stability of six TomoTherapy units was analyzed to investigate changes in performance over time and with system upgrades. Energy and output were tracked using monitor chamber signal, onboard megavoltage computed tomography (MVCT) detector profile, and external ion chamber measurements. The systems (and monitoring periods) include three Hi‐Art (67, 61, and 65 mos.), two TomoHDA (31 and 26 mos.), and one Radixact unit (11 mos.), representing approximately 10 years of clinical use. The four newest systems use the Dose Control Stability (DCS) system and Fixed Target Linear Accelerator (linac) (FTL). The output stability is reported as deviation from reference monitor chamber signal for all systems and/or from an external chamber signal. The energy stability was monitored using relative (center versus off‐axis) MVCT detector signal (beam profile) and/or the ratio of chamber measurements at 2 depths. The clinical TomoHDA data were used to benchmark the Radixact stability, which has the same FTL but runs at a higher dose rate. The output based on monitor chamber data of all systems is very stable. The standard deviation of daily output on the non‐DCS systems was 0.94–1.52%. As expected, the DCS systems had improved standard deviation: 0.004–0.06%. The beam energy was also very stable for all units. The standard deviation in profile flatness was 0.23–0.62% for rotating target systems and 0.04–0.09% for FTL. Ion chamber output and PDD ratios supported these results. The output stability on the Radixact system during extended treatment delivery (20, 30, and 40 min) was comparable to a clinical TomoHDA system. For each system, results are consistent between different measurement tools and techniques, proving not only the dosimetric stability, but also these quality parameters can be confirmed with various metrics. The replacement history over extended time periods of the major dosimetric components of the different delivery systems (target, linac, and magnetron) is also reported.


| INTRODUCTION
TomoTherapy is a helical radiation therapy delivery system developed at the University of Wisconsin 1 . There have been three major system iterations incorporating several hardware and software changes since its clinical inception in 2002. The initial delivery system was called Hi-Art. The second generation system, referred to as TomoHDA, was capable of delivering both helical (H) and static gantry "direct" (D) treatments, as well as advanced (A) dynamic jaw planning capabilities. Radixact is the newest generation of TomoTherapy delivery systems and represents a redesign of both the gantry and the treatment planning system. The principle dosimetric change on the Radixact is its capability to treat at higher dose rate of  The output stability is reported as deviation from reference monitor chamber signal for all systems, and from external ion chamber measurements for four systems. The energy stability was monitored using the relative (center versus off-axis) MVCT detector profiles and/or the ratio of chamber measurements at two depths (PDD ratio). Re-baselining of expected values following machine service was only performed after independent validation of a given metric (output or energy) with an ion chamber measurement. Additional ion chamber measurements were used to investigate the output stability of the RF chain over extended treatment deliveries on an TomoHDA and Radixact system.

2.A | Internal monitor chamber and exit detector data
The internal monitor chamber and MVCT detector signals were used to assess output and energy stability, respectively. Of the two monitor chambers, only the primary monitor chamber signal (dose1) was used for this study. For the Hi-Art systems (SN 04, 96, and 103 non-DCS), the average dose1 signals were obtained from field service records of a weekly "J48 RotVar" QA procedure. The J48 Rot-Var procedure is a 200 s nonmodulated rotating delivery with the largest jaw setting (nonclinical, 48 mm nominal jaw width). After the DCS installation on Hi-Art SN 103, the average dose1 signal during a J7 RotVar (clinical beam, 10 mm field width, rotation variation) procedure was tracked because that was the field size used for the weekly QA for this unit. The stability of the output is independent T A B L E 1 Systems studied and monitoring periods. All systems, except the Radixact research unit are clinical treatment units. The Hi-Art SN 103 is listed twice because DCS was added and the data for this machine is considered with and without DCS.  The monitor chamber data and exit detector data were manually processed using MATLAB TM . The monitor chamber output signal is reported as a difference from the average signal over the monitoring period. The extensive monitor chamber data archives included data generated prior to, and during, field service measurements. Therefore, monitor chamber signal analysis was limited to the data that were within AE3% of the average, since larger variations were from the machine being run under nonclinical delivery conditions. In addition to the monitor chamber analysis, the daily output on all machines was verified on all clinical days to be within AE3% of the baseline in accordance with AAPM guidelines 9 .
To evaluate beam energy, the exit detector central channels (80-570) were extracted from each 200 s RotVar. The normalized ratio was then computed using the first dataset as a reference. The average Root Mean Square (RMS) across each channel of the normalized ratio was then computed and reported for each run.
For the TomoHDA and Radixact, the same monitor chamber and lateral profile at the exit detector was acquired using vendor-supplied diagnostics and analysis software, TomoTherapy Quality Assurance (TQA) 10 . The TQA software offers various modules to capture and analyze machine data. The "Daily QA" module is a 5-min rotational procedure consisting of several tests, one of which approximates the J48 RotVar. The monitor chamber data and output ratio (relative to the calibrated baseline value) and RMS of the exit detector channels relative to baseline values are acquired and computed by the TQA software. These data were captured on each clinical day as part of morning QA.

2.C | Service records
Major service upgrades relating to dosimetric stability were tallied for the delivery systems in this study. Magnetron, linac, and target replacements were considered dosimetrically significant. Routine work (e.g., addition of dielectric gas), MLC system work, and mechanical repairs were not recorded. Complete service records were available for the TomoHDA and Radixact systems. The service history for the Hi-Art systems was assembled from paper records in F I G . 1. Exit Detector signals from 175 J48 RotVar procedures on SN 04. The RMS difference in the profile with respect to baseline is used a surrogate of beam energy. The characteristic shape of the flattening filter-free beam is obscured in the center detector data due to the deliberate offset of the CT detector focal point from the beam source point leading to a lack of buildup in the center channels.
the clinics and service records provided by Accuray TM , but unfortunately did not include the complete duration of this output and energy study. The study periods are summarized with the dosimetric study periods in Table 1 and with the service data in Table 5.

| RESULTS
As expected, the stability of the daily output has significantly improved with DCS. The standard deviation (r) of the output for non-DCS systems was less than 2%, and for DCS system was less than 1%. The percent of readings within 2r is greater than 90% for all units. This result is supported by both ion chamber measurements and monitor chamber analysis. The results for all six systems are summarized in Table 2. Figures 2 and 3 are histograms of the output for non-DCS and DCS systems, respectively, baselined at 100%. The output was tracked for over 60 months for each non-DCS and over 30 months for the DCS systems. The effect of DCS is apparent in that the daily output ranged from AE3% in the older systems, and the introduction of DCS decreased the daily variation to less than 1%.
The Radixact data were not included in Fig. 3 because of the limited data history available as compared to the other systems.
Energy stability also improved with the FTL and DCS. As shown in Table 3, the average RMS flatness value for rotating target systems ranged from 0.23% to 0.62%, while for fixed target systems the RMS flatness ranged from 0.04% to 0.09%. This difference is also illustrated in Fig. 4, where the normalized exit detector ratio is plotted over all datasets for a rotating target system (SN 04) and a fixed target system (SN 103). As a rotating target thins, the energy decreases and becomes more forward peaked, causing the shoulders of the exit detector ratio to drop 11 . The Radixact system appears to be as stable as the TomoHDA systems when considering RMS flatness. The average ion chamber PDD 20/10 ratio difference yielded similar variation in energy stability over time. However, the improved energy stability with the fixed target observed with the exit detector data is not appreciated with the ion chamber PDD 20/10 ratio due to lack of statistical power. The cone ratio data is the key driver in concluding the energy has improved with the FTL, as illustrated in  Table 4. The CoV of the output at both depths (10 cm and 20 cm) was less than 0.2%, which is to be expected as both systems are using DCS. The CoV of T A B L E 2 Output stability. The standard deviation of output and (% within 2r of mean) for all machines for which the data was available. An A1SL was used as an external ion chamber to measure output in solid water phantom. The monitor chamber signal is from the primary internal chamber (dose1).  (67) the ratios of the 60 s averaged readings at the two depths demonstrate the energy stability during these extended deliveries.
The service records from the machines are summarized in

| DISCUSSION
Since its clinical inception in 2002, TomoTherapy has established itself as a versatile and robust treatment delivery system 12,13 . TomoTherapy was conceived as a specialty machine, combining IMRT and IGRT 1 .
Currently, TomoTherapy is used to treat a wide range of plans, from static gantry 3D radiation therapy to complex multitarget IMRT. This is due in part to the introduction of new features, but also because system reliability and stability have increased over time. As expected, the addition of the DCS has improved daily and intrafraction output stability significantly. The standard deviation in the monitor chamber daily output varies less than 0.5%. This stability allows for the delivery of longer complex treatment plans (e.g., TBI and total skin treatments 14,15 ) without the risk of treatment interruption. The dosimetric The service record history does not cover the complete duration of this output and energy study due to limited availability of records for the Hi-Art systems.
can be delivered faster. This is because the gantry speed is set after the optimal fluence for the given parameter set is determined.

| CONCLUSION
TomoTherapy treatment systems have matured from a novel concept of IMRT delivery to fully integrated delivery systems in clinical practice. This long-term review of the output and energy over multiple generations of the delivery systems has demonstrated high dosimetric stability. The fundamental redesign of the target and introduction of DCS has also led to improved reliability. Proof of equivalent energy and output stability on the Radixact system as monitored with a variety of metrics was presented, and found to be consistent with existing systems. Confidence in the overall dosimetric stability of the Radixact delivery system is important as potential advances of a higher dose rate system can only be realized with a stable system.

ACKNOWLEDG MENTS
The authors thank Sabrina Hoffman, Mariajose Bedoya, and Vimal Desai from the University of Wisconsin Department of Medical Physics for their help with data collection on the alpha Radixact unit.
We also thank Dan Lucas from Accuray TM for his help compiling the service records.

CONFLI CT OF INTEREST
The work on the alpha Radixact research system is supported in part by a University of Wisconsin and Accuray TM research agreement.