Investigating output and energy variations and their relationship to delivery QA results using Statistical Process Control for helical tomotherapy

Physica Medica 38 (2017) 105–110

Contents lists available at ScienceDirect

Physica Medica journal homepage: http://www.physicamedica.com

Technical note

Investigating output and energy variations and their relationship to delivery QA results using Statistical Process Control for helical tomotherapy Diana Binny a,b,⇑, Emilio Mezzenga c, Craig M. Lancaster a, Jamie V. Trapp b, Tanya Kairn b,d, Scott B. Crowe a,b a

Cancer Care Services, Royal Brisbane and Women’s Hospital, Brisbane, Australia Queensland University of Technology, Brisbane, Australia Medical Physics Unit, Istituto Scientifico Romagnolo per lo Studio e la Cura dei Tumori (IRST) IRCCS, Meldola, Italy d Genesis Cancer Care Queensland, Brisbane, Australia b c

a r t i c l e

i n f o

Article history: Received 28 November 2016 Received in Revised form 4 May 2017 Accepted 4 May 2017 Available online 24 May 2017 Keywords: Tomotherapy Quality assurance Output and energy variations Statistical Process Control

a b s t r a c t The aims of this study were to investigate machine beam parameters using the TomoTherapy quality assurance (TQA) tool, establish a correlation to patient delivery quality assurance results and to evaluate the relationship between energy variations detected using different TQA modules. TQA daily measurement results from two treatment machines for periods of up to 4 years were acquired. Analyses of beam quality, helical and static output variations were made. Variations from planned dose were also analysed using Statistical Process Control (SPC) technique and their relationship to output trends were studied. Energy variations appeared to be one of the contributing factors to delivery output dose seen in the analysis. Ion chamber measurements were reliable indicators of energy and output variations and were linear with patient dose verifications. Crown Copyright Ó 2017 Published by Elsevier Ltd on behalf of Associazione Italiana di Fisica Medica. All rights reserved.

1. Introduction In helical tomotherapy, slice by slice doses of radiation are delivered with a translational patient movement through the machine bore and a synchronously rotating gantry [1–6]. TomoTherapy Hi-Art TM treatment system (Accuray, Sunnyvale, USA) is a hybrid between a helical CT scanner and a linear accelerator capable of imaging and treatment using helical fan beams [1]. The width of the treatment fan beam is controlled by a set of tungsten jaws and is divided into beamlets by a 64 leaf binary multileaf collimator (MLC), each with a projection of 6.25 mm at the isocentre [7–9] . The intensity of each of the beamlets is optimised by controlling the amount of time its leaf is open during the 51 angular increments. The ratio of maximum to average of non-zero leaf open times is restricted to a particular value (also known as the modulation factor) between unity and five to enable an optimised treatment delivery [8–10]. TomoTherapy is equipped with TomoTherapy Quality Assurance (TQA) modules that monitor the functional status of the treat⇑ Corresponding author at: Department of Radiation Oncology, Royal Brisbane and Women’s Hospital, Cancer Care Services, Level 4, Joyce Tweddell Building, Butterfield Street, Herston, 4029 Queensland, Australia. E-mail address: [email protected] (D. Binny).

ment unit by analysing data from both its software and hardware components [11,12]. Hardware components utilised in the TQA step wedge modules include a 640 channel Hitachi CT (Hitachi, Ltd., Tokyo, Japan) integrated megavoltage on-board system and an aluminium step-wedge placed on the translational treatment couch [4,6,7,12,13]. Optimised and efficient treatment delivery also depends on machine variables like output, energy, geometric accuracy, gantry period and phase angle, pitch, and many other mechanical and dosimetric factors. Long term output consistency for helical tomotherapy is suggested to be within ±2% for Quality Assurance (QA) purposes by a previous study [8]. This tolerance takes into account Type A and Type B uncertainties in the dose distribution over multiple beam directions during patient treatment [4,7,8,14]. Previous studies [8,9,11,15–17] have shown that TomoTherapy Hi-Art suffer from rotational output variations although clinically significant variations from planned doses have not been demonstrated [7,8,10,16,18,19]. Most of the aforementioned studies were conducted on machines prior to the introduction of the Dose Control System (DCS) (Accuray, Sunnyvale, USA) which is designed to use monitor chamber data in a closed-loop feedback system to stabilise the beam output over time [4]. This system aims to stabilise the beam output to within ±0.5% of the nominal dose rate by adjusting the injector current and the pulse amplitude control [4].

http://dx.doi.org/10.1016/j.ejmp.2017.05.052 1120-1797/Crown Copyright Ó 2017 Published by Elsevier Ltd on behalf of Associazione Italiana di Fisica Medica. All rights reserved.

106

D. Binny et al. / Physica Medica 38 (2017) 105–110

Up to now, there are no published studies on the relationship between (i) energy variations measured using helical and static delivery modes and (ii) Delivery quality assurance (DQA) output and machine TQA output for static and helical delivery. In this current work, we have studied the consistency between the energy variations measured using the step wedge TQA module for a static and helical beam using MVCT and ionisation chamber. The relationship between patient DQA and TQA based output variations were also assessed. Statistical Process Control (SPC) analysis technique [20] was employed to monitor and validate comparisons made this in this study. 2. Materials and methods Four TQA modules; ‘‘Basic Dosimetry”, ‘‘System Monitor”, ‘‘Step Wedge Static” and ‘‘Step Wedge Helical” were run routinely on two treatment units (T1 and T2). The Basic Dosimetry module was used to evaluate the beam output and beam shape based on data collected using MVCT by comparing to reference data. The System Monitor module was used to monitor system performance parameters generated from the Step-Wedge Helical and Step-Wedge Static modules. The Step-Wedge Static Module was used to determine variations for energy, output, jaw collimation, couch speed, laser alignment and detector response consistency using a static beam at zero-degree gantry angle. No modulation was performed in this test and comparisons were made with MVCT reference data. The Step-Wedge Helical module was run to collect the same beam and machine data as the Step-Wedge Static with a modulated beam and rotating gantry. The Step-Wedge Static and Step-Wedge Helical TQA modules involve the delivery of a static and helical beam respectively with output variations corresponding to longitudinal movement of the patient couch. An aluminium step wedge is aligned to a planned position on the treatment couch indicated by lasers at the virtual isocentre and sent to the radiation isocentre situated in the treatment bore [21]. In both TQA modules, the photon beam attenuates through its increasingly thicker steps to provide an overview of the machine dosimetric and geometric characteristics using data collected by the MVCT detector. TQA procedures were performed daily on each of the two treatment units: data collected by the on-board MVCT detector were stored online for analysis of short and long term parameter variations from baselines. 2.1. Machine characteristics Dosimetric parameters of the machine that were assessed in this study included rotational and static beam output and energy. Dosimetric comparison between Step Wedge static and helical TQA module results and ionisation chamber measurements in virtual water (Standard Imaging, Inc., Middleton, WI) using static beam were made. Rotational output difference was also compared against results from patient specific pre-treatment verifications. Metrics for output, energy, and gantry period have been presented as percentage variations from baseline values. A list of beam settings for the various TQA modules has been described in Table 1. 2.2. Output variations Output variations using the helical step-wedge module for the two TomoTherapy units were plotted against time and compared with measurements performed with the 0.053 cc Exradin A1SL ionisation chamber (Standard Imaging, Inc., Middleton, WI) in Virtual Water for a static beam arrangement. The chamber was placed at

Table 1 Description of beam settings of the various TQA modules used to assess machine parameters. Parameters

Procedures Step-Wedge Static module

Jaw Setting (cm) MLC Couch Speed (mm/s) No: of Gantry Rotations Beam on Time (sec) Data Compression factor Purpose *

1 Open and Closed 1.5 0 200 10 Output and Energy

Step-Wedge Helical module 1 Open 1 10 200 10 measurements

VW + IC* 5 Open – 0 60 –

VW+IC = Measurements made using ionisation chamber placed in virtual water.

depth of dose maximum along the central axis of the static beam and irradiated at a constant dose rate of 8 Gy/min for 60 s to acquire output (with correction for temperature and pressure). All pre-treatment patient specific verifications (also referred to as Delivery Quality Assurance, DQA) before and after the installation of DCS were analysed. DQA dose point measurements were acquired using the Exradin ionisation chamber placed at the centre of either a cylindrical Virtual Water phantom or an ArcCheck (Sun Nuclear Corporation, Melbourne, Florida, USA) phantom and compared with the dose calculated by the treatment planning system at the same point within the phantom. The tolerance used for ionisation chamber based measurements was ±2% as per the recommendations standards followed by the department [2,8,22,23]. 2.3. Energy variations Varying depth related linear attenuation coefficients were gathered using the on-board detector after irradiating the aluminium step-wedge to measure beam quality. In the helical TQA module, the attenuation on the flat portions of each step of the aluminium wedge is collected when the gantry is at zero-degrees and the measurement is compared to its reference value [24]. A similar approach is applied to obtain a static energy measurement variation except, the centre between each step edge is derived by iterating along the step until the minimum standard deviation is located. A linear fit is acquired from the negative natural log of the data and compared to its Ref. [24]. Percentage variations from reference values were noted for both static and helical TQA modules. In this study, helical and static step-wedge energy variations were compared to PDD20;10 (ratio of percentage depth-dose IC measurements made at 20 and 10 cm along the static beam central axis in virtual water). This has also been referred to as PDD method in this study. Additionally a least square fitted parameter was used to quantify the significance in the energy variations using the two methods [11]. We found a correlation for the energy variation DETQA , based on a four-year retrospective analysis using the helical step-wedge module as below;

DETQA ¼ m:DEPDD þ k

ð1Þ

where DEPDD is the energy variation acquired using an ionisation chamber and Virtual Water at depths 20 and 10 cm, m is the linear fit for the variations between step-wedge helical and PDD20;10 method and k is the intercept. This equation is subject to Type A and Type B uncertainties of the treatment machine and daily TQA procedures (see Table 2 and Fig. 3). The mean standard deviation (SD) of the variations observed using the TQA energy measurement results have also been represented in Table 2. This equation can thus be used to obtain a quantitative comparison of the energy variations between the two methods, to track fluctuations in energy.

D. Binny et al. / Physica Medica 38 (2017) 105–110

2.4. Statistical process control analysis

3. Results

SPC [20,25–28] was applied to this study to obtain control charts to determine an upper control limit (UCL), lower control limit (LCL) and centre line (CL) for a given set of data and are calculated using equation 2. A control chart is used to monitor and control variations in a process by indicating Type A and Type B uncertainties of the data distribution [29,30]. When data fall outside the control limits, Type B uncertainties or errors due to specific causes can occur, while distribution within the control limits can be attributed due to Type A uncertainties or errors that are due to normal causes in the process of measurement.

3.1. Output variations

pffiffi ; CL ¼ X UCL ¼ X þ 3 dmR n

ð2Þ

2

pffiffi LCL ¼ X  3 dmR n 2

X and mR are the average and average moving range of 30 consecutive data points. The constant d2 depends on a continuous set of n measurements. In this study n is 1 therefore d2 is 1.128 [31]. mR is the absolute value between two consecutive points of the dataset [20,30,31]. Process capability indices (cp and cpk) were also used to measure process performance for a given set of action limits defined by the user, also referred to as upper specified limits (USL) and lower specified limits (LSL). cp and cpk values are calculated from a given dataset using the below equation [20,30].

USL  LSL USL  X X  LSL cp ¼ ; cpk ¼ min ; 6r 3r 3r

107

Prior to DCS installation, the maximum output fluctuations measured on both units by using the Step Wedge module were ±3% for the static beam. Fig. 1 shows rotational and static output variations before and after DCS installation. Patient DQA results also changed with the DCS upgrade (p < 0.001, two-tailed t-test). Mean output variations post DCS installation measured with the ionisation chamber in virtual water (VW IC) and DQA dose point were 0.14% and 0.66% respectively. Mean deviations from the reference procedures in the helical output using the step helical wedge test pre DCS upgrade was reduced from +1.58% ± 0.957 (SD) to  0.66% ± 0.947 (SD). From Fig. 1, it can be seen that higher baseline variations in VW IC and DQA dose points can be attributed to the magnetron replacements occurring in that period. No variations were observed in Helical Step Wedge outputs due to baseline resets post magnetron replacement. This trend was not observed prior to DCS upgrade possibly due to higher output fluctuations. A moving average trend line has been represented for DQA dose points on Fig. 1. TQA baselines were reset after every magnetron replacement and variations were within ±0.2% during this period of post DCS installation.

3.2. Energy variations

! ð3Þ

As discussed earlier, X represents the centre line, also known as the average process value and r is the standard deviation of the data distribution. The value for the capability index; cp indicates the amount of dispersion in a given data distribution and the acceptability index; cpk illustrates how close the data points are to the mean value in addition to quantifying its dispersion. A process capability index value of 1 or greater indicates that the QA process has a large portion of results within user specified limits and subsequently a value less than 1 would indicate that the process is outside the user specified limit for a given period. The Anderson-Darling (AD) [32,33] statistic was used in this study to test each of the study groups for normality.

From Fig. 2, it was concluded that TQA static and helical stepwedge energy variation results improved marginally post DCS upgrade. Mean energy variations reduced from 0.14% ± 0.323 (SD) to 0.01% ± 0.246 (SD) post DCS installation. Correlations between the energy variations in helical and static QA modules and DQA output were observed (see Figs. 1 and 2). Table 2 summarises the values for maximum, minimum and median energy variations with standard deviations for the retrospectively assessed period. Fig. 3 represents the energy variation relationship between PDD20;10 and TQA method. The linear fit, m, as defined before, depends on several parameters like output, measurement chamber sensitivity/uncertainty, geometric uncertainties, etc. Therefore, it is imperative to perform this test while minimising all known uncertainties. The PDD based energy check

Fig. 1. Output variations represented for rotational (Helical step-wedge), static (VW IC) and patient specific QA (DQA dose point) beam and magnetron replacement dates pre and post DCS installation.

108

D. Binny et al. / Physica Medica 38 (2017) 105–110

Fig. 2. Energy variations represented for rotational (Helical Step-Wedge) and static (Static Step-Wedge and VW IC) beams pre and post DCS installation.

Table 2 Energy variation (DE) values (%) with their corresponding standard deviation (SD) for TQA static and rotational beams and static IC in VW beam. Energy check procedures

DE(max)

DE(min)

DE(median)

SD

Step-wedge Helical module Step-wedge Static module VW + IC

1.03 1.07 0.80

1.39 0.58 0.76

0.07 0.02 0.04

0.29 0.36 0.30

Fig. 3. Energy variation relationship between helical step-wedge and PDD based method.

agreed with the TQA helical method to within a mean variation of 1.03% ± 0.290 (SD). 4. Statistical process control analysis 4.1. DQA output analysis In this section, we evaluated the DQA measurement process before and after DCS upgrade (See Table 3 and Supplementary material: Fig. 4). Post DCS upgrade an improvement was seen in overall process control. Capability and acceptability indices post upgrade were greater than one and this indicates that the process is operating within specification of ±3%. However, it was noted that the mean variation did not improve post DCS installation.

4.2. TQA output and energy variation analysis TQA output variations for a period in which a magnetron change took place was evaluated (See Fig. 5 and Table 3). SPC analysis reported a Type B uncertainty in the QA process which assisted in investigating this machine behaviour. Post analysis it was found to be a malfunctioning magnetron which was replaced by the engineer and the process was in control (See Supplementary Material: Fig. 6). USL and LSL for IC based measurements were usually set at a higher action limit to MVCT due to larger variability in the measurement process. The process was observed to be control for the helical step-wedge TQA procedure within ±0.5%. As discussed earlier the relationship between energy variations measured using the MVCT detector and IC were compared and

109

D. Binny et al. / Physica Medica 38 (2017) 105–110

Fig. 5. SPC analysis for static VW output measurement variation during the period before and after magnetron replacement. Red circle indicates a higher output during the period which the magnetron was replaced. Post magnetron replacement machine outputs reverted to normal behaviour.

Table 3 SPC analysis for output and energy using IC in VW and MVCT detector for static and helical beams. SPC Parameters

SPC Analysis Output

UCL (%) LCL (%) CL (%)

r USL/LSL cp cpk AD** * **

Energy

DQA (pre DCS)

DQA*

VW + IC*

Helical Step-wedge module*

Static VW*

Helical Step-wedge module*

3.617 0.885 1.366 0.717 ±3% 1.394 0.759 Normal

2.848 0.199 1.325 0.673

0.299 1.420 0.560 0.912 ±2% 0.800 0.576 Not-Normal

0.073 0.005 0.039 0.013 ±0.5% 12.821 11.821 Normal

0.435 0.438 0.001 0.322 ±1% 1.783 1.781 Not-Normal

0.166 0.358 0.096 0.102

1.485 1.152 Normal

3.268 2.954 Normal

Measurement variations post DCS upgrade. Anderson-Darling test for normal data distribution.

found to be linear. As shown in Table 3, this relationship is only valid if the process is within specification limits of ±1%. Higher capability indices in the helical step-wedge testing process indicates that process would be capable at even a narrower specification for both output and energy quality assurance tests.

5. Discussion This study has demonstrated that helical and static TQA output measurements agreed within ±2% when compared to static beam in VW measurements. There were improvements seen in patient DQA dose points post DCS upgrade; this result is consistent with an earlier study performed with a smaller subset of patient verification plans during the initial stages of DCS installation [4]. However other studies [15,16,19,34] state that the effect of overlapping beamlets, size of the measurement chamber/detector, plan complexity, couch offsets etc. using a rotating gantry affect the measurement point more than an output fluctuation within ±3% [35]. The moving average trend line represented on Fig. 1 substantiates this argument even further. From Fig. 1 it was seen that the DQA moving average trend line followed VW IC variations suggesting that changed response in IC or machine behaviour may have contributed to the DQA measurement

capability. We also observed a correlation between energy variations and DQA output dose suggesting that minor fluctuations in energy can affect deliverability. The capability and acceptability indices for the helical step-wedge output tests (Table 3) indicate that the user specified limits may have been too large. However, it should be noted that step-wedge modules (static or helical) do not test for outputs at the isocentre but report variations seen in monitor chamber readings for those beam settings. Therefore, it is important to have an independent output measurement and quality control system using the IC at isocentre to correlate with these results on a regular basis. SPC analysis of the QA process can be helpful in identifying and reducing special causes of variation. TQA procedures discussed in this study have also shown capability of detecting energy variations to within ±0.5% when compared to the PDD method employed in the department using a static beam in VW. Equation 1 quantitatively describes the relationship between energy variations for the TQA and PDD20;10 method based on a four-year retrospective data on the two treatment units. Based on SPC analysis this equation was considered valid for energy variations seen within user specification limit of ±1% and were not subject to any higher alterations and therefore must be only seen as a guide while making independent IC measurements to assess beam quality. SPC analysis on machine behaviour has proven to be beneficial

110

D. Binny et al. / Physica Medica 38 (2017) 105–110

in tracking Type B uncertainties to stabilise a process to function within its specified control limits. Further studies would need to be carried out on the IEC X, Y and Z offset of the step helical wedge module to assess the capability of the MVCT detector to detect and quantify offsets in any given direction using SPC. 6. Conclusions The TQA tool in TomoTherapy is an efficient system that provides a reliable overview of the machine dosimetric status. However, it cannot fully replace independent dose measurements and can be seen as a tool to reduce its frequency. Energy variations measured using two independent detectors; IC and MVCT measurements agreed within ±1%. Output variations detected using the two detectors agreed within ±2%. SPC analysis of machine QA is recommended not as a replacement but for data monitoring and to enhance quality improvement in treatment machines. Acknowledgements The authors would like to thank Jong Gi Lee from Alpha XRT and Steven Sylvander from RBWH Medical Physics for their indispensable input during this study. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ejmp.2017.05. 052. References [1] Mackie TR, Balog J, Ruchala K, Shepard D, Aldridge S, Fitchard E, et al. Tomotherapy. Semin Radiat Oncol: Elsevier 1999:108–17. [2] Klein EE, Hanley J, Bayouth J, Yin F-F, Simon W, Dresser S, et al. Task Group 142 report: Quality assurance of medical acceleratorsa). Med Phys 2009;36:4197–212. [3] Meeks SL, Harmon Jr JF, Langen KM, Willoughby TR, Wagner TH, Kupelian PA. Performance characterization of megavoltage computed tomography imaging on a helical tomotherapy unit. Med Phys 2005;32:2673–81. [4] Moutrie ZR, Lancaster CM, Yu L. First experiences in using a dose control system on a TomoTherapy HiArt II. J Appl Clin Med Phys 2015;16. [5] Yang JN, Mackie TR, Reckwerdt P, Deasy JO, Thomadsen BR. An investigation of tomotherapy beam delivery. Med Phys 1997;24:425–36. [6] Balog J, Mackie TR, Pearson D, Hui S, Paliwal B, Jeraj R. Benchmarking beam alignment for a clinical helical tomotherapy device. Med Phys 2003;30:1118–27. [7] Fenwick J, Tome W, Jaradat H, Hui S, James J, Balog J, et al. Quality assurance of a helical tomotherapy machine. Phys Med Biol 2004;49:2933. [8] Flynn R, Kissick M, Mehta M, Olivera G, Jeraj R, Mackie T. The impact of linac output variations on dose distributions in helical tomotherapy. Phys Med Biol 2007;53:417. [9] Kissick MW, Fenwick J, James JA, Jeraj R, Kapatoes JM, Keller H, et al. The helical tomotherapy thread effect. Med Phys 2005;32:1414–23. [10] Binny D, Lancaster CM, Harris S, Sylvander SR. Effects of changing modulation and pitch parameters on tomotherapy delivery quality assurance plans. J Appl Clin Med Phys 2015;16.

[11] Choi HH, Ho JP, Yang B, Cheung KY, Yu SK. Technical note: Correlation between TQA data trends and TomoHD functional status. J Appl Clin Med Phys 2014;15. [12] Coevoet M, Denis J, Cravens B, Krumbach T, Figueredo J, Olivera G, et al. Tomotherapy quality assurance (TQA): a fast and comprehensive software tool. World Congress on Medical Physics and Biomedical Engineering, September 7–12, 2009, Munich, Germany: Springer; 2009. p. 678. [13] Balog J, Olivera G, Kapatoes J. Clinical helical tomotherapy commissioning dosimetry. Med Phys 2003;30:3097–106. [14] Kirkup L, Frenkel RB. An introduction to uncertainty in measurement: using the GUM (guide to the expression of uncertainty in measurement). Cambridge University Press; 2006. [15] Staton RJ, Langen KM, Kupelian PA, Meeks SL. Dosimetric effects of rotational output variation and x-ray target degradation on helical tomotherapy plans. Med Phys 2009;36:2881–8. [16] Takahashi Y, Hui SK. Impact of very long time output variation in the treatment of total marrow irradiation with helical tomotherapy. Radiat Oncol 2013;8:1. [17] Broggi S, Cattaneo GM, Molinelli S, Maggiulli E, Del Vecchio A, Longobardi B, et al. Results of a two-year quality control program for a helical tomotherapy unit. Radiother Oncol 2008;86:231–41. [18] Althof V, van Haaren P, Westendorp R, Nuver T, Kramer D, Ikink M, et al. A quality assurance tool for helical tomotherapy using a step-wedge phantom and the on-board MVCT detector. J Appl Clin Med Phys 2012;13. [19] Hui SK, Kapatoes J, Fowler J, Henderson D, Olivera G, Manon RR, et al. Feasibility study of helical tomotherapy for total body or total marrow irradiation. Med Phys 2005;32:3214–24. [20] Montgomery DC. Statistical Quality Control. New York: Wiley; 2009. [21] Boswell S, Tomé W, Jeraj R, Jaradat H, Mackie TR. Automatic registration of megavoltage to kilovoltage CT images in helical tomotherapy: an evaluation of the setup verification process for the special case of a rigid head phantom. Med Phys 2006;33:4395–404. [22] Langen KM, Papanikolaou N, Balog J, Crilly R, Followill D, Goddu SM, et al. QA for helical tomotherapy: Report of the AAPM Task Group 148a). Med Phys 2010;37:4817–53. [23] Millar M, Cramb J, Das R. ACPSEM Supplement. ACPSEM Position Paper: Recommendations for the Safe Use of External Beams and Sealed Brachytherapy Sources in Radiation Oncology. 1997. [24] TomoTherapy Quality Assurance TQATM 107594 A ed: 2001–2012 Accuray Inc. p. 300. [25] Gérard K, Grandhaye JP, Marchesi V, Kafrouni H, Husson F, Aletti P. A comprehensive analysis of the IMRT dose delivery process using statistical process control (SPC). Med Phys 2009;36:1275–85. [26] Pawlicki T, Chera B, Ning T, Marks LB. The systematic application of quality measures and process control in clinical radiation oncology. Semin Radiat Oncol: Elsevier 2012:70–6. [27] Pawlicki T, Dunscombe PB, Mundt AJ, Scalliet P. Quality and Safety in Radiotherapy. Taylor & Francis; 2010. [28] Pawlicki T, Whitaker M, Boyer AL. Statistical process control for radiotherapy quality assurance. Med Phys 2005;32:2777–86. [29] Breen SL, Moseley DJ, Zhang B, Sharpe MB. Statistical process control for IMRT dosimetric verification. Med Phys 2008;35:4417–25. [30] Sanghangthum T, Suriyapee S, Srisatit S, Pawlicki T. Retrospective analysis of linear accelerator output constancy checks using process control techniques. J Appl Clin Med Phys 2013;14:147–60. [31] Wheeler DJ, Chambers DS. Understanding Statistical Process Control. SPC press; 1992. [32] Nelson LS. The Anderson-Darling test for normality. J Qual Technol 1998;30:298. [33] Pettitt A. Testing the normality of several independent samples using the Anderson-Darling statistic. Appl Stat 1977;156–161. [34] Binny D, Lancaster CM, Kairn T, Trapp JV, Crowe SB. Monitoring Daily QA3 constancy for routine quality assurance on linear accelerators. Phys Med 2016;32:1479–89. [35] Yu L, Poole C, Lancaster C, Sylvander S. Towards online patient imaging during helical radiotherapy. Australas Phys Eng Sci Med 2015;38:119–28.