Linear accelerator simulation for photon beam modelling in radiotherapy

The method automates photon beam model updates using physical models and quality assurance measurements to address inefficiencies in current photon beam modeling, enhancing accuracy and reducing manual recalculations, particularly for small beam sizes.

WO2026236828A1PCT designated stage Publication Date: 2026-11-19LINAX TECHNOLOGIES LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/CA2026/050670
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-14
Filing Date
2026-05-07
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Current methods for photon beam modeling in radiotherapy are limited by the inability to adapt to changes in beam energy, require extensive manual recalculations, lack inclusion of beam flatness and symmetry factors, and struggle with small beam sizes, leading to inefficiencies and measurement uncertainties.

Method used

A method that constructs an initial photon beam model based on physical models of a medical linear accelerator, updates it using quality assurance measurements, and automates the process to reflect actual beam properties, allowing for accurate dose calculations and predictions.

Benefits of technology

Enables quick and automated updates of photon beam models, reducing the need for manual recalculations and physical measurements, and improving accuracy for all beam sizes, including small beams, while maintaining consistency with commissioning measurements.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CA2026050670_19112026_PF_FP_ABST
    Figure CA2026050670_19112026_PF_FP_ABST
Patent Text Reader

Abstract

A photon beam model for radiotherapy patient dose calculations is built from physical models of how a medical linear accelerator forms its beam, with model parameters set to match beam profiles measured at installation. After the accelerator enters clinical service, a processor receives routine quality assurance measurements, compares them to the original measurements, and adjusts the model parameters when differences are found, producing an updated model used to calculate patient dose. A dose kernel for these calculations is obtained by taking a dose distribution for a photon beam in a medium, calculating the total energy released per unit mass distribution from primary and scattered fluence components, and deconvolving the dose distribution with that distribution to produce a kernel that, when convolved with it, reproduces the dose distribution.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] LINEAR ACCELERATOR SIMULATION FOR PHOTON BEAM MODELLING IN RADIOTHERAPY

[0002] BACKGROUND OF THE DISCLOSURE

[0003] 5 1. Field of Disclosure

[0004] The present disclosure relates generally radiotherapy and in particular to method and system for determining photon beam doses for radiotherapy.

[0005] 2. Description of Related Art

[0006] The current state of the art uses photon beam energy spectrum to create a beam primary total energy released per unit mass (TERMA), which is composed of the TERMA from both the primary photon beam and scattered photon beam. The TERMA functions are typically obtained by Monte Carlo calculations at a fixed beam energy, and do not allow for any variation of the 15 beam’s energy. To produce a dose distribution, the total TERMA is convolved with a dose kernel. This dose kernel is also made at a fixed beam energy. The current state of the art uses an energy spread model to calculate dose kernels, where the response of an energy impulse is calculated using Monte Carlo methods.

[0007] 20

[0008] This method is limited since it cannot be applied to an electron beam of any beam energy, which can be possible as the accelerator environment changes overtime. To reflect a change in beam energy, the method requires substantial calculations to recompute the initial TERMA and dose kernels using Monte Carlo methods. These calculations are typically done using manual configuration of the Monte Carlo calculation environment, and so it is not easily automated or quickly updated.

[0009] The current state of the art is also limited because factors that affect the beam 30 flatness and symmetry are not included in the photon beam model. As a result, the photon beam model cannot easily be adapted or updated based on changes in the linear accelerator configuration and quality assurance measurements.

[0010] 3 193937 KLW 30722816 1The current state of the art also requires extensive beam measurements to fully characterize and validate the photon beam model. Typically, measurements must be made using radiation detectors in a water medium that mimics human tissue. Ion chambers of small sizes are the most typical detector.

[0011] 5 Measurements are typically made with very large field sizes (up to a 40 cm x 40 cm beam size), with repeated measurements at smaller field sizes. As well, measurements are made at various depths with the goal of mapping the dose distribution of any size and depth that the linear accelerator can produce. This is a long and laborious process and requires a large water tank. Large water tanks are inconvenient due to their size and large amount of water needed to create the necessary phantom.

[0012] The current state of the art is also limited for beam models of very small photon beams. Photon beam models have difficulty at small beam sizes (a few mm) 15 since the TERMA due to the primary and scatter photon beams are not well separated, and since the beam penumbra effects are also not well separated. To compensate for the poor properties of the beam model at small field sizes, many radiotherapy centers rely on physical measurements of the small beams to validate the beam model. However, this is also challenging since the typical 20 instrumentation available in hospital-based radiotherapy centers is not well suited for small beam measurements, and so any measurements will have significant measurement uncertainty.

[0013] In the current state of the art, the full photon beam model is calculated by adding a model of the scatter from the treatment head by using an empirical model where the scatter intensity is assumed to have an analytic function with a maximum at the beam center, and whose energy fluence is taken to be a fixed value that is related to the energy fluence of the primary source. This method is limited since the scatter distribution can be described more accurately, and 30 as well the relationship between the scatter energy fluence and the primary fluence depends on factors such as the position within the radiation beam.

[0014] 3 193937 KLW 30722816 1The current state of the art is also limited in that for any material changes to the linear accelerator performance, such as occurs when components age or fail, the current beam models are difficult to adapt to quality assurance beam data, and so either the linear accelerator must be adjusted to return all 5 measurements to match measurements made at the initial installation, or a new beam model must be created to account for radiation beam property changes. Both processes can require a significant amount of labor and are not easily automated to improve operating efficiency.

[0015] SUMMARY

[0016] According to a first embodiment of the present disclosure there is a method of maintaining a photon beam model for radiotherapy patient dose calculations, the method comprising: constructing an initial photon beam model based on physical models of a medical linear accelerator beam formation process, 15 wherein the initial photon beam model may comprise physical models including components of the medical linear accelerator and model parameters matched to base beam profile measurements obtained at machine installation; obtaining, by a processor, quality assurance measurements of the medical linear accelerator after the medical linear accelerator enters clinical service; 20 comparing, by the processor, the quality assurance measurements to the base beam profile measurements; adjusting, by the processor, the model parameters of the photon beam model to match the quality assurance measurements when the quality assurance measurements differ from the base beam profile measurements, thereby producing an updated photon beam model; and utilising the updated photon beam model for radiotherapy patient dose calculations.

[0017] The adjusting step may further comprise automatically updating the primary photon beam spectra used in the photon beam model by accounting for 30 changes in the physical configuration of the medical linear accelerator which affect the electron beam energy and current striking a bremsstrahlung target. The photon beam model may comprise a scatter model that selects a scatter spectrum derived from a primary spectrum matched to a bremsstrahlung beam

[0018] 3 193937 KLW 30722816 1calculated using the physical models. The physical models may be configured to calculate quickly to automate the photon beam modelling process and produce the updated photon beam model automatically when the quality assurance measurements are obtained, without requiring Monte Carlo 5 recalculation.

[0019] The method may further comprise calculating a patient-specific radiation treatment plan dose using the updated photon beam model, wherein the updated photon beam model reflects an actual dose distribution that the medical linear accelerator produces on a day of the quality assurance measurements, and wherein patient dose on a day of treatment may be predicted using the most recent quality assurance measurements. Utilising the updated photon beam model may comprise performing patient-specific quality assurance verification without requiring a separate radiation beam phantom 15 and verifying commissioning measurements by determining whether the initial photon beam model is consistent with the commissioning measurements. The method may further comprise determining a root cause of deviations between the quality assurance measurements and the base beam profile measurements based on the physical models, and conducting a cost-benefit analysis for 20 intervention required to correct out-of-tolerance quality assurance measurements.

[0020] Constructing the initial photon beam model may comprise simulating photon beam characteristics of the medical linear accelerator based on physical principles of a beam forming process, and the physical models may not depend on empirical models. The quality assurance measurements may comprise routine measurements including at least one of beam profiles, depth doses, beam symmetry, beam flatness, isocenter measurements, x-ray beam quality, and physical characteristics of beam generation components including currents 30 and voltages. The physical models may utilise linear accelerator configuration parameters including service mode parameters.

[0021] 3 193937 KLW 30722816 1According to a further embodiment of the present disclosure there is a method of determining a dose kernel for radiotherapy dose calculations, the method comprising: obtaining a dose distribution fora photon beam of a medical linear accelerator in a medium; calculating a total energy released per unit mass 5 (TERMA) distribution for the photon beam, wherein the TERMA distribution may comprise a primary energy fluence component and a secondary scatter energy fluence component; performing a deconvolution of the dose distribution with the TERMA distribution to obtain the dose kernel; and storing the dose kernel for use in radiotherapy dose calculations, wherein the dose kernel, when convolved with the TERMA distribution, produces the dose distribution.

[0022] The updated photon beam model may be capable of modelling any beam profile including asymmetric beam profiles and may predict results of partial beam maintenance procedures including beam steering adjustments and jaw 15 calibration adjustments. Step (a) may comprise constructing the initial photon beam model using a partial set of beam measurements, wherein the physical models enable creation of a complete beam model characterising beam properties for all field sizes and measurement depths and predicting any measurement that would be performed in a physical water tank. The updated 20 photon beam model may predict dose properties of radiation beams with small beam diameters based on partial measurements made at large field sizes. The method may further comprise using the updated photon beam model to generate a virtual water tank simulation having equivalent measurement capabilities to a physical water tank, wherein the virtual water tank simulation replicates measurements of small beam diameters based on partial measurements made at large field sizes, and may be quickly updated using routine quality assurance measurements without requiring complete recommissioning.

[0023] 30 BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In drawings which illustrate embodiments of the disclosure wherein similar characters of reference denote corresponding parts in each view,

[0025] 3 193937 KLW 30722816 1Figure 1 is a schematic view of a medical linear accelerator technology arrangement of the prior art.

[0026] Figure 2 is a schematic view of a further embodiment of a prior art medical linear accelerator arrangement.

[0027] 5 Figure 3 is a schematic view of a further embodiment of a prior art medical linear accelerator arrangement.

[0028] Figure 4 is a detailed side view of the photon beam production area of prior art medical linear accelerators.

[0029] Figure 5 is a side view shows a photon beam source for use in the present disclosure according to a first embodiment.

[0030] Figure 6 is an illustration of the beam originating form scatter of the primary beam from the beam source of Figure 5.

[0031] Figure 7 is a side view of a photon beam source for use in method of the present disclosure according to a further embodiment.

[0032] 15 Figure 8 is are graphs of the properties of the secondary (scattered) radiation beam, and that these are fitted to analytic functions.

[0033] Figure 9 is a schematic view of head geometry of the phone beam source of Figure 5.

[0034] Figure 10 are graphs of the scatter source representations from the source 20 of Figure 9.

[0035] Figure 11 is a side view of a photon beam source for use in the methods of the present disclosure according to a further embodiment.

[0036] Figure 12 are graphs of the photon beam penumbra arising from the geometric penumbra due to the beam spot size, and the transmission penumbra due to the rounded leaf ends of the multi leaf collimator.

[0037] Figure 13 is a graph showing penumbra for fixed jaw of obtained using the method of the present disclosure.

[0038] Figure 14 is a side view of the penumbra for MCL jaws according to the 30 present disclosure.

[0039] Figure 15 is an illustration of the process of calculating patient dose where the beam TERMA is convolved with a dose kernel to produce a dose distribution.

[0040] 3 193937 KLW 30722816 1Figure 16 is an illustration of a method of calculating dose kernels, where a known dose distribution is deconvolved with the TERMA distribution to calculate the dose kernel according to an embodiment of the present disclosure showing the geometric parameters that 5 contribute to the linear accelerator isocentre, the location where rotations of the three axes coincide.

[0041] Figure 17 is a side view of a medical linear accelerator for use in the method of the present disclosure according to an embodiment.

[0042] Figure 18 are graphs of the beam profile and the depth dose as measured from the device of Figure 13.

[0043] Figure 19 are profiles and charts beam profiles that can arise when the linear accelerator is not configured correctly. The beam profiles may show an asymmetry, of may be less uniform (less flat) if the beam energy is too low or too high.

[0044] 15 Figure 20 is a perspective view of an apparatus for use with the method of the present disclosure.

[0045] Figure 21 is an illustration of the isocenter of the apparatus of Figure 20. Figure 22 are graphs of the process of maintaining medical linear accelerators after routine quality assurance measurements according to prior art 20 methods.

[0046] Figure 23 is an illustration of a method for updating the radiation beam model in a treatment planning system based on quality assurance measurements and a physical beam model according to an embodiment of the present disclosure.

[0047] Figure 24 is an illustration of a shows the current inventions method of updating a beam model for improved patient dose verification according to an embodiment of the present disclosure.

[0048] Figure 25 is a schematic view of a system according to an embodiment of the present disclosure.

[0049] 30

[0050] DETAILED DESCRIPTION

[0051] The present disclosure provides a method of maintaining a photon beam model for radiotherapy patient dose calculations through measurements of routine

[0052] 3 193937 KLW 30722816 1medical linac quality assurance quantities. In order to use physical measurements of beam properties to quickly update the beam model, the method constructs a beam model based on the physical processes which the medical linac employs in the beam formation process. In particular, the present 5 disclosure provides an improved beam modelling process which can be automated more easily than existing methods by employing a physical model of the accelerator beamline. In embodiments of the present disclosure, a processor receives quality assurance measurements, performs calculations to compare the measurements to baseline data, and determines adjustments to the beam model parameters. The adjustments determined by the processor may be outputted for execution by a person or a machine or may be automatically applied by the processor to update the beam model. Employing a physical model allows a beam modelling process that can utilize partial beam measurements, such as radiation beam quality assurance measurements to 15 automatically update the beam model, ensuring that it accurately represents the linear accelerator state, and predict a complete complement of beam profiles such as performed at linear accelerator installation and commissioning. As well the present disclosure allows modelling of beam properties that are not currently used in photon beam modelling, such as beam symmetry and flatness.

[0053] 20

[0054] Referring to Figure 1 a common arrangement of the medical linear accelerator prior art is illustrated where a high energy photon beam useful for radiotherapy is created using a standing wave accelerating waveguide which is fed from a klystron RF source and triode gun electron beam source. The RF source is powered by a high voltage modulator that uses a dequing method to regulate the high voltage pulses. The accelerated electron beam is redirected towards a patient using a bending magnet where the electron beam goes through a 270- degree deflection and crosses back over its initial path. The electron beam is converted to a photon beam through the bremsstrahlung process and a suitable 30 target. Steering coils are included to direct the electron beam to the target such that it is incident at the correct position and angle so that the resulting photon beam is symmetric. The apparatus is gantry mounted so that it can direct the beam at the patient from any angle.

[0055] 3 193937 KLW 30722816 1ln particular as utilized in the system of Figure 1, a three-phase electrical feed from the facility is converted to a high voltage through a High Voltage Power Supply (HVPS) 101, which then charges a pulse forming network (PFN) 106.

[0056] 5 The voltage of the PFN is regulated by the charging choke 102, dequing system 103, and charging diode 105. A high voltage pulse is created by discharging the PFN through a thyratron 104 and pulse transformer 107, which energizes a klystron 108. The klystron then amplifies the RF power coming from an RF driver 109, which is transmitted through a transmission waveguide 110 to an accelerating waveguide 111. The accelerating waveguide then increases the energy of electrons injected into it from a triode electron gun 112 to create a high energy electron beam 118. The high energy electron beam is redirected towards the patient by a bending magnet 113. The high energy electron beam strikes a target 114. The position and angle of the electron beam on the target 15 can be controlled by steering coils 117. The photon beam is collimated by jaws 115 and a multileaf (MLC) collimator 119.

[0057] Figure 2 shows an alternative beam creating arrangement used in medical linear accelerators, where the voltage on the PFN 106 is controlled by a flyback 20 transformer 203 and charge regulator 202 where a high energy photon beam useful for radiotherapy is created using a travelling wave accelerator that is fed from a magnetron RF source and a diode gun. The RF source is also powered by a high voltage modulator, which uses a flyback transformer to modulate the high voltage. The accelerated beam is redirected towards the patient using a series of deflecting magnets that produce a slalom trajectory of the electron beam path without the electron beam crossing over the initial path. The electron beam is converted to a photon beam through the bremsstrahlung process and a suitable target. Steering coils are included to direct the electron beam to the target such that it is incident at the correct position and angle so that the 30 resulting photon beam is symmetric. The apparatus is gantry mounted so that it can direct the beam at the patient from any angle.

[0058] 3 193937 KLW 30722816 1In the system of Figure 2, the PFN, when discharged through the thyratron 104 and pulse transformer 107, energizes a magnetron 204 which feeds a travelling wave accelerator 205. This accelerates electron coming from a diode electron gun 210 to create a high energy electron beam 118. This beam is redirected 5 towards the patient through a series of slalom magnets 208 towards the target 114 which converts the electron beam energy to a bremsstrahlung photon beam 116. The position and angle high energy beam 17 on the target 13 can be controlled by the steering coils 117. The photon beam is collimated by jaws 115 and a multileaf collimator 119.

[0059] Figure 3 shows a third alternative beam creating arrangement used in medical linear accelerators, where the RF output of a magnetron 204 is sent to a standing wave accelerator 111. In particular, in the system of Figure 3, a high energy photon beam useful for radiotherapy is created using a standing wave 15 accelerator that is fed from a magnetron RF source and a triode gun. In this embodiment, the linear accelerator is shorter, and so cannot produce high beam energy. Because the accelerator waveguide is shorter, it is pointed directly at the patient, and so it does not need to be redirected towards the patient. This embodiment includes other standard embodiments such as a high 20 voltage modulator, photon beam bremsstrahlung target, beam steering and beam collimation system. The apparatus is gantry mounted so that it can direct the beam at the patient from any angle. In this case the accelerator is pointed directly at the patient, and a redirecting magnet is not needed.

[0060] Those skilled in the art will also know that there are multiple other ways of using linear accelerator technology to assemble a system that can be installed in the limited space of a hospital and produce a megavoltage photon beam. These separate embodiments can be described in a similar way as described here, and the beam modelling method described in the current disclosure can apply 30 to those embodiments as well.

[0061] Each component listed in Figures 1, 2 and 3 have physical models that can be used to accurately predict the electron beam properties based on the physical

[0062] 3 193937 KLW 30722816 1principles of the device, such as found within Karzmark et al. (Medical Electron Accelerators, McGrawiHill, 1993, ISBN 0-07-105410-3) which is hereby incorporated by reference. In particular, the electron beam energy E can be determined through the load line equation, which for a standing wave 5 accelerator is:

[0063]

[0064] In the load line equation Z is the linear accelerator shunt impedance, P is the input RF power from the microwave source (klystron or magnetron), p is the coupling coefficient between the feed waveguide and accelerating waveguide, and / is the electron beam current. Similar load line equations exist for travelling wave accelerators. Those skilled in the art will recognize that there are other variations of these equations and other calculation methods (such as finite element calculations) that can improve the accuracy of the result.

[0065] 15 Physical descriptions of the other common components shown in Figures 1 - 3, are available in the public literature. For example, models of the electron gun are also described in Karzmark as well as in Beck (An Introduction to Physical Electronics, Edward Armold Publishers, 1972) and Anderson (JACMP, 2015) which is hereby incorporated by reference.

[0066] 20

[0067] In another example the physical model of the bending magnet can be shown represented by a linear relationship between the bending magnet current and the pass-through electron beam energy (Anderson, JACMP, 2015; Carlone, Med Phys, 2023). In yet another example, the klystron can be modelled by the method described by Anderson (JACMP, 2015).

[0068] In yet another example, magnetrons can be modelled using the performance chart provided by the magnetron manufacturer and are also described in detail in Collins (Microwave Magnetrons, McGraw Hill Book Company, 1948). Other 30 theories of operation that can be used to provide physical models are also known for the following methods:

[0069] 3 193937 KLW 30722816 1• Resonant charging of the PFN (Glasoe and Lebacqz, Pulse Generators,

[0070] McGraw Hill Book Company, 1948)

[0071] • Flyback transformer (McLyman, Transformer Design Handbook, 4thEdition, CRC Press, 2011)

[0072] 5 • Discharging of the PFN (Glasoe and Lebacqz, Pulse Generators, McGraw Hill Book Company, 1948)

[0073] • Steering of the electron beam (Anderson, JACMP, 2015)

[0074] • Energy switch (Tanabe, Nucl. Instrum. Meth. B10 / 11, 871-76, 1985)

[0075] • RF feedback in the travelling wave accelerator (G Saxon 1954 Proc. Phys.

[0076] Soc. B 67 705)

[0077] • RF Frequency control system and heat load (Pei Shi-Lun et al 2012 Chinese Phys. C 36555)

[0078] By using these physical models, a complete model of the electron beam energy and current can be constructed that will reflect the electron beam physical 15 properties that reach the bremsstrahlung target.

[0079] Those skilled in the art will recognize that there are numerous other previously disclosed models describing medical linear accelerator technology that can be used to improve the model accuracy that predicts the electron beam energy, 20 current, position, and angle of incidence that strikes the bremsstrahlung target.

[0080] In commercially available medical linear accelerators, the configuration and setup of a particular beam is configured in a special console mode often referred to as “service mode”. This is a special console user interface where the parameters that control the physical operation of linear accelerator components, such as the ones described herein, are determined and stored such that they can be retrieved at a later time in a “clinical mode”, which is reserved for patient treatments. In the current Art, after setting the service mode parameters and locking them down to prevent changes, users will measure the 30 resulting radiation beam properties very carefully so that quality control measures can be applied to ensure that the radiation beam used in the clinical mode has not changed from that was configured in service mode. In the present

[0081] 3 193937 KLW 30722816 1disclosure, where physical models of the machine components are used to create a complete model of the linear accelerator electron beam characteristics present at the target, it is also possible to introduce additional calibration models such that the service mode parameters that are employed in a particular 5 commercially available medical linear accelerator can be mapped to the physical model of the component in the linear accelerator beam generation system. For example, there are numerous voltages and currents that must be set for the components in the medical linac to operate at the desired level. In the service mode, a user may input a digital value to represent the actual voltage or current used in the medical linac. The service mode input may have a range of 1 - 10 in arbitrary units. The physical component may then change current, for instance, between 0 - 300 mA, depending on the control electronics provided in the commercial system. In the current disclosure, calibration models may be employed in addition to the physical models so that the service mode 15 user inputs, in the units described in the user interface, can be mapped to the physical component with an appropriate physical parameter that is controlled to the desired value for optimal modelling of the medical linac functionality.

[0082] Figure 4 shows a common arrangement for the x-ray beam forming area of 20 medical linear accelerators in which a high energy electron beam is converted to a photon beam by striking the electron beam on a suitable target. The photon beam is collimated through a primary collimator and a series of secondary collimators. The intensity of the photon beam can be made uniform by attenuating the intensity using a flattening filter. Alternatively, an un-flattened photon beam can also be used by removing the flattening filter. A beam monitor is also used to quantify the photon beam dose so that a pre-determined dose can be administered to the patient. The high energy electron beam 118 is incident on a target 114, which converts the electron beam energy to photon beam energy through a bremsstrahlung process. The photon beam is 30 collimated through a primary collimator 401. The intensity of the photon beam is highly variable due to the bremsstrahlung process, and so a flattening filter 402 is used to make the intensity uniform. Alternatively, the flattening filter 402 may be omitted and the raw beam is used to irradiate the patient. The photon

[0083] 3 193937 KLW 30722816 1beam 116 is collimated to an appropriate aperture shape through collimating jaws 115 or a multileaf collimator (MLC) 119.

[0084] In current methods, the physical properties of the photon beam 116 must be 5 determined and used in a treatment planning system to configure the dose calculation algorithm. The standard method of modelling the radiation beam is to use the photon beam spectrum from a typical or generic medical linear accelerator using Monte Carlo calculations at a fixed electron beam energy, and it is assumed that this beam energy is held constant for all patient treatments. This information is used to calculate the primary beam parameters needed for photon beam modelling such as the mean energy as a function of position within the photon beam. The user supplies beam profile data for photon beam and manually adjusts parameters within the beam modelling systems to match the beam data. If the beam energy were to change, then a new photon 15 beam spectrum would have to be re-computed using Monte Carlo calculations.

[0085] This method is problematic since it is hard to automate this process to keep the model accurate as the physical properties of the medical linear accelerator change.

[0086] 20 In the present disclosure, the photon beam spectra is calculated using the method described by Findlay (Nuclear Instruments and Methods in Physics Research A276 (1989) pages 598-601) which has the benefit of having similar accuracy as Monte Carlo calculations, but can also be updated easily since Findlay’s method utilizes the beam energy and target configuration as model inputs, and then performs an analytic calculation to determine the photon beam spectra. Findlay’s method has been compared to Monte Carlo calculation methods and shown to compare well. The benefit of using Findlay’s method is that if the configuration of the medical linear accelerator changes, the physical model can be updated automatically using measurements of the linear 30 accelerator component operating values and determine the effect on the electron beam energy and current striking the bremsstrahlung target. Using the method of Findlay to calculate the effect on the photon beam spectra then ensures that it remains accurate over the lifetime of the linear accelerator, even

[0087] 3 193937 KLW 30722816 1if the operating point of the linear accelerator changes. Since Findlay’s method can be adapted to accept the angle of incidence of the electron beam 118 upon the target 114, it can be used to predict the asymmetry in the bremsstrahlung beam fluence 116. Also, Findlay’s method can be combined with a model of a 5 flattening filter 402 where the attenuation of the photon beam 116 through the flattening filter 402 is energy dependent, and so the variation in intensity of the photon beam 116 will depend on the energy of the electron beam 118 and the physical shape of the flattening filter 402. Findlay’s method can thus be adapted to simulate the photon beam 116 “flatness”, which is defined as the percentage of variation of the dose measured across the width of the radiation beam 116, excluding the penumbra region, and “symmetry” which is the percentage difference in dose measured at two positions equal distance from the central axis 118, but on opposite sides. Those skilled in the art will recognize that there are numerous formulas that have been used to quantify beam flatness and 15 symmetry, and that the methods presented in the current disclosure can be adapted to simulate any of these methods.

[0088] Figure 5 shows a model for secondary (scattered) radiation from the x-ray beam forming area. Scatter originates from the flattening filter area 402, or primary 20 collimator area 401 and is projected through jaws 119 to create a scatter intensity 501 as shown in Figure 6. The TERMA distribution due to the head scatter 504 can be calculated by integrating the summation points 503 over the field size defined by the MLC beam aperture 119.

[0089] In the current state of the art, the full photon beam model is calculated by adding a model of the scatter from the treatment head by using an empirical model where the scatter intensity is assumed to be a broad function that is peaked at the center, and reducing to 0 over a distance that is an input to the scatter model. This method is limited since the scatter distribution is empirically 30 derived, and so the relationship between the scatter energy fluence and the primary fluence, which depends on factors such as the position within the radiation beam which is not included in the model.

[0090] 3 193937 KLW 30722816 1The present disclosure improves the current state of the art by employing an improved scatter model, where the scatter intensity is calculated using a method like the one used by Sikora (PhD thesis, University of Bergen, 2011) entitled “Virtual Source Modelling of Proton Beams for Monte Carlo Based 5 Radiation Therapy Treatment Planning” which is incorporated herein by reference. In Sikora’s method, the scatter spectrum, mean energy as a function of beam position, and particle fluence intensity were calculated using Monte Carlo calculations for a standard beam energy. It was shown that a good representation of the secondary beam properties could be obtained by fitting them into an analytic function. To utilize the Sikora method in the current disclosure, this method is extended by computing the scatter intensity distribution and spectrum over a range of beam energies large enough to cover radiotherapy applications, and for many combinations of target and primary collimator materials.

[0091] 15

[0092] Figure 7 shows the radiotherapy photon beam generation area limited to the electron beam 118 striking the target 114, being collimated by a primary collimator 401 and attenuated by a flattening filter 402, producing the primary beam, which has a specific energy fluence. The current disclosure calculates 20 the primary energy fluence for a large combination of electron beam energies and target, primary collimator and flattening filter material combinations as shown in Figure 8. For each combination, an analytic fit to the secondary (scatter) spectrum, mean energy and particle fluence is calculated and stored. In the current disclosure, as the beam energy changes depending on the operating state of the medical linear accelerator, the appropriate secondary beam properties are selected by matching the new primary beam spectrum to the one used to create the analytic fit.

[0093] In order to combine the primary energy fluence and secondary (scatter) energy 30 fluence into a TERMA profile, it is necessary to know the relative contribution to dose of both primary and scatter. The question of primary and secondary (scatter) dose contribution was considered by Yang (Yang, A three-source model for the calculation of head scatter factors, Med. Phys. 29 (9), September

[0094] 3 193937 KLW 30722816 12002, which is hereby incorporated by reference), who investigated the total dose variation as the field size changed by adjusting the collimating jaws 115 or the multileaf collimator 119 aperture. Yang did not consider a method to determine the relative contributions of primary and scatter (secondary) energy 5 fluence.

[0095] Figure 7 shows a geometry to calculate the scatter dose 504 for measurements made in air along the central axis 701, where the scatter dose at plane 702, which has a distance d from the target 114.

[0096] In relation to Figure 7, consider two measurement planes at different distances d along the central axis 701 from the target 114. We consider a reference plane 0 at position do, and a measurement plane 1 at position di.

[0097] The ratio of the primary dose on plane 0 at distance do from the target 114 (Po) to plane 1 (Pi) at distance di along the central axis 601 is

[0098] = pQ= pA^)2

[0099]

[0100] 15 The scatter dose 504 is obtained by integrating the intensity of the scatter source 501 seen at the plane a distance d from the target 114. 501 has an intensity distribution that varies as a function of distance rfrom the central axis plane. The scatter source is located a distance s from the target. Referring to the scatter source intensity distribution, 501, as S(r), and the scatter dose measurement 504 on the central axis 601 as S / W:

[0101] SM = a2f S(r) [^7777] dr, where a = d - s.

[0102]

[0103] Assume that S(r) is peaked at r = 0 and diminishes as r — ► Further assume that the scatter source can be represented as a series of uniform intensity disks 801 as shown in Figure 8. Each disk 801 can be referred to as sz.

[0104] 25 In this case:

[0105] n

[0106] S(r) =

[0107] i = l

[0108] n sw

[0109] SM = a2[ —5 — ■ — 7 dr

[0110] 4-i J la2+ r2l

[0111]

[0112] i=l -sw

[0113] The integration limits, sw, are the extents of the scatter source visible at the point on the central axis at distance d from the target, which is labelled sw. By

[0114] 3 193937 KLW 30722816 1similar triangles:

[0115] sw= 0 (fl)'

[0116] Further assuming that for each scatter disk of uniform intensity, the change in scatter summation, XF=isi / is small seen at different distances d from the source. Applying 5 these to the scatter integration:

[0117] SM = 2a arctan ( / 7Jsr where each S / is taken as constant in the

[0118]

[0119] integration.

[0120] For a second and third plane at distances di and d2:

[0121] n

[0122] ( f \ v

[0123] SM-L = 2a±arctan J y Sj

[0124] 10 SMi = 2a-> arctan (— s.-

[0125]

[0126] 2 2\d2J

[0127] i=ll

[0128] So, the ratio of scatter measured at two distances di and d2 from the target is: SM± I / a„i \ \ arctan( yddy-)

[0129] SMz \a2J arctanfd-}

[0130]

[0131] \d2J

[0132] since the scatter summation,

[0133]

[0134] cancels out.

[0135] In an experiment with three measurements, at three different distances from the target 114: a reference plane at distance do, and two other planes at distances di and d2,

[0136] The ratio of scatter dose at plane 2 to plane 1 , will be expressed as

[0137] SM2_ B

[0138] SMI A

[0139] Likewise, the ratio of scatter dose at plane 0 to plane 1 will be expressed as SM0> C

[0140] 20

[0141] SMi A

[0142] The quantities B / A and C / A are known based on the theory explained here. The total signal measured, M, is the sum of the primary dose plus scatter dose: M = P + SM

[0143] The ratio of the measurement at plane 0 to plane 1 is:

[0144] M0PQ+ SMQ + SM±- j

[0145] 25

[0146]

[0147] MrPr+ SMi ?!+ SMr= measurement 1

[0148] Similarly, the ratio of the measurement at plane 2 to plane 1 is:

[0149] 3 193937 KLW 30722816 1

[0150]

[0151] With the known quantities’ B / A and C / A, the two equations above have two unknowns, Pi and SMi, and so they can be determined independently.

[0152] There are many ways to improve the theory presented here such that an 5 improved accuracy can be obtained when making in air measurements as described above. For instance, the scatter source function 510 may be assumed to have a particular form, such as considered by Yang, who suggested that the function

[0153] A „

[0154] S(r) = - e~kr

[0155]

[0156] can be a reasonable representation of the scatter source strength. In this case, the scatter dose measurement along the central axis can be represented as:

[0157] sw A

[0158] I r

[0159] a2+ r2

[0160]

[0161] This equation can be evaluated numerically. Alternatively, S(r) can be modelled using Monte Carlo methods, as was done by Sikora, and the numerical 15 representation of S(r) can be used directly, along with a numerical evaluation of the scatter integration.

[0162] As well, the method can be extended from 1 dimension, as described here, to two dimensions where the jaws are located at different height. As well, as suggested by Yang, the scatter source S(r) can also be divided into sources such as from the primary collimator and flattening filter, which originate from different distances sform the target.

[0163] Finally, the method can be applied in situations of different scatter situations, such as in the presence of a flattening filter, and where the flattening filter has been removed. In these two situations, differential scatter contributions can be 25 held constant. For instance, the scatter contribution from the primary collimator is present when the flattening filter is removed, so by repeating the method under the two scatter situations, the contributions from the primary collimator and flattening filter can be resolved independently.

[0164] 3 193937 KLW 30722816 1Those skilled in the art will also know that there are multiple other ways to improve the precision of the method to separate in air scatter dose form in air primary dose. These separate embodiments can be applied in a similar way as described here, and the in air primary dose to scatter dose separation method 5 described in the current disclosure can apply to those embodiments as well.

[0165] Figure 11 shows the relationship between the photon beam 116 originating from the target 114 and the geometric penumbra 703 and the attenuation penumbra 704. The geometric penumbra arises from the different ray paths of photons with the beam spot size 701 as they pass the MLC 119. The attenuation penumbra is due to the attenuation of the photon beam in the MLC at its rounded ends. The size of the penumbra is a function of the overall field size and is reduced as the field size becomes smaller. As shown in Figure 12 the penumbra 707 for a mid-size field size 708 is smaller than the penumbra 705 15 for the large field size 706. For the very small field size 710, the two penumbras 709 begin to interfere with each other which affects the overall heigh of the profile and renders it smaller than for the larger field sizes 706 and 708.

[0166] This effect was discussed by Das (Med Phys. 2022;49:4043-4055 which is hereby incorporated by reference) and shown to have significant effect on the 20 beam dosimetry for small fields, beginning at field sizes of 1 cm or less where the penumbra contributes significantly to the dose so that the dose is substantially reduced at field sizes less than 1 cm. This was verified by measurements as well as Monte Carlo calculations. Measurements of dose at small field sizes is challenging and requires special equipment for small field dosimetry that is not always available in hospital environments and is more difficult to implement than large field size measurements.

[0167] Figure 10 shows realistic beam profiles for field sizes ranging from 3 cm to 40 cm as defined by jaws 115 similar to those shown in Figure 1. The penumbra 30 1001 is represented by the distance between the position where the dose is 80% of the central dose and the position where the dose is 20% of the central dose 1002. For realistic profiles the penumbra depends very little on the field size as shown in the following table:

[0168] 3 193937 KLW 30722816 1FS D80-D20 (cm) D80-D20 fit (cm) error (cm) error (mm)

[0169] 3 0.51179868 0 508858588 0 0029 0 03

[0170] 4 0.510455653 0.514450967 -0.0040 -0.04

[0171] 6 0 522857143 0.524576623 -0.0017 -002

[0172] 8 0 532658283 0 533371226 -0.0007 -0.01

[0173] 10 0 546860158 0.540932072 0 0059 0.06

[0174] 20 0 559388434 0.563635347 -0.0042 -0.04

[0175] 30 0.572431264 0 570089259 0 0023 0 02

[0176] 40 0.571920405 0 572455938 -0 0005 -0.01

[0177]

[0178] The penumbra (D80 - D20) changed from 0.512 cm at a 3 cm field size to 0.572 cm at a 40 cm field size. This data can be fit to this polynomial:

[0179] Penumbra = 2.027 x 10“6FS3- 0.0002029FS2+ 0.0069375 FS +

[0180] 5 0.489817

[0181] Using this formula, a maximum error or 0.06 mm occurs for the penumbra. Thus, to determine the penumbra originating from jaws 115, it is only necessary to determine the penumbra at a single field size such as the largest, and by applying a correction to the penumbra to correct for field size effects.

[0182] 10 Figure 11 shows the region of the photon beam 1101 that forms the penumbra 1102 from an MLC 119, which has a curved surface. Since the surface curvature remains the same as the MLC 119 is change in field size, the region 1102 that forms the penumbra is also predictable. The ray line 1103 just touches the curved MLC 119 surface, where the ray line 1104 reaches the 15 maximum attenuation. The following table plots the MLC penumbra 1102 for a realistic MLC shape.

[0183] FS MLC penumbra 1102 (cm)

[0184] 3.0 1.44

[0185] 4 1.45

[0186] 6 1.43

[0187] 8 1.49

[0188] 10 1.60

[0189] 20 1.96

[0190]

[0191] 3 193937 KLW 30722816 130 2.27

[0192] 40 2.77

[0193]

[0194] The penumbra for the MLC can thus be determined by a simple ray tracing algorithm and using a beam attenuation function to model the dose response. The penumbra can be determined for any field size, including very small sizes 5 less than a cm, to large field sizes. Alternatively, a more complex ray tracing algorithm, such as that employed by Hirayama (Med Phys. 2026;53:e70220, which is hereby incorporated by reference), which also accounts for secondary scatter ray tracing. This may improve the precision of the penumbra. Those skilled in the art will recognize that many variations and improved ray tracing algorithms may be employed to calculate the penumbra due to attenuation of the curved portion.

[0195] The total beam model TERMA in the current disclosure consists of three parts, the primary fluence, the secondary (scatter) fluence, and a model of the 15 penumbra due to the MLC jaw face. All three of these components can be evaluated using analytic expressions as described herein, which allows the photon beam model of the current disclosure to be adapted automatically when updated beam information is measured by a user through quality assurance measurements, for example. Further, to determine the three components of the 20 dose, measurements at larger field sizes can be used to fully characterize the secondary (scatter) dose as well as the penumbra due to jaws 115 and MLC 119 at any field size. By characterizing the scatter dose as large field sizes which are easier to measure, and by employing the methods described herein where the dose due to the primary beam and scatter beam are added to obtain the complete TERMA, it is possible to predict dose for any field size since the primary dose does not depend on field size, and the scatter dose and penumbra dependance on field size is described herein. This represents an advantage over current methods where small beams must be characterized by measurements, which are difficult to implement in hospital settings. Thus, the 30 beam model described herein allows users to fully characterize the beam

[0196] 3 193937 KLW 30722816 1dosimetry for an arbitrarily small beam size by making measurements are larger beam sizes.

[0197] For a complete beam model, those skilled in the Art will recognize that 5 additional parameters are needed to predict dose outside the beam area. In this case, fixed jaw attenuation factors, or MLC interleaf transmission factors are used to calculate the out of field dose. These methods can be used within the present disclosure without changing the nature of this disclosure.

[0198] Additionally, those skilled in the art will recognize that additional modelling is needed to properly model the dose build up region of megavoltage radiation beam. Electron contamination dose from scattered electrons in the treatment head have been shown to affect the dose build up region. Incorporating these methods to model the dose build up region can be added to the present 15 disclosure without changing the nature of this disclosure.

[0199] Figure 15 shows the current state of the art for converting a TERMA profile to a dose distribution for a megavoltage radiation beam. A TERMA profile 801 is convolved 802 with a dose kernel 803 to create a dose distribution 804. This 20 method depends on the quality of the dose kernel 803. The state of the art for calculating dose kernels is to use Monte Carlo calculation methods as the EDKnrc function supplied by the National Research Council of Canada (EGSnrc Code System described in NRC Report PIRS-701). A well-known problem with this method is that the ray lines for photon beams diverge at positions away from the central axis of the radiation beam. To account for this divergence in dose calculations, the dose kernel should also be tilted to match the divergence. However, this is computationally difficult since it does not lend itself to fast convolution calculation methods.

[0200] 30 Figure 16 shows an improved method of calculating the dose kernel 803. In the current disclosure, the dose kernel 803 is obtained by deconvolving 901 the known dose distribution 804 (which can be calculated by Monte Carlo means, obtained from measurement, or any other means of establishing the dose

[0201] 3 193937 KLW 30722816 1distri bution) with the TERMA distribution 801 to obtain the dose kernel 803. This method has the benefit of forcing or ensuring that the dose kernel will convolve the TERMA to exactly match the target dose distribution, whether it be calculated by Monte Carlo, or measure experimentally. This method can be 5 applied to a water phantom, or it can be applied to any medium such as bone, lung, muscle, or metals used in prosthetic devices or medical procedures. This method also has the additional benefit of determining an average dose kernel for an entire field size, which will include the effect of the beam divergence and kernel tilt, and so the dose kernel will be dependent on the field size of the radiation beam. In the beam model, the appropriate dose kernel is selected depending on the field size.

[0202] Figures 17 and 18 show the measurement process for performing quality assurance measurements of radiotherapy linear accelerators where two 15 important beam parameters. In particular, Figures 17 and 18 show a typical set of beam measurements that are made to quantify radiotherapy beams for treatment planning at the time of machine installation, or for quality assurance measurements done after the machine has entered clinical service to verify that the medical linear accelerator has not changed configuration since the 20 measurements were made at the time of installation. The medical linear accelerator system 1001 goes through a comprehensive process where a detector 1003 is used with a phantom 1002 as shown in Figure 17. The detector is usually an ion chamber; however, many other types of detectors can be used including arrays of ion chambers, diodes or arrays of diodes, diamond detectors, optical dosimetry, film, or any other dose measuring device. The phantom is usually a water tank; however, it can also be a solid device that has the radiological properties of water, or any other medium that has an accepted use for simulating dose deposition in tissues. As illustrated in Figure 18, the most common measurements are 1 ) a dose profile 706 where the detector 1003 30 is scanned in a direction perpendicular to the beam direction and the detector depth in the phantom is held constant; and 2) a depth dose 1004 where the detector is scanned in a direction parallel to the beam direction and the detector depth in the phantom 1002 varies. These profiles are measured for many field

[0203] 3 193937 KLW 30722816 1sizes and detector depths in the phantom to obtain a complete characterization of the radiation beam. These beam properties are the ones used to adjust the beam model in a treatment planning system.

[0204] 5 Figure 19 shows some mechanisms that can result in medical linear accelerator beam profiles having properties that differ from the ones obtained at the time of installation. The steering coil 117 may have an incorrect beam current 1101, which can result in a beam deflection 1102 that directs the electron beam 118 to the target 114 at an incorrect angle of incidence. In this case the dose profile 1103 will be asymmetric, which is not desirable. Asymmetric profiles can also result from a mechanical misalignment of the accelerating waveguide and the bending magnet, target and flattening filter. It can also result from misplacement of the steering coils. Alternatively, the energy of the electron beam 118 may be lower or larger than at the time of machine installation. In the case where the 15 beam energy is too low, the dose profile 1104 will dip at the middle of the profile, and the depth dose 1105 will shift to a smaller value (as compared to the sashed line, which is the profile at installation). In the case where the beam energy is too large, the dose profile 1106 will bulge in the middle, and the dose profile 1107 will shift to a larger value (as compared to the sashed line, which is the 20 profile at installation).

[0205] Those skilled in the art will recognize that the descriptions of the cause of beam profile deviations described in this disclosure may not be complete, and that there may be other reasons why medical linear accelerator beam profile measurements may change overtime. These can include changes in the high voltage system, RF system, ion source, beam optics, mechanical changes to the machine, and a multitude of other reasons.

[0206] Figure 16 shows how medical linear accelerators are positioned about a patient.

[0207] 30 The medical linac 1601 is mounted on a gantry such that it can rotate about a patient that is supported by a bed 1602. As the gantry 1601 is rotated about the axis 1603, its beam will intersect at a point 1604 commonly known as the beam isocenter. The patient bed 1602, as well as the beam collimator 1605, also

[0208] 3 193937 KLW 30722816 1rotate about axes 1606 and 1607. It is desired for the three axes 1603, 1606 and 1607 to intersect at a common isocenter point 1604. However, in practice, the isocenter point 1604 will not be an infinitely small point, since the medical linac bends and deforms as it rotates due to the force of gravity. The patient’s 5 bed and collimator likewise do not rotate about an infinitely small point. Hence, the intersection of all rotation axes is a volume 1608 that has finite dimensions and size. The isocenter size is commonly described at the radius of a sphere that encompasses the intersection of the three axes 1603, 1606 and 1607. In the current state of the art, the misalignment of the radiation beam due to the finite size of the isocenter is not accounted for.

[0209] Misalignment of the radiation beam isocenter can naturally fit into the beam model of the present disclosure since this represents a numeric offset to the beam position, which can be described by an analytic model, and included in 15 the beam model of this disclosure. This has particular importance at small beam sizes (1 cm or less) where the typical isocenter size (~ 1mm) has an important effect on the resulting dosimetry.

[0210] Figure 22 shows the process for maintaining medical linear accelerators after 20 installation according to conventional methods. At periodic frequencies, the beam profile measurements such as profiles 1103, 1104 or 1106 are measured and compared to the dose profile 706 that was measured at the time of the machine installation. Also, depth dose profiles 1105 and 1107 are also measured at periodic frequencies and compared to the depth dose measured at the time of installation. If there are significant differences in beam profiles are detected, the common practice is to adjust the medical linear accelerator operating parameters or repair the system to return the system to the same state as at the time of installation. The linac isocenter is also measured at periodic frequencies, however there is no simple method to adjust the isocenter 30 since this is related to the mechanics of the medical linac. If the isocenter is determined to be significantly different from that was determined at the time of machine installation, then the only action is mechanical repair of the medical linear accelerator.

[0211] 3 193937 KLW 30722816 1The current methods have undesirable properties. The first is that the process of adjusting or repairing the linear accelerator so that the quality assurance measurement match those at the time of installation may require much effort, 5 time and cost.

[0212] The second is that it requires a manual assessment of the beam properties and manual adjustment of the medical linear accelerator in the case where it is judged to be required. This requires a user with a high degree of skill and knowledge about medical linear accelerators to competently carry out the adjustment.

[0213] The third is that variations in radiation beam quality may occur for several physical reasons, and users with varying levels of knowledge about the beam 15 forming process may not respond with similar interventions, depending on the skill level of the person interpreting the beam profile measurements. This can cause an uneven response to quality assurance failures, or longer than necessary repair times or other down times needed to return out of specification medical linear accelerators to clinical service.

[0214] 20

[0215] The fourth is that it is common to reduce beam quality variations to within an acceptable range. For instance, the American Association of Physicists in Medicine (AAPM) has published report 142 (Medical Physics, 36, 4197-4212 (2009) https: / / doi.orq / 10.1118Z1.3190392 ISBN: 978-1-888340-88-4, which is hereby incorporated within by reference) which specifics acceptable tolerance values for beam medical linear accelerator beam flatness, symmetry, depth dose, isocenter, field size and other quantities commonly associated with the radiation beam quality. However, when reducing measured beam profiles to within acceptable tolerances, residual errors are inevitable, and it is difficult to 30 assess the cumulative beam quality error that results when all beam quality measures are considered together. Those skilled in the Art will recognise that there are many other undesirable properties of the methods described in Figure 17 that are not disclosed herein.

[0216] 3 193937 KLW 30722816 1Figure 23 shows a new method for updating a radiation beam model in a treatment planning system based on quality assurance measurements and a physical beam model. At the time of machine installation, the base beam 5 profiles 706 and 1004 are measured and stored, typically using a water tank system. Additional measurements, such as isocenter measurements can also be included in this measurement set. With a suitable set of beam profile measurements, an overall physical beam model can be constructed by using the models described in the current disclosure to match the measured beam properties. This may be a manual or automated process. As well, the modelling process may be done in a single step, or may be done recursively, where the model is constructed, and then one or more iterations are used to obtain the best model fit. The beam model then consists of two components: 1) the physical models used to describe each component in the medical linac, and 2) 15 the model parameters used to match the model to the measured beam profiles.

[0217] The model parameters may be further separated into two components: 1) a base value that matches profile measurements made at machine installation, and 2) calibration parameters that can be adjusted to alter the base model to fit it to new measurements reflecting a change in the machine state.

[0218] 20

[0219] After the machine enters clinical service, routine quality assurance measurements are made at a prescribed frequency, for instance such as described in the AAPM report 142. The quality assurance measurements are received by a processor, which compares the measurements to the base measurements such as scans 1103, 1105 and 1107, or other measurements. In the case where these measurements show beam characteristics that differ from the base measurements, the processor calculates adjustments to the model parameters obtained at the beam modelling step so that a new beam model that matches the quality assurance measurements can be obtained. The 30 processor may output the calculated adjustments for review and execution by a user, may transmit the adjustments to the medical linear accelerator or treatment planning system for machine execution, or may automatically apply the adjustments to produce the updated beam model. Since the beam model is

[0220] 3 193937 KLW 30722816 1based on physical models that can be calculated directly and quickly, it is possible to automate this step and quickly produce an updated beam model based on new quality assurance measurements.

[0221] 5 Figure 24 shows the current disclosure method of updating a beam model for improved patient dose verification. Using the procedure described in Figure 23, once new radiation beam model parameters have been obtained, these parameters can be used to calculate the radiation dose that will match the dose measurements made at the most recent quality assurance measurements.

[0222] Figure 25 shows a schematic view of a system according to an embodiment of the present disclosure. The system comprises a detector 2501 configured to obtain quality assurance measurements of the medical linear accelerator, a processor 2502 in communication with the detector 2501 , an input device 2503 15 for receiving user commands and configuration data, a display 2504 for presenting information to a user, a beam interface 2505 for communicating with the medical linear accelerator or treatment planning system, and memory 2506 for storing computer program instructions and data. The processor 2502 receives the quality assurance measurements from the detector 2501, 20 compares the measurements to base beam profile measurements stored in memory 2506, and calculates adjustments to the photon beam model parameters when the quality assurance measurements differ from the base measurements. The processor 2502 may output the calculated adjustments to the display 2504 for review by a user or transmit the adjustments via the beam interface 2505 to the medical linear accelerator or treatment planning system for automatic implementation.

[0223] More generally, in this specification, the term “processor” is intended to broadly encompass any type of device or combination of devices capable of performing 30 the functions described herein, including (without limitation) other types of microprocessors, microcontrollers, other integrated circuits, other types of circuits or combinations of circuits, logic gates or gate arrays, or programmable devices of any sort, for example, either alone or in combination with other such

[0224] 3 193937 KLW 30722816 1devices located at the same location or remotely from each other. Additional types of processor(s) will be apparent to those ordinarily skilled in the art upon review of this specification, and substitution of any such other types of processor(s) is considered not to depart from the scope of the present 5 disclosure as defined herein. In various embodiments, the processor 2502 can be implemented as a single-chip, multiple chips and / or other electrical components including one or more integrated circuits and printed circuit boards.

[0225] Computer code comprising instructions for the processor(s) to carry out the various embodiments, aspects, features, etc. of the present disclosure may reside in the memory 2506, within the processor 2502 or be obtained from outside the system. The code may be broken into separate routines, products, etc. to carry forth specific steps disclosed herein. In various embodiments, the processor 2502 can be implemented as a single-chip, multiple chips and / or 15 other electrical components including one or more integrated circuits and printed circuit boards. The processor 2502 together with a suitable operating system may operate to execute instructions in the form of computer code and produce and use data. By way of example and not by way of limitation, the operating system may be Windows-based, Mac-based, or Unix or Linux-based, 20 among other suitable operating systems. Operating systems are generally well known and will not be described in further detail here. In particular, the processor 2502 may include functions of voice and text recognition and natural language processing. Optionally, the processor 2502 may include adaptive algorithms including by way of non-limiting example, machine learning algorithms programmed to communicate with an ability to interpret answers from users and provide further analysis and interpretation thereof. Such algorithms may include regression analysis, natural language processing, or any other machine learning methods as are available.

[0226] 30 Memory 2506 may include various tangible, non-transitory computer-readable media including Read-Only Memory (ROM) and / or Random-Access Memory (RAM). As is well known in the art, ROM acts to transfer data and instructions uni-directionally to the processor 2502, and RAM is used typically to transfer

[0227] 3 193937 KLW 30722816 1data and instructions in a bi-directional manner. In the various embodiments disclosed herein, RAM includes computer program instructions that when executed by the processor 2502 cause the processor 2502 to execute the program instructions described in greater detail below. More generally, the term 5 “memory” as used herein encompasses one or more storage mediums and generally provides a place to store computer code (e.g., software and / or firmware) and data. It may comprise, for example, electronic, optical, magnetic, or any other storage or transmission device capable of providing the processor 2502 with program instructions. Memory 2506 may further include a floppy disk, CD-ROM, DVD, magnetic disk, memory chip, ASIC, FPGA, EEPROM, EPROM, flash memory, optical media, or any other suitable memory from which processor 2502 can read instructions in computer programming languages.

[0228] The processor 2502 may also interface with the medical linear accelerator via 15 the beam interface 2505 to provide adjusted beam model parameter values thereto as set out in the present method. Alternatively, the processor 2502 may present to a user, on the display 2504, the amended values for manual input.

[0229] Those skilled in the art will recognize that many other measurement methods 20 can be used to obtain the information acquired in Figure 18. For instance, direct fluence measurements using an Electronic Portal Imaging Device (EPID) can also provide beam profile information needed to complete the currently disclosed procedure. Other devices such as linear arrays used to measure beam profiles may also be used. Furthermore, the arrangement shown in Figure 7, in combination with profile measurements, can be used to characterize the dose distribution that the linear accelerator makes based on the beam specification of the day of the quality assurance measurements.

[0230] With an updated beam model that matches the most recent quality assurance 30 measurements, the method can be used to calculate the dose in a water tank in conjunction with the processor based on a partial set of measurements, and fully characterize the photon beam, from small beams, for instance 1 cm beam width, up to the largest beams the linac can produce. Thus, the method can be

[0231] 3 193937 KLW 30722816 1used to simulate water tank measurement, eliminating the need for a water tank since the same dose distributions may be obtained as those that would have been obtained using a water tank and a complete set of scans. As well, any further dose calculations for a specific patient radiation treatment plan using 5 this new beam model will then reflect the actual dose distribution that the linear accelerator makes based on the beam specification of the day of the quality assurance measurements.

[0232] By way of non-limiting example, the partial beam measurements can be a depth dose measurement for a 10 x 10 beam, in plane and transverse profiles for 40 cm field sizes at the depth of maximum dose, head scatter factors, and in air measurements described in Figure 9. However, alternate partial measurement may also exist that can achieve the purpose described here, which can include routine quality assurance measurements.

[0233] 15

[0234] In the current state of the art, it is difficult to assess the cumulative effect of all beam quality discrepancies. The current disclosure improves the ability to assess the cumulative effect by providing a beam model that can be matched to any beam, including asymmetric beams, and linear accelerators with a large 20 isocenter. By using an updated beam model that matches all beam quality measurements, any re-calculations of dose on a test plan or patient plan will include the cumulative effect of all beam quality defects.

[0235] This method has several benefits over the current state of the art. In the current state of the art, to verify that a radiation treatment plan for a particular patient will be correctly delivered to the patient, a second radiation treatment plan is recalculated on a standard shaped phantom (as opposed to the patient CT). This standard phantom is then irradiated with the patient plan loaded, and the measured dose is compared to the predicted dose. In the case where the 30 comparison shows significant differences, the user can effect some change to either the medical linear accelerator or the treatment plan and repeat the process until the plan is judged to meet the clinical intention.

[0236] 3 193937 KLW 30722816 1In a separate practice of the current state of the art, the planned dose is delivered without a phantom, and the dose fluence is measured on a device such as an electronic portal imaging device (EPID). The measured fluence can be compared to the planned fluence in a similar manner.

[0237] 5

[0238] In a third practice of the current state of the art, the actual MLC positions are acquired during a trial of the patient treatment delivery, and these positions are used to compute the patient dose and compared to the treatment plan using a standard dose algorithm such as Monte Carlo methods.

[0239] To assess a radiation treatment plan for a particular patient, the current state of the art has several problems. First, they require an additional irradiation, either to a separate phantom, or to a fluence detector, orjusta trial run. Second, in the case where significant differences between the planned dose and the 15 measured dose are observed, the method is not adapted to tell the user the cause of the discrepancy. Third, in the case where significant differences between the planned dose and the measured dose are observed, the method is not able to tell the user the clinical impact of the differences. Fourth, in the instance where the patient plan is recalculated using a Monte Carlo dose 20 calculation, if the parameters used in the Monte Carlo beam model, such as the beam spectrum, do not match the actual linear accelerator operating parameters, the method will be inaccurate.

[0240] The proposed disclosure has an important benefit that a separate irradiation on a phantom is not required to assess the dose the patient will receive, which saves the user time and effort.

[0241] Second, in instances where there are significant differences observed between the treatment plan obtained using the baseline beam model and the one using 30 the beam model updated with quality assurance measurements, the user can determine what the root cause of the dose difference is. With the knowledge of the root cause of the failure, the user can make improved decisions regarding

[0242] 3 193937 KLW 30722816 1linear accelerator quality assurance process and repairs, which will result in savings to the user.

[0243] Third, at the time of the quality assurance measurement, the user can assess 5 the clinical importance / effect of the cumulative effect of all any deviations of quality assurance by calculating patient dose calculations with an updated beam model. The user can then take proactive measures to react to deviations in quality assurance including cost benefit analyses of any course of action.

[0244] Fourth, the current method incorporates the most recent linear accelerator beam quality assessment, thus ensuring that the patient dose verification is done using a beam model that is accurate and up to date with the most recent quality assurance measurements. Furthermore, there is an additional advantage in that the number of measurements needed using the current 15 method is much less than those needed for water tank measurements to fully characterize all field sizes from very small to very large.

[0245] The current disclosure improves on the state of the art by using a beam model that is based on a physical model of the linear accelerator. The current 20 disclosure uses a physical model that is derived from the machine configuration parameters such as those parameters used to configure the linear accelerator in service mode at its installation or subsequent maintenance. The physical model is calibrated so that the operating currents, voltages, powers, frequencies, or other physical properties of the machine operation can be reflected in the model and used to optimize the model accuracy. As well, materials and physical dimensions of parts that are used to construct the linear accelerator are also included in the model, which are helpful to calibrate the model against the real life linear accelerator performance. The resulting model is able to compute the electron beam energy, spectrum, position, angle of 30 incidence and current at the bremsstrahlung target. A second aspect of the improved model utilizes a bremsstrahlung model of the target that accepts as inputs electron beam energy, spectrum, position, angle of incidence and current in order to predict the primary and secondary photon beam spectrum as a

[0246] 3 193937 KLW 30722816 1function of position within the photon beam. The current disclosure employs an analytic model of the bremsstrahlung production (Reported by D. J.S. Findlay, Nuclear Instruments and Methods in Physics Research A276 (1989) pages 598-601) which is incorporated herein by reference that utilizes the electron 5 beam energy, electron beam angle of incidence, target material, and target thickness to calculate the bremsstrahlung beam fluence and energy spectrum that emerges from the target. Since the electron beam energy and orientation relative to the target are included in the beam model the model can include the effect of beam misalignment and energy changes, which produce symmetry and flatness changes to the photon beam.

[0247] The current disclosure improves the accuracy of the beam model by using a physical model that improves the separation of the primary component of the bremsstrahlung beam from the scatter component. The disclosure invention 15 uses a method like the one described by Sikora (PhD thesis, University of Bergen, 2011) which is incorporated herein by reference, where Monte Carlo calculations are used to calculate the scatter beam spectra and intensity. The scattercomponentof the bremsstrahlung beam is modelled using analytic fitting to the scatter spectrum. The current disclosure improves on Sikora's method 20 by including more beam energies that cover the entire range of beam energies used in modern radiotherapy (4 MV to 18 MV, however other beam energies could also be included using the same method) and producing an analytic fit to the spectra and intensity results which can be used to calculate the scatter component of any beam energy and any target and flattening filter materials. Critical to the improvement to Sikora is the interpolation of the analytical fit parameters to the primary spectrum calculated using the Findlay method. In this way, the TERMA due to the scatter spectrum can be calculated quickly and that is also correctly reflects the energy fluence of the scatter beam which depends on the photon beam energy.

[0248] 30

[0249] The current disclosure also improves the beam modelling process by employing a novel method to calculate the relative contribution to the total beam energy fluence from the primary photon beam and the scatter component. The current

[0250] 3 193937 KLW 30722816 1state of the art does not allow for separation of primary and scatter radiation and instead must use a large amount of input data to create a model that fits all conditions. The current disclosure uses a method where radiation beam measurements are made in air, which removes the phantom scatter source 5 since measurements are made without a phantom. By varying the distance of the detector from the target, the current disclosure can resolve and separate the primary and scatter radiation energy since the primary component will very as the inverse of the distance squared (1 / r2) but the scatter component varies differently since it originates from a source located at a different position than the primary and with a large enough size such that the scatter contribution must be calculated by integrating over the large size. The different responses of the primary and scatter radiation at the detector at different detector distances are exploited in this disclosure in order to resolve the relative contribution of primary and scatter to the total energy measure at the detector. Further, by making 15 measurements with different scatter sources, such as in the presence of a flattening filter or without a flattening filter, the scatter dose dependance on field size can be further characterized and used to improve accuracy of the primary and scatter energy separation in the beam model.

[0251] 20 The current disclosure further improves the accuracy of the beam model by adding a third component to our model, which is a model of the jaw face. The model of the jaw face produces a model of the penumbra region of the photon beam, which is dominant at small field sizes, but contributes less to the total beam model at larger field sizes. The penumbra, when defined as the distance between the 80% and 20% beam point is invariant with beam size for fixed jaws, meaning that the penumbra can be fully characterized by measurement at a single beam size, such as the largest beam size. The penumbra due to multi leaf collimators (MLC) varies with field size, but in a predicable manner since the penumbra is primarily due to the attenuation of the primary and scatter 30 radiation due to the rounded MLC leaf ends. Exploiting this, a generalized model of the MLC penumbra can be obtained using only a few penumbra measurements.

[0252] 3 193937 KLW 30722816 1The present disclosure improves the accuracy of the beam model by including the center beam position in the model. It is known that the alignment of the radiation beam varies with geometric settings of the linear accelerator, especially the gantry position, since the effect of gravity causes sag and other 5 mechanical distortions as the gantry rotates around the patient. This is quantified using a metric referred to as the radiation isocenter, which is the size of a sphere which encloses the intersection points of radiation beams from all beam angles. There are three isocenters related to medical linear accelerators: the gantry isocenter, the collimator isocenter, and the patient support couch isocenter. Current beam models do not include the effect of mechanical misalignments due to the machine isocenter, which can be up to 1 mm in size for modern medical linear accelerators. This can be a problem for very small beam sizes on the order of a few cm, where a misalignment of a mm can represent an important effect on the dose calculation.

[0253] 15

[0254] The current disclosure improves the beam modelling process by four inputs to the beam model: the primary and scatter radiation components, which are calculated based on a physical model of the linear accelerator and a process to resolve the relative contribution of each component, a model of the 20 penumbra, to accurately model small field sizes, and an alignment model to include the effect of misalignment of the beam for small fields. Because the current disclosure splits the beam model into primary and scatter components, the beam model is very accurate at small field sizes where the scatter component is reduced to zero. However, the primary component of the photon beam still contributes substantial dose at very small field sizes, but this must be modulated by the model of the penumbra, which can have a significant effect at small beam sizes. At larger field sizes, the scatter model maintains accuracy of the calculation. By including the jaw end model, a single beam model is accurate for very small field sizes (less than a cm) to large field sizes (up to 40 30 cm diameter).

[0255] The current disclosure also improves on the accuracy of the beam model by creating dose kernels that are forced to agree with Monte Carlo calculations or

[0256] 3 193937 KLW 30722816 1any beam profile measurements. The current disclosure calculates dose kernels by doing a Monte Carlo calculation in any medium and performing an inverse convolution with the TERMA that is calculated using the bremsstrahlung primary and scatter described in the current disclosure. Alternatively, the 5 inverse convolution can be performed on measured beam profiles. By utilizing the end dose Monte Carlo calculation or the measured profile, the inverse convolution (deconvolution) will calculate a dose kernel that must result in the same dose calculation as a pure Monte Carlo calculation. The medium where dose is calculated is typically water, however any medium can used to calculate dose kernels that are accurate for any materials relevant to radiotherapy planning, such as tissue materials (muscle, fat), bones, lung material or metals used in implant devices such as hip / knee implants or screws used for spinal surgery. The current disclosure also improves on the dose calculation by calculating dose kernels that account for photon beam divergence. At larger 15 field sizes, the orientation of the dose kernel with the photon beam ray coming from the target is affected by the beam divergence. Accounting for this divergence by tilting the dose kernel is computationally intensive, and results in slower dose calculations. By obtaining dose kernels by inverse convolution of the dose with TERMA, the resulting kernel will necessarily produce an average 20 kernel that is representative of the kernel tilt. By using many field sizes to produce a field size dependent dose kernel, the resulting kernels will necessarily include the effect of beam divergence and improve the accuracy of the dose calculation.

[0257] The state-of-the-art beam models are based on empirical models, so any changes in the photon beam measurements can only be implemented by complete recommissioning of the photon beam. This is a time-consuming process. The current disclosure also improves on the state of the art since the model can be easily adapted using measurements of the beam state to 30 maintain the accuracy of the beam model which other beam models are not configured to do. The method is well suited to automation since partial beam measurements, such as those obtained during routine quality assurance and machine maintenance procedures as described above can be used as inputs

[0258] 3 193937 KLW 30722816 1to a processor executing a beam matching procedure that calculates quickly and does not rely on Monte Carlo methods to calculate the beam spectrum. The processor receives the measurement inputs, performs the necessary comparisons and calculations, and outputs the resulting adjustments. The 5 adjustments may be provided to a user for manual implementation, transmitted to the medical linear accelerator or associated systems for automated machine implementation, or automatically applied by the processor to update the beam model without user intervention. Measurements of beam quality, such as beam symmetry, and flatness can be used to determine the root cause changes of the linear accelerator configuration that results in the observed beam quality. Because the current disclosure uses a physical beam model instead of an empirical beam model, the root cause changes can be incorporated into a beam model variation that will accurately predict the observed beam quality. This improvement allows radiotherapy processes to include feedback from routine 15 quality assurance measurements to update beam models so that it continually reflects ongoing measurements of the beam model properties that are done in routine quality assurance measurements, for instance.

[0259] Since the beam model is a physical beam model, updating the beam model 20 based on quality assurance data for a reduced or partial measurement data set (not a complete set of measurements for all field sizes and measurements depths) can be used to calculate the complete beam data set that would otherwise be measured in a full water tank measurement scenario. The accuracy of the calculations will be determined by the accuracy of the physics incorporated into the beam model. By utilizing a complete set of physical models, any desired accuracy can be achieved by adding physical model complexity.

[0260] The current disclosure also includes an improved method for verifying treatment 30 doses received by patients. Since the current disclosure allows for the beam model to be updated on a routine basis using routine partial beam data measurements, calculations for radiotherapy treatment plans using the most up to date beam model will more accurately reflect the actual dose received by a

[0261] 3 193937 KLW 30722816 1patient from radiotherapy at their treatment since the beam model reflects the state of the radiotherapy beam at the latest quality assurance measurement, which can be arbitrarily close to the patient treatment.

[0262] 5 The current disclosure also provides an improved method for maintaining the radiation beam model against linear accelerator performance variation by using partial quality assurance measurements instead of relying on a complete set of measurements performed at all field sizes and measurement depths.

[0263] 10 While specific embodiments have been described and illustrated, such embodiments should be considered illustrative only and not as limiting the disclosure as construed in accordance with the accompanying claims.

[0264] 15

[0265] 3 193937 KLW 30722816 1

Claims

What is claimed is:

1. A method of maintaining a photon beam model for radiotherapy patient dose calculations, the method comprising:constructing an initial photon beam model based on physical models of a medical linear accelerator beam formation process, wherein the initial photon beam model comprises physical models including components of the medical linear accelerator and model parameters matched to base beam profile measurements obtained at machine installation;obtaining, by a processor, quality assurance measurements of the medical linear accelerator after the medical linear accelerator enters clinical service;comparing, by the processor, the quality assurance measurements to the base beam profile measurements;adjusting, by the processor, the model parameters of the photon beam model to match the quality assurance measurements when the quality assurance measurements differ from the base beam profile measurements, thereby producing an updated photon beam model; andutilizing the updated photon beam model for radiotherapy patient dose calculations.

2. The method of claim 1, wherein the adjusting step further comprises automatically updating the primary photon beam spectra used in the photon beam model by accounting for changes in the physical configuration of the medical linear accelerator which affect the electron beam energy and current striking a bremsstrahlung target.

3. The method of claim 1, wherein the photon beam model comprises a scatter model that selects a scatter spectrum derived from a primary spectrum matched to a bremsstrahlung beam calculated using the physical models.

4. The method of claim 1, wherein the physical models are configured to calculate quickly to automate the photon beam modelling process and produce the updated photon beam model automatically when the quality assurance measurements are obtained, without requiring Monte Carlo recalculation.

5. The method of claim 1, further comprising calculating a patient-specific radiation treatment plan dose using the updated photon beam model, wherein the updated photon beam model reflects an actual dose distribution that the medical linear accelerator produces on a day of the quality assurance measurements, and wherein patient dose on a day of treatment is predicted using the most recent quality assurance measurements.

6. The method of claim 1 , wherein utilising the updated photon beam model comprises performing patient-specific quality assurance verification without requiring a separate radiation beam phantom, and verifying commissioning measurements by determining whether the initial photon beam model is consistent with the commissioning measurements.

7. The method of claim 1, further comprising determining a root cause of deviations between the quality assurance measurements and the base beam profile measurements based on the physical models, and conducting a costbenefit analysis for intervention required to correct out-of-tolerance quality assurance measurements.

8. The method of claim 1, wherein constructing the initial photon beam model comprises simulating photon beam characteristics of the medical linear accelerator based on physical principles of a beam forming process, and wherein the physical models do not depend on empirical models.

9. The method of claim 1, wherein the quality assurance measurements comprise routine measurements including at least one of beam profiles, depth doses, beam symmetry, beam flatness, isocenter measurements, x-ray beam quality, and physical characteristics of beam generation components including currents and voltages.

10. The method of claim 1, wherein the physical models utilise linear accelerator configuration parameters including service mode parameters.

11. A method of determining a dose kernel for radiotherapy dose calculations, the method comprising:obtaining a dose distribution for a photon beam of a medical linear accelerator in a medium;calculating a total energy released per unit mass (TERMA) distribution for the photon beam, wherein the TERMA distribution comprises a primary energy fluence component and a secondary scatter energy fluence component;performing a deconvolution of the dose distribution with the TERMA distribution to obtain the dose kernel; andstoring the dose kernel for use in radiotherapy dose calculations, wherein the dose kernel, when convolved with the TERMA distribution, produces the dose distribution.

12. The method of claim 1, wherein the updated photon beam model is capable of modelling any beam profile including asymmetric beam profiles, and predicts results of partial beam maintenance procedures including beam steering adjustments and jaw calibration adjustments.

13. The method of claim 1 , wherein step (a) comprises constructing the initial photon beam model using a partial set of beam measurements, wherein the physical models enable creation of a complete beam model characterising beam properties for all field sizes and measurement depths, and predicting any measurement that would be performed in a physical water tank.

14. The method of claim 1, wherein the updated photon beam model predicts dose properties of radiation beams with small beam diameters based on partial measurements made at large field sizes.

15. The method of claim 1, further comprising using the updated photon beam model to generate a virtual water tank simulation having equivalent measurement capabilities to a physical water tank, wherein the virtual water tank simulation replicates measurements of small beam diameters based on partial measurements made at large field sizes, and is quickly updated using routine quality assurance measurements without requiring complete recommissioning.