JP2011212418A - Corpuscular beam irradiation instrument and corpuscular beam therapeutic instrument - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a corpuscular beam therapeutic instrument which realizes accurate irradiation of beams in raster scanning and hybrid scanning by eliminating the influence of hysteresis of a scanning electromagnet.SOLUTION: The corpuscular beam irradiation instrument has a scanning power source 4 outputting exciting current of a scanning electromagnet 3, and an irradiation controller 5 controlling the scanning power source 4. The irradiation controller 5 has a scanning electromagnet command value learning generator 37 having a mathematic model 60 generating a command value of the exciting current, evaluating the result of a run-through as a series of irradiation actions by the command value of the exiting current output to the scanning power source 4, updating the mathematic model 60 on the basis of the result of evaluation, accumulating the experiences of the run-through and outputting the command value of the exiting current generated by the mathematic model 60 on the basis of the accumulated run-through experiences to the scanning power source 4.

Description

  The present invention relates to a particle beam therapy apparatus used for medical use and research, and more particularly to a scanning particle beam irradiation apparatus and a particle beam therapy apparatus such as raster scanning.

  In general, a particle beam therapy system is connected to a beam generator for generating a charged particle beam, an accelerator for accelerating the generated charged particle beam, and a charge emitted after being accelerated to energy set by the accelerator. A beam transport system that transports a particle beam, and a particle beam irradiation device that is installed downstream of the beam transport system and that irradiates a target with a charged particle beam. The particle beam irradiation device is large, so that the charged particle beam is scattered and expanded by a scatterer, and the expanded charged particle beam is matched to the shape of the irradiation target to form an irradiation field, and to match the shape of the irradiation target There are scanning irradiation methods (spot scanning, raster scanning, etc.) in which an irradiation field is formed by scanning a thin pencil beam.

  In simple terms, spot scanning is a method of forming an irradiation field by irradiating a particle beam in a small spot shape like point drawing. That is, beam supply (spotting), beam stop, and movement are repeated. This is an irradiation method with a high degree of freedom in which the irradiation dose can be changed for each spot position, and has attracted much attention in recent years.

  In short, raster scanning is a method of forming an irradiation field by continuously irradiating a particle beam as in a single stroke. That is, it is a method of moving a region where the target dose is constant at a constant speed while continuing to irradiate the beam. Since it is not necessary to repeatedly supply / stop the beam, there is an advantage that the treatment time can be shortened.

  An irradiation method intermediate between spot scanning and raster scanning has also been proposed. The beam continues to irradiate like raster scanning, and the beam irradiation position moves one after another between spot positions like spot scanning. It is intended to incorporate the advantages of both spot scanning and raster scanning. In this specification, this intermediate irradiation method will be referred to as hybrid scanning.

  In the broad irradiation method, an irradiation field that matches the shape of the affected area is formed using a collimator or a bolus. An irradiation field that matches the shape of the affected area is formed to prevent unnecessary irradiation of normal tissue. This is the most widely used irradiation method. However, it is necessary to manufacture a bolus for each patient or to deform the collimator according to the affected area.

  On the other hand, the scanning irradiation method is a highly flexible irradiation method in which a collimator and a bolus are unnecessary. However, since these components that prevent irradiation of normal tissues other than the affected part are not used, higher beam irradiation position accuracy than that of the broad irradiation method is required.

  In the particle beam therapy system, various inventions for increasing the accuracy of the irradiation position and the irradiation dose have been made. Patent Document 1 discloses the following invention for the purpose of providing a particle beam therapy apparatus capable of accurately irradiating an affected area. In the invention of Patent Document 1, the scanning amount of the charged particle beam by the scanning device and the beam position of the charged particle beam detected by the beam position detector at that time are stored in a storage device, and the stored scanning amount and beam position are stored. The scanning amount of the scanning device is set according to the beam position based on the treatment plan information by the control device. Since the relationship between the scanning amount obtained by actual irradiation and the beam position is stored in the storage device, it can be expected that the affected area is irradiated accurately.

  Patent Document 2 discloses the following invention for the purpose of providing a particle therapy apparatus that ensures high safety and can irradiate a charged particle beam with high accuracy. The invention of Patent Document 2 supplies a charged particle beam emitted from a charged particle beam generator to a scanning electromagnet that scans an irradiation surface perpendicular to the beam traveling direction, and irradiates the charged particle beam that has passed through the scanning electromagnet. Based on the position on the surface and the dose, the emission amount of the charged particle beam from the charged particle beam generator is controlled. Specifically, among a plurality of regions formed by dividing on the irradiation surface, the supply of the charged particle beam to the region that has reached the target dose is stopped, and the charged particles are applied to other regions that have not reached the target dose. Supply the beam. In this way, high safety is expected by comparing the irradiation dose in each region with the target dose and performing ON / OFF control (supply / stop) of the emission amount of the charged particle beam.

  Patent Document 3 discloses the following invention with respect to the problem that the hysteresis characteristic between the current of the scanning electromagnet and the magnetic field reduces the accuracy of the beam irradiation position. The invention of Japanese Patent Application Laid-Open No. H10-228561 is a first calculator that calculates the current value of the scanning electromagnet that does not consider the influence of hysteresis corresponding to the beam irradiation position based on the irradiation plan, and the scanning electromagnet calculated by the first calculator. And a second calculation means for correcting the current value in consideration of the influence of hysteresis, and the irradiation control device controls the current of the scanning electromagnet based on the calculation result of the second calculation means. Thus, by performing the correction calculation so as to eliminate the influence of the hysteresis on the second calculation means, that is, by having a mathematical model representing the hysteresis characteristics in the second calculation means, the beam irradiation position can be calculated by calculation. The improvement of accuracy is expected.

JP 2005-296162 A JP 2008-272139 A JP 2007-132902 A

  In the invention disclosed in Patent Document 1, a conversion table is created based on actual data of the scanning amount and beam position of a charged particle beam obtained by actual irradiation, and a scanning electromagnet is created using this conversion table. The set current value is calculated.

  However, actually, as shown in Patent Document 3, there is a hysteresis characteristic between the current of the scanning electromagnet and the magnetic field, and the current value decreases when the current value increases. Sometimes it is a different magnetic field. That is, even if the current value of the scanning electromagnet at a certain moment is known, the exact value of the magnetic field cannot be specified only from the information. Therefore, the invention disclosed in Patent Document 1 has a problem that the affected area cannot be accurately irradiated due to the influence of hysteresis of the electromagnet.

  In the invention disclosed in Patent Document 2, the emission amount of the charged particle beam is ON / OFF controlled (supplied / stopped) so that the irradiation dose in each defined region becomes the target dose.

However, a plurality of regions formed by being divided on the irradiation surface described in the invention disclosed in Patent Document 2 are regions (excitations) in an excitation current space defined by the excitation current range of the corresponding scanning electromagnet. Region) and does not match the region (irradiation region) in the actual irradiation space. This is because unless the hysteresis of the scanning electromagnet is taken into consideration, the excitation area and the irradiation area do not exactly correspond to each other. Therefore, even in devices and methods that attempt to increase the safety by managing the irradiation dose in units of excitation areas in this way, the effect of managing the dose in a small area is demonstrated unless the influence of the hysteresis of the scanning electromagnet is excluded. Can not. That is, there is a problem that the accuracy of the beam irradiation position is deteriorated due to hysteresis of the scanning electromagnet.

  In the invention disclosed in Patent Document 3, a mathematical model of hysteresis is created inside the calculation means, and the current value of the scanning electromagnet is corrected by calculation.

  However, even if hysteresis is taken into account, there are still some problems with the method of consideration as in the invention of Patent Document 3. The first problem is that it is actually quite difficult to correct the hysteresis characteristic with high accuracy using an arithmetic method. For example, the curve representing the hysteresis characteristics between the current and the magnetic field has various modes depending on not only the amplitude of the input (current) but also the speed of changing the input (current) and the pattern to be changed. Representing this complex hysteresis phenomenon by an arithmetic method, that is, a mathematical model, has been devised for many years in many fields, but it is still difficult in practice. The second problem lies in the method for detecting the beam irradiation position. In many of the conventional techniques, as in the invention disclosed in Patent Document 3, the beam irradiation position is detected only by one or a plurality of beam position monitors. The beam position monitor knows the beam irradiation position only after the charged particle beam is irradiated. Therefore, when the beam deviates from the target and irradiates normal tissue etc., the beam can only be stopped, and the beam irradiation position cannot be controlled to the correct irradiation position that should have been originally irradiated. There was a point.

  The present invention has been made to solve the above-described problems, and obtains a particle beam irradiation apparatus that eliminates the influence of hysteresis of a scanning electromagnet and realizes high-accuracy beam irradiation in raster scanning or hybrid scanning. With the goal.

  A scanning power source that outputs an excitation current of the scanning electromagnet and an irradiation control device that controls the scanning power source are provided. The irradiation control device has a mathematical model for generating a command value of the excitation current, evaluates the result of the run-through that is a series of irradiation operations by the command value of the excitation current output to the scanning power supply, and based on the result of the evaluation A scanning electromagnet command value learning generator is provided that updates the mathematical model, accumulates the experience of run-through, and outputs the command value of the excitation current generated by the mathematical model based on the accumulated experience of run-through to the scanning power supply.

  The particle beam irradiation apparatus according to the present invention evaluates the run-through result, accumulates the experience of run-through, and outputs the command value of the excitation current generated by the mathematical model based on the accumulated experience of the run-through to the scanning power supply. Therefore, it is possible to eliminate the influence of the hysteresis of the scanning electromagnet and realize high-accuracy beam irradiation in raster scanning or hybrid scanning.

It is a schematic block diagram of the particle beam therapy system in Embodiment 1 of this invention. It is a block diagram of the irradiation control apparatus of FIG. It is a block diagram of the other irradiation control apparatus of FIG. It is a figure which shows the several area | region defined in the magnetic field space. It is a figure which shows the example of the score table | surface in the case of learning irradiation. It is a block diagram of the irradiation control apparatus in Embodiment 2 of this invention. It is a block diagram of the other irradiation control apparatus in Embodiment 2 of this invention. It is an example of the mathematical model which produces | generates the command electric current in Embodiment 3 of this invention. It is a block diagram of the irradiation control apparatus in Embodiment 3 of this invention.

Embodiment 1 FIG.
FIG. 1 is a schematic configuration diagram of a particle beam therapy system according to Embodiment 1 of the present invention. The particle beam treatment apparatus includes a beam generation device 51, an accelerator 52, a beam transport device 53, a particle beam irradiation device 54, a treatment planning device 55, and a data server 56. The treatment planning device 55 is not considered as a component part of the particle beam treatment device, and can be prepared independently. The beam generator 51 generates charged particle beams by accelerating charged particles generated by an ion source. The accelerator 52 is connected to the beam generator 51 and accelerates the generated charged particle beam. The beam transport device 53 transports the charged particle beam emitted after being accelerated to the energy set by the accelerator 52. The particle beam irradiation device 54 is installed downstream of the beam transport system 53 and irradiates the irradiation target 15 with a charged particle beam.

The treatment planning device 55 can make several treatment plans such as irradiation conditions based on the three-dimensional data of the affected part 15 that is an irradiation target, and simulate particle beam treatment. The treatment plan finally selected by the doctor performing the particle beam therapy is converted into a code for driving the particle beam therapy apparatus. For example, in the case of spot scanning, the driving code is the coordinates X _j and Y _j of the spot for each irradiation layer (layer) Z _i (subscript i is the layer number) (subscript j is a spot). Number) and the target dose D _j to be irradiated to the spot, and in the case of raster scanning or the like, the irradiation trajectory for each irradiation layer (layer) Z _{i is} expressed as the irradiation position X _k , Y _k for each sampling period. Time series data represented by (subscript k is a sequence number). Here, the target dose to be irradiated to the spot is because the upper layer is affected by the lower layer irradiation due to the characteristics of the Bragg peak, and is not the target dose necessary for the location. The data server 56 stores treatment plan data generated for each patient by the treatment plan device 55 and a code for driving.

  The particle beam irradiation device 54 includes a beam transport duct 2 that transports the incident charged particle beam 1a incident from the beam transport device 53, and incident charged particle beams in the X direction and the Y direction that are perpendicular to the incident charged particle beam 1a. Scanning electromagnets 3a and 3b that scan 1a, magnetic field sensors 20a and 20b that detect magnetic fields generated by the scanning electromagnets 3a and 3b, a magnetic field data converter 21, a beam position monitor 7, and a position data converter 8, A dose monitor 11, a dose data converter 12, an irradiation control device 5, and a scanning power supply 4 are provided. The magnetic field sensors 20a and 20b are magnetic field sensors having a pickup coil, for example. As shown in FIG. 1, the traveling direction of the incident charged particle beam 1a is the Z direction.

  The scanning electromagnet 3a is an X-direction scanning electromagnet that scans the incident charged particle beam 1a in the X direction, and the scanning electromagnet 3b is a Y-direction scanning electromagnet that scans the incident charged particle beam 1a in the Y direction. The magnetic field sensor 20a is an X-direction magnetic field sensor that detects a magnetic field in the X direction, and the magnetic field sensor 20b is a Y-direction magnetic field sensor that detects a magnetic field in the Y direction. The magnetic field data converter 21 converts the electrical signal of the sensor representing the magnetic field detected by the magnetic field sensors 20a and 20b into a measured magnetic field Bs of digital data. The beam position monitor 7 detects the passing position of the outgoing charged particle beam 1b deflected by the scanning electromagnets 3a and 3b. The position data converter 8 calculates the irradiation position on the irradiation layer (layer) from the electrical signal of the sensor representing the passing position detected by the beam position monitor 7, and generates the measurement position coordinate Ps of the digital data. The dose monitor 11 detects the dose of the outgoing charged particle beam 1b. The dose data converter 12 converts the electrical signal of the sensor representing the dose detected by the dose monitor 11 into a measured dose Ds of digital data.

The irradiation control device 5 outputs command currents Ix _k and Iy _k which are command values of excitation currents to the scanning power supply 4, and controls the irradiation position in each irradiation layer (layer) Z _i . The scanning power supply 4 outputs an excitation current that is actually driven to the scanning electromagnets 3a and 3b based on the command currents Ix _k and Iy _k output from the irradiation control device 5.

FIG. 2 is a configuration diagram of the irradiation control device 5. The irradiation control device 5 includes an inverse mapping generator 30, an inverse mapping calculator 22, a command value output unit 25, a command evaluator 33, a command updater 34, a scanning electromagnet command value series generator 35, a beam A supply command output unit 26. The scanning electromagnet command value series generator 35 has a command value storage device 36. The command value output unit 25, the command evaluator 33, the command updater 34, and the scanning electromagnet command value series generator 35 constitute a scanning electromagnet command value learning generator 37. The scanning electromagnet command value learning generator 37 evaluates the result of the run-through that is a series of irradiation operations by the command current I _k = (Ix _k , Iy _k ) output to the scanning power supply 4, and the evaluation result is a predetermined condition. the command current if not satisfied _{I k = (Ix k, Iy} k) is executed run through updates the evaluation of the command current result satisfies a predetermined condition _{I k = (Ix k, Iy} k) scanning Output to power supply 4.

  The operation of the irradiation control device 5 will be described. The irradiation of the particle beam therapy system can be broadly divided into test irradiation during calibration and main irradiation during treatment. In general, test irradiation at the time of calibration is irradiation for so-called calibration, and is performed only when calibration is necessary without a patient. Test irradiation is performed with various values of the control input (current Ix) to the X direction scanning electromagnet 3a and the control input (current Iy) to the Y direction scanning electromagnet 3b, and the beam irradiation position at that time is measured. The trial irradiation at the time of calibration in the first embodiment is performed in the same manner as in the prior art, but the charged particle beam is supplied and stopped in a spot scanning manner, and at the time of the trial irradiation, the measurement position coordinates Ps (xs, ys) of the beam are measured. In addition, the magnetic field sensors 20a and 20b measure the measurement magnetic field Bs (Bxs, Bys). By performing the trial irradiation in a spot scanning manner, the measurement position coordinates Ps (xs, ys) of the beam can be clearly measured. The relationship between the measurement magnetic field Bs (Bxs, Bys) of the scanning electromagnet 3 and the beam measurement position coordinates Ps (xs, ys) at this time is a mathematical model of the inverse mapping calculator 22 generated by the inverse mapping generator 30. Realize.

A command current I _l = (Ix _l , Iy _l ) (subscript l represents the spot number of trial irradiation) is prepared in advance for trial irradiation (step S001). The command current I _l = (Ix _l , Iy _l ) for trial irradiation is prepared so that the irradiated spot is irradiated within the irradiation range assumed by the particle beam irradiation apparatus 54. Command value output unit 25 outputs command current I ₁ = (Ix ₁ , Iy ₁ ) to scanning power supply 4 (step S002). The scanning power supply 4 controls the scanning electromagnet 3 in accordance with the command current I ₁ = (Ix ₁ , Iy ₁ ) (Step S003).

The beam supply command output unit 26, in synchronization with the command value output unit 25, waits for a settling time sufficient for the magnetic fields of the scanning electromagnets 3a and 3b to settle, and then supplies the beam generation unit 51 with a beam supply command. The command S _start is output. The beam generator 51 starts irradiation with a charged particle beam. After irradiation time T _on course necessary for trial irradiation, and outputs the beam stop command S _stop instructing to stop the beam on the beam generator 51, a beam generating device 51 stops the irradiation of the charged particle beam.

The magnetic fields of the scanning electromagnets 3a and 3b controlled by the command current I ₁ = (Ix ₁ , Iy ₁ ) are measured by the magnetic field sensors 20a and 20b. The measured magnetic field B ₁ = (Bx ₁ , By ₁ ) measured for each trial irradiation spot is input to the inverse mapping generator 30 via the magnetic field data converter 21.

The irradiation position coordinates P ₁ = (X ₁ , Y ₁ ) of the outgoing charged particle beam 1 b scanned by the scanning electromagnet 3 for each spot are calculated by the beam position monitor 7. The irradiation position coordinates P ₁ = (X ₁ , Y ₁ ) are input to the inverse mapping generation period 30 via the position data converter 8.

The inverse mapping generator 30 stores the measurement magnetic field B _l = (Bx _l , By _l ) and irradiation position coordinates P _l = (X _l , Y _l ) of all spots in a memory which is a built-in storage device ( Step S004).

The inverse mapping generator 30 creates a mathematical model based on the stored measurement magnetic field B ₁ = (Bx ₁ , By ₁ ) and irradiation position coordinates P ₁ = (X ₁ , Y ₁ ), and reverses the created mathematical model. The data is stored in the mapping calculator 22 (step S005).

  The mathematical model of the inverse mapping calculator 22 is realized using a polynomial as a preferred example. Unlike the conventional conversion table, the reverse mapping calculator 22 will be described. Under the assumption that the specifications of the scanning electromagnet 3, the specifications of the scanning power supply 4, and the irradiation beam specifications (irradiation energy, incident beam position, etc.) are constant, the magnetic field B (Bx, By) of the scanning electromagnet 3 is determined. For example, since the beam irradiation position coordinate P (x, y) is uniquely determined, a physical phenomenon related to the relationship between the magnetic field B and the beam irradiation position coordinate P can be regarded as a two-input two-output normal map.

However, at the time of the main irradiation in the treatment, the target irradiation position coordinates P _obj = (Px _obj , Py _obj ) of the beam are given first, and the target irradiation position coordinates P _obj = (Px _obj , Py _obj ) of the beam. The magnetic field B (Bx, By) of the scanning electromagnet 3 must be controlled so as to realize the above. That is, in the main irradiation in the treatment, the magnetic field of the scanning electromagnet 3 is realized so as to realize the target irradiation position coordinate P _obj = (Px _obj , Py _obj ) from the target irradiation position coordinate P _obj = (Px _obj , Py _obj ) of the beam. Estimated value B _est = (▲ Bx ▼, ▲ By ▼) of B (Bx, By) must be calculated (for the explanation of (Bx) and (By), see equations (1) and (2)) . Thus, it should be noted that this irradiation requires a reverse mapping from the position to the magnetic field. Therefore, the inverse mapping calculator 22 is required to obtain the estimated value B _est of the magnetic field B (Bx, By).

  An outline of a method for realizing the mathematical model of the inverse mapping calculator 22 by a polynomial will be described. Here, the polynomial is a polynomial generally defined in mathematics, and is defined as “an expression consisting only of the sum and product of constants and indefinite elements”. Specifically, for example, it is as shown in the following mathematical formula.

ここで、ｍ_００、ｍ_０１、ｍ_０２、ｍ_１０、ｍ_１１、ｍ_２０、ｎ_００、ｎ_０１、ｎ_０２、ｎ_１０、ｎ_１１、ｎ_２０は未知パラメータ定数である。また、Ｐｘ_ｏｂｊ、Ｐｙ_ｏｂｊは多項式の不定元に相当する。また、数式（１）の左辺（Ｂの上に小⌒が付いたＢｘ、これを▲Ｂｘ▼で示す。）はＢｘの推定値であり、数式（２）の左辺（Ｂの上に小⌒が付いたＢｙ、これを▲Ｂｙ▼で示す。）はＢｙの推定値であることを示す。磁場Ｂ（Ｂｘ，Ｂｙ）の推定値Ｂ_ｅｓｔは（▲Ｂｘ▼，▲Ｂｙ▼）である。

_{Here, m 00, m 01, m} 02, m 10, m 11, m 20, n 00, n 01, n 02, n 10, n 11, n 20 is an unknown parameter constant. Px _obj and Py _obj correspond to indefinite _elements of the polynomial. The left side of equation (1) (Bx with a gavel on B, which is indicated by Bx) is the estimated value of Bx, and the left side of equation (2) (the gavel on B) "By", which is indicated by (By), indicates an estimated value of By. The estimated value B _est of the magnetic field B (Bx, By) is (Bx, By).

The unknown parameter constants of the polynomial are obtained by the least square method or the like based on the measurement magnetic field B ₁ = (Bx ₁ , By ₁ ) and the irradiation position coordinates P ₁ = (X ₁ , Y ₁ ) for trial irradiation.

The estimated value B _est of the magnetic field B = (Bx, By) that realizes the target irradiation position coordinates P _obj = (Px _obj , Py _obj ) is obtained by Expressions (1) and (2) substituting the obtained unknown parameter constants. Ask.

In the prior art, the relationship between the control input (command current I ₁ = (Ix ₁ , Iy ₁ )) to the scanning electromagnet 3 for calibration and the irradiation position coordinates P ₁ = (X ₁ , Y ₁ ) of the beam, It was a method of creating a conversion table and storing this conversion table in the scanning electromagnet command value generator 6.

That is, the control input (command current Ix _l ) to the X-direction scanning electromagnet 3a is from the x coordinate (Px _obj ) of the target irradiation position P _obj of the beam, and the control input (command current Iy _l ) to the Y-direction scanning electromagnet 3b is It was obtained independently from the y coordinate (Py _obj ) of the target irradiation position P _obj of the beam.

However, actually, the control input (command current Ix ₁ ) to the X-direction scanning electromagnet 3a is the x-coordinate and y-coordinate of the beam irradiation position P, and the control input (command current to the Y-direction scanning electromagnet 3b). Iy _l ) also affects both the x-coordinate and y-coordinate of the beam irradiation position P, that is, there is an interference term, so that the irradiation position accuracy deteriorates in the method using the conversion table obtained independently.

The particle beam irradiation apparatus 54 according to the first embodiment obtains the interference term by calculating the estimated value B _est of the magnetic field B = (Bx, By) for realizing the target irradiation position coordinate P _obj = (Px _obj , Py _obj ). Since the mathematical model considered is realized in the inverse mapping calculator 22, unlike the conventional case, the irradiation position accuracy of the outgoing charged particle beam 1 b can be improved.

  Next, the main irradiation in the particle beam therapy system according to the first embodiment will be described. The main irradiation is divided into learning irradiation for optimizing control of the beam irradiation position and dose, and therapeutic irradiation for irradiating the irradiation target 15 of the patient with the beam. Learning irradiation is performed according to the following procedure.

Of the treatment plans created by the treatment planning device 55 for a certain irradiation object 15, the one finally selected by the doctor is converted into a code for driving the particle beam treatment device, and the irradiation control device 5 Sent (step S101). Here, the learning irradiation and the treatment irradiation are assumed to be raster scanning, and the driving code is the irradiation trajectory for each irradiation layer (layer) Z _i (subscript i is the layer number), and the irradiation position P _k for each sampling period. In the following description, it is assumed that the time series data is represented by (X _k , Y _k ) (subscript k is a sequence number).

The command current I _k = (Ix _k , Iy _k ) (the subscript k is a sequence number) for learning irradiation is created by the method described later (step S102). In the learning irradiation, the run-through is performed according to the command current I _k = (Ix _k , Iy _k ) for the learning irradiation in a state where there is no patient.

The command current I _k = (Ix _k , Iy _k ) for the learning irradiation may be anything as long as it is a candidate or an initial value of learning. Here, an initial value is obtained by a method according to the prior art. Command current for learning ^{_{irradiation, a I k = (a Ix}} k, a Iy k) ( where, a is the number of learning, if the initial value a = 0 to.) Represents a, commanded by learning Specify that the current is updated.

Further, the scanning electromagnet command value series generator 35 stores the command current ^a I _k = (
^a Ix _k, and stores ^a Iy _k). The command current ^a I _k = ( ^a Ix _k , ^a Iy _k ) is a control input for each sampling period, where the subscript k represents the sequence number. Run-through of the initial command current ⁰ I _k = ( ⁰ Ix _k , ⁰ Iy _k ) is performed in the absence of a patient (step S103).

The irradiation start signal St generated by the signal generator 29 is sent to the beam supply command output unit 26 and the scanning electromagnet command value series generator 35 in response to the learning irradiation start instruction from the operator of the particle beam therapy system. The scanning electromagnet command value series generator 35 outputs the command current ⁰ I _k = ( ⁰ Ix _k , ⁰ Iy _k ) for the ^0th learning in the order of the sequence number for each sampling period.

The beam supply command output unit 26 receives the main irradiation start signal St and outputs a beam supply command S _start for generating a beam to the beam generator 51. The beam generator 51 starts irradiation with a charged particle beam.

The passage position of the outgoing charged particle beam 1 b is detected by the beam position monitor 7, and ^a P _k that is the measurement position coordinate Ps calculated by the position data converter 8 is input to the command evaluator 33. The command evaluator 33 compares the irradiation position P _k (X _k , Y _k ), which is the target irradiation position, with ^a P _k ( ^a X _k , ^a Y _k ), which is the measurement position coordinate Ps, and the initial command current ⁰ The run-through scoring of I _k = ( ⁰ Ix _k , ⁰ Iy _k ) is performed (step S104). The run-through scoring method will be described later.

The signal generator 29 transmits a main irradiation end signal Se at the time when the main irradiation ends. The time at which the main irradiation is completed is a time elapsed by sampling period × k (total number of sequences) from the time at which the main irradiation is started. The beam supply command output unit 26 receives the main irradiation end signal Se and outputs a beam stop command S _stop to stop the beam to the beam generator 51. The beam generator 51 receives the beam stop command S _stop and stops the charged particle beam 1a (step S105).

Based on the scoring result that is the evaluation result output from the command evaluator 33, the command updater 34 is a partial sequence of the learning a = 0th command current ^a I _k = ( ^a Ix _k , ^a Iy _k ). Is determined as a change target, and the sequence is changed (step S106). For example, learning a = 0-th command current _{^{0 I k = (0 Ix k}} , 0 Iy k) third sequence ⁰ of _{^{I 3 = (0 Ix 3,}} 0 Iy 3), 0 from Ix ₃ ⁰ Ix ₃ + [Delta] I of To change a little.

  Next, run-through is performed again using a command current in which a part of the sequence is slightly changed in the absence of a patient (step S107). That is, step S103 to step S105 are executed as the second run-through.

It is assumed that the scoring result has changed from J point by J + ΔJ point by slightly changing a part of the sequence. Then, it can be seen that the third sequence ⁰ Ix ₃ of the command current may be updated using information of ΔJ / ΔI. Similar to a general learning function, if ΔI is positive and the scoring result is bad, updating such as making ΔI negative may be performed. This work should be performed for all sequences whose scoring results are affected. With this update, learning is counted once (a is incremented (a + 1)).

The command updater 34 creates the updated learning a = 1st command current ^a I _k = ( ^a Ix _k , ^a Iy _k ) (step S108). Confirm that the second run-through scored better than the previous run-through, and continue learning. The learning is repeated until a preset condition (such as a passing score) is satisfied. The command current ^a I _k = ( ^a Ix _k , ^a Iy _k ) that has finally been learned is stored in the command value storage device 36. In addition, if the scoring result is worse than the previous run-through as a result of learning, a device such as reducing the update speed (the amount to be updated at one time) or updating the command current as described above. do.

Treatment irradiation is performed according to the following procedure. A main irradiation start signal St is sent to the beam supply command output unit 26 and the scanning electromagnet command value series generator 35 in response to a therapeutic irradiation start instruction from the operator of the particle beam therapy system. The electromagnet command value series generator 35 outputs the command current ^a I _k = ( ^a Ix _k , ^a Iy _k ) for which learning has been completed in order of sequence number for each sampling period.

The beam supply command output unit 26 receives the main irradiation start signal St and outputs a beam supply command S _start for generating a beam to the beam generator 51. The beam generator 51 starts irradiation with a charged particle beam (step S109).

The signal generator 29 transmits a main irradiation end signal Se at the time when the main irradiation ends. The time at which the main irradiation is completed is a time elapsed by sampling period × k (total number of sequences) from the time at which the main irradiation is started. The beam supply command output unit 26 receives the main irradiation end signal Se and outputs a beam stop command S _stop to stop the beam to the beam generator 51. The beam generator 51 receives the beam stop command S _stop and stops the charged particle beam 1a (step S110).

Next, the run-through scoring method will be described. The most direct scoring method (first scoring method) is a code for driving for raster scanning, that is, an irradiation trajectory for each irradiation layer (layer) Z _{i and} an irradiation position P _k (X for each sampling period). _k , Y _k ) (subscript k is a sequence number) and irradiation positions ^a P _k ( ^a X _k , ^a Y _k ) (a The number of times of learning) is compared and the following evaluation function is considered. Learning ends when the value of this evaluation function reaches a predetermined value (when a predetermined condition is satisfied).

The run-through scoring can also be performed by the following method (second scoring method), focusing on the irradiation dose. In the run-through scoring, the target dose Di and the measured dose Ds are compared for each of a plurality of small regions defined in the magnetic field space as shown in FIG. 4, and the scoring is performed according to the score table T shown in FIG. For the evaluation of run-through, for example, an evaluation function is defined by adding a score for each of a plurality of small regions defined in the magnetic field space, and the evaluation is performed based on the score of the evaluation function. It is determined that the run-through in which the score of the evaluation function is higher is superior to that in which the score is lower. FIG. 4 is a diagram showing a magnetic field small region _{Si, j} defined in the magnetic field space (Bx, By), and FIG. 5 is a diagram showing an example of a score table T at the time of learning irradiation. The target dose Di corresponding to the small area is calculated and given by the treatment planning apparatus. The measurement dose Ds is obtained from the measurement result of the beam position monitor 7, the time when the charged particle beam 1b passes through the small area, and the like.

In the second run-through scoring method, the signals input to the command evaluator 33 are different as shown in FIG. Accordingly, step S104 is also different. FIG. 3 is a configuration diagram of an irradiation control apparatus that employs the second run-through scoring method. The target dose Di and the measured dose Ds ( ^a D _k ) are input to the command evaluator 33b of the scanning electromagnet command value learning generator 37b of the irradiation controller 5b. In step S104, the dose is detected by the dose monitor 11, and the measured dose Ds converted by the dose data converter 12 is input to the command evaluator 33b. The command evaluator 33b compares the target dose Di and the measured dose Ds, and performs a run-through scoring of the initial command current ⁰ I _k = ( ⁰ Ix _k , ⁰ Iy _k ).

In FIG. 4, (B ₀ , B ₁ ) in the left column of the table simply shows that the X component Bx of the magnetic field B satisfies the relationship of B ₀ ≦ Bx <B _1. B _m−1 , B _m ) simply shows that Bx satisfies the relationship B _m−1 ≦ Bx <B _m . (B ₀ , B ₁ ) in the upper part of the table simply shows that the Y component By of the magnetic field B satisfies the relationship of B ₀ ≦ By <B _1. Similarly, (B _m−1 , B _m ) simply shows that By satisfies the relationship B _m−1 ≦ By <B _m . The region S _0,0 is a region satisfying the relationship of B ₀ ≦ Bx <B ₁ and B ₀ ≦ By <B ₁ , and the region S _{m−1, m−1} is B _m−1 ≦ Bx <B _m and B _This is a region satisfying the relationship of _m−1 ≦ By <B _m .

  The run-through scoring is performed in the entire magnetic field space corresponding to the irradiable range of the particle beam irradiation device 54. As a result, the irradiation target 15 of the patient is managed with the target dose Di, and the dose corresponding to the normal tissue that is not the irradiation target 15 is managed with zero dose. Therefore, the dose management of the charged particle beam of the irradiation target 15 and the non-irradiation target is accurate. Can be done well.

If the magnetic field of the scanning electromagnet 3 is determined, the irradiation position of the charged particle beam is uniquely determined. In other words, the magnetic field of the scanning electromagnet 3 and the irradiation position of the charged particle beam are in a one-to-one relationship. Therefore, it is not affected by the hysteresis of the scanning electromagnet as in the region defined in the command value current space of the prior art. The relationship between the magnetic field by the magnetic field sensor 20 obtained by irradiating the charged particle beam at the time of calibration and the beam position by the beam position monitor 7 is the same as that in the case of scanning charged particles as in the calibration. The relationship between the magnetic field at the time of irradiation and the beam position agrees very well. Since the actual irradiation space can be obtained from the emission position of the charged particle beam, the passing position in the beam position monitor 7, and the positional relationship between the particle beam irradiation device 54 and the irradiation target 15, the area of the actual irradiation space is defined as a magnetic field space. This mapping relationship is almost the same even during the main irradiation. Therefore, the run-through scoring is performed for each of the plurality of small magnetic fields S _{i, j} defined in the magnetic field space, and the dose management of the charged particle beam is performed for each of the small magnetic fields S _{i, j.} Dose management in the irradiation space can be performed with high accuracy.

The score table T at the time of learning irradiation shown in FIG. 5 is an example of a deduction method. The dose error de is a difference obtained by subtracting the target dose Di from the area measurement dose Dss. The area measurement dose Dss is the actual irradiation dose of the magnetic field small area S _{i, j} defined in the magnetic field space (Bx, By), and is based on the measurement magnetic field Bs measured by the magnetic field sensor 20 and the measurement dose Ds measured by the dose monitor 11. create. Δd is a dose error width, and is set to a predetermined value which is an allowable range. The absolute value of the change rate of the score when the measured dose Ds exceeds the target dose Di is larger than the absolute value of the change rate of the score when the measured dose Ds is less than the target dose Di. As a result, correction when the measured dose Ds exceeds the target dose Di can be performed quickly and accurately.

Since the particle beam irradiation apparatus 54 to which the first scoring method of the first embodiment is applied uses the actual irradiation position as an evaluation function, the command current to the scanning electromagnet 3 ^a I _k = ( ^a Ix _k , ^a Iy _k ) Can be learned to a suitable one. Further, the particle beam irradiation apparatus 54 to which the second scoring method of the first embodiment is applied uses the target dose Di and the region measurement dose Dss for each small magnetic field region _{Si, j} defined in the magnetic field space (Bx, By). By assigning a score and defining the evaluation function based on this score, the command current ^a I _k = ( ^a Ix _k , ^a Iy _k ) to the scanning electromagnet 3 can be learned appropriately. Therefore, a highly accurate and highly safe particle beam irradiation apparatus can be provided.

In the first scoring method, the particle beam irradiation apparatus 54 uses the actual irradiation position reflecting the influence of hysteresis generated between the current of the scanning electromagnet 3 and the magnetic field as an evaluation function, and the command current ^a I to the scanning electromagnet 3. _{Since k} = ( ^a Ix _k , ^a Iy _k ) is learned as a suitable one, the influence of hysteresis generated between the current of the scanning electromagnet 3 and the magnetic field can be eliminated. Further, in the second scoring method, the particle beam irradiation device 54 performs dose management of the charged particle beam for each magnetic field small region S _{i, j} defined in the magnetic field space (Bx, By). It is possible to eliminate the influence of hysteresis occurring between the magnetic field and the magnetic field. Therefore, it is possible to eliminate the influence of the hysteresis of the scanning electromagnet and realize high-precision beam irradiation.

  The magnetic field sensor 20 may be a magnetic field sensor having a Hall element. By using the Hall element, the absolute value of the magnetic field generated by the scanning electromagnet 3 can be measured, and it is not necessary to perform calculations such as integration of the voltage measured by the pickup coil. Therefore, the magnetic field data converter 21 can be simplified and downsized.

  The most desirable magnetic field sensor 20 includes both a big-up coil and a Hall element. This is because it is possible to incorporate both the advantage that the absolute value of the magnetic field of the Hall element can be measured and the advantage that the change amount of the magnetic field of the pickup coil can be measured without hysteresis.

  In the conventional particle beam irradiation apparatus, the beam irradiation position is detected only by one or a plurality of beam position monitors, and the feedback control of the charged particle beam is performed based on the measurement position coordinates. Disposing a large number of objects that block charged particle beams such as a position monitor leads to scattering and expansion of the beam, and there is a problem that a desired beam spot diameter cannot be obtained.

In the particle beam irradiation apparatus 54 according to the first embodiment, the scanning electromagnet command value series generator 35 receives the command current ^a I _k = ( ^a Ix _k , ^a Iy _k ) for which learning has been completed at the time of treatment irradiation. Each time, the irradiation position and irradiation dose of the charged particle beam are controlled by outputting them in the order of the sequence number. Therefore, during the main irradiation, the beam position monitor 7 is moved by a moving device (not shown), and the emitted charged particle beam 1b is the beam. The position monitor 7 may not be passed. By doing so, the outgoing charged particle beam 1b can be prevented from being scattered and expanded by the beam position monitor 7. Thereby, the beam spot diameter can be reduced. Therefore, when it is better to irradiate with a small beam diameter, treatment can be performed with an appropriate spot diameter.

Further, the command value output unit 25 generates a command current I _k = (Ix _k , Iy _k ) based on the estimated value B _est of the magnetic field B (Bx, By) of the scanning electromagnet 3 and executes this as an initial value. You can also learn by running through.

  In addition, although the example which has the reverse mapping generator 30 and the reverse mapping calculator 22 in the irradiation control apparatus 5 was demonstrated, even if it is a case where the reverse mapping generator 30 and the reverse mapping calculator 22 are not provided, the hysteresis of the scanning electromagnet Needless to say, it is possible to eliminate the influence and realize high-accuracy beam irradiation in raster scanning or hybrid scanning.

As described above, according to the particle beam irradiation apparatus 54 of the first embodiment, the irradiation power control apparatus 5 includes the scanning power supply 4 that outputs the excitation current of the scanning electromagnet 3 and the irradiation control apparatus 5 that controls the scanning power supply 4. evaluates the results of the run-through a series of irradiation operation by the command value I _k of the output excitation current to the scanning power source 4, the command value I _k of the exciting current when the result of the evaluation does not satisfy a predetermined condition update run-through is performed, because the result of the evaluation had a command value scanning electromagnet command value learning generator 37 for outputting a I _k to the scanning power source 4 of the exciting current satisfies a predetermined condition, the scanning electromagnet command value learning Based on the result of the run-through, the generator 37 can learn the command value I _k of the excitation current to the scanning power supply 4 to a suitable value, eliminate the influence of the hysteresis of the scanning magnet, and perform raster scanning or hybrid scanning. High-accuracy beam irradiation can be realized in the ning.

According to the particle beam therapy system of the first embodiment, a beam generator 51 that generates a charged particle beam, an accelerator 52 that accelerates a charged particle beam generated by the beam generator 51, and a charge that is accelerated by the accelerator 52. A beam transport device 53 that transports the particle beam, and a particle beam irradiation device 54 that scans the charged particle beam transported by the beam transport device 53 with the scanning electromagnet 3 and irradiates the irradiation target 15. Includes a scanning power supply 4 that outputs an excitation current of the scanning electromagnet 3 and an irradiation control device 5 that controls the scanning power supply 4. The irradiation control device 5 includes a command value I _{k of} the excitation current output to the scanning power supply 4. a series of irradiation operation by evaluate the results of the run-through, run-through is the execution result of the evaluation is to update the command value I _k of the exciting current If does not satisfy the predetermined condition, evaluation Since the result had the command value scanning electromagnet command value learning generator 37 for outputting a I _k to the scanning power source 4 of the exciting current satisfies a predetermined condition, the scanning electromagnet command value learning generator 37 based on the result of the run-through Therefore, the command value I _k of the excitation current to the scanning power source 4 can be learned to a suitable value, the influence of hysteresis of the scanning electromagnet is eliminated, and high-precision particles are used using high-precision beam irradiation in raster scanning and hybrid scanning. Line therapy can be realized.

Embodiment 2. FIG.
In the first embodiment, the charged particle beam is irradiated in the learning irradiation of the main irradiation, but learning is performed without performing the charged particle beam irradiation, and the command current I _k = (Ix _k , Iy _k ) is optimized. be able to. This will be described below. In the particle beam therapy system according to the second embodiment, two types of run-through scoring methods (third scoring method and fourth scoring method) can be used as the run-through scoring for learning irradiation. In the third scoring method, a driving code for raster scanning, that is, an irradiation trajectory for each irradiation layer (layer) Z _i , an irradiation position P _k (X _k , Y _k ) (subscript) for each sampling period. The letter k is an estimated value B _est of the magnetic field corresponding to the sequence number (B _k = (Bx _k , By _k )), and the measured magnetic field ^a B for each sampling period when actually running through. _k (is a number of times of ^{_{learning) (a Bx k, a by}} k) by comparing the, is to consider the following evaluation function. The magnetic field estimated value B _k is calculated by the inverse mapping calculator 22.

The fourth scoring method is a method of scoring according to the score table T shown in FIG. However, since charged particle beam irradiation is not performed during learning irradiation, the dose error de is different from the first embodiment by subtracting the target dose Di from the area dose calculation value Dsc. The area dose calculation value Dsc is derived from the measured magnetic field ^a B _k ( ^a Bx _k , ^a By _k ) measured by the magnetic field sensor 20 for each of the output magnetic particles S _{i, j} defined in the magnetic field space (Bx, By). The dose is integrated and determined by the time during which the beam 1b stays.

FIG. 6 is a block diagram of an irradiation control apparatus according to Embodiment 2 of the present invention, in which a third scoring method is employed as a scoring for learning irradiation run-through. The signal input to the command evaluator 33 is different from the irradiation control device of the first embodiment. The command evaluator 33c of the scanning electromagnet command value learning generator 37c of the irradiation control device 5c has an estimated magnetic field value B _k = (Bx _k , By _k ) and a measured magnetic field ^a B _k ( ^a Bx _k , ^a By _k ). Entered. Irradiation control device 5 in Embodiment 2
The operation of c will be described. Trial irradiation at the time of calibration is the same as steps S001 to S005 in the first embodiment. Learning irradiation is basically the same as step S101 to step S108 in the first embodiment, but step S104 is different because charged particle beam irradiation is not performed. Step S201 is executed instead of step S104 in the first embodiment.

Magnetic fields generated by the scanning electromagnets 3a and 3b controlled by the command current I _k = (Ix _k , Iy _k ) are measured by the magnetic field sensors 20a and 20b. The measured magnetic field ^a B _k ( ^a Bx _k , ^a By _k ) measured by the magnetic field sensor 20 and converted by the magnetic field data converter 21 is input to the command evaluator 33c. The command evaluator 33c compares the estimated magnetic field value B _k = (Bx _k , By _k ) calculated by the inverse mapping calculator 22 for each sequence number with the measured magnetic field ^a B _k ( ^a Bx _k , ^a By _k ). Then, the run-through scoring of the initial command current ⁰ I _k = ( ⁰ Ix _k , ⁰ Iy _k ) is performed (step S201).

  Next, the irradiation control device and operation when the fourth scoring method is employed as the learning irradiation run-through scoring will be described. FIG. 7 is a block diagram of another irradiation control apparatus according to Embodiment 2 of the present invention, in which a fourth scoring method is adopted as scoring for learning irradiation run-through. The signal input to the command evaluator 33 is different from the irradiation control device of the first embodiment. The target dose Di and the area dose calculation value Dsc are input to the command evaluator 33d of the scanning electromagnet command value learning generator 37d of the irradiation control device 5d. The operation of the irradiation control device 5d in the second embodiment will be described. Trial irradiation at the time of calibration is the same as steps S001 to S005 in the first embodiment. Learning irradiation is basically the same as step S101 to step S108 in the first embodiment, but step S104 is different because charged particle beam irradiation is not performed. Step S202 is executed instead of step S104 in the first embodiment.

The charged particle beam 1 controlled by the command current I _k = (Ix _k , Iy _k ) is integrated by the time spent in each magnetic field small region S _{i, j} , and the region dose calculation value Dsc is calculated. 33d. The command evaluator 33c compares the target dose Di with the area dose calculated value Dsc, and performs run-through scoring of the initial command current ⁰ I _k = ( ⁰ Ix _k , ⁰ Iy _k ) (step S202).

Since the particle beam irradiation apparatus 54 to which the third scoring method of the second embodiment is applied uses the measurement magnetic field as an evaluation function, the command current ^a I _k = (= ^a Ix _k , ^a Iy _k ) can be learned to a suitable one. In addition, the particle beam irradiation apparatus 54 to which the fourth scoring method of the second embodiment is applied has an area dose calculation value Dsc calculated for each magnetic field small area _{Si, j} defined in the magnetic field space (Bx, By). By assigning a score from the target dose Di and defining an evaluation function based on this score, the command current ^a I _k = ( ^a Ix _k , ^a Iy _k ) to the scanning electromagnet 3 is preferable without irradiation with a charged particle beam. You can learn something. Therefore, a highly accurate and highly safe particle beam irradiation apparatus can be provided.

The particle beam irradiation device 54 learns the command current ^a I _k = ( ^a Ix _k , ^a Iy _k ) to the scanning electromagnet 3 as ^a suitable one using the measurement magnetic field as an evaluation function in the third scoring method. The influence of the hysteresis generated between the current of the scanning electromagnet 3 and the magnetic field can be eliminated. Further, in the fourth scoring method, the particle beam irradiation device 54 performs dose management of the charged particle beam for each magnetic field small region S _{i, j} defined in the magnetic field space (Bx, By). It is possible to eliminate the influence of hysteresis occurring between the magnetic field and the magnetic field. Therefore, it is possible to eliminate the influence of the hysteresis of the scanning electromagnet and realize high-precision beam irradiation.

In addition, since the run-through of learning irradiation can be performed without irradiating a charged particle beam,
Unnecessary energy consumption can be suppressed.

Embodiment 3 FIG.
The learning function can be interpreted as "a function that approaches a more ideal solution for one task". In the first and second embodiments, the command current I _k that realizes more ideal irradiation with respect to the problem of how to make actual irradiation approach one treatment plan data generated for each patient by the treatment planning device 55. The function of generating = (Ix _k , Iy _k ) has been described.

  A more advanced learning function (or more intelligent learning function) in control engineering is interpreted as “a function for deriving a more ideal solution for unknown future problems by accumulating past experience”. The Therefore, in the third embodiment, a particle beam irradiation apparatus and a particle beam therapy apparatus having a more advanced learning function by further applying the learning function described in the first and second embodiments will be described.

In the first embodiment, the learning irradiation is run-through scoring, and the command current I _k itself is updated to a more optimal one based on the scoring result. This method for one treatment planning data, although more command current I _k to realize an ideal illumination can be generated, wherein the experience can not be reflected to other treatment planning data.

In the third embodiment, learning irradiation is scored for run-through as in the first or second embodiment. However, based on the run-through scoring result, a mathematical model for generating the command current I _k = (Ix _k , Iy _k ) is provided instead of updating the command current I _k = (Ix _k , Iy _k ) itself. The parameters of the mathematical model are updated to be more optimal. The details are described below.

FIG. 8 is a diagram illustrating an example of a mathematical model for generating the command current I _k = (Ix _k , Iy _k ) in the third embodiment. Based on FIG. 8, a mechanism for generating the command current I _k = (Ix _k , Iy _k ) will be described.

A feedforward type NN (neural network, hereinafter referred to as “NN”) 60 having a single hidden layer is an example of a mathematical model for generating the command current I _k = (Ix _k , Iy _k ). The input layer 61 is an input unit of the NN 60, and corresponds to the irradiation position P _k (X _k , Y _k ) for each sampling period, which is a target in the third embodiment of the present invention. _{X k} is input to the input layer 61a, _{Y k} is input to the input layer 61b. The output layer 63 is an output unit of the NN 60, and the command current I _k = (Ix _k , Iy _k ) corresponds to the third embodiment of the present invention. Ix _k is output to the output layer 63a, and Iy _k is output to the output layer 63b. The hidden layer 62 is a basis function (activation function) of the NN 60, and weights are given to input signals from the input layers 61a and 61b by the plurality of hidden layers 62a to 62n, and output to the output layers 63a and 63b. To do.

In the particle beam therapy system, the charged particle beam 1 is deflected by the scanning electromagnet 3 and its irradiation position is determined. That is, if a series of command currents I _k = (Ix _k , Iy _k ) to the scanning electromagnet 3 is determined, the beam irradiation position coordinates P = (X, Y) are uniquely determined, and thus a series of command currents I _k = A physical phenomenon from (Ix _k , Iy _k ) to the irradiation position coordinate P of the beam can be regarded as a 2-input 2-output system having dynamics called hysteresis characteristics. What is going to be realized by NN60 is to mathematically realize a two-input two-output inverse system having the dynamics of this hysteresis characteristic.

In the first and second embodiments, the update in the learning irradiation is the command current I _k = (Ix _k , Iy _k ) itself. On the other hand, in the third embodiment, the weight in the hidden layer 62 of the NN 60 is updated. FIG. 9 shows an irradiation control device 5 (5e) provided with a scanning electromagnet command value series generator 35b having NN60. In the irradiation control device 5e, the inverse map generator 30, the inverse mapping calculator 22, the command value output unit 25, and the command updater 34 are deleted from the irradiation control device 5a of the first embodiment, and a scanning electromagnet command value learning generator is obtained. In this example, the parameter updater 38 is provided in 37e, and the NN 60 is mounted on the scanning electromagnet command value series generator 35. The irradiation control device 5e is a case where the scoring method for learning irradiation run-through is the first scoring method. The parameter updater 38 updates the weight in the hidden layer 62 that is a parameter of the NN 60 so that the evaluation result of the command evaluator 33 becomes a predetermined condition. The irradiation control device 5 provided with the scanning electromagnet command value series generator 35b having the NN 60 can use the second to fourth scoring methods as the learning irradiation run-through scoring methods. In this case, the command evaluators 33b to 33d may be used instead of the command evaluator 33a of FIG. In the case of the third scoring method, the irradiation control device 5 may include the inverse mapping generator 30 and the inverse mapping calculator 22, and the magnetic field estimation value B _k may be calculated by the inverse mapping calculator 22.

  With the configuration as described above, in the particle beam therapy system according to Embodiment 3 of the present invention, experience of learning irradiation run-through (learning run-through) with respect to treatment plan data performed in the past is accumulated. When a lot of learning run-through experience is accumulated, it is not necessary to perform a learning run-through every time for new treatment plan data, and it is difficult to be affected by the hysteresis of the scanning magnet, and highly accurate irradiation is performed. A particle beam therapy system can be obtained.

  Note that the learning algorithms described in Embodiments 1 to 3 are only examples, and other algorithms implemented in other technical fields such as a steepest descent method and a genetic algorithm may be applied.

Further, the evaluation function in the run-through evaluation is not limited to that described here, and other evaluation functions may be used. Weighting may be applied to each sequence number k in Equation (3) and Equation (4). In addition, in the case of assigning points for each of a plurality of small regions defined in the magnetic field space, an evaluation function obtained by adding weights w _{i, j} to the points Sci _{, j} of each small region may be used. In these cases, important positions and regions can be optimized with high accuracy.

  The particle beam irradiation apparatus and the particle beam therapy apparatus according to the present invention can be suitably applied to a particle beam therapy apparatus used for medical use or research.

DESCRIPTION OF SYMBOLS 1 Charged particle beam 1a Incident charged particle beam 1b Outgoing charged particle beam 3 Scanning electromagnet 3a X direction scanning electromagnet 3b Y direction scanning electromagnet 4 Scanning power supply 7 Beam position monitor 11 Dose monitor 15 Irradiation target 20 Magnetic field sensor 20a X direction electromagnet magnetic field sensor 20b Magnetic field sensor for Y-direction electromagnet 22 Reverse mapping calculator 33 Command evaluator 34 Command updater 35 Scanning electromagnet command value series generator 35a Scanning electromagnet command value series generator 35b Scanning electromagnet command value series generator 37 Scanning electromagnet command value learning Generator 37a Scanning magnet command value learning generator 37b Scanning magnet command value learning generator 37c Scanning magnet command value learning generator 37d Scanning magnet command value learning generator 37e Scanning magnet command value learning generator 38 Parameter updater 51 Beam generator 52 Accelerator 53 Beam Transporter 5 Particle beam irradiation apparatus 60 NN (Neural Network) Ps measurement position coordinates
^a P _k irradiation position P _k irradiation position Ds Measurement dose ^a D _k Measurement dose Di Target dose Bs Measurement magnetic field
^a B _k measured magnetic field B _Est magnetic field estimated value B _k magnetic field estimated value Dss area measured dose Dsc area dose calculated value de dose error T score table S _{i, j} magnetic field small area I _k command current

Claims (7)

A scanning electromagnet having hysteresis for scanning the charged particle beam accelerated by the accelerator; a scanning power source for outputting an excitation current for driving the scanning electromagnet;
A particle beam irradiation device comprising an irradiation control device for controlling the scanning power source,
The irradiation control device includes:
A mathematical model for generating a command value of the excitation current, and evaluating a run-through result that is a series of irradiation operations according to the command value of the excitation current output to the scanning power source, and based on the result of the evaluation A scanning electromagnet command value learning generator for updating the mathematical model, accumulating the run-through experience, and outputting the excitation current command value generated by the mathematical model based on the accumulated run-through experience to the scanning power supply A particle beam irradiation apparatus comprising: The scanning electromagnet command value learning generator is
The mathematical model for generating a command value of the excitation current;
A learning function that evaluates a result of run-through that is a series of irradiation operations based on a command value of the excitation current output to the scanning power supply, updates the mathematical model based on the result of the evaluation, and accumulates the experience of the run-through Means,
2. A scanning magnet command value series generator for outputting a command value of the excitation current generated by the mathematical model to the scanning power supply based on the latest experience accumulated by the learning function means. The particle beam irradiation apparatus described.   The particle beam irradiation apparatus according to claim 1, wherein the mathematical model generates a command value of the excitation current based on a target irradiation position coordinate in an irradiation target irradiated with the charged particle beam.   The mathematical model is a reverse mapping when the mapping from the command value of the excitation current having the dynamics of the hysteresis characteristics of the scanning electromagnet to the irradiation position of the charged particle beam is a normal mapping. The particle beam irradiation apparatus according to claim 1 or 2.   The particle beam irradiation apparatus according to claim 1, wherein the mathematical model is a feedforward neural network.   6. The particle beam irradiation apparatus according to claim 5, wherein the feedforward neural network has a single hidden layer. A beam generator for generating a charged particle beam, an accelerator for accelerating the charged particle beam generated by the beam generator, a beam transport device for transporting a charged particle beam accelerated by the accelerator, and the beam transport device A charged particle beam transported by a scanning electromagnet and a particle beam irradiation device for irradiating an irradiation target,
The particle beam irradiation apparatus according to any one of claims 1 to 6, wherein the particle beam irradiation apparatus is the particle beam irradiation apparatus.

Images