KR20200134092A - System and method for respiration-gated radiotherapy evaluation using machine learning - Google Patents

Abstract

An embodiment of the present invention relates to a system for respiratory-gated radiotherapy evaluation using machine learning. The system includes: a signal acquisition unit acquiring a first beam signal including irradiation information and a first respiratory signal from a radiotherapy system during respiratory-gated radiotherapy; a learning unit machine-learning a respiratory-gated radiotherapy evaluation model based on the acquired first respiratory signal and the acquired first beam signal; and a determination unit determining whether the respiratory-gated radiotherapy is suitable for a patient based on the respiratory-gated radiotherapy evaluation model and a second beam signal and a second respiratory signal acquired by the signal acquisition unit.

Description

Respiratory-linked radiation therapy evaluation system using machine learning and its method {SYSTEM AND METHOD FOR RESPIRATION-GATED RADIOTHERAPY EVALUATION USING MACHINE LEARNING}

Embodiments of the present invention relate to a system and method for evaluating respiratory interlocking radiation therapy using machine learning.

One of the important factors to be aware of when treating patients with 4D radiation therapy or respiratory synchrotron radiation therapy is that it should be done with the premise that long-term movements are constant during radiation therapy. There may be various types of motor elements that affect organ movements of the human body, but among them, maintaining consistent breathing is a very important factor in radiation therapy.

Radiation treatment using a medical radiation therapy device may be the most important factor in intensively irradiating the tumor site with radiation and delivering a minimum dose to surrounding normal tissues. In particular, in radiation therapy for moving organs, a technique of controlling the beam according to the fluctuation of the geometrical movement of the organ that occurs during radiation treatment is indispensable.

Various devices are used to fulfill the prerequisites related to breathing during the radiation treatment as described above, but most of these devices use only guide signals, so that the patient cannot check the patient's own breathing status, and only breathes as intuitively practiced. There is a situation.

To this end, the conventional radiation treatment technology mainly uses a gating method of controlling and treating a medical radiation treatment device using a Real-time position management (RPM) system. However, the success or failure of the treatment of this technology lies in the stability of the breathing of radiation therapy patients, and the problem of respiration-tuned radiation therapy of patients with unstable breathing increases the treatment time. In addition, since the RPM system sets a phase of breathing with only a breathing cycle and applies it to respiration-tuned radiation therapy, there is no consideration of the breathing pattern or the volume of breath, so the accuracy of the treatment is very poor.

Particularly, body stereotactic radiation treatment for moving organs has a significantly longer treatment time compared to other radiation treatments, and the risk of side effects can also be very high, so in order to solve this problem, the patient is educated to have a stable breathing cycle and volume and after practicing breathing. Radiotherapy was performed. However, the conventional radiation therapy system does not have a feedback function that informs the patient how effective such training or breathing exercises have been for radiation therapy, or accurately informs the patient of the improvement, and there is no means of evaluating whether the respiratory-linked radiation therapy itself is suitable for the patient or not. Since it does not exist, there is a problem that the efficiency and accuracy of treatment are poor.

The above-described background technology is technical information possessed by the inventors for derivation of the present invention or acquired during the derivation process of the present invention, and is not necessarily known to be publicly known prior to filing the present invention.

Embodiments of the present invention intend to provide a system and method for evaluating respiration-linked radiation therapy that can evaluate and give feedback on how efficient and accurate whether respiratory-linked radiation therapy is suitable for each patient using machine learning.

An embodiment of the present invention is a signal acquisition unit for acquiring a first beam signal including a first respiration signal and radiation irradiation information during respiration-linked radiation treatment from a radiation treatment system, the acquired first respiration signal and 1 A learning unit for machine learning a respiratory-linked radiation therapy evaluation model based on a beam signal, a second breathing signal and a second beam signal acquired by the signal acquisition unit, and the respiratory-linked radiation therapy evaluation model. It provides a system for evaluating respiratory-linked radiation therapy using machine learning, including a determination unit that determines whether or not the respiration-linked radiation therapy is suitable for Korea.

In an embodiment of the present invention, the learning unit may generate evaluation data for evaluating the respiratory-linked radiation therapy using the acquired first respiration signal and the first beam signal.

In one embodiment of the present invention, the learning unit is to generate the respiratory-linked radiation therapy evaluation model that derives the treatment efficiency, treatment accuracy, and equipment accuracy of the respiratory-linked radiation therapy as result values using the generated evaluation data. I can.

In one embodiment of the present invention, the learning unit generates the evaluation data using any one of respiratory cycle information, average amplitude information, standard deviation information, and peak value information derived from the first breathing signal. Thus, it is possible to machine learning the respiratory-linked radiation therapy evaluation model.

In an embodiment of the present invention, the learning unit uses one of the breathing period information, the average amplitude information, the standard deviation information, and the peak value information to perform a baseline shift of the breathing signal. ) Can be created.

In one embodiment of the present invention, the learning unit includes an expected margin given when generating a planning target volume of a patient during a radiation treatment plan, and a real margin derived from the baseline shift. ) Can be compared to derive an evaluation result of the treatment accuracy.

In one embodiment of the present invention, the first beam signal includes reference radiation irradiation information set in advance in response to the planned breathing cycle of the patient during a radiation treatment plan, and the actual breathing of the patient acquired in a simulated treatment or an actual treatment. It may include information about actual irradiation conducted in response to the period.

In one embodiment of the present invention, the learning unit calculates a reference duty factor from the reference radiation irradiation information, calculates an actual duty ratio from the actual radiation irradiation information, and then generates the evaluation data, and the Using the evaluation data, it is possible to derive a result value for the respiration-linked radiation treatment efficiency.

In an embodiment of the present invention, the learning unit calculates a radiation irradiation time error using the reference radiation irradiation information and the actual radiation irradiation information, and calculates a result value for the equipment accuracy using the radiation irradiation time error. Can be derived.

In an embodiment of the present invention, obtaining a first beam signal including a patient's first respiration signal and radiation irradiation information during respiration-linked radiation treatment from a radiation treatment system, the obtained first respiration signal, and the first Machine learning a respiration-linked radiotherapy evaluation model based on a beam signal, acquiring a second respiration signal and a second beam signal of an existing patient or a new patient, the second respiration signal and the second beam signal, It provides a method for evaluating respiratory-linked radiation therapy using machine learning, comprising the step of determining whether a patient is appropriate for respiratory-linked radiation therapy based on the respiratory-linked radiation therapy evaluation model.

In one embodiment of the present invention, the step of machine learning the respiratory-linked radiation therapy evaluation model includes evaluation data for evaluating the respiratory-linked radiation therapy using the acquired first breathing signal and the first beam signal. And, using the evaluation data, the respiratory-linked radiation therapy evaluation model can be generated.

In one embodiment of the present invention, the respiration-linked radiation treatment evaluation model may derive the treatment efficiency, treatment accuracy, and equipment accuracy of the respiratory-linked radiation treatment as a result value using the generated evaluation data.

In one embodiment of the present invention, the step of machine learning the respiratory-linked radiation therapy evaluation model includes respiratory cycle information derived from the first respiration signal, average amplitude information, standard deviation information, and peak value information. The evaluation data may be generated using any one of them, and the respiratory-linked radiation therapy evaluation model may be machine-learned using the evaluation data.

In one embodiment of the present invention, the step of machine learning the respiratory-linked radiation therapy evaluation model uses any one of the respiratory cycle information, the average amplitude information, the standard deviation information, and the peak value information. Thus, a baseline shift of the breathing signal can be generated.

In one embodiment of the present invention, the respiration-linked radiation therapy evaluation model compares the expected margin given when the patient's treatment target application volume is generated during the radiation treatment plan, and the actual margin derived from the baseline shift. It may be a model that derives evaluation results.

In one embodiment of the present invention, the first beam signal includes reference radiation irradiation information set in advance in response to the planned breathing cycle of the patient during a radiation treatment plan, and the actual breathing of the patient acquired in a simulated treatment or an actual treatment. It may include information about actual irradiation conducted in response to the period.

In an embodiment of the present invention, the machine learning of the respiratory-linked radiation therapy evaluation model includes calculating a reference duty factor from the reference radiation irradiation information, and calculating an actual duty ratio from the actual radiation irradiation information. After calculation, the evaluation data may be generated, and a result value for the respiratory-linked radiation treatment efficiency may be derived using the evaluation data.

In an embodiment of the present invention, the machine learning of the respiration-linked radiation treatment evaluation model comprises: calculating a radiation exposure time error using the reference radiation radiation information and the actual radiation exposure information, and the radiation exposure time error It is possible to derive a result value for the equipment accuracy by using.

An embodiment of the present invention provides a computer program stored in a medium to execute any one of the above-described methods using a computer.

Other aspects, features, and advantages other than those described above will become apparent from the following drawings, claims, and detailed description of the invention.

In the system and method for evaluating respiration-linked radiation therapy using machine learning according to embodiments of the present invention, after generating a respiration-linked radiation treatment evaluation model using a first respiration signal and a first beam signal, a new second respiration signal and a Using the second beam signal and the respiration-linked radiation treatment evaluation model, the patient's radiation-respiratory-linked radiation therapy can be automatically evaluated. The system and method for evaluating respiration-linked radiation therapy using machine learning according to the embodiments of the present invention can provide a patient with an evaluation result of whether or not respiration-linked radiation therapy is appropriate, and through this, a radiation plan is provided with radiation therapy suitable for the patient. Can be changed to improve the efficiency and accuracy of radiation treatment.

1 is a block diagram schematically illustrating a system for evaluating respiratory interlocking radiation therapy using machine learning according to an embodiment of the present invention.
2 is a diagram illustrating the radiation treatment system of FIG. 1.
3 is a diagram sequentially showing a method for evaluating respiratory interlocking radiation therapy using machine learning according to an embodiment of the present invention.
FIG. 4 is a schematic diagram for explaining a method of evaluating respiration-linked radiation therapy of FIG. 3.
5 is a diagram sequentially illustrating the machine learning steps of FIG. 3.
6 is a diagram for explaining a method of generating evaluation data for evaluating treatment efficiency in a method for evaluating respiration-linked radiation therapy.
7 and 8 are diagrams for explaining a method of generating evaluation data for evaluation of treatment accuracy in a method of evaluating respiration-linked radiation therapy.
9 and 10 are diagrams for explaining a method of generating evaluation data for evaluating equipment accuracy in a method for evaluating respiration-linked radiation therapy.

Since the present invention can apply various transformations and have various embodiments, specific embodiments are illustrated in the drawings and will be described in detail in the detailed description. Effects and features of the present invention, and a method of achieving them will be apparent with reference to the embodiments described later in detail together with the drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various forms.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, and when describing with reference to the drawings, the same or corresponding components are assigned the same reference numerals, and redundant descriptions thereof will be omitted. .

In the following embodiments, terms such as first and second are not used in a limiting meaning, but are used for the purpose of distinguishing one component from another component.

In the following examples, the singular expression includes the plural expression unless the context clearly indicates otherwise.

In the following embodiments, terms such as include or have means that the features or elements described in the specification are present, and do not preclude the possibility of adding one or more other features or elements in advance.

In the following embodiments, when a part such as a film, a region, or a component is on or on another part, not only the case directly above the other part, but also another film, region, component, etc. are interposed therebetween. This includes cases where there is.

In the drawings, components may be exaggerated or reduced in size for convenience of description. For example, the size and thickness of each component shown in the drawings are arbitrarily shown for convenience of description, and the present invention is not necessarily limited to what is shown.

When a certain embodiment can be implemented differently, a specific process order may be performed differently from the described order. For example, two processes described in succession may be performed substantially simultaneously, or may be performed in an order opposite to the described order.

In the following embodiments, when a film, a region, a component, etc. are connected, not only the film, the region, and the components are directly connected, but other films, regions, and components are interposed between the film, the region, and the components. It includes the case of being connected indirectly. For example, in this specification, when a film, region, component, etc. are electrically connected, not only the film, region, component, etc. are directly electrically connected, but also other films, regions, components, etc. are interposed therebetween. Indirect electrical connection is also included.

1 is a block diagram schematically illustrating a breathing-linked radiation treatment evaluation system 1 using machine learning according to an embodiment of the present invention, and FIG. 2 is a diagram illustrating the radiation treatment system 2 of FIG. 1 .

1 and 2, the breathing-linked radiation therapy evaluation system 1 using machine learning according to an embodiment of the present invention includes a signal acquisition unit 110, a learning unit 131, and a determination unit 133. Can include. In addition, the respiratory interrelated radiation therapy evaluation system 1 may further include a memory 150. The respiratory-linked radiation therapy evaluation system (1) using machine learning receives breathing signals and beam signals from the radiation treatment system (2) that performs radiation treatment plans and treatments for patients, and uses them as a basis for machine learning. Radiation therapy evaluation models can be created. In other words, the respiratory-linked radiation therapy evaluation system (1) automatically predicts the respiratory-linked radiation treatment results for the patient by applying the new patient's breathing signal or the existing patient's new breathing signal to the learned respiratory-linked radiation therapy evaluation model. It is characterized by evaluating. This will be described in more detail as follows.

In detail, with the advancement of radiation treatment technology such as radiation surgery, intensity-controlled radiation treatment, and image-guided radiation treatment, it is possible to ensure that the minimum dose is delivered to the normal breakfast around the target, and most of the dose is intensively irradiated to the tumor site. Became. Meanwhile, radiation therapy for tumors in the lung, liver, and abdomen affected by respiratory movements is also continuously increasing.

However, radiation treatment for organs that have movement due to breathing such as lungs and liver, due to the movement of the treatment organs, has poor precision and accuracy in terms of radiation delivery technology, so it is not possible to focus the radiation dose on a desired location or There are limitations to be investigated. In order to overcome this, in the field of radiation therapy, radiation therapy is performed using a breathing method.

The success or failure of respiratory-linked radiation therapy lies in controlling the beam according to fluctuations in the geometrical movement of organs that occur during or between treatments. Particularly, in the case of particle ray radiation therapy, which has recently been in the spotlight, the introduction of respiration-linked radiation therapy is indispensable because it is significantly affected by the change in radiation dose due to the movement of the treatment organ compared to the conventional treatment. In the conventional radiation therapy technology for this, a gating method that tracks only the movement of an extracorporeal marker based on a linear accelerator is used, and in this method, respiration synchronization is performed using a Real-time Position Management (RPM) system.

However, the success or failure of the treatment of this technology lies in the stability of the breathing of radiation therapy patients, and the problem of respiration-tuning radiation therapy of patients with unstable breathing increases the treatment time. Particularly, since the treatment time is very long in body part radiation therapy and body part radiation surgery, it is very important to educate the patient to have a regular and stable respiratory cycle and volume during radiation treatment and to practice breathing.

However, it cannot be said that such respiratory-linked radiation therapy is effective for all patients. If the patient's breathing regularity is poor, radiation therapy for organs caused by movement has a worse effect, and thus, respiratory-linked radiation therapy may be unsuitable for such patients. Conventionally, since there is no system for evaluating whether respiration-linked radiation therapy is suitable for a patient or not, there is a problem of performing respiratory-linked radiation therapy collectively even for unsuitable patients.

In order to solve such a problem, in the breathing-linked radiation therapy evaluation system 1 using machine learning according to an embodiment of the present invention, breathing from the first breath or simulated breath (actual breath) for respiratory-linked radiation therapy of a patient By acquiring information and beam information, and generating a respiratory-linked radiation therapy evaluation model, the respiratory-linked radiation therapy evaluation for a patient can be automatically performed. Through this, the respiratory-linked radiation therapy evaluation system (1) can automatically evaluate the suitability of respiratory-linked radiation therapy, rather than evaluating the patient's respiratory-linked radiation therapy according to the capabilities and experience of the medical staff participating in the radiation treatment plan. Yes, it can provide accurate and quick feedback.

Hereinafter, each component of the respiratory interlocking radiation therapy evaluation system 1 using machine learning of the present invention will be described in more detail.

The signal acquisition unit 110 may acquire a beam signal including a respiratory signal and radiation irradiation information of a patient during respiratory-linked radiation therapy from the radiation treatment system 2. Here, the respiration signal and the beam signal may include both a signal obtained to generate a respiratory-linked radiation treatment evaluation model and a signal obtained to evaluate an actual respiratory-linked radiation treatment. However, hereinafter, for convenience of explanation, the breathing signal and the beam signal acquired to generate the respiratory-linked radiation treatment evaluation model in this specification are defined as the first breathing signal and the first beam signal, and the actual respiratory-linked radiation treatment A new signal obtained for evaluation will be defined as a second respiration signal and a second beam signal.

Here, the radiation treatment system 2 is a system for performing simulated treatment or actual treatment according to a radiation treatment plan, and in particular, an imaging unit 203 and a breathing signal conversion unit ( 211), and includes a radiation irradiation unit 250 for irradiating radiation in connection with the patient's breathing signal. In addition, although the radiation treatment system 2 is not shown, a control unit (not shown), an input/output unit 205, a memory (not shown), and a program storage unit (not shown) capable of controlling components in the radiation treatment system 2 City) and the like may be further provided.

The control unit (not shown) is a kind of central processing unit and controls the entire process of irradiating radiation in conjunction with the patient's breath in the radiation treatment system 2. That is, the control unit (not shown) drives the control software installed in the program storage unit (not shown), controls the imaging unit 203 to take a breathing image of the patient, and captures the patient's breathing image obtained by the imaging unit 203. It is possible to control the radiation irradiation unit 250 to obtain a respiration signal by image processing and tracking analysis of the respiration image, and to irradiate a radiation beam in connection with the respiration signal.

Meanwhile, the input/output unit 205 may perform a function of providing a breathing guide signal to a patient during respiratory-linked radiation therapy. The input/output unit 205 may be configured with a touch-sensitive display controller or various input/output controllers. For example, the touch-sensitive display controller provides an output interface and an input interface between a device and a user. The touch-sensitive display controller transmits and receives electrical signals to and from the controller. In addition, the touch-sensitive display controller displays a visual output to the user, and the visual output may include text, graphics, images, video, and combinations thereof. The input/output unit 205 may be, for example, a predetermined display member such as an organic light emitting display (OLED) or liquid crystal display (LCD) capable of recognizing a touch. In FIG. 2, as an example of the input/output unit 205, a glasses-type goggle display and a monitor are shown, but the present invention is not limited thereto, and various types of display devices such as PDA and smart phone may be applied.

The memory (not shown) performs a function of temporarily storing data, etc. processed by the control unit (not shown).

The program storage unit (not shown) is a task in which the radiation treatment system 2 captures the patient's breathing image and obtains the patient's breathing signal from it, the work of irradiating radiation in conjunction with the patient's breathing signal, and a guide signal to the patient. It is equipped with control software that performs tasks and the like.

The respiration signal conversion unit 211 acquires a respiration signal according to time t by image processing and tracking analysis of the patient's respiration image acquired by the imaging unit 203. To this end, a predetermined marker 230 may be disposed on the chest or abdomen of the patient, and the movement of the marker 230 among the breathing images of the patient acquired by the imaging unit 203 is image-processed and traced to analyze the respiratory signal. Can be obtained. At this time, a patient signal of the patient acquired by the respiration signal conversion unit 211 may be displayed on the input/output unit 205 in the form of a sine wave as shown in FIG. 8.

The radiation irradiation unit 250 may irradiate radiation in response to the first respiration signal obtained from the respiration signal conversion unit 211 under control of a control unit (not shown). The radiation irradiation unit 250 may generate a first beam signal including radiation beam irradiation information and transmit it to the respiratory-linked radiation treatment evaluation system 1. At this time, the first beam signal is the reference radiation irradiation information set in advance in response to the patient's planned breathing cycle in the radiation treatment plan, and the actual radiation irradiation information performed in response to the actual breathing cycle of the patient acquired in the simulation or actual treatment. It may include. In other words, the first beam signal may be generated by including information on ideal radiation irradiation generated by a radiation treatment plan and information on radiation irradiation actually irradiated when performing treatment. Meanwhile, here, the simulated treatment may include a case where the patient first attempts to breathe to establish a radiation treatment plan, and a case where the patient attempts to breathe according to the practiced breathing cycle after having the patient perform practice breathing.

Returning to the description of the respiration-linked radiation treatment evaluation system 1 using machine learning again, the signal acquisition unit 110 of the respiration-linked radiation treatment evaluation system 1 is the first respiration signal from the respiration signal conversion unit 211 Is provided, and the first beam signal may be provided from the radiation irradiation unit 250. Alternatively, the radiation treatment system 2 provides the first respiration signal and the first beam signal to an external server (not shown), and the external server (not shown) stores the first respiration signal and the first beam signal. The data may be provided to the signal acquisition unit 110 of the respiratory interlocking radiation treatment evaluation system 1. Like the first beam signal, the first respiration signal may include a reference respiration signal according to a radiation treatment plan and an actual respiration signal obtained in a simulated treatment or an actual treatment.

The learning unit 131 may machine learn a breathing-linked radiation treatment evaluation model based on the acquired first respiration signal and the first beam signal.

The determination unit 133 is based on the second respiration signal and the second beam signal obtained by the signal acquisition unit 110 and the generated respiration-linked radiation treatment evaluation model, whether or not the patient is suitable for respiration-linked radiation therapy. Can judge.

The learning unit 131 and the determination unit 133 may include a device capable of processing data, such as a processor 130. As shown in FIG. 1, as an embodiment, the respiratory interlocking radiation therapy evaluation system 1 using machine learning includes one processor 130 to perform the functions of the learning unit 131 and the determination unit 133. I can. However, the technical idea of the present invention is not limited thereto, and the breathing-linked radiation therapy evaluation system 1 using machine learning so that the learning unit 131 and the determination unit 133 can perform functions in separate processors Of course, it can include more than one processor.

Here, the'processor' may refer to a data processing device embedded in hardware having a circuit physically structured to perform a function represented by a code or instruction included in a program. As an example of such a data processing device built into the hardware, a microprocessor, a central processing unit (CPU), a processor core, a multiprocessor, and an application-specific integrated (ASIC) Circuit) and processing devices such as a Field Programmable Gate Array (FPGA) may be covered, but the scope of the present invention is not limited thereto.

The memory 150 performs a function of temporarily or permanently storing data processed by the respiratory-linked radiation therapy evaluation system 1 using machine learning. The memory 150 may include a magnetic storage medium or a flash storage medium, but the scope of the present invention is not limited thereto.

Hereinafter, a method of evaluating respiration-linked radiation therapy using machine learning will be described in more detail with reference to FIGS. 3 to 10.

FIG. 3 is a diagram sequentially showing a method of evaluating respiration-linked radiation therapy using machine learning according to an embodiment of the present invention, and FIG. 4 is a schematic diagram illustrating a method of evaluating respiration-linked radiation therapy of FIG. 3. 5 is a diagram sequentially showing the steps of machine learning of FIG. 3, and FIG. 6 is a diagram illustrating a method of generating evaluation data for evaluating treatment efficiency in a method for evaluating respiration-linked radiation therapy. 7 and 8 are diagrams for explaining a method of generating evaluation data for evaluation of treatment accuracy in a method of evaluating respiration-linked radiation therapy, and FIGS. 9 and 10 are diagrams for evaluating equipment accuracy in a method of evaluating respiration-linked radiation therapy It is a diagram for explaining a method of generating evaluation data for.

First, referring to FIGS. 3 and 4, a method for evaluating respiratory-linked radiation therapy using machine learning according to an embodiment of the present invention includes a first respiratory signal of a patient during respiratory-linked radiation therapy from the radiation therapy system 2, and Acquiring a first beam signal including radiation irradiation information (S100), machine learning a respiratory-linked radiation therapy evaluation model based on the acquired first respiration signal and the first beam signal (S200), an existing patient or After acquiring the second respiration signal and the second beam signal of the new patient, using the respiration-linked radiation treatment evaluation model, determining whether or not the respiration-linked radiation treatment is suitable for the patient (S300). Thereafter, the respiration-linked radiation treatment evaluation method may provide the determination result to the radiation treatment system 2 (S400).

This will be described in more detail as follows.

First, when the radiation treatment system 2 generates a reference breathing signal having an ideal breathing cycle for the patient according to the radiation treatment plan, and generates a beam signal including reference radiation irradiation information corresponding thereto, the signal acquisition unit 110 ) Obtains a first beam signal including a reference respiration signal and reference radiation irradiation information from the radiation treatment system 2. In addition, when the radiation treatment system 2 performs the simulated treatment of the patient or performs the actual first treatment, the signal acquisition unit 110 includes the actual respiration signal and the actual radiation irradiation information from the radiation treatment system 2 A first beam signal may be obtained (S100).

Next, the learning unit 131 machine learns the respiration-linked radiation treatment evaluation model based on the acquired first respiration signal and the first beam signal (S200). The learning unit 131 learns a respiration-linked radiation therapy evaluation model based on machine learning, and machine learning is a high-level abstraction through a combination of several nonlinear transducers, such as large amounts of data or complex data. It is defined as a set of machine learning algorithms that attempt to summarize the core content or function in the interior. The learning unit 131 includes deep neural networks (DNN), convolutional neural networks (CNN), reccurent neural networks (RNN), and deep trust neural networks among models of deep learning. Networks, DBN) may be used.

As an embodiment, referring to FIG. 5, the learning unit 131 uses the first respiration signal S11 and the first beam signal S12 as the first signal S1 to evaluate respiration-linked radiation therapy. The evaluation data may be generated, and a respiratory-linked radiation treatment evaluation model (M) may be generated using the evaluation data (S201). In other words, the learning unit 131 generates evaluation data, which is a feature vector for evaluating the suitability of respiration-linked radiation therapy using the input first respiration signal S11 and the first beam signal S12, Using the evaluation data, it is possible to create a respiratory-linked radiation therapy evaluation model (M). Here, the respiratory-linked radiation treatment evaluation model (M) may be a model that derives the treatment efficiency (S210), treatment accuracy (S220), and equipment accuracy (S230) of the respiratory-linked radiation treatment as result values using the generated evaluation data. have.

More specifically, in order to derive the above-described result value, the learning unit 131 includes respiratory cycle information, average amplitude information, standard deviation information, peak value information, and error bar from the first breathing signal S11. error bar) information can be derived and used to generate evaluation data. In addition, the learning unit 131 is a beam signal, that is, from the reference radiation irradiation information and the actual radiation irradiation information, the total radiation setting time, the total radiation treatment time, the radiation beam irradiation time (beam on timing), the radiation beam stop time (beam off timing) ) Can be derived and used to generate evaluation data.

6, the learning unit 131 calculates a reference duty factor (Dref) from the above-described reference radiation irradiation information as shown in Equation 1 below (S211), and the actual duty ratio from the actual radiation irradiation information (real duty factor, Dreal) is calculated (S212).

[Equation 1]

Here, the total set time refers to the total planned treatment time according to the radiation treatment plan, and the set time to which the beam is applied refers to the time that the beam is irradiated according to the radiation treatment plan, specifically, the time that the beam is irradiated during the total set time. do.

In irradiating the radiation beam according to the radiation treatment plan, the most ideal plan is to continuously irradiate the beam for the entire set time. However, as described above, in the case of organs that are moved by the patient's breath, the beam must be irradiated or stopped according to the breathing cycle in order to accurately irradiate the beam to the target site, and through this, the ratio of the time the beam is irradiated out of the total set time. Phosphorus-based duty ratio (Dref) can be calculated. The reference duty ratio Dref may be set in advance by the radiation treatment plan, but the learning unit 131 may generate the reference duty ratio through learning using the planned breathing cycle among the patient's first breathing signals.

Referring to Equation 1 again, the total treatment time means the total treatment time performed in the simulated treatment or the actual treatment, and the treatment time applied to the beam is the time when the actual beam is irradiated, specifically, the beam is It means the time to be investigated.

Even if the patient's ideal respiratory cycle is trained and guided according to the radiation treatment plan, the respiratory cycle may become irregular during simulation or actual treatment. Since the radiation treatment system 2 irradiates the patient with radiation in conjunction with breathing, the time at which the actual radiation beam is irradiated due to an irregular actual breathing cycle is different from the planned irradiation time. The actual duty ratio (Dreal) can be calculated as the ratio of the time the beam is irradiated out of the total treatment time.

Next, the learning unit 131 may generate evaluation data, which is a treatment efficiency evaluation index, as shown in Equation 2 by using the reference duty ratio Dref and the actual duty ratio Dreal (S213).

[Equation 2]

The learning unit 131 may derive a result value for the respiration-linked radiation treatment efficiency using the evaluation data. That is, the learning unit 131 may compare the reference range with the calculated treatment efficiency evaluation data TE (S214) to derive a result value for evaluating the efficiency of the radiation treatment. In this case, the reference range is It may be set in advance by, but after receiving feedback from the patient's actual radiation treatment fit result, it may be learned and reset.

As an embodiment, when the treatment efficiency evaluation data TE is less than a first reference value (eg, 90%) set in advance, the learning unit 131 indicates that the respiratory-linked radiation therapy is inappropriate for the patient. Can be derived (S215). The learning unit 131 is a warning signal when the treatment efficiency evaluation data TE is equal to or greater than a pre-set first reference value (eg, 90%) and less than a second reference value (eg, 95%). May be derived as a result value (S216), and if it is greater than a preset second reference value (eg, 95%), a result value indicating that the corresponding respiratory-linked radiation therapy is suitable may be derived (S317). In other words, the learning unit 131 may generate a breathing-linked radiation therapy evaluation model having the above algorithm.

7 and 8, the learning unit 131 provides at least one of breathing period information, average amplitude information, standard deviation information, peak value information, and error bar information of a breathing signal. Using the baseline shift of the breathing signal can be generated as the evaluation data (S222). Here, the baseline of the breathing signal is the tendency of the breathing signal over time, for example, a line connecting the phases of the breathing signal at the same point in the repetitive cycle, or the peak value of the breathing cycle. It can be a line connecting them. The baseline shift means the degree to which the baseline is inclined, and may mean that the respiration signal over time is unstable. Or, as another embodiment, as shown in FIG. 8, when using a deep inspiration breath hold (DIBH) method, which is one of the breath-linked radiation treatment methods, the baseline is the patient's breathing. It may also be a line connecting values in a state of reference.

As described above, the respiration signal is derived by image processing and tracking analysis of the respiration image acquired by the imaging unit 203 by photographing the marker 230, and the indication that the baseline of the respiration signal is inclined is the indicator 230 This could mean that the location of has changed. The change in the position of the marker 230 may mean sagging of the patient's organs due to breathing. Therefore, in respiratory-linked radiation therapy, the above-described shift of the baseline means that the position of the target site is changed.

In this case, the learning unit 131 may first acquire and store an expected margin (EM) given when generating a planning target volume of a patient during a radiation treatment plan in advance (S221). The radiation treatment system (2) sets an expectation margin (EM) to provide margin so that the dose irradiated to the critical target volume (CTV) is not omitted even if an error occurs during radiation treatment in the radiation treatment plan. , When setting a moving tumor (internal target volume, ITV), an expectation margin (EM) may be set (see P1, P2, Fig. 8) to provide a margin.

The learning unit 131 may derive a result value for treatment accuracy by comparing the expected margin EM and a real margin RM derived from the baseline shift (S223). As shown in FIG. 8, the expected margin EM may be a preset range consisting of a first margin value P1 and a second margin value P2, and the actual margin RM derived from the baseline shift. In other words, if the degree to which the baseline is shifted exceeds the expected margin (EM), a result value indicating that the respiratory-linked radiation therapy is inappropriate for the patient may be derived (S224). Here, the expected margin (EM) may be composed of a first margin value (P1) and a second margin value (P2), which are ranges previously set by the medical staff, but after that, the learning unit 131 is suitable for respiration-linked radiation therapy of the patient. It is also possible to learn the range of the expected margin by receiving the result feedback. In addition, the learning unit 131 compares the expected margin (EM) with the actual margin (RM) (S323), and if the actual margin (RM) is less than the expected margin (EM), the corresponding respiratory-linked radiation therapy is appropriate for the patient. Results can be derived.

Meanwhile, referring to FIGS. 9 and 10, the learning unit 131 generates evaluation data by calculating a radiation irradiation time difference by using the reference radiation radiation information and the actual radiation radiation information (S231). Specifically, the reference radiation irradiation information provided from the radiation treatment plan includes reference beam irradiation time (Beam on _ref ) information set in correspondence with the reference respiratory cycle (bold arrow, see FIG. 10), and the actual radiation irradiation information is As applied radiation irradiation information, it includes information about the actual beam irradiation time (Beam on _real ) (thin arrow, see FIG. 10). The learning unit 131 may generate machine accuracy (MA) data based on a difference between the reference beam irradiation time and the actual beam irradiation time, as shown in Equation 3 below.

[Equation 3]

The equipment accuracy evaluation data (MA) represents an error caused by the difference between the planned irradiation time and the actual irradiation time, and can be used as an index to evaluate how accurate the equipment is.

The learning unit 131 may generate evaluation data MA for equipment accuracy, compare it with a preset error range (S223), and derive a result value indicating that the equipment accuracy is high and radiation treatment is suitable if it is smaller than this (S223). S225). However, the learning unit 131 may compare the equipment accuracy evaluation data MA with a preset error range and, if it is out of the range, may derive a result value indicating that the equipment accuracy is low and the respiratory-linked radiation therapy is inappropriate (S324).

Thereafter, the determination unit 133 acquires a second respiration signal and a second beam signal of a new patient or a new second respiration signal and a second beam signal of an existing patient, and the second respiration signal and the second beam signal, and the Using the respiratory-linked radiation treatment evaluation model, it is possible to determine whether the patient is suitable for respiratory-linked radiation treatment (S300, see FIG. 3), and provide the determination result to the radiation treatment system 2 (S400). The radiation treatment system 2 may completely reestablish or review the radiation treatment plan in response thereto. At this time, the radiation treatment system 2 may change the plan to a plan not to perform respiratory-linked radiation treatment. Alternatively, the radiation therapy system 2 may perform breathing training of the patient again according to other respiratory-linked radiation therapy.

As described above, in the system and method for evaluating respiration-linked radiation therapy using machine learning according to the embodiments of the present invention, after generating a respiration-linked radiation treatment evaluation model using the first respiration signal and the first beam signal, Using the second respiration signal and the second beam signal and the respiration-linked radiation treatment evaluation model, the patient's radiation-respiratory-linked radiation treatment can be automatically evaluated. The system and method for evaluating respiration-linked radiation therapy using machine learning according to the embodiments of the present invention can provide a patient with an evaluation result of whether or not respiration-linked radiation therapy is appropriate, and through this, a radiation plan is provided with radiation therapy suitable for the patient. Can be changed to improve the efficiency and accuracy of radiation treatment.

The embodiment according to the present invention described above may be implemented in the form of a computer program that can be executed through various components on a computer, and such a computer program may be recorded in a computer-readable medium. In this case, the medium may store a program executable by a computer. Examples of media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magnetic-optical media such as floptical disks, and And a ROM, RAM, flash memory, and the like, and may be configured to store program instructions.

Meanwhile, the computer program may be specially designed and configured for the present invention, or may be known and usable to those skilled in the computer software field. Examples of the computer program may include not only machine language codes such as those produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like.

The specific implementations described in the present invention are examples, and do not limit the scope of the present invention in any way. For brevity of the specification, descriptions of conventional electronic configurations, control systems, software, and other functional aspects of the systems may be omitted. In addition, the connection or connection members of the lines between the components shown in the drawings exemplarily represent functional connections and/or physical or circuit connections, and in an actual device, various functional connections that can be replaced or additionally It may be referred to as a connection, or circuit connections. In addition, if there is no specific mention such as "essential", "important", etc., it may not be an essential component for the application of the present invention.

Therefore, the spirit of the present invention is limited to the above-described embodiments and should not be defined, and all ranges equivalent to or equivalently changed from the claims to be described later as well as the claims to be described later are the scope of the spirit of the present invention. It will be said to belong to.

1: Respiratory interlocking radiation therapy evaluation system
2: radiation therapy system
110: signal acquisition unit
131: Learning Department
133: judgment unit

Claims (19)

A signal acquisition unit for acquiring a first beam signal including a first respiration signal and radiation irradiation information during respiration-linked radiation treatment from the radiation treatment system;
A learning unit for machine learning a respiratory interlocking radiation therapy evaluation model based on the acquired first respiration signal and the first beam signal; And
A determination unit that determines whether or not the respiration-linked radiation treatment is suitable for the patient based on the second respiration signal and the second beam signal obtained by the signal acquisition unit and the respiration-linked radiation treatment evaluation model. Respiratory interlocking radiation therapy evaluation system using learning. The method of claim 1,
The learning unit generates evaluation data for evaluating the respiratory-linked radiation therapy using the acquired first respiration signal and the first beam signal. The method of claim 2,
The learning unit uses the generated evaluation data to generate the respiratory-linked radiation treatment evaluation model that derives the treatment efficiency, treatment accuracy, and equipment accuracy of the respiratory-linked radiation therapy as a result value. Evaluation system. The method of claim 3,
The learning unit generates the evaluation data using any one of breathing cycle information, average amplitude information, standard deviation information, and peak value information derived from the first breathing signal to generate the respiratory-linked radiation therapy evaluation model. Machine learning, respiratory-linked radiation therapy evaluation system using machine learning. The method of claim 4,
The learning unit generates a baseline shift of the breathing signal using any one of the breathing period information, the average amplitude information, the standard deviation information, and the peak value information, using machine learning. Respiratory interlocked radiation therapy evaluation system. The method of claim 5,
The learning unit evaluates the treatment accuracy by comparing an expected margin given when generating a planning target volume of a patient during a radiation treatment plan and a real margin derived from the baseline shift. Respiratory-linked radiation therapy evaluation system using machine learning to derive results. The method of claim 3,
The first beam signal is the reference radiation irradiation information set in advance in response to the planned respiratory cycle of the patient during the radiation treatment plan, and the actual radiation irradiation performed in response to the actual respiratory cycle of the patient acquired in a simulated treatment or an actual treatment. Respiratory interlocking radiation therapy evaluation system using machine learning, including information. The method of claim 7,
The learning unit calculates a reference duty factor from the reference radiation irradiation information, calculates an actual duty ratio from the actual radiation irradiation information, generates the evaluation data, and uses the evaluation data to generate the respiration-linked radiation. Respiratory-linked radiation therapy evaluation system using machine learning that derives results for treatment efficiency. The method of claim 7,
The learning unit calculates a radiation irradiation time error using the reference radiation irradiation information and the actual radiation irradiation information, and uses the radiation irradiation time error to derive a result value for the equipment accuracy, breathing interlocking using machine learning Radiotherapy evaluation system. Acquiring a first beam signal including a first respiration signal and radiation irradiation information of a patient during respiratory-linked radiation treatment from a radiation treatment system;
Machine learning a respiratory interlocking radiation therapy evaluation model based on the acquired first respiration signal and the first beam signal;
Acquiring a second respiration signal and a second beam signal of an existing patient or a new patient;
Determining whether or not the respiration-linked radiation treatment is suitable for the patient based on the second respiration signal and the second beam signal, and the respiration-linked radiation treatment evaluation model; containing, respiration-linked radiation treatment using machine learning Assessment Methods. The method of claim 10,
The step of machine learning the respiratory-linked radiation therapy evaluation model,
Machine learning to generate evaluation data for evaluating the respiratory-linked radiation therapy using the acquired first respiration signal and the first beam signal, and generating the respiration-linked radiation therapy evaluation model using the evaluation data A method of evaluating respiratory-linked radiation therapy using. The method of claim 11,
The respiratory-linked radiation therapy evaluation model uses the generated evaluation data to derive treatment efficiency, treatment accuracy, and equipment accuracy of the respiratory-linked radiation therapy as result values, a method of evaluating respiratory-linked radiation therapy using machine learning. The method of claim 12,
The step of machine learning the respiratory-linked radiation therapy evaluation model,
The evaluation data is generated using any one of respiratory cycle information, average amplitude information, standard deviation information, and peak value information derived from the first respiration signal, and the respiratory-linked radiation using the evaluation data A method of evaluating respiratory-linked radiation therapy using machine learning, which is a machine learning treatment evaluation model. The method of claim 13,
The step of machine learning the respiratory-linked radiation therapy evaluation model,
Respiratory interlocking radiation using machine learning to generate a baseline shift of the breathing signal using any one of the breathing period information, the average amplitude information, the standard deviation information, and the peak value information Methods of evaluation of treatment. The method of claim 14,
The respiratory-linked radiation therapy evaluation model is a model that derives the evaluation result of the treatment accuracy by comparing the expected margin given when the patient's treatment target application is generated during the radiation treatment plan and the actual margin derived from the baseline shift. A method of evaluating respiratory-associated radiation therapy using learning. The method of claim 12,
The first beam signal is the reference radiation irradiation information set in advance in response to the planned respiratory cycle of the patient during the radiation treatment plan, and the actual radiation irradiation performed in response to the actual respiratory cycle of the patient acquired in a simulated treatment or an actual treatment. A method of evaluating respiratory-linked radiation therapy using machine learning, including information. The method of claim 16,
The step of machine learning the respiratory-linked radiation therapy evaluation model
Calculate a reference duty factor from the reference radiation irradiation information, calculate the actual duty ratio from the actual radiation irradiation information, generate the evaluation data, and use the evaluation data to determine the respiratory-linked radiation treatment efficiency. A method of evaluating respiratory-linked radiation therapy using machine learning to derive results for The method of claim 16,
The step of machine learning the respiratory-linked radiation therapy evaluation model,
Respiratory-linked radiation therapy evaluation using machine learning to calculate a radiation irradiation time error using the reference radiation irradiation information and the actual radiation irradiation information, and to derive a result value for the equipment accuracy using the radiation irradiation time error Way. A computer program stored in a medium to execute any one of the methods of claim 10 to 18 using a computer.

Images