阅读说明：本技术 机器人放射治疗设备剂量预测方法及装置 (Dosage prediction method and device for robot radiotherapy equipment ) 是由 刘博� 李晗 周付根 于 2021-10-11 设计创作，主要内容包括：本发明公开了一种机器人放射治疗设备剂量预测方法及装置,该方法包括：模型训练,根据患者的医学影像,建立患者模体；根据机器人放射治疗设备治疗头参数,利用放疗剂量第一计算方法计算患者模体单射野的剂量分布H；根据机器人放射治疗设备治疗头参数,利用放疗剂量第二计算方法计算患者模体单射野的剂量分布L；以L和患者的医学影像为输入,以H为输出,送入深度学习神经网络训练得到放疗剂量预测网络；剂量预测,利用放疗剂量第二计算方法计算得到任一患者模体单射野的剂量分布L*；获取放疗剂量预测网络输出端预测得到的任一患者模体单射野的剂量分布H*。利用深度学习网络实现了利用一种算法预测另一种算法输出计算结果的技术效果。(The invention discloses a dose prediction method and a device for robot radiotherapy equipment, wherein the method comprises the following steps: model training, namely establishing a patient model body according to the medical image of the patient; calculating the dose distribution H of the single radiation field of the phantom of the patient by utilizing a first calculation method of a radiotherapeutic agent according to the parameters of a treatment head of the robot radiotherapy equipment; calculating the dose distribution L of the single radiation field of the phantom of the patient by utilizing a second calculation method of the radiotherapeutic agent according to the parameters of the treatment head of the robot radiotherapy equipment; inputting the L and the medical image of the patient and outputting the H, and sending the L and the medical image into a deep learning neural network for training to obtain a radiotherapy dose prediction network; dose prediction, namely calculating the dose distribution L of a single radiation field of any patient phantom by using a second calculation method of the radiotherapeutic dose; and obtaining the dose distribution H of the single radiation field of any patient phantom predicted by the output end of the radiotherapy dose prediction network. The technical effect of predicting the calculation result output by one algorithm and the other algorithm by using the deep learning network is achieved.)

1. A method of dose prediction for a robotic radiation therapy device, comprising:

the training of the model is carried out,

establishing a patient phantom according to a medical image of a patient, wherein the phantom is constructed according to a mapping relation between a medical image pixel value and a physical material and electron density;

calculating the dose distribution H of the single radiation field of the phantom of the patient by utilizing a first calculation method of a radiotherapeutic agent according to the parameters of a treatment head of the robot radiotherapy equipment;

calculating the dose distribution L of the single radiation field of the patient phantom by using a second calculation method of the radiotherapeutic agent according to the parameters of the treatment head of the robot radiotherapy equipment;

the treatment head parameters comprise treatment head coordinates, target point coordinates, collimator size and radiation dosage;

inputting the L and the medical image of the patient as input, and outputting the H to a deep learning neural network for training to obtain a radiotherapy dose prediction network;

the prediction of the dosage is carried out,

establishing any patient phantom according to the medical image of any patient;

the medical image of any patient is sent into the input end of the radiotherapy dose prediction network, and the dose distribution L of the single radiation field of any patient phantom is calculated by utilizing a second calculation method of the radiotherapy dose^*；

Obtaining theDose distribution H of any patient model body single-field predicted by radiotherapy dose prediction network output end^*。

2. The method of claim 1, wherein the first calculation method of a radiotherapeutic agent is a monte carlo simulation method.

3. The method of claim 1, wherein the second calculation of the radiotherapeutic agent is a ray tracing method.

4. The method of claim 1, wherein the deep learning neural network is HD U-Net.

5. The method of claim 1, further comprising, before said training into the deep-learning neural network to obtain the radiation therapy dose prediction network: and calibrating the L and the H.

6. A robotic radiation treatment device dose prediction apparatus, comprising:

model training module for

Establishing a patient phantom according to a medical image of a patient, wherein the phantom is constructed according to a mapping relation between a medical image pixel value and a physical material and electron density;

calculating the dose distribution H of the single radiation field of the phantom of the patient by utilizing a first calculation method of a radiotherapeutic agent according to the parameters of a treatment head of the robot radiotherapy equipment;

calculating the dose distribution L of the single radiation field of the patient phantom by using a second calculation method of the radiotherapeutic agent according to the parameters of the treatment head of the robot radiotherapy equipment;

the treatment head parameters comprise treatment head coordinates, target point coordinates, collimator size and radiation dosage;

inputting the L and the medical image of the patient as input, and outputting the H to a deep learning neural network for training to obtain a radiotherapy dose prediction network;

a dose prediction module for

Establishing any patient phantom according to the medical image of any patient;

the medical image of any patient is sent into the input end of the radiotherapy dose prediction network, and the dose distribution L of the single radiation field of any patient phantom is calculated by utilizing a second calculation method of the radiotherapy dose^*；

Obtaining the dose distribution H of the single radiation field of any patient model body predicted by the output end of the radiotherapy dose prediction network^*。

7. A computing device, comprising: processor and memory storing a program, wherein the processor implements the method of any one of claims 1 to 5 when executing the program.

8. A computer-readable storage medium having a program stored thereon, wherein the program when executed implements the method of any of claims 1-5.

Technical Field

The invention relates to the technical field of radiotherapy dose calculation, in particular to a dose prediction method for robot radiotherapy equipment.

Background

Malignant tumors seriously harm human health, patients with seven-stage tumors need to receive radiation therapy, and physicians can make radiation therapy plans before radiation therapy. In the process of making a radiation treatment plan, dose calculation is an important link, and the efficiency of dose calculation determines the quality of making the radiation treatment plan. In the prior art, dose calculation methods fall into three major categories: factor-based algorithms, model-based algorithms, and monte carlo simulations. The factor-based algorithm uses a semi-empirical approach to solve the problems of tissue heterogeneity and surface curvature based on effective spatial dose measurements, has the advantage of fast computation speed, and does not require the differentiation of the subsequent energy transfer of photons and electrons in the patient. Factor-based algorithms are less accurate for heterosomes with energies greater than 6MV, where the scattering contribution is small and the effect of photon-induced electron motion can locally lead to high dose variations. The model-based algorithm calculates the patient dose distribution through the primary particle continuity and dose kernel, also with higher computational efficiency, and is more accurate than factor-based algorithms, especially in heterogeneous media. However, model-based algorithms still rely on approximations and only partially deal with the physics of microscopic absorption of energy delivered by a radiation field in the microscopic field. The Monte Carlo simulation calculates the dose distribution according to the physical process of particles in a computer-simulated substance, has higher precision in the field of dose calculation, and is often used for verifying the accuracy of other dose calculation algorithms, but the Monte Carlo dose calculation consumes more time and is difficult to meet the timeliness requirement in practical application. In summary, the current dose calculation method is difficult to achieve high precision and high efficiency at the same time.

Disclosure of Invention

In view of this, the invention provides a dose prediction method for a robot radiotherapy device, which introduces a deep learning neural network, and greatly improves the calculation efficiency while ensuring the accuracy of dose calculation so as to alleviate the defects of the prior art.

In a first aspect, the present invention provides a dose prediction method for a robotic radiation therapy device, comprising: model training, namely establishing a patient phantom according to a medical image of a patient, wherein the phantom is constructed according to a mapping relation between a pixel value of the medical image and a physical material and electron density; calculating the dose distribution H of the single radiation field of the phantom of the patient by utilizing a first calculation method of a radiotherapeutic agent according to the parameters of a treatment head of the robot radiotherapy equipment; calculating the dose distribution L of the single radiation field of the phantom of the patient by utilizing a second calculation method of the radiotherapeutic agent according to the parameters of the treatment head of the robot radiotherapy equipment; the treatment head parameters comprise treatment head coordinates, target point coordinates, collimator size and radiation dosage; inputting the L and the medical image of the patient and outputting the H, and sending the L and the medical image into a deep learning neural network for training to obtain a radiotherapy dose prediction network; dose prediction, namely establishing a phantom of any patient according to the medical image of any patient; the medical image of any patient is sent into the input end of the radiotherapy dose prediction network, and the dose distribution L of the single radiation field of any patient phantom is calculated by utilizing a second calculation method of the radiotherapy dose; and obtaining the dose distribution H of the single radiation field of any patient phantom predicted by the output end of the radiotherapy dose prediction network.

Further, the first calculation method of the radiotherapeutic dose is a monte carlo simulation method.

Further, the second calculation method of the radiotherapeutic dose is a ray tracing method.

Further, the deep learning neural network is HD U-Net.

Further, before the deep learning neural network training is carried out to obtain the radiotherapy dose prediction network, the method further comprises the following steps: and calibrating L and H.

In a second aspect, the present invention provides a dose prediction apparatus for a robotic radiation therapy device, comprising: the model training module is used for establishing a patient phantom according to a medical image of a patient, and the phantom is constructed according to the mapping relation between the pixel value of the medical image and the physical material and the electron density; calculating the dose distribution H of the single radiation field of the phantom of the patient by utilizing a first calculation method of a radiotherapeutic agent according to the parameters of a treatment head of the robot radiotherapy equipment; calculating the dose distribution L of the single radiation field of the phantom of the patient by utilizing a second calculation method of the radiotherapeutic agent according to the parameters of the treatment head of the robot radiotherapy equipment; the treatment head parameters comprise treatment head coordinates, target point coordinates, collimator size and radiation dosage; inputting the L and the medical image of the patient and outputting the H, and sending the L and the medical image into a deep learning neural network for training to obtain a radiotherapy dose prediction network; the dose prediction module is used for establishing any patient phantom according to the medical image of any patient; the medical image of any patient is sent into the input end of the radiotherapy dose prediction network, and the dose distribution L of the single radiation field of any patient phantom is calculated by utilizing a second calculation method of the radiotherapy dose; and obtaining the dose distribution H of the single radiation field of any patient phantom predicted by the output end of the radiotherapy dose prediction network.

In a third aspect, the invention provides a computing device comprising: a processor and a memory storing a program, the processor implementing the method of the first aspect when executing the program.

In a fourth aspect, the present invention provides a computer readable storage medium having stored thereon a program which, when executed, performs the method of the first aspect.

The invention has the following beneficial effects:

the technical scheme provided by the invention can have the following beneficial effects: the dose prediction method for the robot radiotherapy equipment is provided, and a deep learning neural network is trained by using medical images of a patient and dose distribution output by two dose calculation methods, so that the neural network can predict the dose distribution obtained by the other dose calculation method when the medical images of the patient and the dose distribution calculated by one dose calculation method are input. Therefore, the technical effect of predicting the output calculation result of the high-precision algorithm by using the low-precision algorithm is achieved, and the technical problem that the high precision and the high efficiency are difficult to achieve simultaneously in the prior art is solved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are one embodiment of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.

FIG. 1 is a flowchart illustrating a dose prediction method for a robotic radiation therapy device according to a first embodiment of the present invention;

FIG. 2 is a schematic structural diagram of a deep learning neural network HD U-Net according to a dose prediction method for a robotic radiation therapy device of the present invention;

FIG. 3 is a schematic structural diagram of a dose prediction device of a robotic radiation therapy device according to a second embodiment of the present invention;

fig. 4 is a schematic structural diagram of a computing device according to a third embodiment of the present invention.

Detailed Description

To make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and the described embodiments are some, but not all embodiments of the present invention.

The first embodiment is as follows:

fig. 1 is a flowchart illustrating a dose prediction method of a robotic radiation therapy device according to a first embodiment of the present invention, and as shown in fig. 1, the method includes the following two steps.

Step S101: and training a radiotherapy dose prediction model. Specifically, a patient phantom is established according to a medical image of a patient, and the phantom is established according to the mapping relation between the pixel value of the medical image and the physical material and the electron density; calculating the dose distribution H of the single radiation field of the phantom of the patient by utilizing a first calculation method of a radiotherapeutic agent according to the parameters of a treatment head of the robot radiotherapy equipment; calculating the dose distribution L of the single radiation field of the phantom of the patient by utilizing a second calculation method of the radiotherapeutic agent according to the parameters of the treatment head of the robot radiotherapy equipment; and (3) inputting the L and the medical image of the patient and outputting the H, and sending the L and the medical image into a deep learning neural network for training to obtain a radiotherapy dose prediction network. The treatment head parameters comprise the coordinates of the treatment head, the coordinates of a target point, the size of a collimator and the dosage of a radiation field.

Alternatively, the patient phantom may be three-dimensional data of a medical image of the patient. In engineering practice, medical image data are two-dimensional data, and three-dimensional data of medical images need to be established through a three-dimensional reconstruction method.

It should be noted that the first calculation method and the second calculation method of the radiation dose are both used for calculating the radiation dose distribution. The second calculation method has a feature of high calculation efficiency but low calculation accuracy, and on the contrary, the first calculation method has a feature of low calculation efficiency but high calculation accuracy. The deep learning neural network obtained by training is used for rapidly predicting the H obtained by the first calculation method by using the L obtained by the second calculation method. Optionally, the medical image of the patient is a CT image. In the field of radiation dose technology, a radiation field refers to the irradiation range of radioactive rays on the body surface of a patient, and a single radiation field refers to only one radiation field in radiotherapy. In engineering practice, the robot radiotherapy system has the characteristic of multi-field focusing irradiation, and the total dose field is obtained by superposing hundreds of groups of single-field dose distributions. The radiotherapy dose prediction network trained in the embodiment is based on a single radiation field, and the dose distribution calculation of multiple radiation fields only needs to superpose the obtained single radiation field distribution. A treatment head of a robot radiotherapy device is a radiotherapy device which can be used for radiotherapy of various tumors.

In an alternative embodiment, the first calculation method of the radiotherapeutic dose is a monte carlo simulation method. The Monte Carlo simulation calculates the dose distribution according to the physical process of particles in a computer simulation substance, has higher precision in the field of radiotherapy, but has lower calculation efficiency, the time required for calculating a single field is about 20 minutes, and if the calculation is carried out on multiple fields, the calculation time is difficult to meet the requirement of radiotherapy.

In an alternative embodiment, the second calculation method of the radiotherapeutic dose is a ray tracing method. The ray tracing method has simple algorithm, and the radiation particles radiated from the treatment head of the robot radiation treatment equipment are regarded as a plurality of rays, and the energy of each ray is propagated in a respective independent ray tube. And tracking the propagation of each ray, and adopting a vector superposition method to radiate the dose distribution. The ray tracing method has high calculation efficiency, usually tens of milliseconds, but has low calculation precision, and is difficult to meet the requirements of radiotherapy.

In an alternative embodiment, the deep learning neural network is HD U-Net. Fig. 2 is a schematic diagram of a deep learning neural network structure of a dose prediction method for a robotic radiotherapy apparatus according to a first embodiment of the present invention, and as shown in fig. 2, the HD U-Net includes five layers to reduce Feature size, and local and global characteristics are learned by using 2 × 2 × 2 Pooling layers (Pooling) between each layer and finally reducing the Feature Map (Feature Map) to 6 × 4 × 6 at the bottom layer. In each layer, a convolution kernel of size 3 × 3 × 3 is used, and Zero Padding (Zero Padding) is used to maintain the size of the feature. In the first half of the HD U-Net, 4 feature maps (filters) are generated per convolution step. In the remaining half, the number of signatures convolved per layer increases by 4 from bottom to top, except for the last convolution step. The last convolution step generates a channel as the final output. Since the output channel of the radiotherapy dose prediction network needs to have the same dimension as the input, the encoding and decoding network is an ideal architecture. The HD U-Net has the characteristics of automatic feature extraction, three-dimensional space information extraction, automatic optimization and the like, and is suitable for dose prediction. It includes a coding stage for extracting the data of input channel hierarchically and a decoding stage for reconstructing the dimension required by output channel, and the skipped connection is used between the convolution down-sampling stage and the deconvolution up-sampling stage of network. The skip connection connects the output of the early volume block with the input of the later volume block in the network. This doubles the size of the volume block, but reduces redundancy within the network and helps mitigate information loss caused by the multi-scale structure of the network architecture.

In an optional embodiment, before entering the deep learning neural network training to obtain the radiotherapy dose prediction network, the method further comprises: and calibrating L and H. It should be noted that calibration of the radiation treatment dose distribution allows calculation of the absorbed dose and dose equivalent.

In an alternative embodiment, the number of model trainings is at least 1. It should be noted that, in order to obtain high prediction accuracy, a deep learning neural network generally needs to be trained many times. Optionally, the termination condition of the model training is determined by the numerical characteristics of the loss function.

Step S102: dose prediction was performed using a predictive model. Specifically, according to the medical image of any patient, establishing any patient phantom; the medical image of any patient is sent into the input end of the radiotherapy dose prediction network, and the dose distribution L of the single radiation field of any patient phantom is calculated by utilizing a second calculation method of the radiotherapy dose; and obtaining the dose distribution H of the single radiation field of any patient phantom predicted by the output end of the radiotherapy dose prediction network.

In one embodiment, the Monte Carlo method takes 20 minutes to calculate, the ray-tracing method takes only 0.05s, and the radiation dose prediction network takes 0.4s to predict the high-accuracy dose for the same set of test data. This embodiment can therefore greatly reduce the time required for dose calculation.

Example two:

the embodiment of the present invention provides a dose prediction device for a robot radiotherapy apparatus, which is mainly used for executing the dose prediction method for a robot radiotherapy apparatus provided by the above-mentioned contents of the embodiment of the present invention, and the following describes the dose prediction device for a robot radiotherapy apparatus provided by the embodiment of the present invention in detail.

Fig. 3 is a schematic structural diagram of a dose prediction device of a robotic radiation therapy device according to a second embodiment of the present invention. As shown in fig. 3, the robotic radiation therapy device dose prediction apparatus 200 includes the following modules:

the model training module 201 is used for establishing a patient phantom according to a medical image of a patient, wherein the phantom is constructed according to a mapping relation between a pixel value of the medical image and a physical material and electron density; calculating the dose distribution H of the single radiation field of the phantom of the patient by utilizing a first calculation method of a radiotherapeutic agent according to the parameters of a treatment head of the robot radiotherapy equipment; calculating the dose distribution L of the single radiation field of the phantom of the patient by utilizing a second calculation method of the radiotherapeutic agent according to the parameters of the treatment head of the robot radiotherapy equipment; the treatment head parameters comprise treatment head coordinates, target point coordinates, collimator size and radiation dosage; inputting the L and the medical image of the patient and outputting the H, and sending the L and the medical image into a deep learning neural network for training to obtain a radiotherapy dose prediction network;

a dose prediction module 202 for creating any patient phantom based on medical images of any patient; the medical image of any patient is sent into the input end of the radiotherapy dose prediction network, and the dose distribution L of the single radiation field of the phantom of any patient is calculated by utilizing a second calculation method of the radiotherapy dose^*(ii) a Obtaining the dose distribution H of any patient model body single-field predicted by the output end of the radiotherapy dose prediction network^*。

Example three:

the embodiment of the invention also provides the computing equipment. As shown in fig. 4, the city area correlation calculation apparatus 300 of this embodiment includes: a processor 301, a memory 302, and programs stored in the memory 302 and executable on the processor 301. The processor 301, when executing the program, implements the steps in the various robotic radiation treatment device dose prediction method embodiments described above, such as steps S101 and S102 shown in fig. 1. Alternatively, the processor 301, when executing the program, implements the functions of the modules in the above-described embodiments of the apparatus, such as the modules in fig. 3, to implement the robotic radiation therapy device dose prediction apparatus.

Illustratively, the program may be partitioned into one or more modules that are stored in the memory 302 and executed by the processor 301 to implement the present invention. The one or more modules may be a series of program instruction segments capable of performing certain functions, which are used to describe the execution of the program in a computing device. For example, the program may be partitioned into a model training module and a dose prediction module.

The specific functions of each module are as follows: the model training module is used for establishing a patient phantom according to a medical image of a patient, and the phantom is constructed according to the mapping relation between the pixel value of the medical image and the physical material and the electron density; calculating the dose distribution H of the single radiation field of the phantom of the patient by utilizing a first calculation method of a radiotherapeutic agent according to the parameters of a treatment head of the robot radiotherapy equipment; calculating the dose distribution L of the single radiation field of the phantom of the patient by utilizing a second calculation method of the radiotherapeutic agent according to the parameters of the treatment head of the robot radiotherapy equipment; the treatment head parameters comprise treatment head coordinates, target point coordinates, collimator size and radiation dosage; inputting the L and the medical image of the patient and outputting the H, and sending the L and the medical image into a deep learning neural network for training to obtain a radiotherapy dose prediction network; the dose prediction module is used for establishing any patient phantom according to the medical image of any patient; the medical image of any patient is sent into the input end of the radiotherapy dose prediction network, and the dose distribution L of the single radiation field of any patient phantom is calculated by utilizing a second calculation method of the radiotherapy dose; and obtaining the dose distribution H of the single radiation field of any patient phantom predicted by the output end of the radiotherapy dose prediction network.

The computing device can be a single chip microcomputer system, a desktop computer, a notebook computer, a palm computer, a cloud server and other computing devices. The computing device may include, but is not limited to, a processor, a memory. Those skilled in the art will appreciate that the schematic diagrams are merely examples and do not constitute a limitation of computing devices, and may include more or fewer components than those shown, or some components may be combined, or different components, e.g., the computing devices may also include input-output devices, etc.

The Processor may be a Micro Control Unit (MCU), a Central Processing Unit (CPU), or other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component, etc. The general purpose processor may be a microprocessor or the processor may be any conventional processor or the like that is the control center for the computing device and that connects the various parts of the overall computing device using various interfaces and lines.

The memory can be used for storing the programs and/or modules, and the processor can realize various functions of the dose prediction method and the device of the robot radiotherapy equipment by operating or executing the programs and/or modules stored in the memory and calling the data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, a phonebook, etc.) created according to the use of the cellular phone, and the like. In addition, the memory may include high speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), at least one magnetic disk storage device, a Flash memory device, or other volatile solid state storage device.

Example four:

the modules integrated by the dose prediction means of the robotic radiation therapy device, if implemented in the form of software functional units and sold or used as separate products, can be stored in a computer readable storage medium. Based on such understanding, all or part of the flow of the method according to the embodiments of the present invention may also be implemented by a computer program, which may be stored in a computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method embodiments may be implemented. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, usb disk, removable hard disk, magnetic disk, optical disk, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution medium, and the like. It should be noted that the computer readable medium may contain content that is subject to appropriate increase or decrease as required by legislation and patent practice in jurisdictions, for example, in some jurisdictions, computer readable media does not include electrical carrier signals and telecommunications signals as is required by legislation and patent practice.

Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present invention, which are used for illustrating the technical solutions of the present invention and not for limiting the same, and the protection scope of the present invention is not limited thereto, although the present invention is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the embodiments of the present invention, and they should be construed as being included therein. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.