Radiotherapy dose distribution prediction method and system based on deep learning

Abstract

The invention provides a radiation therapy dose distribution prediction method and system based on deep learning, and relates to the technical field of medical image processing, and the method comprises the steps: collecting a data set containing a CT scanning image of a patient, radiation therapy plan parameters and actual dose distribution, taking an electron density map and a three-dimensional flux map as the input of a training deep learning model, and carrying out the prediction of the radiation therapy dose distribution. Taking the endangered organ mask image, the radiotherapy target area mask image and the three-dimensional dose distribution image corresponding to the CT scanning image of the patient as the output of the model, and training to obtain an optimal dose distribution prediction model; and predicting the three-dimensional dose distribution of the new patient by using the trained dose distribution prediction model. According to the invention, by using the data of the patient and the radiotherapy plan information in the past, the dose distribution can be predicted more quickly and accurately, the delineation dependence of the doctor on the region of interest is reduced, and the treatment efficiency and the result stability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method and system for predicting radiation therapy dose distribution based on deep learning. Background Art

[0002] In tumor radiotherapy, about 70% of cancer patients need radiotherapy during treatment, and dose calculation is a key step in developing a treatment plan. Accurate dose calculation is essential to ensure that patients receive appropriate radiation doses and minimize damage to normal tissues. A high-quality radiotherapy often needs to meet the four principles of dosimetry: high radiation dose to the tumor target area, low radiation dose to the organs at risk around the tumor, accurate dose to the tumor area, and uniform dose distribution within the tumor area. To achieve this goal, the dose distribution of the target area must be accurately obtained before radiotherapy.

[0003] Among them, Intensity Modulated Radiotherapy (IMRT) is an advanced radiotherapy technology that achieves better tumor treatment effects while reducing damage to surrounding healthy tissues by precisely controlling the dose distribution of radiation. IMRT uses a multi-leaf collimator (MLC) to change the intensity and shape of each beam of radiation, so that tumors in different locations can receive different doses of radiation, thereby achieving "intensity modulation" of the dose.

[0004] Traditional physical dose calculation models, such as AAA (Anisotropic Analytical Algorithm) and Monte Carlo algorithm, although they have good performance in dose calculation accuracy, usually require a long calculation time, especially for complex treatment plans. Deep learning models can achieve accuracy comparable to traditional physical algorithms with shorter inference time under the same computing power.

[0005] In tumor radiotherapy, when designing a radiotherapy plan, the radiotherapist needs to outline the normal tissue in the images scanned during the simulation positioning phase. This is the premise and basis of radiotherapy. However, this step is quite time-consuming and highly dependent on the doctor's experience level. Doctors are prone to making mistakes during the high-intensity outlining work. Radiotherapy plan design is mainly based on a large amount of prior knowledge, using the anatomical information of medical images and the clinician's knowledge and experience in radiation rules and algorithms to achieve dose distribution, incident intensity and angle control. In recent years, deep learning-based methods have used prior knowledge in different ways. They can automatically learn personalized image features for different areas of the body and different patients from raw data (CT images, target areas and contours of organs at risk, etc.).

[0006] For example, in the three-dimensional dose distribution prediction method for intensity modulated radiotherapy plan based on deep network learning disclosed in patent CN110085298A, it is proposed to collect effective intensity modulated radiotherapy plan data to form a case database; extract the three-dimensional anatomical structure and dose characteristics of each person's area of ​​interest from the case database, and build a deep convolutional neural network to learn the model between anatomy and dose, and finally predict the dose distribution of new patients. For example, in a three-dimensional dose prediction method for radiotherapy based on deep learning disclosed in patent CN112837782B, by collecting past patient CT data and corresponding dose distribution maps, a U-Net neural network model is established for learning, so that the CT data of the patient to be predicted is input into the U-Net neural network model to obtain a predicted three-dimensional dose distribution map. For example, in a radiation therapy dose prediction system and method based on artificial intelligence disclosed in patent CN116779173B, a 3D-Unet deep learning model is established for learning, and a two-stage module is constructed, which is divided into a first model training module and a second model training module. The first model training module inputs the patient's organs at risk, target area, and distance map from the target area to the outside, outputs the intermediate dose distribution, and inputs the second model. The second model training module outputs the final dose distribution.

[0007] The above-mentioned learning methods based on deep learning must collect historical data for learning. However, in clinical practice, linear accelerators need to be calibrated once in a while, and doctors’ treatment experience and plans are also being optimized, so the proposed technology should be continuously learned. In addition, as some methods consider more prior information, such as the patient’s organs at risk and target areas, they ignore that the dose distribution is the result of energy deposition from the interaction between photons and the human body. Therefore, the flux based on the radiotherapy plan is also an important prior knowledge for dose distribution calculation, which many methods do not consider. Finally, when deep learning is used for new patients, some methods rely on the delineation of organs at risk and target areas based on prior information, but the delineation of these areas of interest often takes doctors a lot of time and is often prone to errors. Sometimes, in order to automatically obtain these prior areas, it is necessary to design a multi-stage model to achieve this, which increases the time for dose distribution calculation. Summary of the invention

[0008] In view of the shortcomings of the above-mentioned prior art, the purpose of the present invention is to propose a method and system for predicting radiotherapy dose distribution based on prior information and deep learning. By utilizing previous patient data and radiotherapy plan information, a faster and more accurate dose distribution prediction can be achieved, reducing the doctor's reliance on the delineation of the region of interest, and improving treatment efficiency and result stability.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] Based on the above objectives, in a first aspect, the present invention provides a method for predicting radiation therapy dose distribution based on deep learning, comprising the following steps:

[0011] Collect data sets containing patient CT scan images, radiotherapy planning parameters, and actual dose distribution;

[0012] Perform non-uniform tissue correction on the patient's CT scan image, convert the CT value into an electron density map of relative electron density values, and obtain a three-dimensional flux map of irradiation from the radiotherapy plan parameters;

[0013] The electron density map and the 3D flux map are used as the input for training the deep learning model, and the mask map of the organ at risk, the target mask map of the radiotherapy, and the 3D dose distribution map corresponding to the patient's CT scan image are used as the output of the model;

[0014] Dividing the data set into a training set, a validation set, and a test set according to a set ratio; training a deep learning model based on the data set, and adjusting training parameters according to the training results to obtain an optimal dose distribution prediction model;

[0015] The trained dose distribution prediction model is used to predict the three-dimensional dose distribution of new patients. After the tumor radiotherapy is completed, the result data is imported into the database and the model parameters are fine-tuned and trained.

[0016] As a further solution of the present invention, converting the CT value into an electron density map of relative electron density values ​​comprises the following steps:

[0017] Acquire a CT scan image of a patient, and perform image preprocessing on the CT scan image of the patient;

[0018] Using the calibration data of the CT scanner, the conversion relationship between CT value and relative electron density is established;

[0019] According to the conversion relationship, the CT value of each voxel is converted into a relative electron density, and each voxel of the CT scan image is traversed to convert the CT value into a relative electron density value;

[0020] The converted relative electron density values ​​are reorganized into the same three-dimensional structure as the original CT image to obtain the electron density map.

[0021] As a further embodiment of the present invention, obtaining a three-dimensional flux map of irradiation from radiotherapy planning parameters comprises the following steps:

[0022] Read the radiotherapy plan file and extract the key parameters of the radiotherapy plan, including beam, photon energy, dose rate, irradiation angle, leaf position (MLC), collimator, and source-to-skin distance (SSD);

[0023] Obtain the spatial resolution and size of the patient's CT scan image, and obtain the geometric and physical parameter information of the angle, position, energy, and dose rate of each radiation field;

[0024] The flux distribution of each field in three-dimensional space is calculated using the geometric parameters and physical parameters of the field, and a three-dimensional flux map is constructed based on the flux superposition of the field parameters;

[0025] The irradiated area of ​​each radiation field is projected into the patient's three-dimensional CT image space, and the flux of each voxel is calculated using the photon transport model.

[0026] As a further solution of the present invention, when obtaining the three-dimensional flux map of irradiation from the radiotherapy plan parameters, the three-dimensional flux map is calculated based on the axis-aligned bounding box projection of ray tracing.

[0027] As a further solution of the present invention, when training a deep learning model based on the data set, the model parameters are adjusted through a back-propagation algorithm to minimize the error between the predicted organs at risk, target areas and doses and the actual ones, and the model is continuously optimized using cross-validation, and the optimal dose distribution prediction model is obtained by adjusting the learning rate.

[0028] As a further solution of the present invention, when using the trained dose distribution prediction model to predict the three-dimensional dose distribution of a new patient, the gamma pass rate is used as an evaluation index to evaluate the consistency between the calculated dose distribution and the actual measured dose distribution to determine whether the evaluation point passes.

[0029] As a further solution of the present invention, when judging whether the evaluation point is passed, if δ(r e ,r r )γ(r e ,r r )≤1, it means that the evaluation point has passed, otherwise it has not passed; where δ(r e ,r r ) represents r e and r r The dose difference at e ,r r ) represents r e and r r The spatial distance at which e represents any point in the evaluation dose distribution, r r represents any point in the reference dose distribution.

[0030] As a further solution of the present invention, when the gamma pass rate is used as the evaluation index, the calculation of the gamma pass rate includes the following steps:

[0031] For each evaluation point re , calculate the dose difference δ(r e ,r r )=D e (r e )-D r (r r ) and spatial distance r(r e ,r r )=|r e -r r |; Among them, D e (r e ) represents the evaluation dose distribution, D r (r r ) represents the reference dose distribution;

[0032] Calculate each evaluation point r e Gamma value in γ(r e ,r r ) indicates whether the evaluation point passes the threshold function;

[0033] Count the proportion of gamma values ​​of 1 among all evaluation points involved in the calculation, that is, the gamma pass rate.

[0034] As a further scheme of the present invention, the calculation of the gamma pass rate includes a comparison of the dose difference and the spatial distance, which meets the requirements of industry standards (such as AAPM TG-218 and YY / T 0889-2013). The gamma pass rate requirements are set as: the composite field (3% / 2mm, 10% dose threshold) is higher than 90%, and the single field (3% / 3mm, 10% dose threshold) is higher than 93%.

[0035] In a second aspect, the present invention provides a radiation therapy dose distribution prediction system based on deep learning, comprising:

[0036] A data acquisition module, used to collect data sets including patient CT scan images, radiotherapy planning parameters and actual dose distribution;

[0037] An image processing module is used to perform non-uniform tissue correction on the patient's CT scan image and convert the CT value into an electron density map of relative electron density values;

[0038] A flux calculation module, used for obtaining a three-dimensional flux map of irradiation from radiotherapy plan parameters;

[0039] A model training module, for dividing the data set into a training set, a validation set and a test set according to a set ratio, using the electron density map and the three-dimensional flux map as inputs for training the deep learning model, and using the mask map of organs at risk, the mask map of the target area of ​​radiotherapy and the three-dimensional dose distribution map corresponding to the patient's CT scan image as outputs of the model; training the deep learning model based on the data set, and adjusting the training parameters according to the training results to obtain an optimal dose distribution prediction model;

[0040] A dose prediction module is used to predict the three-dimensional dose distribution of a new patient using the trained dose distribution prediction model;

[0041] The model adjustment module is used to import the result data into the database and perform fine-tuning training on the model parameters after the tumor radiotherapy is completed.

[0042] In another aspect of the present invention, a computer device is provided, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, any one of the above-mentioned methods for predicting radiation therapy dose distribution based on deep learning according to the present invention is executed.

[0043] In another aspect of the present invention, a computer-readable storage medium is provided, storing computer program instructions, which, when executed, implement any of the above-mentioned methods for predicting radiation therapy dose distribution based on deep learning according to the present invention.

[0044] Compared with the prior art, the method and system for predicting radiation therapy dose distribution based on deep learning proposed in the present invention have the following beneficial effects:

[0045] 1. The present invention introduces the three-dimensional flux map of radiotherapy as a priori information input into the dose calculation model. Compared with the original technology that only considers human anatomical information and tumor area, it adds information on the actual irradiation conditions of the radiotherapy equipment, which can make the input information obtained by the deep learning model more comprehensive, and thus the accuracy of the results can be more precise.

[0046] 2. During the model training process, the areas of organs at risk and target areas are used as the output of the deep learning model for learning, rather than the original technology as the input of the model, so that the model can learn the characteristics of these different human tissues and reduce dependence on input during use.

[0047] 3. While collecting previous patient data for learning, subsequent continuous learning is added. Compared with the original technology that only obtains one model, which leads to insufficient generalization ability, the present invention can continuously optimize according to different clinical scenarios, with higher and higher accuracy.

[0048] In summary, the radiotherapy dose distribution prediction method and system based on deep learning of the present invention significantly improves the quality and effect of radiotherapy through high-precision, high-efficiency and personalized prediction means, provides patients with more accurate and reliable treatment plans, and also provides clinicians with powerful auxiliary tools, promoting the advancement and development of radiotherapy technology.

[0049] These and other aspects of the present application will be more concise and understandable in the following description of the embodiments. It should be understood that the above general description and the following detailed description are only exemplary and explanatory and cannot limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the following briefly introduces the drawings required for use in the exemplary embodiments or related technical descriptions. The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0051] Figure 1 This is a flowchart of a method for predicting radiation therapy dose distribution based on deep learning according to an embodiment of the present invention.

[0052] Figure 2 This is a flowchart of converting CT values ​​into relative electron density values ​​in a method for predicting radiation therapy dose distribution based on deep learning in an embodiment of the present invention.

[0053] Figure 3 This is a flow chart of obtaining a three-dimensional flux map of irradiation from radiotherapy plan parameters in a method for predicting radiotherapy dose distribution based on deep learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] Below, the present application is further described in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form a new embodiment.

[0055] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the embodiments of the present invention are further described in detail below in combination with specific embodiments and with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0056] It should be noted that all expressions using "first" and "second" in the embodiments of the present invention are intended to distinguish two non-identical entities or non-identical parameters with the same name. It can be seen that "first" and "second" are only for the convenience of expression and should not be understood as limitations on the embodiments of the present invention. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, other steps or units inherent to a process, method, system, product or device that includes a series of steps or units.

[0057] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0058] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0059] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0060] Since existing target-range dose distribution learning methods consider more prior information, such as the patient's organs at risk and target area, but ignore that dose distribution is the result of energy deposition from the interaction between photons and the human body, the flux based on the radiotherapy plan is also an important prior knowledge for dose distribution calculation, which many methods do not consider. Finally, when deep learning is used for new patients, some methods rely on the delineation of organs at risk and target areas based on prior information, but the delineation of these areas of interest often takes doctors a lot of time and is often prone to errors. Sometimes, in order to automatically obtain these prior areas, it is necessary to design a multi-stage model to achieve this, which increases the time for dose distribution calculation.

[0061] The present invention proposes a method and system for predicting radiation therapy dose distribution based on deep learning. The deep learning model is based on previous patient data for learning. It not only uses the organs at risk, target areas and other areas of previous patients, but also uses the radiotherapy plan to generate the patient's three-dimensional flux map as important prior information for the input of the deep learning model. In actual use of this technology, the organs at risk and target areas of previous patients are used as model outputs for learning, thereby avoiding dependence on doctors to outline areas of interest when using this technology, saving time and avoiding unstable results. In addition, after each treatment, the patient's online data is used to train the dose distribution calculation model, allowing the model to continue online learning and continuously improve accuracy.

[0062] See also Figure 1 As shown, an embodiment of the present invention provides a method for predicting radiation therapy dose distribution based on deep learning, the method comprising the following steps:

[0063] Step S10, collecting a data set including the patient's CT scan image, radiotherapy plan parameters and actual dose distribution.

[0064] In this step, a large number of data sets containing patient CT scan images, radiotherapy plan parameters and actual dose distribution are collected; the CT images are standardized, including denoising and contrast enhancement, to ensure the consistency of the input data.

[0065] Step S20, performing non-uniform tissue correction on the patient's CT scan image, converting the CT values ​​into an electron density map of relative electron density values, and obtaining a three-dimensional flux map of irradiation from the radiotherapy plan parameters.

[0066] In this step, see Figure 2 As shown, converting the CT value into an electron density map of relative electron density values ​​includes the following steps:

[0067] Step S201, obtaining a CT scan image of a patient, and performing image preprocessing on the CT scan image of the patient;

[0068] Step S202: using the calibration data of the CT scanner to establish a conversion relationship between the CT value and the relative electron density;

[0069] Step S203, converting the CT value of each voxel into a relative electron density according to the conversion relationship, traversing each voxel of the CT scan image, and converting the CT value into a relative electron density value;

[0070] Step S204: reorganize the converted relative electron density values ​​into a three-dimensional structure identical to the original CT image to obtain an electron density map.

[0071] In this embodiment, see Figure 3 As shown, obtaining a three-dimensional flux map of irradiation from radiotherapy planning parameters includes the following steps:

[0072] Step S211, read the radiotherapy plan file, and extract the key parameters of the radiotherapy plan parameters including beam, photon energy, dose rate, irradiation angle, leaf position (MLC), collimator, and source-to-skin distance (SSD).

[0073] Step S212: Acquire the spatial resolution and size of the patient's CT scan image, and acquire the geometric and physical parameter information of the angle, position, energy, and dose rate of each radiation field.

[0074] Step S213: using the geometric parameters and physical parameters of the field, calculate the flux distribution of each field in the three-dimensional space, and obtain a constructed three-dimensional flux map based on the flux superposition of the field parameters.

[0075] Step S214: project the irradiation area of ​​each radiation field into the three-dimensional CT image space of the patient, and calculate the flux of each voxel using a photon transmission model.

[0076] Among them, the three-dimensional flux of irradiation is obtained from the radiotherapy planning parameters (such as beam angle, collimator opening degree, etc.) so as to be input into the deep learning model.

[0077] In this embodiment, the patient's scanned CT image is corrected for non-uniform tissue. The correction process is to convert the CT value into a relative electron density value. In clinical practice, when a CT machine is used for radiotherapy planning, the relationship data between the CT value and its relative electron density must be measured, which can be obtained from the CT-relative electron density (CT-RED) curve obtained from the radiotherapy system. Therefore, the CT image value is first converted here.

[0078] The radiotherapy planning file contains the parameters of the treatment process. Some radiotherapy systems can directly obtain the three-dimensional irradiation flux map. If it cannot be directly exported from the system, when the three-dimensional irradiation flux map is obtained from the radiotherapy planning parameters, the three-dimensional flux map is calculated based on the axis-aligned bounding box projection (AABB) of ray tracing.

[0079] Step S30: using the electron density map and the three-dimensional flux map as inputs for training the deep learning model, and using the mask map of organs at risk, the target area mask map of radiotherapy, and the three-dimensional dose distribution map corresponding to the patient's CT scan image as outputs of the model.

[0080] Step S40, dividing the data set into a training set, a validation set and a test set according to a set ratio; training a deep learning model based on the data set, and adjusting the training parameters according to the training results to obtain an optimal dose distribution prediction model.

[0081] In this step, when training the deep learning model based on the data set, the model parameters are adjusted through the back propagation algorithm to minimize the error between the predicted organs at risk, target areas and doses and the actual ones, and the model is continuously optimized by cross-validation, and the optimal dose distribution prediction model is obtained by adjusting the learning rate.

[0082] Step S50: Use the trained dose distribution prediction model to predict the three-dimensional dose distribution of a new patient. After the tumor radiotherapy is completed, the result data is imported into the database, and the model parameters are fine-tuned and trained.

[0083] In this step, when the trained dose distribution prediction model is used to predict the three-dimensional dose distribution of a new patient, the gamma pass rate is used as an evaluation index to evaluate the consistency between the calculated dose distribution and the actual measured dose distribution and to determine whether the evaluation point passes.

[0084] When judging whether the evaluation point is passed, if δ(r e ,r r )γ(r e ,r r )≤1, it means that the evaluation point has passed, otherwise it has not passed; where δ(r e ,r r ) represents r e and r r The dose difference at e ,r r ) represents r e and r r The spatial distance at which e represents any point in the evaluation dose distribution, r r represents any point in the reference dose distribution.

[0085] When the gamma pass rate is used as the evaluation indicator, the calculation of the gamma pass rate includes the following steps:

[0086] For each evaluation point r e , calculate the dose difference δ(r e ,r r )=D e (r e )-D r (r r ) and spatial distance r(r e ,r r )=|r e-r r |; Among them, D e (r e ) represents the evaluation dose distribution, D r (r r ) represents the reference dose distribution;

[0087] Calculate each evaluation point r e Gamma value in γ(r e ,r r ) indicates whether the evaluation point passes the threshold function;

[0088] Count the proportion of gamma values ​​of 1 among all evaluation points involved in the calculation, that is, the gamma pass rate.

[0089] Among them, the calculation of gamma pass rate includes the comparison of dose difference and spatial distance, which meets the requirements of industry standards (such as AAPMTG-218 and YY / T 0889-2013), namely: TG-218 issued by the American Association of Physics in Medicine (AAPM) and the pharmaceutical industry standard of the People's Republic of China YY / T 0889-2013 "Performance and Test Methods of Intensity Modulated Radiation Therapy Planning System". The gamma pass rate requirements are set as: composite field (3% / 2mm, 10% dose threshold) is higher than 90%, and single field (3% / 3mm, 10% dose threshold) is higher than 93%.

[0090] The dose distribution prediction model is integrated into the radiotherapy planning system to realize automatic dose calculation; after each patient is treated, the patient's data is added to the database, the model is fine-tuned and trained, and the model is continuously optimized and improved. More comprehensive clinical prior information is used for model learning, so that the accuracy can be higher. Because the present invention takes into account that the dose is inseparable from the actual radiotherapy irradiation process, it supplements important prior information, which can make the dose distribution calculation more accurate; it also has a faster calculation speed. The input in the present invention only needs to input CT images and three-dimensional flux maps of radiotherapy plans, which are naturally obtained during the radiotherapy process. There is no need for doctors to manually outline or other methods to obtain the process of outlining the organs at risk and the target area, so the speed is faster. Moreover, the method of the present invention is continuously optimized, and the model is continuously fine-tuned and trained using the patient's online data, getting rid of the problem that various methods are fixed and no longer optimized.

[0091] The radiotherapy dose distribution prediction method based on deep learning of the present invention introduces the three-dimensional flux map of radiotherapy as a priori information input into the dose calculation model. Compared with the original technology that only considers human anatomical information and tumor area, it adds information on the actual irradiation conditions of the radiotherapy equipment, which can make the input information obtained by the deep learning model more comprehensive, and thus the accuracy of the results can be more precise.

[0092] During the model training process, the areas of organs at risk and target areas are used as the output of the deep learning model for learning, rather than the original technology as the input of the model, so that the model can learn the characteristics of these different human tissues and reduce the dependence on input during use. While collecting previous patient data for learning, subsequent continuous learning is added. Compared with the original technology that only obtains one model, which leads to insufficient generalization ability, the present invention can continuously optimize according to different clinical scenarios, and the accuracy is getting higher and higher.

[0093] It should be noted that the above figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0094] It should be understood that, although described in a certain order, these steps are not necessarily performed in sequence in the above order. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, a part of the steps of the present embodiment may include a plurality of steps or a plurality of stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a part of the steps or stages in other steps or other steps.

[0095] In a second aspect of the embodiments of the present invention, the present invention further provides a radiation therapy dose distribution prediction system based on deep learning, comprising:

[0096] A data acquisition module, used to collect data sets including patient CT scan images, radiotherapy planning parameters and actual dose distribution;

[0097] An image processing module is used to perform non-uniform tissue correction on the patient's CT scan image and convert the CT value into an electron density map of relative electron density values;

[0098] A flux calculation module, used for obtaining a three-dimensional flux map of irradiation from radiotherapy plan parameters;

[0099] A model training module, for dividing the data set into a training set, a validation set and a test set according to a set ratio, using the electron density map and the three-dimensional flux map as inputs for training the deep learning model, and using the mask map of organs at risk, the mask map of the target area of ​​radiotherapy and the three-dimensional dose distribution map corresponding to the patient's CT scan image as outputs of the model; training the deep learning model based on the data set, and adjusting the training parameters according to the training results to obtain an optimal dose distribution prediction model;

[0100] A dose prediction module is used to predict the three-dimensional dose distribution of a new patient using the trained dose distribution prediction model;

[0101] The model adjustment module is used to import the result data into the database and perform fine-tuning training on the model parameters after the tumor radiotherapy is completed.

[0102] Through the above detailed steps and examples, the deep learning-based radiotherapy dose distribution prediction system of the present invention is used to execute the steps of the deep learning-based radiotherapy dose distribution prediction method in the above embodiment, which will not be repeated here.

[0103] The deep learning-based radiotherapy dose distribution prediction system of the present invention significantly improves the quality and effect of radiotherapy through high-precision, high-efficiency and personalized prediction methods, provides patients with more accurate and reliable treatment plans, and also provides clinicians with powerful auxiliary tools, promoting the advancement and development of radiotherapy technology.

[0104] According to a third aspect of an embodiment of the present invention, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the method of any one of the above embodiments is implemented.

[0105] The computer device includes a processor and a memory, and may also include: an input system and an output system. The processor, memory, input system and output system may be connected via a bus or other means, and the input system may receive input digital or character information, and generate signal input related to the migration of radiation therapy dose distribution prediction based on deep learning. The output system may include display devices such as display screens.

[0106] As a non-volatile computer-readable storage medium, the memory can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as the program instructions / modules corresponding to the method for predicting radiation therapy dose distribution based on deep learning in the embodiment of the present application. The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the use of the method for predicting radiation therapy dose distribution based on deep learning, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the local module via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0107] The processor may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor is generally used to control the overall operation of the computer device. In this embodiment, the processor is used to run the program code stored in the memory or process data. The processors of multiple computer devices of the computer device of this embodiment execute various functional applications and data processing of the server by running non-volatile software programs, instructions and modules stored in the memory, that is, implementing the steps of the radiation therapy dose distribution prediction method based on deep learning of the above method embodiment.

[0108] It should be understood that, to the extent that they do not conflict with each other, all the embodiments, features and advantages described above for the deep learning-based radiotherapy dose distribution prediction method according to the present invention are also applicable to the deep learning-based radiotherapy dose distribution prediction and storage medium according to the present invention.

[0109] It will also be appreciated by those skilled in the art that various exemplary logic blocks, modules, circuits and algorithm steps described in conjunction with the disclosure herein can be implemented as electronic hardware, computer software or a combination of the two. In order to clearly illustrate this interchangeability of hardware and software, a general description has been given to the functions of various schematic components, blocks, modules, circuits and steps. Whether this function is implemented as software or hardware depends on specific applications and the design constraints imposed on the entire system. Those skilled in the art can implement the function in various ways for each specific application, but this implementation decision should not be interpreted as causing a departure from the disclosed scope of the embodiments of the present invention.

[0110] Finally, it should be noted that the computer-readable storage medium (e.g., memory) herein may be a volatile memory or a non-volatile memory, or may include both a volatile memory and a non-volatile memory. As an example and not by way of limitation, a non-volatile memory may include a read-only memory (ROM), a programmable ROM (PROM), an electrically programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), or a flash memory. A volatile memory may include a random access memory (RAM), which may act as an external cache memory. As an example and not by way of limitation, RAM may be obtained in a variety of forms, such as synchronous RAM (DRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct Rambus RAM (DRRAM). The storage devices of the disclosed aspects are intended to include, but are not limited to, these and other suitable types of memory.

[0111] The various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure herein may be implemented or executed using the following components designed to perform the functions herein: a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination of these components. A general purpose processor may be a microprocessor, but alternatively, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, for example, a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP, and / or any other such configuration.

[0112] The above is an exemplary embodiment disclosed by the present invention, but it should be noted that, without departing from the scope disclosed by the embodiment of the present invention, various changes and modifications may be made. The functions, steps and / or actions of the method according to the disclosed embodiment described herein do not need to be performed in any particular order. In addition, although the elements disclosed by the embodiment of the present invention may be described or required in individual form, it may also be understood as multiple unless explicitly limited to the singular.

[0113] It should be understood that, as used herein, the singular form "a" or "an" is intended to include the plural form as well, unless the context clearly supports an exception. It should also be understood that, as used herein, "and / or" refers to any and all possible combinations of one or more of the items listed in association. The serial numbers of the embodiments disclosed in the above embodiments of the present invention are for description only and do not represent the advantages and disadvantages of the embodiments.

[0114] A person skilled in the art should understand that the discussion of any of the above embodiments is only exemplary and is not intended to imply that the scope of the disclosure of the embodiments of the present invention is limited to these examples; under the idea of ​​the embodiments of the present invention, the technical features in the above embodiments or different embodiments can also be combined, and there are many other changes in different aspects of the above embodiments of the present invention, which are not provided in detail for the sake of simplicity. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present invention should be included in the protection scope of the embodiments of the present invention.

Claims

1. A method for predicting radiation therapy dose distribution based on deep learning, characterized in that: The method comprises the following steps: Collect data sets containing patient CT scan images, radiotherapy planning parameters, and actual dose distribution; Perform non-uniform tissue correction on the patient's CT scan image, convert the CT value into an electron density map of relative electron density values, and obtain a three-dimensional flux map of irradiation from the radiotherapy plan parameters; The electron density map and the 3D flux map are used as the input for training the deep learning model, and the mask map of the organ at risk, the target mask map of the radiotherapy, and the 3D dose distribution map corresponding to the patient's CT scan image are used as the output of the model; Dividing the data set into a training set, a validation set, and a test set according to a set ratio; training a deep learning model based on the data set, and adjusting training parameters according to the training results to obtain an optimal dose distribution prediction model; The trained dose distribution prediction model is used to predict the three-dimensional dose distribution of new patients. After the tumor radiotherapy is completed, the result data is imported into the database and the model parameters are fine-tuned and trained.

2. The method for predicting radiation therapy dose distribution based on deep learning according to claim 1, characterized in that: The CT value is converted into an electron density map of relative electron density values, including the following steps: Acquire a CT scan image of a patient, and perform image preprocessing on the CT scan image of the patient; Using the calibration data of the CT scanner, the conversion relationship between CT value and relative electron density is established; According to the conversion relationship, the CT value of each voxel is converted into a relative electron density, and each voxel of the CT scan image is traversed to convert the CT value into a relative electron density value; The converted relative electron density values ​​are reorganized into the same three-dimensional structure as the original CT image to obtain the electron density map.

3. The method for predicting radiation therapy dose distribution based on deep learning according to claim 1, characterized in that: Obtaining a three-dimensional flux map of irradiation from radiotherapy planning parameters includes the following steps: Read the radiotherapy plan file and extract the key parameters of the radiotherapy plan, including radiation field, photon energy, dose rate, irradiation angle, leaf position, radiation field shape, and source-to-skin distance; Obtain the spatial resolution and size of the patient's CT scan image, and obtain the geometric and physical parameter information of the angle, position, energy, and dose rate of each radiation field; The flux distribution of each field in three-dimensional space is calculated using the geometric parameters and physical parameters of the field, and a three-dimensional flux map is constructed based on the flux superposition of the field parameters; The irradiated area of ​​each radiation field is projected into the patient's three-dimensional CT image space, and the flux of each voxel is calculated using the photon transport model.

4. The method for predicting radiation therapy dose distribution based on deep learning according to claim 3, characterized in that: When a three-dimensional flux map of irradiation is obtained from the radiation therapy planning parameters, the three-dimensional flux map is calculated based on the axis-aligned bounding box projection of ray tracing.

5. The method for predicting radiation therapy dose distribution based on deep learning according to claim 1, characterized in that: When training the deep learning model based on the data set, the model parameters are adjusted through the back propagation algorithm to minimize the error between the predicted organs at risk, target areas and doses and the actual ones. The model is continuously optimized using cross-validation, and the optimal dose distribution prediction model is obtained by adjusting the learning rate.

6. The method for predicting radiation therapy dose distribution based on deep learning according to claim 1, characterized in that: When using the trained dose distribution prediction model to predict the three-dimensional dose distribution of a new patient, the gamma pass rate is used as an evaluation index to assess the consistency between the calculated dose distribution and the actual measured dose distribution and to determine whether the evaluation point passes.

7. The method for predicting radiation therapy dose distribution based on deep learning according to claim 6, characterized in that: When judging whether the evaluation point is passed, if δ(r e , r r )γ(r e , r r )≤1, it means that the evaluation point has passed, otherwise it has not passed; where δ(r e , r r ) represents r e and r r The dose difference at e , r r ) represents r e and r r The spatial distance at which e represents any point in the evaluation dose distribution, r r represents any point in the reference dose distribution.

8. The method for predicting radiation therapy dose distribution based on deep learning according to claim 7, characterized in that: When the gamma pass rate is used as the evaluation indicator, the calculation of the gamma pass rate includes the following steps: For each evaluation point r e , calculate the dose difference δ(r e , r r )=D e (r e )-D r (r r ) and spatial distance r(r e , r r )=|r e -r r |; Among them, D e (r e ) represents the evaluation dose distribution, D r (r r ) represents the reference dose distribution; Calculate the gamma value of each evaluation point re in γ(r e , r r ) indicates whether the evaluation point passes the threshold function; Count the proportion of gamma values ​​of 1 among all evaluation points involved in the calculation, that is, the gamma pass rate.

9. A radiation therapy dose distribution prediction system based on deep learning, characterized in that: Used to perform the radiotherapy dose distribution prediction method based on deep learning as described in any one of claims 1 to 8, the system comprising: A data acquisition module, used to collect data sets including patient CT scan images, radiotherapy planning parameters and actual dose distribution; An image processing module is used to perform non-uniform tissue correction on the patient's CT scan image and convert the CT value into an electron density map of relative electron density values; A flux calculation module, used to obtain a three-dimensional flux map of irradiation from radiotherapy plan parameters; A model training module, for dividing the data set into a training set, a validation set and a test set according to a set ratio, using the electron density map and the three-dimensional flux map as inputs for training the deep learning model, and using the mask map of organs at risk, the mask map of the target area of ​​radiotherapy and the three-dimensional dose distribution map corresponding to the patient's CT scan image as outputs of the model; training the deep learning model based on the data set, and adjusting the training parameters according to the training results to obtain an optimal dose distribution prediction model; The dose prediction module is used to predict the three-dimensional dose distribution of new patients using the trained dose distribution prediction model.

10. The radiation therapy dose distribution prediction system based on deep learning according to claim 9, characterized in that: The radiation therapy dose distribution prediction system also includes a model adjustment module, which is used to import the result data into the database after the tumor radiation therapy is completed, and to perform parameter fine-tuning training of the model.

Images