A radiotherapy plan generation method, device, terminal equipment and storage medium

Abstract

The application discloses a radiotherapy plan generation method and device, a terminal equipment and a storage medium, wherein the method comprises the following steps: acquiring CT image data and dose distribution data of a patient; inputting the CT image data and the dose distribution data into a trained radiotherapy machine parameter calculation model, so that the radiotherapy machine parameter calculation model calculates and outputs MLC positions, Jaw positions and MU values of each control point of a radiotherapy machine according to the CT image data and the dose distribution data; and generating a radiotherapy machine operation file according to the MLC positions, the Jaw positions and the MU values of all the control points; wherein the radiotherapy machine operation file is used for directly inputting a radiotherapy machine to control the radiotherapy machine to execute a radiotherapy plan. The application can automatically generate a radiotherapy plan which can be directly executed by the radiotherapy machine according to the CT image data and the dose distribution data, thereby saving labor cost and improving radiotherapy plan generation efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of radiotherapy, and in particular to a radiotherapy plan generation method and device, a terminal device and a storage medium. BACKGROUND

[0002] At present, the most commonly used in the design of radiotherapy plan is still the TPS treatment planning system. The obtained CT image or dose prediction scheme is input into the TPS, and the prescription dose of the doctor is taken as the target. The radiotherapy plan parameters that can be directly used and meet the clinical requirements, that is, the final usable treatment scheme, are obtained through the continuous modification, trial and error and reverse optimization of the physicist. Moreover, because of the difference in professional ability of the physicist and the use of different equipment, the learning and training period of the radiotherapy plan design is relatively long, and it is difficult to improve the design level of the dose planner in a short time, so that the design quality and consistency of the treatment plan cannot be guaranteed.

[0003] In the field of automatic radiotherapy research, many dose prediction models have been developed, which can automatically complete the dose prediction part of the radiotherapy plan. This saves part of the time of the radiotherapy process to a certain extent. However, only dose prediction is not enough, because it cannot be directly delivered to the radiotherapy machine. The radiotherapy machine cannot directly recognize the dose prediction model or the dose distribution map. Only by using specific parameters such as arm position, collimator angle, treatment window angle, leaf position sequence, sub-field weight and the like in the form of corresponding files can the radiotherapy machine be input to the clinical use, and this process still needs the participation of the physicist, which requires a lot of time and labor cost.

[0004] In summary, the prior art only completes the automation of the dose prediction part, and cannot generate a radiotherapy plan that can be directly used in the radiotherapy machine, which needs human participation and is low in efficiency. SUMMARY

[0005] The present application provides a radiotherapy plan generation method, device, terminal device and storage medium to solve the technical problem of low efficiency caused by the inability of the prior art to generate a radiotherapy plan that can be directly used in the radiotherapy machine.

[0006] In order to solve the above technical problems, the present application provides a radiotherapy plan generation method, comprising:

[0007] obtaining CT image data and dose distribution data of a patient;

[0008] inputting the CT image data and the dose distribution data into the trained radiotherapy machine parameter calculation model, so that the radiotherapy machine parameter calculation model calculates and outputs MLC position, Jaw position and MU value of each control point of the radiotherapy machine according to the CT image data and the dose distribution data;

[0009] generating a radiotherapy machine operation file according to the MLC position, the Jaw position and the MU value of all control points; wherein the radiotherapy machine operation file is used for directly inputting the radiotherapy machine to control the radiotherapy machine to execute the radiotherapy plan.

[0010] As a preferred solution, the radiotherapy machine parameter calculation model comprises a preprocessing module, an encoding module, a processing module and a decoding module.

[0011] The preprocessing module is configured to preprocess the dose distribution data.

[0012] The encoding module is configured to generate anatomical space dose distribution data according to the CT image data and the preprocessed dose distribution data; wherein the anatomical space dose distribution data is used to describe the spatial distribution of dose in the patient's body.

[0013] The processing module is configured to generate dose projection data of each control point according to the anatomical space dose distribution data.

[0014] The decoding module is configured to decode the dose projection data of each control point to obtain the MLC position, the Jaw position and the MU value of each control point.

[0015] As a preferred solution, the preprocessing of the dose distribution data comprises:

[0016] matching the CT image data and the dose distribution data to determine the isocenter point of the CT image data and the dose distribution data;

[0017] cropping the dose distribution data according to the preset size requirement with the isocenter point as the center of the cropping area;

[0018] resampling the cropped dose distribution data according to the preset dimension size to obtain the preprocessed dose distribution data.

[0019] As a preferred solution, the generation of the anatomical space dose distribution data according to the CT image data and the preprocessed dose distribution data comprises:

[0020] matching and fusing the CT image data and the dose distribution data to determine the anatomical space position information corresponding to each voxel in the dose distribution data in the CT image data and the dose information corresponding to each voxel in the dose distribution data.

[0021] generate the anatomic spatial dose distribution data according to the anatomic spatial position information and the dose distribution information corresponding to each voxel.

[0022] As a preferred solution, the generating of the dose projection data of each control point according to the anatomic spatial dose distribution data comprises:

[0023] For each control point, all the anatomic spatial dose distribution data in the projection direction of the ray beam path of the radiotherapy machine at the control point are projected on the control point to obtain the dose projection data of the control point.

[0024] As a preferred solution, the division process of the control points comprises:

[0025] obtain anatomic spatial data of a patient under a radiotherapy machine;

[0026] divide the patient body space corresponding to the anatomic spatial data into a plurality of sub-space regions around the longitudinal axis of the patient, and set a control point for each sub-space region.

[0027] As a preferred solution, the obtaining of the dose distribution data comprises:

[0028] obtain CT image data, OAR mask image data and PTV image data of a patient;

[0029] input the CT image data, OAR mask image data and PTV image data into a trained dose distribution prediction model, so that the dose distribution prediction model predicts the dose distribution according to the CT image data, OAR mask image data and PTV image data, and outputs dose distribution data.

[0030] On the basis of the above-mentioned embodiments, another embodiment of the present application provides a radiotherapy plan generation device, comprising a data acquisition module, a radiotherapy machine parameter calculation module and a radiotherapy machine operation file generation module.

[0031] The data acquisition module is configured to acquire CT image data and dose distribution data of a patient.

[0032] The radiotherapy machine parameter calculation module is configured to input the CT image data and dose distribution data into a trained radiotherapy machine parameter calculation model, so that the radiotherapy machine parameter calculation model calculates and outputs the MLC position, Jaw position and MU value of each control point of the radiotherapy machine according to the CT image data and dose distribution data.

[0033] The radiotherapy machine operation file generation module is configured to generate a radiotherapy machine operation file according to the MLC position, the Jaw position and the MU value of all control points, wherein the radiotherapy machine operation file is configured to be directly input into a radiotherapy machine to control the radiotherapy machine to execute the radiotherapy plan.

[0034] On the basis of the above-mentioned embodiments, a terminal device is provided in another embodiment of the application, which comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the radiotherapy plan generation method described in the above-mentioned embodiments of the application when executing the computer program.

[0035] On the basis of the above-mentioned embodiments, a storage medium is provided in another embodiment of the application, which comprises a stored computer program, wherein the storage medium controls a device where the storage medium is located to execute the radiotherapy plan generation method described in the above-mentioned embodiments of the application when the computer program is running.

[0036] Compared with the prior art, the embodiments of the application have the following beneficial effects:

[0037] The CT image data and the dose distribution data of a patient are acquired, and the CT image data and the dose distribution data are input into a trained radiotherapy machine parameter calculation model, so that the radiotherapy machine parameter calculation model calculates and outputs the MLC position, the Jaw position and the MU value of each control point of a radiotherapy machine according to the CT image data and the dose distribution data, and generates a radiotherapy machine operation file according to the MLC position, the Jaw position and the MU value of all control points, wherein the radiotherapy machine operation file is configured to be directly input into a radiotherapy machine to control the radiotherapy machine to execute a radiotherapy plan. The embodiments of the application can automatically generate a radiotherapy plan which can be directly input into a radiotherapy machine and executed according to CT image data and dose distribution data, realize the automation of the most time-consuming and labor-consuming process in the radiotherapy plan formulation, save the labor cost, and improve the radiotherapy plan generation efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 is a flowchart of a radiotherapy plan generation method provided by an embodiment of the application;

[0039] Figure 2 is an architecture diagram of a pre-module and a post-module of a dose distribution prediction model provided by an embodiment of the application;

[0040] Figure 3 is an architecture diagram of a radiotherapy machine parameter calculation model provided by an embodiment of the application;

[0041] Figure 4 is a control point subspace region division schematic diagram provided by an embodiment of the application;

[0042] Figure 5 Figure 1 is a structural schematic diagram of a radiotherapy plan generation device according to an embodiment of the present application. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0044] Embodiment one

[0045] Please refer to Figure 1 Figure 2 is a flowchart of a radiotherapy plan generation method according to an embodiment of the present application, which comprises the following steps.

[0046] S1, acquiring CT image data and dose distribution data of a patient.

[0047] In a preferred embodiment, the dose distribution data is acquired, comprising:

[0048] acquiring CT image data, OAR mask image data and PTV image data of the patient;

[0049] inputting the CT image data, OAR mask image data and PTV image data into a trained dose distribution prediction model, so that the dose distribution prediction model predicts the dose distribution according to the CT image data, OAR mask image data and PTV image data, and outputs dose distribution data.

[0050] It should be noted that the dose distribution data is used to describe the 3D dose distribution in the patient's body space. The dose distribution prediction model takes the CT image data, OAR mask image data and PTV image data of the patient as input, and takes the dose distribution data as output.

[0051] The dose distribution prediction model is a two-layer U-net architecture, which is composed of a front module, a processing block and a rear module, wherein the front module and the rear module are both U-net architectures.

[0052] The front module uses the CT image, OAR mask image and PTV image of the patient as input, and can predict a rough 3D dose distribution.

[0053] The processing block stacks and integrates the rough 3D dose output by the front module as a channel with the input multiple channels together as the input of the rear module.

[0054] The post-module is used to further fine-tune the prediction result of the pre-module, and outputs the final prediction result, i.e., the dose distribution data.

[0055] Reference is made to Figure 2 The architecture diagram of the pre-module and the post-module of the dose distribution prediction model provided by an embodiment of the present application is shown in FIG. 1. The U-net is a kind of full convolutional neural network, which is composed of a down-sampling path and an up-sampling path as well as skip concatenation, so as to bring the features of lower layers to the up-sampling path to preserve more details. The 3D-U-net has one down-sampling path and one up-sampling path. Each path has 5 levels (with dimensions of 128x128x128, 64x64x64, 32x32x32, 16x16x16, 8x8x8) respectively. The number of channels gradually expands from 16 to 256 in the down-sampling path. The down-sampling path is composed of similar modules: 3x3x3 convolutional layer, instance normalization (IN) layer and rectified linear unit (RELU). The down-sampling adopts 3x3x3 convolutional layer with a step of 2. After each down-sampling operation, the resolution of the feature image is reduced by half. The number of feature maps is doubled. The resolution of the feature image is gradually restored by using the up-sampling path to generate a high-resolution predicted dose image. The up-sampling module is mainly composed of repeated up-sampling layers, skip concatenation, 3x3x3 convolutional layer, IN layer and RELU. In order to prevent the loss of features in the down-sampling process, the original image of the corresponding size of the encoding layer of the down-sampling is transmitted to the decoding layer block of the up-sampling through the way of skip connection, because the earlier high-resolution feature images from the up-sampling module can help to extract the position information. These skip concatenations are to supplement the details lost in the down-sampling operation. We use trilinear interpolation scaling and 3x3x3 convolutional layer to up-sample the feature images. On the basis of the decoding module, the feature images are gradually up-sampled to the original resolution, and the predicted dose image is generated after 1x1x1 layer convolution. Since the dose value cannot be negative, the negative prediction value will be replaced by 0, and the final output is a 3D dose distribution data with a dimension of 128x128x128.

[0056] In the training process of the dose distribution prediction model, CT image data, OAR mask image data and PTV image data of a plurality of patients are collected as sample data, and corresponding dose distribution data is collected, from which spatial dose information is extracted, pre-processed into a NumPy array as sample labels. The CT image data is first cropped between -1024 and 1500HU, and then normalized by 1000HU. The OAR image and the PTV image are marked using a binary mask. The model outputs a dose image normalized to 70Gy.

[0057] The U-net architecture is deeper than most other 3D dose prediction models, and due to the increased difficulty of optimization with network depth, we used deep supervision for model optimization. The mean absolute error (MAE) between the predicted dose and the ground truth dose was used as the loss function. The loss value between the predicted dose result and the label is calculated as follows:

[0058]

[0059] where D Q (i) represents the dose value of the i-th voxel predicted by the front module, D H (i) represents the dose value of the i-th voxel predicted by the back module; GT(i) represents the true dose value of the i-th voxel; N represents the total number of voxels; a represents a weight factor, and the value is 0.5.

[0060] S2, input the CT image data and dose distribution data into the trained radiotherapy machine parameter calculation model, so that the radiotherapy machine parameter calculation model calculates and outputs the MLC position, Jaw position and MU value of each control point of the radiotherapy machine according to the CT image data and dose distribution data.

[0061] In a preferred embodiment, the control point division process comprises:

[0062] Obtaining the anatomical space data of the patient under the radiotherapy machine;

[0063] Dividing the patient body space corresponding to the anatomical space data into a plurality of sub-space regions around the longitudinal axis of the patient, and setting a control point for each sub-space region.

[0064] It should be noted that the division of the control points is first around the longitudinal axis of the patient, that is, the axis pointing from the head of the patient to the foot, and the number of control points optimized according to the double-arc treatment is generally from 179° counterclockwise to 181°, and then from 181° clockwise back to 179°, 178 control points per single-arc, a total of 2X178 control points, and the angle distribution between each control point is about 2°, but the control point position is basically fixed, and then a control point is set for each sub-space region. Please refer to Figure 4 A control point sub-space region division schematic diagram provided by an embodiment of the present application is shown in the figure, which divides 180 sub-space regions with the patient in a supine position lying above the gantry angle of the treatment machine at 0 degrees as the starting point, that is, every 2 degrees is a sub-space region, and the schematic diagram only draws the control points around the gantry 90°.

[0065] In the training process of the radiotherapy machine parameter calculation model, CT image data and dose distribution data of a plurality of patients are collected as samples, and the MLC position, Jaw position and MU value of each control point of the corresponding radiotherapy machine are collected and converted into a three-dimensional numpy array as sample labels. The input data is normalized using the min-max normalization technique, which scales the values to the range of 0 to 1. The output normalization includes scaling the MLC and Jaw to values between -1 and 1, and scaling the MU to values between 0 and 1. The L1 loss of the MLC position, Jaw position and MU is calculated respectively. After weighting the MLC loss by a factor of 100, the losses are summed, and when we study the model, we emphasize learning this property to protect the patient's critical organs and ensure the patient's health after the operation. The learning rate is adjusted according to the single-cycle strategy, and the Lion optimization model is used, with an initial learning rate of 10-4 and a weight decay of 10-3.

[0066] The hyperparameter adjustment is used to determine the optimal value of the model hyperparameters. The loss function is calculated as follows:

[0067] F total = 100F MLC + F Jaw + F MU .

[0068] In the formula, F total represents the total loss function value; F MLC represents the loss function value of the MLC decoder head; F Jaw represents the loss function value of the Jaw decoder head; and F MU represents the loss function value of the MU decoder head.

[0069] In a preferred embodiment, the radiotherapy machine parameter calculation model comprises a preprocessing module, an encoding module, a processing module and a decoding module.

[0070] The preprocessing module is configured to preprocess the dose distribution data.

[0071] The encoding module is configured to generate anatomical space dose distribution data based on the CT image data and the preprocessed dose distribution data; wherein the anatomical space dose distribution data is used to describe the spatial distribution of the dose in the patient's body.

[0072] The processing module is configured to generate dose projection data for each control point based on the anatomical space dose distribution data.

[0073] The decoding module is configured to decode the dose projection data for each control point to obtain the MLC position, Jaw position and MU value for each control point.

[0074] Please refer toFigure 3 An architecture diagram of a radiotherapy machine parameter calculation model provided by an embodiment of the present application.

[0075] It should be noted that the input of the radiotherapy machine parameter calculation model is CT image data and dose distribution data. The original size of the dose distribution data is 512mm*512mm*512mm, the image dimension is (128, 128, 128), and the original size of the voxel is 4mm*4mm*4mm. The CT image data has a layer thickness of 3mm and an in-plane resolution of 0.81mm.

[0076] In a preferred embodiment, the pre-processing of the dose distribution data includes:

[0077] The CT image data and the dose distribution data are matched to determine the isocenter of the CT image data and the dose distribution data.

[0078] The dose distribution data is cropped according to a preset size requirement with the isocenter as the center of the cropping region.

[0079] The cropped dose distribution data is resampled according to a preset dimension size to obtain pre-processed dose distribution data.

[0080] In this embodiment, the CT image data and the dose distribution data are matched to determine the isocenter of the CT image data and the dose distribution data. Then, the dose distribution data is cropped according to a preset size requirement with the isocenter as the center of the cropping region, and the preset size is 400mm*400mm*400mm. Next, the cropped dose distribution data is resampled according to a preset dimension size, and the preset dimension size is (128, 128, 320). The size of the voxel after resampling is 3.125mm*3.125mm*1.250mm. The number of voxels of the pre-processed dose distribution data is increased, and the dose information data contained in the increased voxels is obtained through three-dimensional linear interpolation.

[0081] In a preferred embodiment, the generation of the anatomical space dose distribution data from the CT image data and the pre-processed dose distribution data includes:

[0082] The CT image data and the dose distribution data are matched and fused to determine the corresponding anatomical space position information of each voxel in the dose distribution data in the CT image data and the corresponding dose information in the dose distribution data.

[0083] The anatomical space dose distribution data is generated according to the corresponding anatomical space position information and dose distribution information of each voxel.

[0084] It should be noted that the encoding module is mainly composed of 3D convolution layer, BatchNorm3D, 3D Max pooling layer and activation function Relu layer, which is used to extract spatial dose information, i.e. anatomical spatial dose distribution data.

[0085] In a preferred embodiment, the generating dose projection data of each control point according to the anatomical spatial dose distribution data comprises:

[0086] For each control point, all the anatomical spatial dose distribution data in the projection direction of the ray beam path of the radiotherapy machine at the control point are projected on the control point to obtain the dose projection data of the control point.

[0087] In this embodiment, the projection operation is performed at each control point position, and the projection direction is along the ray beam path (the direction of the treatment beam generated by the accelerator during treatment) to project at the perspective angle of the field direction, and the 3D dose distribution data is reconstructed into 2D dose distribution data at each control point by performing the projection operation along the corresponding direction at each control point.

[0088] In order to prevent the opposite control points (rotating 1π) from being represented identically, such as the front and back of the patient (0 degrees and 180 degrees), a three-dimensional tensor gradient field is introduced to simulate the absorption of the light beam when passing through the patient, and at each control point, the same direction as the projection is along the axis parallel to the beam path from 1 to 0, by multiplying the dose with the gradient field tensor, the same side voxel is given a greater weight than the opposite side voxel, and the 2D image generated at each control point is stacked in the order of the control point to form the final input array of the decoder.

[0089] S3, generating a radiotherapy machine operation file according to the MLC position, the Jaw position and the MU value of all control points; wherein the radiotherapy machine operation file is used to directly input the radiotherapy machine to control the radiotherapy machine to execute the radiotherapy plan.

[0090] The decoding module is composed of three independent decoder heads, which are respectively used to calculate the MLC position, the Jaw position and the MU value.

[0091] The MLC and Jaw decoder heads both follow the same design principle: 8 convolution blocks with residual connections to gradually reduce the data to the shape of the expected output: MLC 2X 178x60x2, Jaw 2X 178x2x1, and the data is processed using 3D convolution layer, BatchNorm3D layer and LeakyRelu activation function.

[0092] The MU decoder head consists of six convolutional blocks, followed by an average pooling layer and an output convolutional layer. A tanh output activation function is used for the MLC and Jaw decoders, and for the MU decoder, the output of the final convolutional layer is used directly, with dimensions: 2 X 178 X 1 X 1. A threshold is set for the MU decoder to prevent negative outputs.

[0093] Finally, the data output by the decoder is saved in DICOM format as a DICOM RT plan file for subsequent secondary dose review and plan quality confirmation before delivery to a treatment system device (such as a LINAC).

[0094] Embodiment Two

[0095] Please refer to Figure 2 A structure diagram of a radiotherapy plan generation device provided by an embodiment of the present application, comprising: a data acquisition module, a radiotherapy machine parameter calculation module, and a radiotherapy machine operation file generation module.

[0096] The data acquisition module is configured to acquire CT image data and dose distribution data of a patient.

[0097] The radiotherapy machine parameter calculation module is configured to input the CT image data and the dose distribution data into a trained radiotherapy machine parameter calculation model, so that the radiotherapy machine parameter calculation model calculates and outputs MLC positions, Jaw positions, and MU values of each control point of a radiotherapy machine according to the CT image data and the dose distribution data.

[0098] The radiotherapy machine operation file generation module is configured to generate a radiotherapy machine operation file according to the MLC positions, the Jaw positions, and the MU values of all control points, wherein the radiotherapy machine operation file is used to directly input a radiotherapy machine to control the radiotherapy machine to perform a radiotherapy plan.

[0099] Embodiment Three

[0100] Correspondingly, an embodiment of the present application provides a terminal device, which comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the radiotherapy plan generation method provided in the above-mentioned embodiments of the present application when executing the computer program.

[0101] Embodiment Four

[0102] Correspondingly, an embodiment of the present application provides a storage medium, which comprises a stored computer program, wherein the storage medium controls a device where the storage medium is located to execute the radiotherapy plan generation method provided in the above-mentioned embodiments of the present application when the computer program runs.

[0103] It should be noted that the apparatus embodiments described above are merely illustrative, and the units described as separate units can or can not be physically separate, and the units shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment according to actual needs. In addition, the connection between the modules in the apparatus embodiment provided by the present application indicates that there is a communication connection between them, which can be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor.

[0104] Those skilled in the art can clearly understand that, for the convenience and brevity, the specific working process of the apparatus described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0105] The terminal device can be a desktop computer, a notebook computer, a palm computer, a cloud server and other computing devices. The terminal device can include, but is not limited to, a processor and a memory.

[0106] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor and the like. The processor is the control center of the device, and connects all parts of the device through various interfaces and lines.

[0107] The memory can be used to store the computer program, and the processor realizes various functions of the device by running or executing the computer program stored in the memory and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required by a function, and the like; and the data storage area can store data created according to the use of the mobile phone and the like. In addition, the memory can include a high-speed random access memory, and can also include a nonvolatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0108] The storage medium is a storage medium, and the computer program is stored in the storage medium. When the processor executes the computer program, the steps of each method embodiment described above can be realized. The computer program includes computer program code, which can be in the form of source code, object code, an executable file, or some intermediate form, etc. The computer readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. It should be noted that the content included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0109] The above is the preferred embodiment of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which are also considered within the scope of protection of the present application.

Claims

1. A radiation therapy plan generation method, characterized by, The application relates to a method for generating a radiotherapy machine operation file, comprising the following steps: acquiring CT image data and dose distribution data of a patient; inputting the CT image data and the dose distribution data into a trained radiotherapy machine parameter calculation model, so that the radiotherapy machine parameter calculation model calculates and outputs MLC positions, Jaw positions and MU values of the radiotherapy machine at each control point according to the CT image data and the dose distribution data; generating a radiotherapy machine operation file according to the MLC positions, the Jaw positions and the MU values at all control points; wherein the radiotherapy machine operation file is used for directly inputting a radiotherapy machine to control the radiotherapy machine to execute a radiotherapy plan; wherein the radiotherapy machine parameter calculation model comprises a preprocessing module, an encoding module, a processing module and a decoding module; the encoding module mainly comprises a 3D convolution layer, a BatchNorm3D, a 3D Max pooling layer and an activation function Relu layer; the preprocessing module is used for matching the CT image data and the dose distribution data, determining isocenter points of the CT image data and the dose distribution data, cropping the dose distribution data according to a preset size requirement with the isocenter points as a cropping region center, and resampling the cropped dose distribution data according to a preset dimension size to obtain preprocessed dose distribution data; the encoding module is used for matching and fusing the CT image data and the preprocessed dose distribution data, determining corresponding anatomic spatial position information of each voxel in the dose distribution data in the CT image data and corresponding dose information in the dose distribution data, and generating anatomic spatial dose distribution data according to the corresponding anatomic spatial position information and dose distribution information of each voxel; wherein the anatomic spatial dose distribution data is used for describing spatial distribution of dose in the patient body; the processing module is used for generating dose projection data of each control point according to the anatomic spatial dose distribution data; the decoding module is used for decoding the dose projection data of each control point to obtain MLC positions, Jaw positions and MU values of each control point.

2. The radiation therapy plan generation method of claim 1, wherein, the method for generating dose projection data of each control point according to the anatomic spatial dose distribution data comprises the following steps: for each control point, projecting all anatomic spatial dose distribution data in a projection direction of a ray beam path of the radiotherapy machine at the control point on the control point to obtain dose projection data of the control point.

3. The radiation therapy plan generation method of claim 1, wherein, the division process of the control point comprises the following steps: acquiring anatomic spatial data of a patient under a radiotherapy machine; dividing a patient body space corresponding to the anatomic spatial data into a plurality of sub-space regions around a longitudinal axis of the patient, and setting a control point for each sub-space region.

4. The radiation therapy plan generation method of claim 1, wherein, the method for acquiring dose distribution data comprises the following steps: acquiring CT image data, OAR mask image data and PTV image data of a patient; Input the CT image data, OAR mask image data and PTV image data into the trained dose distribution prediction model, so that the dose distribution prediction model predicts the dose distribution according to the CT image data, OAR mask image data and PTV image data, and outputs dose distribution data.

5. A radiotherapy plan generation apparatus, characterized by, Comprise: Data acquisition module, radiotherapy machine parameter calculation module and radiotherapy machine operation file generation module; The data acquisition module is used for acquiring CT image data and dose distribution data of a patient; The radiotherapy machine parameter calculation module is used for inputting the CT image data and dose distribution data into a trained radiotherapy machine parameter calculation model, so that the radiotherapy machine parameter calculation model calculates and outputs MLC position, Jaw position and MU value of each control point of a radiotherapy machine according to the CT image data and dose distribution data; The radiotherapy machine operation file generation module is used for generating a radiotherapy machine operation file according to the MLC position, Jaw position and MU value of all control points; wherein the radiotherapy machine operation file is used for directly inputting a radiotherapy machine to control the radiotherapy machine to execute a radiotherapy plan; Wherein, the radiotherapy machine parameter calculation model comprises: a preprocessing module, an encoding module, a processing module and a decoding module; the encoding module mainly consists of a 3D convolution layer, a BatchNorm3D, a 3D Max pooling layer and an activation function Relu layer; The preprocessing module is used for matching the CT image data and the dose distribution data, determining the isocenter points of the CT image data and the dose distribution data; taking the isocenter points as the center of the cropping region, cropping the dose distribution data according to the preset size requirement; resampling the cropped dose distribution data according to the preset dimension size to obtain preprocessed dose distribution data; The encoding module is used for matching and fusing the CT image data and the preprocessed dose distribution data, determining the corresponding anatomical spatial position information of each voxel in the CT image data and the corresponding dose information in the dose distribution data in the dose distribution data; generating anatomical spatial dose distribution data according to the corresponding anatomical spatial position information and dose distribution information of each voxel; wherein the anatomical spatial dose distribution data is used for describing the spatial distribution of dose in the patient's body; The processing module is used for generating dose projection data of each control point according to the anatomical spatial dose distribution data; The decoding module is used for decoding the dose projection data of each control point to obtain the MLC position, Jaw position and MU value of each control point.

6. A terminal device, characterized by comprising: The storage medium comprises a stored computer program, wherein the storage medium controls the device where the storage medium is located to execute the radiotherapy plan generation method according to any one of claims 1 to 4 when the computer program runs.

7. A storage medium, characterized by The storage medium comprises a stored computer program, wherein the storage medium controls the device where the storage medium is located to execute the radiotherapy plan generation method according to any one of claims 1 to 4 when the computer program runs.

Images