WO2021052150A1 - 放疗计划推荐方法、装置、电子设备及存储介质 - Google Patents

放疗计划推荐方法、装置、电子设备及存储介质 Download PDF

Info

Publication number
WO2021052150A1
WO2021052150A1 PCT/CN2020/112367 CN2020112367W WO2021052150A1 WO 2021052150 A1 WO2021052150 A1 WO 2021052150A1 CN 2020112367 W CN2020112367 W CN 2020112367W WO 2021052150 A1 WO2021052150 A1 WO 2021052150A1
Authority
WO
WIPO (PCT)
Prior art keywords
radiotherapy
map
image
training
measurement
Prior art date
Application number
PCT/CN2020/112367
Other languages
English (en)
French (fr)
Inventor
王季勇
毋戈
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2021052150A1 publication Critical patent/WO2021052150A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
    • G16H20/40—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to mechanical, radiation or invasive therapies, e.g. surgery, laser therapy, dialysis or acupuncture
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
    • G06F16/53—Querying
    • G06F16/532—Query formulation, e.g. graphical querying
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks

Definitions

  • This application relates to the field of artificial intelligence technology, and in particular to a method, device, electronic device, and storage medium for recommending a radiotherapy plan.
  • Radiotherapy therapy generally involves a doctor reading the patient’s CT images before starting the treatment, making a radiotherapy plan based on the CT images, and generating a radiotherapy measurement chart through the radiotherapy plan.
  • radiotherapy physicists usually make radiotherapy plans manually.
  • a conventionally difficult radiotherapy plan still requires a physicist's 2 to 3 hours of work time, so if the method is based on AI It will be a very meaningful innovation to help radiotherapy physicists predict and analyze and improve the efficiency of formulating radiotherapy plans.
  • Existing technologies for automatically formulating a radiotherapy plan often use an algorithm to calculate a radiotherapy plan based on CT images.
  • there are also products that initialize radiotherapy plans through image search technology but the sample size of the image search database is small, and it is still difficult to cover the diversity of patient organs and tumors.
  • the sample search library is simply expanded, the search speed based on image content is relatively slow, so the search with a large sample size will reduce the speed to be difficult to use.
  • the present application provides a radiotherapy plan recommendation method, device, electronic equipment, and storage medium to solve the problems of the prior art that it is difficult to recommend a reasonable radiotherapy plan and the speed reduction caused by simply expanding the sample search database.
  • the first aspect of this application is to provide a radiotherapy plan recommendation method applied to electronic equipment, including:
  • the radiotherapy metrology map sample library including at least a radiotherapy metrology map
  • the second aspect of the present application is to provide a radiotherapy plan recommendation device, including:
  • the sample library building module is used to build a sample library of radiotherapy metering charts.
  • the radiotherapy metering chart samples include at least radiotherapy metering charts;
  • Image acquisition module for acquiring CT images
  • the metering map output module is used to process the CT image with a radiotherapy metering prediction network model generated by pre-training to obtain a radiotherapy metering map corresponding to the CT image;
  • the matching module is used to match the radiotherapy metering chart with the radiotherapy metering chart samples in the radiotherapy metering chart sample library to obtain a matching sample corresponding to the radiotherapy metering chart;
  • the recommendation module is used to obtain the radiotherapy plan corresponding to the matched sample as the recommended radiotherapy plan of the CT image.
  • the third aspect of the present application is to provide an electronic device, the electronic device includes a processor and a memory, the memory includes a radiotherapy plan recommendation program, the radiotherapy plan recommendation program is used by the processor Implementation of the above-mentioned recommended method of radiotherapy plan.
  • the fourth aspect of the present application is to provide a computer-readable storage medium, the computer-readable storage medium includes a radiotherapy plan recommendation program, when the radiotherapy plan recommendation program is executed by a processor, the above is achieved The recommended method of radiotherapy plan.
  • This application obtains the recommended radiotherapy plan by matching the radiotherapy meter. Since the radiotherapy meter contains the radiation metering required for radiotherapy but does not include the patient’s anatomical information, the matching CT image is changed to the matching radiotherapy meter, eliminating the patient Redundant content such as anatomical information simplifies the problem, improves the matching speed, facilitates matching in a large sample size, improves matching accuracy, and enhances reliability.
  • the radiotherapy plan is obtained, there is no need to upload any modal medical images such as CT required to make the radiotherapy plan, which avoids patient data leakage, and the corresponding radiotherapy plan obtained through matching is a manually formulated radiotherapy plan. Excluding inferior radiotherapy plans, its rationality and reliability are higher than those calculated by automated algorithms.
  • Figure 1 is a schematic flow diagram of the recommended method of radiotherapy plan provided by this application.
  • Figure 2 is a schematic diagram of the network structure of the radiotherapy measurement prediction network model in this application.
  • Figure 3 is a schematic diagram of the structure of the radiotherapy plan recommendation device in this application.
  • FIG. 1 is a schematic flow diagram of the radiotherapy plan recommendation method provided by this application. As shown in Figure 1, the radiotherapy plan recommendation method provided by this application includes the following steps:
  • Step S1 construct a sample library of radiotherapy measurement maps.
  • the sample library of radiotherapy images includes radiotherapy measurement map samples and corresponding radiotherapy plans.
  • the radiotherapy plans are artificially formulated and optimized radiotherapy plans and do not include inferior radiotherapy plans, which improves the recommendation. The reliability and rationality of the radiotherapy plan.
  • Step S2 Obtain a CT image.
  • acquiring a CT image includes: acquiring an original CT image; performing normalization processing on the original image according to the following formula to obtain a processed CT image: (CT value+1024)/2048.
  • step S3 the CT image is processed using the radiotherapy measurement prediction network model generated by pre-training to obtain a radiotherapy measurement map corresponding to the CT image.
  • Step S4 matching the radiotherapy metering chart with the radiotherapy metering chart samples in the radiotherapy metering chart sample library to obtain a matching sample corresponding to the radiotherapy metering chart.
  • the radiotherapy measurement map contains the radiation measurement required for radiotherapy but does not include the patient's anatomical information
  • the matching CT image is changed to the matching radiotherapy measurement map, which removes redundant content such as patient anatomy information, which simplifies the problem and improves
  • the matching speed is improved, it is convenient to realize the matching under a large sample size, the matching accuracy is improved, and the reliability is enhanced.
  • Step S5 Obtain a radiotherapy plan corresponding to the matched sample as a recommended radiotherapy plan for the CT image.
  • the recommended radiotherapy plan After obtaining the recommended radiotherapy plan corresponding to the CT image, the recommended radiotherapy plan is loaded on the CT image for reference, and the patient's radiotherapy plan is further adjusted and improved by the doctor or expert.
  • the method further includes: acquiring the radiotherapy corresponding to the CT image An outline map, where the radiotherapy outline map includes a target area outline map and/or an organ outline map; the radiotherapy outline map is processed using a radiotherapy measurement prediction network model generated by pre-training to obtain the radiotherapy measurement corresponding to the CT image Figure.
  • the target area outline map and the organ outline map can be input into the radiotherapy measurement prediction network model at the same time as the CT image to obtain the radiotherapy measurement map corresponding to the CT image.
  • the target area outline map and organ outline map can be uploaded to the cloud with the derived radiotherapy measurement map.
  • the radiotherapy measurement map is used to search for matching radiotherapy measurement map samples in the radiotherapy measurement map sample library, target area outline map and organ outline map It is used to update the radiotherapy measurement prediction network model, and can be used with the radiotherapy measurement chart for reference by doctors or physicists.
  • FIG. 2 is a schematic diagram of the network structure of the radiotherapy measurement prediction network model in this application.
  • the radiotherapy measurement prediction network model is an end-to-end V-net model, with one end input passing through The normalized CT image, one end outputs the corresponding radiotherapy meter
  • the V-net model is a V-shaped convolutional neural network, including an input layer, four encoding layers, four decoding layers, and an output layer , Where each coding layer is used to extract image features, and the current coding layer transfers the extracted image features to the next coding layer and the decoding layer corresponding to the current coding layer, so that the next coding layer can extract more In-depth feature information, and the decoding layer can improve the accuracy of decoding according to the received image feature information.
  • the input layer and the coding layer each include 3 convolution kernels, and the convolution step lengths of the coding layer and the decoding layer are both 2.
  • the number of feature channels corresponding to the input layer, four coding layers, four decoding layers, and output layer are set to 1, 32, 64, 128, 256, 256, 256, 128, 64 and 1, each coding layer uses a deeper convolutional network, thereby improving the convergence speed and accuracy of the model.
  • the V-net model further includes a plurality of modified linear units, and all decoding layers in the model are correspondingly connected with a modified linear unit.
  • the modified linear unit is realized by the ReLU activation function, which improves the training efficiency of the model.
  • the V-net model adopts Group Normalization (GN) processing.
  • GN Group Normalization
  • the output layer adopts a Sigmoid function as the activation function.
  • radiotherapy measurement prediction network model in this application is not limited to the V-net model, and may also be other prediction models.
  • the method before the CT image is processed by the radiotherapy measurement prediction network model generated by pre-training to obtain the radiotherapy measurement map corresponding to the CT image, the method further includes: measuring the radiotherapy measurement The prediction network model is trained.
  • the training step includes: constructing a training sample set, the training samples in the training sample set include CT images and corresponding radiotherapy measurement maps; training the radiotherapy measurement prediction network model using the training samples, Obtain the training radiotherapy measurement map; generate the similarity between the training radiotherapy measurement map and the corresponding radiotherapy measurement map in the training sample, and when the similarity is greater than the preset similarity threshold, end the radiotherapy measurement prediction network model The training process.
  • the use of CT images to train the radiotherapy measurement prediction network model facilitates the acquisition of a large number of training samples and improves the accuracy of the model.
  • matching the radiotherapy metrology chart with the radiotherapy metrology chart samples in the radiotherapy metrology chart sample library includes:
  • using one or more metering chart samples with similarity greater than or equal to a preset similarity threshold as a matching sample corresponding to the radiotherapy metering chart includes: setting a similarity threshold; determining that the corresponding similarity is greater than or equal to One or more meter map samples of the preset similarity threshold are used as matching samples.
  • the similarity threshold is set so that the similarity of the radiotherapy plans corresponding to the two radiotherapy metering charts meets the requirements.
  • the similarity of the radiotherapy plan includes the similarity of the absorbed dose distribution of various parts of the human body.
  • a large sample of the measurement chart can obtain a corresponding relatively similar radiotherapy plan, thereby reducing the adaptability of doctors or experts to the recommended radiotherapy plan, reducing workload and improving work efficiency. Therefore, the measured graph samples with the determined similarity greater than or equal to the similarity threshold can be used as matching samples.
  • the above-mentioned matching of the radiotherapy measurement map is to compare the similarity of the content information of the two radiotherapy measurement maps, and the comparison content is the various features of the image. Since the radiotherapy metering chart contains the radiation metering required for radiotherapy, the radiotherapy plans corresponding to similar radiotherapy metering charts are similar. Therefore, the corresponding radiotherapy plan can be obtained by matching the radiotherapy metering chart recommendation. The doctor or expert can base on this After appropriate adjustments, the patient's radiotherapy plan can be worked out, and work efficiency can be improved accordingly.
  • the commonly used similarity algorithms based on vector models for comparing various features of images include Euclidean distance, Minkowski distance, Manhattan distance, histogram intersection, Mahalanobis distance, etc.
  • Euclidean distance is used for the orthogonality of feature vectors.
  • Mahalanobis distance is used to have statistical characteristics or to perform correlation analysis of samples. In specific operations, appropriate measurement methods should be selected according to different characteristics.
  • generating the similarity between the radiotherapy metrology chart and each metering chart sample in the radiotherapy metrology chart sample library includes:
  • d(X, Y) represents the similarity
  • X represents the feature vector corresponding to the radiotherapy meter
  • Y represents the feature vector corresponding to the sample of the meter.
  • generating the similarity between the radiotherapy metrology chart and each metering chart sample in the radiotherapy metrology chart sample library includes:
  • d(X,Y) represents the similarity
  • x represents the vector element of the feature vector X
  • y represents the vector element of the feature vector Y
  • k represents the index of the vector element
  • n represents the number of vector elements.
  • the radiotherapy plan recommendation method described in this application is applied to electronic equipment, and the electronic equipment may be terminal equipment such as a television, a smart phone, a tablet computer, and a computer.
  • the electronic device includes a processor and a memory for storing a radiotherapy plan recommendation program, and the processor executes the radiotherapy plan recommendation program to implement the following radiotherapy plan recommendation method:
  • Construct a sample library of radiotherapy metrology maps acquire CT images; process the CT images with the radiotherapy metrology prediction network model generated by pre-training to obtain the radiotherapy metrology maps corresponding to the CT images; compare the radiotherapy metrology maps with all Matching metering map samples in the radiotherapy metering map sample library to obtain a matching sample corresponding to the radiotherapy metering map; acquiring a radiotherapy plan corresponding to the matching sample as a recommended radiotherapy plan for the CT image.
  • the electronic device also includes a network interface, a communication bus, and the like.
  • the network interface may include a standard wired interface and a wireless interface
  • the communication bus is used to realize the connection and communication between various components.
  • the memory includes at least one type of readable storage medium, which can be a non-volatile storage medium such as a flash memory, a hard disk, an optical disk, or a plug-in hard disk, etc., and is not limited to this, and can be stored in a non-transitory manner Any device that provides instructions or software and any associated data files to the processor to enable the processor to execute the instructions or software program.
  • the software program stored in the memory includes a radiotherapy plan recommendation program, and the radiotherapy plan recommendation program can be provided to the processor, so that the processor can execute the radiotherapy plan recommendation program and implement the radiotherapy plan recommendation method.
  • the processor may be a central processing unit, a microprocessor, or other data processing chips, etc., and may run a stored program in the memory, for example, the radiotherapy plan recommendation program in this application.
  • the electronic device may also include a display, and the display may also be called a display screen or a display unit.
  • the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an organic light-emitting diode (OLED) touch device, and the like.
  • the display is used to display the information processed in the electronic device and to display the visual work interface.
  • the electronic device may also include a user interface, and the user interface may include an input unit (such as a keyboard), a voice output device (such as a stereo, earphone), and the like.
  • the user interface may include an input unit (such as a keyboard), a voice output device (such as a stereo, earphone), and the like.
  • FIG. 3 is a schematic diagram of the structure of the radiotherapy plan recommendation device in this application.
  • the radiotherapy plan recommendation device in this application includes: a sample library construction module 1, an image acquisition module 2, a metering chart output module 3, and a matching module 4 And recommendation module 5, in which a radiotherapy measurement map sample library is constructed through the sample library construction module 1 to match related radiotherapy measurement maps; after the image acquisition module 2 acquires the CT images, it is generated by the measurement map output module 3 with pre-training
  • the radiotherapy metrology prediction network model of processes the CT image to obtain a radiotherapy metrology map corresponding to the CT image, and then, through the matching module 4, compares the radiotherapy metrology map with the metrology in the radiotherapy metrology map sample library The map samples are matched to obtain a matching sample corresponding to the radiotherapy metering map.
  • the image acquisition module includes: an original image acquisition unit, used to acquire the original CT image, and a normalization unit, used to normalize the original image so as to be input into the radiotherapy measurement prediction model.
  • the conversion unit normalizes the original image by the following formula, (CT value+1024)/2048.
  • the radiotherapy plan recommendation device obtains the recommended radiotherapy plan by matching the radiotherapy meter map. Since the radiotherapy meter map contains the radiation metering required for radiotherapy but does not include the patient's anatomical information, the existing matching CT image is changed to match The radiotherapy measurement chart removes redundant content such as patient anatomy information, simplifies the problem, improves the matching speed, facilitates matching in a large sample size, improves matching accuracy, and enhances reliability. In addition, the corresponding radiotherapy plan obtained through matching is a radiotherapy plan made manually, excluding inferior radiotherapy plans, and its rationality and reliability are higher than those calculated by automated algorithms.
  • the radiotherapy metering prediction network model used in the metering map output module 3 is an end-to-end V-net model.
  • One end inputs the normalized CT image, and the other end outputs the corresponding
  • the V-net model is a V-shaped convolutional neural network, including an input layer, four coding layers, four decoding layers, and an output layer.
  • Each coding layer is used to extract Image features, and the current encoding layer transfers the extracted image features to the next encoding layer and the decoding layer corresponding to the current encoding layer, so that the next encoding layer can extract deeper feature information, and the decoding layer is based on the received
  • the image feature information can improve the accuracy of decoding.
  • the input layer and the coding layer each include 3 convolution kernels, and the convolution step lengths of the coding layer and the decoding layer are both 2.
  • the matching module 4 includes: a similarity generation unit, which generates the similarity between the radiotherapy measurement map and each measurement map sample in the radiotherapy measurement map sample library; and a sorting unit, The metering chart samples in the radiotherapy metering chart sample library are arranged in descending order according to the corresponding similarity; the matching sample determination unit takes one or more metering chart samples whose similarity is greater than or equal to the preset similarity threshold as the radiotherapy The matching sample corresponding to the metering chart.
  • a similarity generation unit which generates the similarity between the radiotherapy measurement map and each measurement map sample in the radiotherapy measurement map sample library
  • a sorting unit The metering chart samples in the radiotherapy metering chart sample library are arranged in descending order according to the corresponding similarity; the matching sample determination unit takes one or more metering chart samples whose similarity is greater than or equal to the preset similarity threshold as the radiotherapy The matching sample corresponding to the metering chart.
  • the matching sample determining unit determines the matching sample in the following manner, including: setting a similarity threshold; and determining one or more meter map samples with corresponding similarity greater than or equal to the preset similarity threshold as the matching sample.
  • the similarity threshold is set so that the similarity of the radiotherapy plans corresponding to the two radiotherapy metering charts meets the requirements.
  • the similarity of the radiotherapy plan includes the similarity of the absorbed dose distribution of various parts of the human body.
  • a large sample of the measurement chart can obtain a corresponding relatively similar radiotherapy plan, thereby reducing the adaptability of doctors or experts to the recommended radiotherapy plan, reducing workload and improving work efficiency. Therefore, the measured graph samples with the determined similarity greater than or equal to the similarity threshold can be used as matching samples.
  • the matching of the radiotherapy measurement map is to compare the similarity of the content information of the two radiotherapy measurement maps, and the comparison content is the various features of the image. Since the radiotherapy metering chart contains the radiation metering required for radiotherapy, the radiotherapy plans corresponding to similar radiotherapy metering charts are similar. Therefore, the corresponding radiotherapy plan can be obtained by matching the radiotherapy metering chart recommendation. The doctor or expert can base on this After appropriate adjustments, a patient-specific radiotherapy plan can be formulated, which improves work efficiency to a certain extent.
  • the similarity acquisition unit includes a feature extraction subunit and a similarity calculation subunit, wherein the feature extraction subunit is used to extract the feature vector of the radiotherapy meter and the sample of the meter respectively, using the feature vector Characterize the corresponding image; the similarity calculation subunit is used to obtain the similarity between the radiotherapy metrology chart and each metering chart sample in the radiotherapy metrology chart sample library.
  • the commonly used similarity algorithms based on vector models for comparing various features of images at present include Euclidean distance, Minkowski distance, Manhattan distance, histogram intersection, Mahalanobis distance, etc. Among them, Euclidean distance is used for feature vector orthogonality irrelevant The Mahalanobis distance is used to analyze the correlation of samples or have statistical characteristics. For specific operations, appropriate measurement methods should be selected according to different characteristics.
  • the similarity calculation sub-unit can obtain the similarity of two radiotherapy measurement graphs through a variety of similarity calculation methods. For example, the similarity between the radiotherapy meter chart and the meter chart sample is obtained by the following formula,
  • d(X, Y) represents the similarity
  • X represents the feature vector corresponding to the radiotherapy meter
  • Y represents the feature vector corresponding to the sample of the meter.
  • d(X,Y) represents the similarity
  • x represents the vector element of the feature vector X
  • y represents the vector element of the feature vector Y
  • k represents the index of the vector element
  • n represents the number of vector elements.
  • the radiotherapy plan recommendation program may also be divided into one or more modules, and one or more modules are stored in the memory and executed by the processor to implement the radiotherapy plan recommendation device in this application.
  • the module referred to in this application refers to a series of computer program instruction segments that can complete specific functions.
  • the radiotherapy plan recommendation program can be divided into: a sample library construction module 1, an image acquisition module 2, a metrology map output module 3, a matching module 4, and a recommendation module 5.
  • the functions or operation steps implemented by the above modules are all similar to the above, and will not be described in detail here. For example, for example:
  • Sample library building module 1 to build a sample library of radiotherapy metering charts
  • Image acquisition module 2 to acquire CT images
  • the metering map output module 3 uses the radiotherapy metering prediction network model generated by pre-training to process the CT image to obtain a radiotherapy metering map corresponding to the CT image;
  • the matching module 4 matches the radiotherapy metering chart with metering chart samples in the radiotherapy metering chart sample library to obtain a matching sample corresponding to the radiotherapy metering chart;
  • the recommendation module 5 obtains the radiotherapy plan corresponding to the matched sample as the recommended radiotherapy plan of the CT image.
  • the computer-readable storage medium may be any tangible medium that contains or stores a program or instruction.
  • the computer-readable storage medium may be non-volatile or volatile, and the program may be It is executed, and the corresponding function is realized through the stored program instruction related hardware.
  • the computer-readable storage medium may be a computer disk, a hard disk, a random access memory, a read-only memory, and so on.
  • the present application is not limited to this, and can be any device that stores instructions or software and any related data files or data structures in a non-transitory manner and can be provided to the processor to enable the processor to execute the programs or instructions therein.
  • the computer-readable storage medium includes a radiotherapy plan recommendation program, and when the radiotherapy plan recommendation program is executed by the processor, the following radiotherapy plan recommendation method is implemented:
  • the specific implementation of the computer-readable storage medium of the present application is substantially the same as the specific implementation of the above-mentioned radiotherapy plan recommendation method, electronic equipment, and radiotherapy plan recommendation device, and will not be repeated here.
  • the disclosed device and method may be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the modules is only a logical function division, and there may be other divisions in actual implementation, such as: multiple modules or components can be combined, or It can be integrated into another system, or some features can be ignored or not implemented.
  • the coupling, or direct coupling, or communication connection between the components shown or discussed may be indirect coupling or communication connection through some interfaces, devices or modules, and may be in electrical, mechanical or other forms. of.
  • the units described above as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • the functional units in the embodiments of the present application can be all integrated into one processing unit, or each unit can be individually used as a unit, or two or more units can be integrated into one unit;
  • the unit can be implemented in the form of hardware, or in the form of hardware plus software functional units.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Urology & Nephrology (AREA)
  • Artificial Intelligence (AREA)
  • Public Health (AREA)
  • Surgery (AREA)
  • Databases & Information Systems (AREA)
  • Primary Health Care (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Epidemiology (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

一种放疗计划推荐方法、装置、电子设备及存储介质，属于人工智能技术领域。其中，方法包括：构建放疗计量图样本库（S1）；获取CT图像（S2）；用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图（S3）；将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本（S4）；获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划（S5）。通过匹配放疗计量图获取推荐的放疗计划，去除了病人解剖学信息等冗余内容，简化了问题，提升了匹配速度，便于实现在大样本量下进行匹配，提升匹配精度，并且推荐得到的放疗计划均非劣质放疗计划，提高可靠性和合理性。

Description

放疗计划推荐方法、装置、电子设备及存储介质
本申请要求于2019年9月17日提交中国专利局、申请号为201910875109.2，发明名称为“放疗计划推荐方法、装置、设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能技术领域，尤其涉及一种放疗计划推荐方法、装置、电子设备及存储介质。
背景技术
放射治疗一般是在治疗开始之前，由医生阅读病人的CT图像，基于CT图像制定放疗计划，通过放疗计划生成放疗计量图。目前，临床上通常由放疗物理师人工制定放疗计划，尽管有一些器官勾画工具辅助，但是，一个常规难度的放疗计划仍旧需要一位物理师2～3小时的工作时间，所以如果基于AI的方法能够帮助放疗物理师预测分析，提升制定放疗计划的效率，将是一个很有意义的创新。现有自动制定放疗计划的技术，往往基于CT图像由算法推算一个放疗计划，发明人意识到由于病人器官和肿瘤的多变，算法很难推算出一个相对合理的放疗计划。另一方面，也有通过图像搜索技术初始化放疗计划的产品，但是其图像搜索库样本量较小，仍旧难以覆盖病人器官和肿瘤的多样性。而如果单纯扩大样本搜索库，因为基于图像内容的搜索相对速度较慢，所以大样本量的搜索就会导致速度降低到难以使用。
发明内容
本申请提供一种放疗计划推荐方法、装置、电子设备及存储介质，以解决现有技术难以推荐出合理的放疗计划以及单纯扩大样本搜索库导致速度降低的问题。
为了实现上述目的，本申请的第一个方面是提供一种放疗计划推荐方法，应用于电子设备，包括：
构建放疗计量图样本库，所述放疗计量图样本库至少包括放疗计量图；
获取CT图像；
用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图；
将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本；
获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划。
为了实现上述目的，本申请的第二个方面是提供一种放疗计划推荐装置，包括：
样本库构建模块，用于构建放疗计量图样本库，放疗计量图样本至少包括放疗计量图；
图像获取模块，用于获取CT图像；
计量图输出模块，用于用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图；
匹配模块，用于将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本；
推荐模块，用于获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划。
为了实现上述目的，本申请的第三个方面是提供一种电子设备，该电子设备包括：处理器和存储器，所述存储器中包括放疗计划推荐程序，所述放疗计划推荐程序被所述处理器执行时实现如上所述的放疗计划推荐方法。
为了实现上述目的，本申请的第四个方面是提供一种计算机可读存储介质，所述计算 机可读存储介质中包括放疗计划推荐程序，所述放疗计划推荐程序被处理器执行时，实现如上所述的放疗计划推荐方法。
相对于现有技术，本申请具有以下优点和有益效果：
本申请通过匹配放疗计量图获得推荐的放疗计划，由于放疗计量图包含了放疗所需的射线计量但不包括病人的解剖学信息，所以，将匹配CT图像改为匹配放疗计量图，去除了病人解剖学信息等冗余内容，简化了问题，提升了匹配速度，便于实现在大样本量下进行匹配，提升匹配精度，增强可靠性。并且，在得到放疗计划时，不需要上传任何制定放疗计划所需要的CT等任何模态医学影像，避免了病人数据泄露，并且，通过匹配得到的对应的放疗计划是经过人工制定的放疗计划，不包括劣质放疗计划，其合理性和可靠性均高于自动化算法推算得到的放疗计划。
附图说明
图1为本申请所提供的放疗计划推荐方法的流程示意图；
图2为本申请中放疗计量预测网络模型的网络结构示意图；
图3为本申请中放疗计划推荐装置的构成示意图。
本申请目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
下面将参考附图来描述本申请所述的实施例。本领域的普通技术人员可以认识到，在不偏离本申请的精神和范围的情况下，可以用各种不同的方式或其组合对所描述的实施例进行修正。因此，附图和描述在本质上是说明性的，仅仅用以解释本申请，而不是用于限制权利要求的保护范围。此外，在本说明书中，附图未按比例画出，并且相同的附图标记表示相同的部分。
图1为本申请所提供的放疗计划推荐方法的流程示意图，如图1所示，本申请所提供的放疗计划推荐方法，包括以下步骤：
步骤S1，构建放疗计量图样本库，放疗图像样本库中包括放疗计量图样本以及对应的放疗计划，其中的放疗计划均是经过人工制定优化的放疗计划，并不包括劣质放疗计划，提高了推荐放疗计划的可靠性和合理性。
步骤S2，获取CT图像。
优选地，获取CT图像包括：获取原始CT图像；根据下式对原始图像做归一化处理，得到处理后的CT图像：(CT值+1024)/2048。
步骤S3，用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图。
步骤S4，将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本。由于放疗计量图包含了放疗所需的射线计量但不包括病人的解剖学信息，所以，将匹配CT图像改为匹配放疗计量图，去除了病人解剖学信息等冗余内容，简化了问题，提升了匹配速度，便于实现在大样本量下进行匹配，提升匹配精度，增强可靠性。
步骤S5，获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划。
获取与CT图像对应的推荐放疗计划之后，将推荐的放疗计划加载于所述CT图像以供参考，由医生或专家对病人的放疗计划进一步调整和完善。
本申请在构建放量计量图样本库时，只需要放疗计划所生成的放疗计量图、对应的放疗计划，不需要和病人直接相关的医学影像，降低了病人数据泄密的可能性。
本申请的一个实施例中，在获取CT图像之后，将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配之前，还包括：获取与所述CT图像对应的放疗勾勒图，所述放疗勾勒图包括靶区勾勒图和/或器官勾勒图；用预先训练生成的放疗计量预测网络模型对所述放疗勾勒图进行处理，得出与所述CT图像对应的放疗计量图。
其中，靶区勾勒图和器官勾勒图可以与CT图像同时输入放疗计量预测网络模型中，以得出与CT图像对应的放疗计量图。并且，靶区勾勒图和器官勾勒图可以与得出的放疗计量图均上传到云端，放疗计量图用来搜索匹配放疗计量图样本库中的放疗计量图样本，靶区勾勒图和器官勾勒图用来更新放疗计量预测网络模型，且可以与放疗计量图一起供医生或物理师参考。
图2为本申请中放疗计量预测网络模型的网络结构示意图，如图2所示，本申请的一个实施例中，所述放疗计量预测网络模型是一个端对端的V-net模型，一端输入经过归一化获得的CT图像，一端输出对应的放疗计量图，所述V-net模型是一个V形的卷积神经网络，包括一个输入层、四个编码层、四个解码层以及一个输出层，其中，每个编码层均用于提取图像特征，并且，当前编码层将提取到的图像特征传递到下一个编码层以及与当前编码层对应的解码层，以便于下一个编码层提取到更深层的特征信息，而解码层根据接收的图像特征信息可以提高解码的精准度。优选地，所述V-net模型中，所述输入层和所述编码层均包括3个卷积核，所述编码层和所述解码层的卷积步长均为2。
如图2所示，在端对端的V-net模型中，将输入层、四个编码层、四个解码层和输出层对应的特征通道数量分别设置为1、32、64、128、256、256、256、128、64和1，每一个编码层均采用了更深的卷积网络，从而提高了模型的收敛速度和精度。
优选地，所述V-net模型还包括多个修正线性单元，模型中的所有解码层均对应连接有一个修正线性单元，修正线性单元通过ReLU激活函数来实现，提高模型的训练效率。
优选地，所述V-net模型采用群组归一化(Group Normalization，GN)处理。
优选地，所述输出层采用Sigmoid函数作为激活函数。
需要说明的是，本申请中的放疗计量预测网络模型并不限于V-net模型，也可以是其他预测模型。
本申请的一个可选实施例中，用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图之前，还包括：对所述放疗计量预测网络模型进行训练，具体地，训练步骤包括：构建训练样本集，所述训练样本集中的训练样本包括CT图像和对应的放疗计量图；利用训练样本对所述放疗计量预测网络模型进行训练，得到训练放疗计量图；生成所述训练放疗计量图与训练样本中相应的放疗计量图间的相似度，当所述相似度大于预设相似度阈值时，结束对所述放疗计量预测网络模型的训练过程。利用CT图像对放疗计量预测网络模型进行训练，便于获取大量的训练样本，提高模型的精度。
本申请的一个可选实施例中，将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配，包括：
生成所述放疗计量图与所述放疗计量图样本库中的各个计量图样本的相似度；将放疗计量图样本库中的计量图样本按照对应的相似度从大到小排列；将相似度大于或等于预设相似度阈值的一个或多个计量图样本作为与所述放疗计量图对应的匹配样本。获取多个匹配样本时，可以得到对应的多个推荐放疗计划，以供医生或专家参考，提高可靠性。
进一步地，将相似度大于或等于预设相似度阈值的一个或多个计量图样本作为与所述放疗计量图对应的匹配样本，包括：设定相似度阈值；确定对应的相似度大于或等于所述预设相似度阈值的一个或多个计量图样本作为匹配样本。其中，相似度阈值的设定以使得两幅放疗计量图对应的放疗计划的相似性满足要求为准，放疗计划的相似性包括人体内各部分的吸收剂量的分布的相似性，通过相似度较大的计量图样本可以得到对应的较为相似的放疗计划，从而减少医生或专家对推荐得到的放疗计划的适应性调整，减少工作量，提高工作效率。因此，确定的相似度大于或等于相似度阈值的计量图样本可以作为匹配样本。
上述对放疗计量图进行匹配，即为对两幅放疗计量图中内容信息进行相似度比较，比较内容即为图像的各类特征。由于放疗计量图中包含了放疗所需要的射线计量，相似的放 疗计量图所对应的放疗计划即为相似，因此，可以通过匹配放疗计量图推荐得到相应的放疗计划，医生或专家在此基础上再进行适当调整，即可制定出病人的放疗计划，相应地提高了工作效率。
目前比较图像的各类特征常用的基于向量模型的相似度算法有欧氏距离法、Minkowski距离、曼哈顿距离、直方图交、马氏距离等，其中，欧氏距离用于特征向量正交无关的图像，马氏距离用于具有统计特性或进行样本的相关性分析，具体操作时应根据不同特征选择合适的度量方法。
本申请的一个可选实施例中，生成所述放疗计量图与所述放疗计量图样本库中的各个计量图样本的相似度，包括：
分别提取所述放疗计量图和计量图样本的特征向量，用特征向量表征对应的图像；
通过下式获取所述放疗计量图与计量图样本的相似度，
Figure PCTCN2020112367-appb-000001
其中，d(X,Y)表示相似度，X表示放疗计量图对应的特征向量，Y表示计量图样本对应的特征向量。
本申请的一个可选实施例中，生成所述放疗计量图与所述放疗计量图样本库中的各个计量图样本的相似度，包括：
分别提取所述放疗计量图和计量图样本的特征向量，用特征向量表征对应的图像；
通过下式获取所述放疗计量图与计量图样本的相似度，
Figure PCTCN2020112367-appb-000002
其中，d(X,Y)表示相似度，X表示放疗计量图对应的特征向量X＝(x 1，x 2，…，x n)，Y表示计量图样本对应的特征向量Y＝(y 1，y 2，…，y n)，x表示特征向量X的向量元素，y表示特征向量Y的向量元素，k表示向量元素的索引，n表示向量元素的数量。
本申请所述放疗计划推荐方法应用于电子设备，所述电子设备可以是电视机、智能手机、平板电脑、计算机等终端设备。
所述电子设备包括：处理器和存储器，用于存储放疗计划推荐程序，处理器执行所述放疗计划推荐程序，实现以下的放疗计划推荐方法：
构建放疗计量图样本库；获取CT图像；用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图；将所述放疗计量图与所述放疗计量图样本库中的计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本；获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划。
所述电子设备还包括网络接口和通信总线等。其中，网络接口可以包括标准的有线接口、无线接口，通信总线用于实现各个组件之间的连接通信。
存储器包括至少一种类型的可读存储介质，可以是闪存、硬盘、光盘等非易失性存储介质，也可以是插接式硬盘等，且并不限于此，可以是以非暂时性方式存储指令或软件以及任何相关联的数据文件并向处理器提供指令或软件程序以使该处理器能够执行指令或软件程序的任何装置。本申请中，存储器存储的软件程序包括放疗计划推荐程序，并可以向处理器提供该放疗计划推荐程序，以使得处理器可以执行该放疗计划推荐程序，实现放疗计划推荐方法。
处理器可以是中央处理器、微处理器或其他数据处理芯片等，可以运行存储器中的存储程序，例如，本申请中放疗计划推荐程序。
所述电子设备还可以包括显示器，显示器也可以称为显示屏或显示单元。在一些实施例中显示器可以是LED显示器、液晶显示器、触控式液晶显示器以及有机发光二极管(Organic Light-Emitting Diode，OLED)触摸器等。显示器用于显示在电子设备中处理的信息以及用于显示可视化的工作界面。
所述电子设备还可以包括用户接口，用户接口可以包括输入单元(比如键盘)、语音输出装置(比如音响、耳机)等。
图3为本申请中放疗计划推荐装置的构成示意图，如图3所示，本申请所述放疗计划推荐装置包括：样本库构建模块1、图像获取模块2、计量图输出模块3、匹配模块4以及推荐模块5，其中，通过样本库构建模块1构建放疗计量图样本库，用来进行相关的放疗计量图的匹配；图像获取模块2获取CT图像之后，通过计量图输出模块3用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图，然后，通过匹配模块4将所述放疗计量图与所述放疗计量图样本库中的计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本，最后，通过推荐模块5获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划。其中，图像获取模块包括：原始图像获取单元，用于获取原始的CT图像，以及归一化单元，用于对原始图像进行归一化处理，以便于输入放疗计量预测模型中，其中，归一化单元通过下式对原始图像进行归一化处理，(CT值+1024)/2048。
所述放疗计划推荐装置通过匹配放疗计量图获得推荐的放疗计划，由于放疗计量图包含了放疗所需的射线计量但不包括病人的解剖学信息，所以，将现有的匹配CT图像改为匹配放疗计量图，去除了病人解剖学信息等冗余内容，简化了问题，提升了匹配速度，便于实现在大样本量下进行匹配，提升匹配精度，增强可靠性。并且，通过匹配得到的对应的放疗计划是经过人工制定的放疗计划，不包括劣质放疗计划，其合理性和可靠性均高于自动化算法推算得到的放疗计划。
本申请的一个可选实施例中，所述计量图输出模块3中所使用的放疗计量预测网络模型是一个端对端的V-net模型，一端输入经过归一化获得的CT图像，一端输出对应的放疗计量图，所述V-net模型是一个V形的卷积神经网络，包括一个输入层、四个编码层、四个解码层以及一个输出层，其中，每个编码层均用于提取图像特征，并且，当前编码层将提取到的图像特征传递到下一个编码层以及与当前编码层对应的解码层，以便于下一个编码层提取到更深层的特征信息，而解码层根据接收的图像特征信息可以提高解码的精准度。优选地，所述V-net模型中，所述输入层和所述编码层均包括3个卷积核，所述编码层和所述解码层的卷积步长均为2。
本申请中，通过将所述V-net模型中的所述输入层、四个编码层、四个解码层和所述输出层对应的特征通道数量分别设置为1、32、64、128、256、256、256、128、64和1，使得每一个编码层采用了更深的卷积网络，从而提高模型的收敛速度和精度。
需要说明的是，所述V-net模型的其他设置以及训练方法等均与上文中放疗计划推荐方法中的相关内容相同，在此不再赘述。
本申请的一个可选实施例中，匹配模块4包括：相似度生成单元，生成所述放疗计量图与所述放疗计量图样本库中的各个计量图样本的相似度；排序单元，将所述放疗计量图样本库中的计量图样本按照对应的相似度从大到小排列；匹配样本确定单元，将相似度大于或等于预设相似度阈值的一个或多个计量图样本作为与所述放疗计量图对应的匹配样本。获取多个匹配样本时，可以得到对应的多个推荐放疗计划，以供医生或专家参考，提高可靠性。
进一步地，匹配样本确定单元通过下述方式确定匹配样本，包括：设定相似度阈值；确定对应的相似度大于或等于所述预设相似度阈值的一个或多个计量图样本作为匹配样本。其中，相似度阈值的设定以使得两幅放疗计量图对应的放疗计划的相似性满足要求为准，放疗计划的相似性包括人体内各部分的吸收剂量的分布的相似性，通过相似度较大的计量图样本可以得到对应的较为相似的放疗计划，从而减少医生或专家对推荐得到的放疗计划的适应性调整，减少工作量，提高工作效率。因此，确定的相似度大于或等于相似度阈值的计量图样本可以作为匹配样本。
本申请中，对放疗计量图的匹配，即为对两幅放疗计量图中内容信息进行相似度比较，比较内容即为图像的各类特征。由于放疗计量图中包含了放疗所需要的射线计量，相似的放疗计量图所对应的放疗计划即为相似，因此，可以通过匹配放疗计量图推荐得到相应的放疗计划，医生或专家在此基础上再进行适当调整，即可制定出针对于患者的放疗计划，一定程度上提高了工作效率。
本申请的一个实施例中，相似度获取单元包括特征提取子单元和相似度计算子单元，其中，特征提取子单元用于分别提取所述放疗计量图和计量图样本的特征向量，用特征向量表征对应的图像；相似度计算子单元，用于获取所述放疗计量图与所述放疗计量图样本库中的各个计量图样本的相似度。
由于目前比较图像的各类特征常用的基于向量模型的相似度算法有欧氏距离法、Minkowski距离、曼哈顿距离、直方图交、马氏距离等，其中，欧氏距离用于特征向量正交无关的图像，马氏距离用于具有统计特性或进行样本的相关性分析，具体操作时应根据不同特征选择合适的度量方法。相似度计算子单元可以通过多种相似度计算方式得到两幅放疗计量图的相似度。例如，通过下式获取所述放疗计量图与计量图样本的相似度，
Figure PCTCN2020112367-appb-000003
其中，d(X,Y)表示相似度，X表示放疗计量图对应的特征向量，Y表示计量图样本对应的特征向量。
或者，通过下式获取所述放疗计量图与计量图样本的相似度，
Figure PCTCN2020112367-appb-000004
其中，d(X,Y)表示相似度，X表示放疗计量图对应的特征向量X＝(x 1，x 2，…，x n)，Y表示计量图样本对应的特征向量Y＝(y 1，y 2，…，y n)，x表示特征向量X的向量元素，y表示特征向量Y的向量元素，k表示向量元素的索引，n表示向量元素的数量。
上述获取相似度的方法，均是基于提取的图像特征进行计算，本申请中，可以根据需求选择相应的计算方式。
在其他实施例中，所述放疗计划推荐程序还可以被分割为一个或者多个模块，一个或者多个模块被存储于存储器中，并由处理器执行，以实现本申请中放疗计划推荐装置的各个功能。本申请所称的模块是指能够完成特定功能的一系列计算机程序指令段。其中，所述放疗计划推荐程序可以被分割为：样本库构建模块1、图像获取模块2、计量图输出模块3、匹配模块4和推荐模块5。上述模块所实现的功能或操作步骤均与上文类似，此处不再详述，示例性地，例如其中：
样本库构建模块1，构建放疗计量图样本库；
图像获取模块2，获取CT图像；
计量图输出模块3，用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图；
匹配模块4，将所述放疗计量图与所述放疗计量图样本库中的计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本；
推荐模块5，获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划。
本申请的一个实施例中，计算机可读存储介质可以是任何包含或存储程序或指令的有形介质，所述计算机可读存储介质可以是非易失性，也可以是易失性，其中的程序可以被执行，通过存储的程序指令相关的硬件实现相应的功能。例如，计算机可读存储介质可以是计算机磁盘、硬盘、随机存取存储器、只读存储器等。本申请并不限于此，可以是以非暂时性方式存储指令或软件以及任何相关数据文件或数据结构并且可提供给处理器以使处理器执行其中的程序或指令的任何装置。所述计算机可读存储介质中包括放疗计划推荐 程序，所述放疗计划推荐程序被处理器执行时，实现如下的放疗计划推荐方法：
构建放疗计量图样本库；
获取CT图像；
用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图；
将所述放疗计量图与所述放疗计量图样本库中的计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本；
获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划。
本申请之计算机可读存储介质的具体实施方式与上述放疗计划推荐方法、电子设备、放疗计划推荐装置的具体实施方式大致相同，在此不再赘述。
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、装置、物品或者方法不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、装置、物品或者方法所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、装置、物品或者方法中还存在另外的相同要素。
在本申请中的几个实施例中，应该理解到，所揭露的设备和方法，可以通过其它的方式实现。以上所描述的设备实施例仅仅是示意性的，例如，所述模块的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，如：多个模块或组件可以结合，或可以集成到另一个系统，或一些特征可以忽略，或不执行。另外，所显示或讨论的各组成部分相互之间的耦合、或直接耦合、或通信连接可以是通过一些接口，设备或模块的间接耦合或通信连接，可以是电性的、机械的或其它形式的。
上述作为分离部件说明的单元可以是、或也可以不是物理上分开的，作为单元显示的部件可以是、或也可以不是物理单元；既可以位于一个地方，也可以分布到多个网络单元上；可以根据实际的需要选择其中的部分或全部单元来实现本实施例方案的目的。
另外，在本申请各实施例中的各功能单元可以全部集成在一个处理单元中，也可以是各单元分别单独作为一个单元，也可以两个或两个以上单元集成在一个单元中；上述集成的单元既可以采用硬件的形式实现，也可以采用硬件加软件功能单元的形式实现。
上述本申请实施例序号仅仅为了描述，不代表实施例的优劣，并非因此限制本申请的专利范围，凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本申请的专利保护范围内。通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在如上所述的一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，或者网络设备等)执行本申请各个实施例所述的方法。

Claims (20)

  1. 一种放疗计划推荐方法，应用于电子设备，包括：
    构建放疗计量图样本库，所述放疗计量图样本库至少包括放疗计量图；
    获取CT图像；
    用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图；
    将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本；
    获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划。
  2. 根据权利要求1所述的放疗计划推荐方法，其中，在获取CT图像之后，将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配之前，还包括：获取与所述CT图像对应的放疗勾勒图，所述放疗勾勒图包括靶区勾勒图和/或器官勾勒图；
    用预先训练生成的放疗计量预测网络模型对所述放疗勾勒图进行处理，得出与所述CT图像对应的放疗计量图。
  3. 根据权利要求1所述的放疗计划推荐方法，其中，所述放疗计量预测网络模型是端对端的V-net模型，包括输入层、四个编码层、四个解码层以及输出层，其中，所述输入层和所述编码层均包括3个卷积核，所述编码层和所述解码层的卷积步长均为2。
  4. 根据权利要求3所述的放疗计划推荐方法，其中，所述输入层、四个编码层、四个解码层和所述输出层对应的特征通道数量分别为1、32、64、128、256、256、256、128、64和1。
  5. 根据权利要求3所述的放疗计划推荐方法，其中，所述V-net模型还包括多个修正线性单元，每个修正线性单元均与一个解码层对应连接。
  6. 根据权利要求3所述的放疗计划推荐方法，其中，所述V-net模型采用群组归一化处理。
  7. 根据权利要求1所述的放疗计划推荐方法，其中，用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图之前，还包括：
    构建训练样本集，所述训练样本集中的训练样本包括CT图像和对应的放疗计量图；
    利用训练样本对所述放疗计量预测网络模型进行训练，得到训练放疗计量图；
    生成所述训练放疗计量图与训练样本中相应的放疗计量图间的相似度，当所述相似度大于预设相似度阈值时，结束对所述放疗计量预测网络模型的训练过程。
  8. 一种放疗计划推荐装置，包括：
    样本库构建模块，用于构建放疗计量图样本库，放疗计量图样本至少包括放疗计量图；
    图像获取模块，用于获取CT图像；
    计量图输出模块，用于用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图；
    匹配模块，用于将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本；
    推荐模块，用于获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划。
  9. 一种电子设备，该电子设备包括：处理器和存储器，所述存储器中包括放疗计划推荐程序，所述放疗计划推荐程序被所述处理器执行时实现如下步骤：
    构建放疗计量图样本库，所述放疗计量图样本库至少包括放疗计量图；
    获取CT图像；
    用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT 图像对应的放疗计量图；
    将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本；
    获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划。
  10. 根据权利要求9所述的电子设备，其中，在获取CT图像之后，将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配之前，还包括：获取与所述CT图像对应的放疗勾勒图，所述放疗勾勒图包括靶区勾勒图和/或器官勾勒图；
    用预先训练生成的放疗计量预测网络模型对所述放疗勾勒图进行处理，得出与所述CT图像对应的放疗计量图。
  11. 根据权利要求9所述的电子设备，其中，所述放疗计量预测网络模型是端对端的V-net模型，包括输入层、四个编码层、四个解码层以及输出层，其中，所述输入层和所述编码层均包括3个卷积核，所述编码层和所述解码层的卷积步长均为2。
  12. 根据权利要求11所述的电子设备，其中，所述输入层、四个编码层、四个解码层和所述输出层对应的特征通道数量分别为1、32、64、128、256、256、256、128、64和1。
  13. 根据权利要求11所述的电子设备，其中，所述V-net模型还包括多个修正线性单元，每个修正线性单元均与一个解码层对应连接。
  14. 根据权利要求11所述的电子设备，其中，所述V-net模型采用群组归一化处理。
  15. 根据权利要求9所述的电子设备，其中，用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图之前，还包括：
    构建训练样本集，所述训练样本集中的训练样本包括CT图像和对应的放疗计量图；
    利用训练样本对所述放疗计量预测网络模型进行训练，得到训练放疗计量图；
    生成所述训练放疗计量图与训练样本中相应的放疗计量图间的相似度，当所述相似度大于预设相似度阈值时，结束对所述放疗计量预测网络模型的训练过程。
  16. 一种计算机可读存储介质，所述计算机可读存储介质中包括放疗计划推荐程序，所述放疗计划推荐程序被处理器执行时，实现如下步骤：
    构建放疗计量图样本库，所述放疗计量图样本库至少包括放疗计量图；
    获取CT图像；
    用预先训练生成的放疗计量预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图；
    将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配，得到与所述放疗计量图对应的匹配样本；
    获取与所述匹配样本对应的放疗计划，作为所述CT图像的推荐放疗计划。
  17. 根据权利要求16所述的计算机可读存储介质，其中，在获取CT图像之后，将所述放疗计量图与所述放疗计量图样本库中的放疗计量图样本进行匹配之前，还包括：获取与所述CT图像对应的放疗勾勒图，所述放疗勾勒图包括靶区勾勒图和/或器官勾勒图；
    用预先训练生成的放疗计量预测网络模型对所述放疗勾勒图进行处理，得出与所述CT图像对应的放疗计量图。
  18. 根据权利要求16所述的计算机可读存储介质，其中，所述放疗计量预测网络模型是端对端的V-net模型，包括输入层、四个编码层、四个解码层以及输出层，其中，所述输入层和所述编码层均包括3个卷积核，所述编码层和所述解码层的卷积步长均为2。
  19. 根据权利要求18所述的计算机可读存储介质，其中，所述输入层、四个编码层、四个解码层和所述输出层对应的特征通道数量分别为1、32、64、128、256、256、256、128、64和1。
  20. 根据权利要求16所述的计算机可读存储介质，其中，用预先训练生成的放疗计量 预测网络模型对所述CT图像进行处理，得出与所述CT图像对应的放疗计量图之前，还包括：
    构建训练样本集，所述训练样本集中的训练样本包括CT图像和对应的放疗计量图；
    利用训练样本对所述放疗计量预测网络模型进行训练，得到训练放疗计量图；
    生成所述训练放疗计量图与训练样本中相应的放疗计量图间的相似度，当所述相似度大于预设相似度阈值时，结束对所述放疗计量预测网络模型的训练过程。
PCT/CN2020/112367 2019-09-17 2020-08-31 放疗计划推荐方法、装置、电子设备及存储介质 WO2021052150A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910875109.2A CN110739043A (zh) 2019-09-17 2019-09-17 放疗计划推荐方法、装置、设备及存储介质
CN201910875109.2 2019-09-17

Publications (1)

Publication Number Publication Date
WO2021052150A1 true WO2021052150A1 (zh) 2021-03-25

Family

ID=69268102

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/112367 WO2021052150A1 (zh) 2019-09-17 2020-08-31 放疗计划推荐方法、装置、电子设备及存储介质

Country Status (2)

Country Link
CN (1) CN110739043A (zh)
WO (1) WO2021052150A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114864070A (zh) * 2022-04-28 2022-08-05 福建自贸试验区厦门片区Manteia数据科技有限公司 放疗信息管理系统

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110739043A (zh) * 2019-09-17 2020-01-31 平安科技（深圳）有限公司 放疗计划推荐方法、装置、设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020080912A1 (en) * 1997-10-15 2002-06-27 Mackie Thomas R. Method and apparatus for calibration of radiation therapy equipment and verification of radiation treatment
CN110085298A (zh) * 2019-04-26 2019-08-02 南方医科大学 基于深度网络学习的调强放疗计划三维剂量分布预测方法
CN110141801A (zh) * 2019-05-17 2019-08-20 天津大学 面向近距离粒子内放疗手术智能剂量规划方法
CN110197709A (zh) * 2019-05-29 2019-09-03 广州瑞多思医疗科技有限公司 一种基于深度学习与先验计划的三维剂量预测方法
CN110739043A (zh) * 2019-09-17 2020-01-31 平安科技（深圳）有限公司 放疗计划推荐方法、装置、设备及存储介质

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102184334B (zh) * 2011-05-19 2013-11-06 南京航空航天大学 一种基于检索的放射治疗计划系统及检索方法
CN106920234B (zh) * 2017-02-27 2021-08-27 北京连心医疗科技有限公司 一种复合式自动放疗计划的方法
CN108744313A (zh) * 2018-06-25 2018-11-06 西安大医数码科技有限公司 放疗计划规划方法、放疗计划系统及放射治疗系统
CN110211664B (zh) * 2019-04-25 2022-11-04 安徽大学 一种基于机器学习自动设计放射治疗方案的系统
CN110232964B (zh) * 2019-06-04 2023-11-14 苏州雷泰智能科技有限公司 基于预测剂量引导和高斯过程优化的计划实现方法及装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020080912A1 (en) * 1997-10-15 2002-06-27 Mackie Thomas R. Method and apparatus for calibration of radiation therapy equipment and verification of radiation treatment
CN110085298A (zh) * 2019-04-26 2019-08-02 南方医科大学 基于深度网络学习的调强放疗计划三维剂量分布预测方法
CN110141801A (zh) * 2019-05-17 2019-08-20 天津大学 面向近距离粒子内放疗手术智能剂量规划方法
CN110197709A (zh) * 2019-05-29 2019-09-03 广州瑞多思医疗科技有限公司 一种基于深度学习与先验计划的三维剂量预测方法
CN110739043A (zh) * 2019-09-17 2020-01-31 平安科技（深圳）有限公司 放疗计划推荐方法、装置、设备及存储介质

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114864070A (zh) * 2022-04-28 2022-08-05 福建自贸试验区厦门片区Manteia数据科技有限公司 放疗信息管理系统

Also Published As

Publication number Publication date
CN110739043A (zh) 2020-01-31

Similar Documents

Publication Publication Date Title
US20230352152A1 (en) Systems and methods for controlling user repeatability and reproducibility of automated image annotation correction
CN109741806B (zh) 一种医学影像诊断报告辅助生成方法及其装置
US10366491B2 (en) Deep image-to-image recurrent network with shape basis for automatic vertebra labeling in large-scale 3D CT volumes
CN109074500B (zh) 用于分割同一患者的医学图像的系统和方法
US20190065897A1 (en) Medical image analysis method, medical image analysis system and storage medium
US20140341449A1 (en) Computer system and method for atlas-based consensual and consistent contouring of medical images
CN110797101B (zh) 医学数据处理方法、装置、可读存储介质和计算机设备
CN109685102A (zh) 胸部病灶图像分类方法、装置、计算机设备及存储介质
US20200202516A1 (en) Prediction system, method and computer program product thereof
WO2021052150A1 (zh) 放疗计划推荐方法、装置、电子设备及存储介质
WO2015128803A2 (en) System and method for auto-contouring in adaptive radiotherapy
CN111785366B (zh) 患者治疗方案的确定方法、装置及计算机设备
CN113628726B (zh) 基于图神经网络的中医辨治推荐系统、方法和电子设备
WO2022178997A1 (zh) 医学影像配准方法、装置、计算机设备及存储介质
CN111860484A (zh) 一种区域标注方法、装置、设备及存储介质
CN111681247A (zh) 肺叶肺段分割模型训练方法和装置
Tang et al. Lesion segmentation and RECIST diameter prediction via click-driven attention and dual-path connection
CN116719904A (zh) 基于图文结合的信息查询方法、装置、设备及存储介质
EP3270308B9 (en) Method for providing a secondary parameter, decision support system, computer-readable medium and computer program product
CN110473636B (zh) 基于深度学习的智能医嘱推荐方法及系统
Permanasari et al. A web-based decision support system of patient time prediction using iterative dichotomiser 3 algorithm
CN110120266B (zh) 骨骼年龄评估方法
KR102514957B1 (ko) 사용자의 치매 정도 결정 방법 및 장치
CN116798565A (zh) 基于人工智能的医学数据处理方法、装置及计算机设备
CN116244492A (zh) 一种医疗导诊方法、装置及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20866624

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 20/07/2022)

122 Ep: pct application non-entry in european phase

Ref document number: 20866624

Country of ref document: EP

Kind code of ref document: A1