A model training method and device, a storage medium and an electronic device

Abstract

The specification discloses a model training method and device, a storage medium and an electronic device. In the model training method provided in the specification, a sample medical image is obtained, the sample medical image containing lesion tissues at different growth stages; a labeled metastasis feature is determined according to the difference between the lesion tissues at different growth stages in each sample medical image; the sample medical image is input into a lesion development prediction model to be trained, an image feature of the sample medical image is extracted through an extraction layer in the lesion development prediction model; a predicted metastasis feature of the lesion tissue is output according to the image feature through an output layer in the lesion development prediction model; and the lesion development prediction model is trained with the difference between the predicted metastasis feature and the labeled metastasis feature as an optimization target.

Description

Technical Field

[0001] This specification relates to the field of computer technology, and in particular to a model training method, apparatus, storage medium, and electronic device. Background Technology

[0002] Cancer is a highly dangerous disease. If metastasis is not treated promptly, there is a high probability of death for the patient. Therefore, assessing the likelihood of metastasis and determining the prognosis are crucial.

[0003] Currently, in clinical treatment, immunohistochemical markers such as Ki67, P53, CEA, and CYFRA21-1 are widely used to assess tumor malignancy and predict metastasis risk. However, these immunohistochemical tests require local tissue biopsy, making them invasive procedures. Furthermore, most immunohistochemical tests can only indicate the likelihood of tumor metastasis and cannot provide a definitive prognosis.

[0004] To address the aforementioned issues, this specification provides a model training method for training a model capable of safely and effectively predicting the risk of diseased tissue. Summary of the Invention

[0005] This specification provides a model training method, apparatus, storage medium, and electronic device to partially solve the aforementioned problems existing in the prior art.

[0006] The following technical solution is adopted in this specification: This manual provides a model training method, including: Acquire sample medical images, which contain diseased tissue at different stages of growth; Based on the differences between lesions at different growth stages in the medical images of each sample, the marked metastasis features are determined; The sample medical image is input into the lesion development prediction model to be trained, and the image features of the sample medical image are extracted through the extraction layer in the lesion development prediction model. The predicted metastasis features of the lesion tissue are output through the output layer of the lesion development prediction model based on the image features. The lesion development prediction model is trained with the goal of minimizing the difference between the predicted metastasis features and the labeled metastasis features.

[0007] Optionally, the lesion development prediction model further includes an attention layer; Before outputting the predicted metastatic features of the lesion tissue, the method further includes: Obtain a preset constraint image, wherein the constraint image contains the region of interest in the sample medical image; Through the attention layer, the attention image of the lesion development prediction model when processing the sample medical image is determined based on the image features; The parameters of the extraction layer and the attention layer in the lesion development prediction model are adjusted with the goal of minimizing the difference between the constraint image and the attention image.

[0008] Optionally, the constraint image is predetermined, specifically including: Identify the diseased tissue regions contained in the medical images of the samples; Extend the edge of the lesion tissue region outward by a specified length to obtain the region of interest, so that the region of interest includes the lesion tissue region and other risk areas surrounding the lesion tissue region; The sample medical image containing the region of interest is used as the constraint image.

[0009] Optionally, the annotation transfer feature includes the annotation transfer direction, and the predicted transfer feature includes the predicted transfer direction; The lesion development prediction model is trained with the goal of minimizing the difference between the predicted metastasis features and the labeled metastasis features. Specifically, this includes: The lesion development prediction model is trained with the goal of minimizing the difference between the predicted transfer direction and the labeled transfer direction.

[0010] Optionally, the predicted metastasis direction of the lesion tissue includes the predicted metastasis direction of the next growth stage of the lesion tissue relative to the current growth stage, and / or the predicted metastasis direction of the final growth stage of the lesion tissue relative to the current growth stage.

[0011] Optionally, the annotation transfer feature includes an annotation transfer method, and the prediction transfer feature includes a prediction transfer method; The lesion development prediction model is trained with the goal of minimizing the difference between the predicted metastasis features and the labeled metastasis features. Specifically, this includes: The lesion development prediction model is trained with the goal of minimizing the difference between the predicted transfer method and the labeled transfer method.

[0012] Optionally, the annotation transfer feature includes an annotation growth stage, and the prediction transfer feature includes a prediction growth stage; The lesion development prediction model is trained with the goal of minimizing the difference between the predicted metastasis features and the labeled metastasis features. Specifically, this includes: The lesion development prediction model is trained with the goal of minimizing the difference between the predicted growth stage and the labeled growth stage.

[0013] This manual provides a risk assessment method, employing methods such as... Figure 1 The method involves pre-training a lesion development prediction model, the method comprising: Acquire medical images containing diseased tissue; The medical image is input into a pre-trained lesion development prediction model, and the image features of the medical image are extracted through the extraction layer in the lesion development prediction model. The predicted metastasis features of the lesion tissue are output through the output layer of the lesion development prediction model based on the image features.

[0014] This specification provides a model training apparatus, including: An acquisition module is used to acquire sample medical images, which contain diseased tissue at different stages of growth; The annotation module is used to determine annotation transfer features based on the differences between lesion tissues at different growth stages in each sample medical image; The input module is used to input the sample medical image into the lesion development prediction model to be trained, and to extract the image features of the sample medical image through the extraction layer in the lesion development prediction model. The output module is used to output the predicted metastasis features of the lesion tissue based on the image features through the output layer of the lesion development prediction model; The training module is used to train the lesion development prediction model with the goal of minimizing the difference between the predicted transfer features and the labeled transfer features.

[0015] This specification provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described model training method.

[0016] This specification provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the model training method described above.

[0017] The above-mentioned technical solutions adopted in this specification can achieve the following beneficial effects: In the model training method provided in this specification, sample medical images are acquired, which contain lesions at different growth stages; based on the differences between lesions at different growth stages in each sample medical image, labeled transfer features are determined; the sample medical images are input into a lesion development prediction model to be trained, and image features of the sample medical images are extracted through the extraction layer in the lesion development prediction model; the predicted transfer features of the lesions are output through the output layer in the lesion development prediction model based on the image features; the lesion development prediction model is trained with the minimum difference between the predicted transfer features and the labeled transfer features as the optimization objective.

[0018] When training the lesion development prediction model using the model training method provided in this specification, medical images of lesions at different growth stages are used as training samples, and the training labels are determined based on the differences between lesions at different growth stages. The sample medical images are used as input to the lesion development prediction model, and the optimization objective is to minimize the difference between the predicted metastasis features output by the lesion development prediction model and the labeled metastasis features. This method can train a lesion development prediction model capable of predicting the metastasis risk of lesions at each stage, with good training results. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this specification and form part of this specification, illustrate exemplary embodiments and are used to explain this specification, but do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating one model training method described in this specification. Figure 2 This is a schematic diagram of the structure of a disease development prediction model in this specification; Figure 3 This is a schematic diagram of a constraint image used in this specification; Figure 4 This is a flowchart illustrating one of the risk assessment methods described in this specification. Figure 5 This is a schematic diagram of a model training device provided in this specification; Figure 6 This is a schematic diagram of a risk assessment device provided in this specification; Figure 7 The corresponding information provided in this specification Figure 1 A schematic diagram of an electronic device. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this specification clearer, the technical solutions of this specification will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of them. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0021] Additionally, it should be noted that all actions involving the acquisition of signals, information, or data in this invention are carried out in compliance with the relevant data protection laws and regulations of the country where the invention is located, and with authorization from the owner of the corresponding device.

[0022] The technical solutions provided in the various embodiments of this specification are described in detail below with reference to the accompanying drawings.

[0023] Figure 1 This is a flowchart illustrating a model training method provided in this specification.

[0024] S100: Acquire a sample medical image, the sample medical image containing diseased tissue at different stages of growth.

[0025] All steps in the model training method provided in this manual can be implemented by any electronic device with computing capabilities, such as a terminal or server.

[0026] In the model training method provided in this specification, the sample medical images can be obtained through any medical means, including but not limited to X-ray imaging, computed tomography (CT), magnetic resonance imaging (MRI), ultrasound images, and radionuclide imaging. Furthermore, the pathological tissues included in the medical images can be lesions occurring in different parts of the human body, including but not limited to gliomas, lung cancer, bone cancer, breast cancer, gastric cancer, and lymphoma. Simultaneously, the symptoms of the lesions include, but are not limited to, nodules, benign tumors, and malignant tumors.

[0027] In the medical field, most diseased tissues are academically defined into different growth stages to distinguish the duration and severity of the disease. In layman's terms, these are descriptions of the early and late stages of various clinical conditions. Taking tumors in cancer as an example, tumor growth is typically divided into five stages: precancerous stage, carcinoma in situ stage, invasive carcinoma stage, metastatic stage, and disseminated stage. Tumors in the disseminated stage are colloquially referred to as late-stage cancer. The size, growth rate, and risk of spread of diseased tissue vary at different growth stages; therefore, diseased tissues at different growth stages need to be considered separately.

[0028] Based on this, in this step, sample medical images containing diseased tissue at different growth stages can be acquired. The acquired sample medical images can contain medical images of all different growth stages of the diseased tissue that needs to be predicted.

[0029] S102: Based on the differences between lesions at different growth stages in the medical images of each sample, determine the labeled metastasis features.

[0030] The model training method provided in this manual is used to train a lesion development prediction model. In practical applications, the input to the lesion development prediction model is a medical image containing lesion tissue, and the output is the metastasis features of the lesion tissue in the medical image. These metastasis features characterize the risk of metastasis in the lesion tissue. It is conceivable that to obtain the metastasis features of lesion tissue, a comprehensive analysis of the same lesion tissue at different growth stages is required. Therefore, in this step, the labeled metastasis features can be determined based on the differences between lesion tissues at different growth stages in each sample medical image.

[0031] It is worth mentioning that, regarding the acquisition of annotation transfer features, the trainer can obtain the annotation transfer features independently based on the acquired sample medical images; alternatively, during the actual medical process, the trainer can directly determine the annotation transfer features in advance based on the medical images generated at each stage of the medical process, and provide the medical images as sample medical images along with the annotation transfer features to the trainer. This manual does not impose specific restrictions on this.

[0032] S104: Input the sample medical image into the lesion development prediction model to be trained, and extract the image features of the sample medical image through the extraction layer in the lesion development prediction model.

[0033] Figure 2 This document illustrates the structural diagram of the lesion development prediction model used in the model training method provided in this manual. (For example...) Figure 2 As shown, a disease progression prediction model may include at least an extraction layer and an output layer.

[0034] The extraction layer is used to extract image features from the input image. During the training of the lesion development prediction model, sample medical images can be input into the model to be trained, and the extraction layer extracts the image features of the sample medical images for use in subsequent steps.

[0035] S106: Through the output layer of the lesion development prediction model, output the predicted metastasis features of the lesion tissue based on the image features.

[0036] The output layer in the lesion development prediction model is used to output the predicted metastasis features of the corresponding lesion tissue based on the image features extracted by the extraction layer. In this step, the output layer can output the predicted metastasis features of the lesion tissue contained in the sample medical image based on the image features of the sample medical image extracted in step S104.

[0037] S108: The lesion development prediction model is trained with the goal of minimizing the difference between the predicted metastasis features and the labeled metastasis features.

[0038] In the model training method provided in this specification, the lesion development prediction model is trained using supervised learning. The training annotations are the labeled transfer features determined in step S102. When training the lesion development prediction model, it is desirable that the output of the lesion development prediction model be as close as possible to the annotations. Based on this, the training objective is to minimize the difference between the predicted transfer features output by the lesion development prediction model and the labeled transfer features.

[0039] When training the lesion development prediction model using the model training method provided in this specification, medical images of lesions at different growth stages are used as training samples, and the training labels are determined based on the differences between lesions at different growth stages. The sample medical images are used as input to the lesion development prediction model, and the optimization objective is to minimize the difference between the predicted metastasis features output by the lesion development prediction model and the labeled metastasis features. This method can train a lesion development prediction model capable of predicting the metastasis risk of lesions at each stage, with good training results.

[0040] Additional, such as Figure 2As shown, the lesion development prediction model used in the model training method provided in this specification may also include an attention layer. Through the attention layer, combined with a pre-set constraint image, the extraction layer can be further optimized. Specifically, a preset constraint image can be obtained, which contains the region of interest in the sample medical image; through the attention layer, based on the image features, the attention image of the lesion development prediction model when processing the sample medical image is determined; with the goal of minimizing the difference between the constraint image and the attention image, the parameters of the extraction layer and the attention layer of the lesion development prediction model are adjusted.

[0041] The constraint image is a highlighted image of the region containing diseased tissue in the sample medical image. Specifically, the constraint image is predetermined, which identifies the diseased tissue region contained in the sample medical image; the edge of the diseased tissue region is extended outward by a specified length to obtain the region of interest, so that the region of interest includes the diseased tissue region and other risk regions surrounding the diseased tissue region; the sample medical image containing the region of interest is used as the constraint image. In the model training method provided in this specification, the region of interest may include, but is not limited to, tumors, peritumoral regions, and surrounding organs, tissues, and air cavities, whether or not diseased. Other risk regions surrounding the diseased tissue region may include, for example, lymph nodes, pleura, and other organ regions that are prone to becoming carriers of diseased tissue metastasis.

[0042] Figure 3 A schematic diagram of the constraint images used in this specification is shown. For example... Figure 3 As shown, this is a constrained image corresponding to a sample medical image of a malignant pulmonary nodule with visceral pleural invasion. Figure 3 In the diagram, the white area in the lower left corner represents the visceral pleura. Irregularly shaped structures that are ligamentous to the visceral pleura are lung nodules, i.e., diseased tissue. The dark area encompassing the diseased tissue is the identified region of interest (ROI). It can be seen that the ROI extends outward beyond the diseased tissue area to ensure complete coverage, making it less susceptible to errors and model fluctuations during training. Furthermore, by extending the ROI outward from the diseased tissue area, it can include other surrounding areas at risk of disease progression and metastasis. This allows for better prediction and management of the risk of metastasis.

[0043] During training, the region of interest (ROI) is the area that the disease progression prediction model is expected to focus on when extracting image features from sample images. Based on this, an attention layer is used to additionally determine an attention image of the sample medical image, based on the image features extracted by the model's extraction layer. Using a constraint image containing the ROI as a reference, the parameters in the extraction and attention layers of the disease progression prediction model are adjusted with the optimization objective of minimizing the difference between the attention image and the constraint image, thereby optimizing the extraction and attention layers. It is important to note that the attention layer is only used during the training phase of the disease progression prediction model; it does not participate in the actual application of the model to predict the metastatic features of diseased tissues.

[0044] by Figure 3 Let's take a constrained image as an example for illustration. Figure 3 The diseased tissue area included is a pulmonary nodule, and the pleura surrounding the nodule is susceptible to erosion and metastasis, which constitutes other high-risk areas. Correspondingly, in areas where [the relevant technology / method] can be used... Figure 3 As a constraint image, the sample medical image should also be an image containing lung nodules in the lesion tissue region. After obtaining the attention image of the sample medical image through the extraction layer and attention layer in the lesion development prediction model, the difference between the constraint image and the attention image can be calculated, and the obtained difference can be used to adjust the parameters of the extraction layer in the model. In fact, this involves using... Figure 3 The constraint image shown guides the model's extraction of features from sample medical images when the lesion area is a pulmonary nodule. This ensures that the extracted image features not only accurately represent the pulmonary nodule region but also additionally include the surrounding pleural region, representing other risk areas. Therefore, based on the image features of the pulmonary nodule region and the surrounding pleural region in the sample medical image, the lesion development prediction model can output more accurate and comprehensive metastasis features, thus better predicting the subsequent development and metastasis of pulmonary nodules.

[0045] Of course, the guidance provided by the lesion development prediction model will also change when the constraint image changes. For example, if the constraint image includes lymph nodes as other risk areas, the extraction layer in the lesion development prediction model will also extract features of the lymph nodes surrounding the lesion tissue area.

[0046] Additionally, the lesion progression prediction model can output more specific predicted metastasis features, and correspondingly, more specific labeled metastasis features can be obtained when acquiring annotations.

[0047] For example, in one specific embodiment, more preferably, the labeled transfer feature includes a labeled transfer direction, and the predicted transfer feature includes a predicted transfer direction. When training the lesion development prediction model, minimizing the difference between the predicted transfer direction and the labeled transfer direction can be used as the optimization objective for training the lesion development prediction model.

[0048] The migration direction is used to characterize the specific direction of the lesion tissue during metastasis. Furthermore, the predicted migration direction of the lesion tissue includes the predicted migration direction of the next growth stage relative to the current growth stage, and / or the predicted migration direction of the final growth stage relative to the current growth stage. That is, when predicting the migration direction of the lesion tissue, the lesion development prediction model can output both the possible migration direction when the lesion tissue enters the next growth stage and the possible migration direction when the lesion tissue reaches the final growth stage. Depending on the output, the pre-determined labeled migration direction can also be different. When the output predicted migration direction is the possible migration direction of the lesion tissue entering the next growth stage, the labeled migration direction can be the direction of difference between the lesion tissue in the sample medical image and the sample medical image of the lesion tissue in the next growth stage; similarly, when the output predicted migration direction is the possible migration direction of the lesion tissue reaching the final growth stage, the labeled migration direction can be the direction of difference between the lesion tissue in the sample medical image and the sample medical image of the lesion tissue in the final growth stage. Furthermore, the lesion development prediction model can output both the predicted metastasis direction of the next growth stage and the predicted metastasis direction of the final growth stage simultaneously, or it can output only one of the two. This manual does not impose any specific restrictions on this.

[0049] For example, in one specific embodiment, more preferably, the annotation transfer feature includes an annotation transfer method, and the prediction transfer feature includes a prediction transfer method. When training the lesion development prediction model, minimizing the difference between the prediction transfer method and the annotation transfer method can be used as the optimization objective when training the lesion development prediction model.

[0050] The term "metastasis mode" is used to characterize how the lesion tissue metastasizes. Metastasis modes typically include, but are not limited to, direct spread, lymphatic metastasis, hematogenous metastasis, and implantation; this specification does not impose specific limitations on these. The metastasis mode indicated refers to the actual metastasis pattern of the lesion tissue in the sample medical images.

[0051] For example, in one specific embodiment, more preferably, the labeled transfer feature includes a labeled growth stage, and the predicted transfer feature includes a predicted growth stage. When training the lesion development prediction model, minimizing the difference between the predicted growth stage and the labeled growth stage can be used as the optimization objective for training the lesion development prediction model.

[0052] The growth stage is used to characterize the stage of a diseased tissue. Different diseased tissues exist in different growth stages in medicine. Taking tumors in cancer as an example, tumor growth is typically divided into five stages: precancerous stage, carcinoma in situ stage, invasive carcinoma stage, metastatic stage, and disseminated stage. Other diseased tissues can also be identified into different growth stages according to professional classifications in the medical field. The labeled growth stage refers to the actual growth stage of the diseased tissue in the sample medical image.

[0053] In conjunction with the model training methods described above, this manual also provides a risk prediction method, such as... Figure 4 As shown.

[0054] Figure 4 This is a flowchart illustrating a risk assessment method provided in this specification.

[0055] S200: Acquire medical images containing diseased tissue.

[0056] The risk assessment method provided in this manual is implemented using the lesion development prediction model trained by the model training method provided in this manual.

[0057] When assessing the risk of diseased tissue, one can first obtain medical images of the diseased tissue that need to be assessed for risk.

[0058] S202: Input the medical image into a pre-trained lesion development prediction model, and extract the image features of the medical image through the extraction layer in the lesion development prediction model.

[0059] In this step, the acquired medical image of the lesion tissue can be input into a pre-trained lesion development prediction model, which is trained using the model training method provided in this manual. Image features of the medical image can be extracted through the extraction layer in the lesion development prediction model.

[0060] S204: Through the output layer of the lesion development prediction model, output the predicted metastasis features of the lesion tissue based on the image features.

[0061] In this step, the image features of the medical image extracted in step S202 can be input into the output layer of the lesion development prediction model, which outputs the predicted metastasis features of the lesion tissue contained in the medical image. The predicted metastasis features output by the trained lesion development prediction model can reflect the possible metastasis of the lesion tissue well.

[0062] The above describes one or more methods for implementing model training and risk prediction in this manual. Based on the same approach, this manual also provides corresponding model training devices and risk prediction devices, such as... Figure 5 , Figure 6 As shown.

[0063] Figure 5 A schematic diagram of a model training device provided in this specification specifically includes: The acquisition module 300 is used to acquire sample medical images, which contain diseased tissues at different growth stages; The annotation module 302 is used to determine the annotation transfer features based on the differences between lesion tissues at different growth stages in each sample medical image; The input module 304 is used to input the sample medical image into the lesion development prediction model to be trained, and extract the image features of the sample medical image through the extraction layer in the lesion development prediction model; Output module 306 is used to output the predicted metastasis features of the lesion tissue based on the image features through the output layer of the lesion development prediction model; Training module 308 is used to train the lesion development prediction model with the goal of minimizing the difference between the predicted transfer features and the labeled transfer features.

[0064] Optionally, the lesion development prediction model further includes an attention layer; The device further includes a constraint module 310, specifically used to acquire a preset constraint image, the constraint image containing the region of interest in the sample medical image; through the attention layer, based on the image features, determine the attention image of the lesion development prediction model when processing the sample medical image; and adjust the parameters of the extraction layer and the attention layer of the lesion development prediction model with the minimum difference between the constraint image and the attention image as the optimization objective.

[0065] Optionally, the constraint module 310 is specifically used to determine the lesion tissue region contained in the sample medical image; extend the edge of the lesion tissue region outward by a specified length to obtain a region of interest, so that the region of interest includes the lesion tissue region and other risk areas around the lesion tissue region; and use the sample medical image containing the region of interest as a constraint image.

[0066] Optionally, the annotation transfer feature includes the annotation transfer direction, and the predicted transfer feature includes the predicted transfer direction; The training module 308 is specifically used to train the lesion development prediction model with the goal of minimizing the difference between the predicted transfer direction and the labeled transfer direction.

[0067] Optionally, the predicted metastasis direction of the lesion tissue includes the predicted metastasis direction of the next growth stage of the lesion tissue relative to the current growth stage, and / or the predicted metastasis direction of the final growth stage of the lesion tissue relative to the current growth stage.

[0068] Optionally, the annotation transfer feature includes an annotation transfer method, and the prediction transfer feature includes a prediction transfer method; The training module 308 is specifically used to train the lesion development prediction model with the goal of minimizing the difference between the predicted transfer method and the labeled transfer method.

[0069] Optionally, the annotation transfer feature includes an annotation growth stage, and the prediction transfer feature includes a prediction growth stage; The training module 308 is specifically used to train the lesion development prediction model with the goal of minimizing the difference between the predicted growth stage and the labeled growth stage.

[0070] Figure 6 This specification provides a schematic diagram of a risk assessment device, which specifically includes: Image acquisition module 400 is used to acquire medical images containing diseased tissue; Image input module 402 is used to input the medical image into a pre-trained lesion development prediction model, and extract the image features of the medical image through the extraction layer in the lesion development prediction model; The feature output module 404 is used to output the predicted metastasis features of the lesion tissue based on the image features through the output layer of the lesion development prediction model.

[0071] This specification also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described... Figure 1The model training method shown.

[0072] This instruction manual also provides Figure 7 The diagram shows a schematic structural representation of the electronic device. Figure 7 At the hardware level, the electronic device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for the business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to achieve the above-mentioned functions. Figure 1 The model training method is shown. Of course, in addition to the software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. In other words, the execution subject of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0073] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using hardware physical modules. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must also be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0074] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0075] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0076] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0077] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0078] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0079] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0080] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0081] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0082] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0083] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0084] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0085] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0086] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0087] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0088] The above description is merely an embodiment of this specification and is not intended to limit this specification. Various modifications and variations can be made to this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of the claims of this specification.

Claims

1. A model training method, characterized in that, The method includes: Acquire sample medical images, which contain diseased tissue at different stages of growth; Based on the differences between lesions at different growth stages in the medical images of each sample, the metastasis features are identified and labeled. The metastasis features are used to characterize the evolution process of the lesions, including the metastasis direction, metastasis mode, and growth stage of the lesions. The sample medical image is input into the lesion development prediction model to be trained, and the image features of the sample medical image are extracted through the extraction layer in the lesion development prediction model. A preset constraint image is obtained, in which the region of interest (ROI) in the sample medical image is marked; based on the image features, the attention image of the disease development prediction model is determined when processing the sample medical image using the attention layer included in the disease development prediction model; the parameters of the extraction layer and the attention layer of the disease development prediction model are adjusted with the optimization objective of minimizing the difference between the constraint image and the attention image; wherein, the preset determination of the constraint image includes: determining the diseased tissue region contained in the sample medical image; extending the edge of the diseased tissue region outward by a specified length to obtain the ROI, so that the ROI includes the diseased tissue region and other risk areas around the diseased tissue region that may be invaded by diseased tissue due to disease deterioration and metastasis; the sample medical image containing the ROI is used as the constraint image; The predicted metastasis features of the lesion tissue are output through the output layer of the lesion development prediction model based on the image features. The lesion development prediction model is trained with the goal of minimizing the difference between the predicted metastasis features and the labeled metastasis features.

2. The method as described in claim 1, characterized in that, The annotation transfer feature includes the annotation transfer direction, and the predicted transfer feature includes the predicted transfer direction; The lesion development prediction model is trained with the goal of minimizing the difference between the predicted metastasis features and the labeled metastasis features. Specifically, this includes: The lesion development prediction model is trained with the goal of minimizing the difference between the predicted transfer direction and the labeled transfer direction.

3. The method as described in claim 2, characterized in that, The predicted metastasis direction of the lesion tissue includes the predicted metastasis direction of the next growth stage of the lesion tissue relative to the current growth stage, and / or the predicted metastasis direction of the final growth stage of the lesion tissue relative to the current growth stage.

4. The method as described in claim 1, characterized in that, The annotation transfer feature includes the annotation transfer method, and the prediction transfer feature includes the prediction transfer method; The lesion development prediction model is trained with the goal of minimizing the difference between the predicted metastasis features and the labeled metastasis features. Specifically, this includes: The lesion development prediction model is trained with the goal of minimizing the difference between the predicted transfer method and the labeled transfer method.

5. The method as described in claim 1, characterized in that, The annotation transfer feature includes an annotation growth stage, and the prediction transfer feature includes a prediction growth stage; The lesion development prediction model is trained with the goal of minimizing the difference between the predicted metastasis features and the labeled metastasis features. Specifically, this includes: The lesion development prediction model is trained with the goal of minimizing the difference between the predicted growth stage and the labeled growth stage.

6. A risk prediction method, characterized in that, The method employs the pre-trained lesion development prediction model as described in any one of claims 1-5, the method comprising: Acquire medical images containing diseased tissue; The medical image is input into a pre-trained lesion development prediction model, and the image features of the medical image are extracted through the extraction layer in the lesion development prediction model. The predicted metastasis features of the lesion tissue are output through the output layer of the lesion development prediction model based on the image features.

7. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method described in any one of claims 1 to 6.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method described in any one of claims 1 to 6.

Images