Breast cancer risk assessment method and assessment model training method and device

Abstract

The invention provides a breast cancer risk assessment method and a training method and device of an assessment model, which can be applied to the field of digital intelligent health management. The evaluation method is executed by an electronic device and comprises the steps that breast cancer disease data are obtained, the breast cancer disease data comprise a focus segmentation image and target multi-dimensional monitoring data, and the target multi-dimensional monitoring data comprise electrocardiogram data, electroencephalogram data, myoelectricity data, sleep data and behavior expression data of a testee; processing the focus segmentation image and the target multi-dimensional monitoring data by using a risk assessment network to obtain a risk assessment probability; inputting the target multi-dimensional monitoring data into a convolutional neural network, and outputting an emotion classification probability; and processing the physiological risk assessment probability and the psychological emotion classification probability based on an adaptive fusion algorithm, and adjusting the weight between the two to obtain a breast cancer risk assessment result.

Description

Technical Field

[0001] The present disclosure relates to the field of digital intelligent health management, and in particular to a breast cancer risk assessment method, and a training method and device for an assessment model. Background Art

[0002] The number of new cancers and malignant tumors worldwide has hit new highs. Relevant data show that breast cancer accounts for the highest proportion of female cancers. Cancer mismanagement usually refers to the failure of cancer patients to receive effective management and tracking during the treatment process. Currently, more than 87% of cancer patients are mismanaged, and compared with receiving out-of-hospital management, the harm caused by mismanagement is often huge. For example, the person who mismanages cannot detect the recurrence of cancer in time and misses the opportunity for treatment. Therefore, there is an urgent need for a solution that can assess the risk of breast cancer. Patients can obtain breast cancer risk assessment results in a timely manner based on monitoring data, consult a doctor or seek medical treatment in a timely manner, and avoid serious consequences. Summary of the invention

[0003] In view of the above problems, the present disclosure provides a breast cancer risk assessment method, and a training method and device for an assessment model.

[0004] According to a first aspect of the present disclosure, a breast cancer risk assessment method is provided, comprising: obtaining breast cancer disease data, the breast cancer disease data comprising lesion segmentation images and target multi-dimensional monitoring data, the target multi-dimensional monitoring data comprising the tester's electrocardiogram data, electroencephalogram data, electromyography data, sleep data and behavioral expression data; using a risk assessment network to process the lesion segmentation images and the target multi-dimensional monitoring data to obtain a physiological risk assessment probability; inputting the target multi-dimensional monitoring data into a convolutional neural network to output a psychological emotion classification probability; processing the physiological risk assessment probability and the psychological emotion classification probability based on an adaptive fusion algorithm to obtain a breast cancer risk assessment result.

[0005] According to an embodiment of the present disclosure, a risk assessment network includes an image classification layer and a risk assessment layer; the risk assessment network is used to process lesion segmentation images and target multi-dimensional monitoring data to obtain a physiological risk assessment probability, including: detecting the lesion morphology in the lesion segmentation image to obtain a disease recurrence probability; inputting the lesion segmentation image into the image classification layer for feature analysis, and outputting a breast classification result; based on the breast classification result, determining the lesion morphology classification probability; inputting the target multi-dimensional monitoring data into the risk assessment layer, and outputting a breast cancer staging result; based on the breast cancer staging result, determining the breast cancer staging category probability; processing the disease recurrence probability, the lesion morphology classification probability and the breast cancer staging category probability according to an adaptive fusion algorithm to obtain a physiological risk assessment probability.

[0006] According to an embodiment of the present disclosure, obtaining breast cancer disease data includes: performing semantic segmentation on an acquired target breast scan image to obtain a lesion segmentation image; and / or performing lesion image entity extraction on an acquired breast knowledge graph to obtain a lesion segmentation image, wherein the breast knowledge graph includes lesion image entities, and the lesion image entities represent the lesion segmentation image obtained by performing semantic segmentation on the target breast scan image.

[0007] According to an embodiment of the present disclosure, a breast knowledge graph is constructed in the following manner: knowledge modeling is performed based on breast cancer disease data to obtain entity relationship information; the breast cancer disease data is preprocessed based on the entity relationship information to obtain preprocessed target data; the target data is processed based on a knowledge extraction mechanism to obtain a breast knowledge graph, wherein the knowledge extraction mechanism includes named entity recognition, relationship extraction, and entity linking.

[0008] According to an embodiment of the present disclosure, the breast cancer risk assessment method also includes: obtaining an initial breast scanning image and initial multi-dimensional monitoring data; normalizing the initial breast scanning image and the initial multi-dimensional monitoring data respectively to obtain a target breast scanning image and target multi-dimensional monitoring data.

[0009] A second aspect of the present disclosure provides a training method for an assessment model for breast cancer risk assessment, wherein the assessment model includes a risk assessment network and a convolutional neural network; the training method includes: obtaining sample breast scan images, sample multi-dimensional monitoring data, and sample breast cancer risk assessment labels; using the risk assessment network to process the sample breast scan images and the sample multi-dimensional monitoring data to obtain sample physiological risk assessment probabilities; inputting the sample multi-dimensional monitoring data into the convolutional neural network to output sample psychological emotion classification probabilities; processing the sample physiological risk assessment probabilities and the sample psychological emotion classification probabilities based on an adaptive fusion algorithm to obtain sample breast cancer risk assessment results; processing the sample breast cancer risk assessment results and the sample breast cancer risk assessment labels according to a loss function to obtain a target loss value; and training the assessment model according to the target loss value to obtain a trained assessment model.

[0010] The third aspect of the present disclosure provides a breast cancer risk assessment device, including: an acquisition module, used to acquire breast cancer disease data, the breast cancer disease data including lesion segmentation images and target multi-dimensional monitoring data, the target multi-dimensional monitoring data including the tester's electrocardiogram data, electroencephalogram data, electromyography data, sleep data and behavioral expression data; a processing module, used to process the lesion segmentation image and the target multi-dimensional monitoring data using a risk assessment network to obtain a physiological risk assessment probability; a psychological emotion classification probability module, used to input the target multi-dimensional monitoring data into a convolutional neural network and output the psychological emotion classification probability; a risk assessment result acquisition module, used to process the physiological risk assessment probability and the psychological emotion classification probability based on an adaptive fusion algorithm to obtain a breast cancer risk assessment result.

[0011] According to the embodiments of the present disclosure, a risk assessment network is used to process lesion segmentation images and target multi-dimensional monitoring data to obtain a physiological risk assessment probability, and the target multi-dimensional monitoring data is input into a convolutional neural network to output a psychological emotion classification probability. When the mood is stable or optimistic, the risk of recurrent breast cancer can be reduced. The psychological emotion classification probability can be used to fine-tune the breast cancer risk assessment result, and an adaptive fusion algorithm is used to process the physiological risk assessment probability and the emotion classification result, thereby obtaining a more accurate breast cancer risk assessment result. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The above and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0013] Figure 1 A flow chart of a method for identifying tumor features in ultrasound images based on information fusion and multi-task learning according to an embodiment of the present disclosure is shown;

[0014] Figure 2 A schematic diagram showing the principle of constructing a breast knowledge graph according to an embodiment of the present disclosure is shown;

[0015] Figure 3 A data flow diagram of an intelligent medical diagnosis management system according to an embodiment of the present disclosure is shown;

[0016] Figure 4 A flowchart of a method for training an assessment model for breast cancer risk assessment according to an embodiment of the present disclosure is shown;

[0017] Figure 5 A flow chart of a method for assessing breast cancer risk according to an embodiment of the present disclosure is shown;

[0018] Figure 6 A structural block diagram of a breast cancer risk device according to an embodiment of the present disclosure is shown;

[0019] Figure 7 A structural block diagram of a training device for an assessment model for breast cancer risk assessment according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0020] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0021] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0022] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0023] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0024] The incidence of new cancers and malignant tumors has hit new highs worldwide. Currently, more than 87% of cancer patients are out of control. Compared with outpatient management, the harm caused by out of control is manifested in the following aspects: the recovery of out of control patients is poor due to the lack of statistics on recovery; out of control patients cannot find and treat in time; cancer recurrence cannot be found in time; pain cannot be managed in time to ease the treatment process; and treatment opportunities are missed due to failure to find in time, etc. Due to the huge harm caused by out of control, it is urgent to improve the survival period and the quality of life after surgery.

[0025] In view of this, the embodiments of the present disclosure provide a breast cancer risk assessment method, an assessment model training method and a device. The method includes: obtaining breast cancer disease data, the breast cancer disease data includes lesion segmentation images and target multi-dimensional monitoring data, the target multi-dimensional monitoring data includes the tester's electrocardiogram data, electroencephalogram data, electromyography data, sleep data and behavioral expression data; using a risk assessment network to process the lesion segmentation image and the target multi-dimensional monitoring data to obtain a physiological risk assessment probability; inputting the target multi-dimensional monitoring data into a convolutional neural network to output a psychological emotion classification probability; processing the physiological risk assessment probability and the psychological emotion classification probability based on an adaptive fusion algorithm to obtain a breast cancer risk assessment result.

[0026] In the technical solution of the present disclosure, the user information (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0027] It should be noted that the sequence numbers of the operations in the following method are only used as representations of the operations for the purpose of description, and should not be regarded as representing the execution order of the operations. Unless explicitly stated, the method does not need to be executed completely in the order shown.

[0028] Figure 1 A flowchart of a method for identifying tumor features in ultrasound images based on information fusion and multi-task learning according to an embodiment of the present disclosure is shown.

[0029] like Figure 1 As shown, the method 100 includes operations S110 to S150.

[0030] In operation S110, breast cancer disease data is acquired, where the breast cancer disease data includes lesion segmentation images and target multi-dimensional monitoring data.

[0031] In operation S120, the lesion segmentation image and the target multi-dimensional monitoring data are processed using a risk assessment network to obtain a physiological risk assessment probability.

[0032] In operation S130, the target multi-dimensional monitoring data is input into a convolutional neural network, and a psychological emotion classification probability is output.

[0033] In operation S140, the physiological risk assessment probability and the psychological emotion classification probability are processed based on an adaptive fusion algorithm to obtain a breast cancer risk assessment result.

[0034] According to the embodiments of the present disclosure, breast cancer disease data can be obtained. It is possible to cooperate with patient-centered specialized institutions (cancer centers, authoritative medical institutions, industry associations, etc.) and cooperate with primary medical institutions to conduct preoperative and postoperative physiological monitoring and clinical examinations or add new data for doctors and patients. Breast cancer disease data can represent data related to breast cancer, and can include lesion segmentation images and target multi-dimensional monitoring data, etc., but is not limited to this. The embodiments of the present disclosure do not limit the specific data of breast disease data.

[0035] According to an embodiment of the present disclosure, the target multi-dimensional monitoring data includes the tester's ECG data, EEG data, EMG data, sleep data, and behavioral expression data. The target multi-dimensional monitoring data can come from monitoring and acquisition terminals and primary medical institutions. The monitoring and acquisition terminals include network dynamic ECG monitors (real-time transmission of ECG data over the network), sleep monitors (nighttime sleep quality assessment), psychological assessment equipment (micro-expressions and physiological electrical assessment), etc. The role of primary medical institutions is to collect and enter clinical data and collect and enter patients' multi-dimensional follow-up data (including out-of-hospital physical examination results, in-hospital clinical test results, color ultrasound, and computed tomography (CT) imaging).

[0036] According to the embodiments of the present disclosure, the risk assessment network can be used to process the lesion segmentation image and the target multi-dimensional monitoring data to obtain the physiological risk assessment probability. The physiological risk assessment probability can be used to make a rough assessment of the risk of a patient suffering from breast cancer.

[0037] According to the embodiments of the present disclosure, the target multi-dimensional monitoring data can be input into the convolutional neural network to output the probability of psychological emotion classification. The convolutional neural network can be used for physiological electrical signal analysis, and a deep belief network can be used to combine EEG signals and other biological signals, as well as convolutional neural networks and long short-term memory networks to extract features and conduct emotion classification research, but it is not limited to this. The embodiments of the present disclosure do not limit the specific model.

[0038] According to the embodiments of the present disclosure, the physiological risk assessment probability and the psychological emotion classification probability can be processed based on the adaptive fusion algorithm to obtain the breast cancer risk assessment result. The physiological risk assessment probability and the psychological emotion classification probability can be evaluated to determine their relative importance in the decision-making process. According to the evaluation results, weights are assigned to the physiological risk assessment probability and the psychological emotion classification probability. For example, the physiological risk assessment probability is 80% for the risk of breast cancer, and the probability of pessimism in the emotion classification is 90%. The physiological risk assessment probability and the psychological emotion classification probability are weighted, and the physiological risk assessment probability accounts for 75%, and the psychological emotion classification probability accounts for 25%. The breast cancer risk assessment result can be calculated, 80%*75%+90%*25%=82.5%. The breast cancer risk assessment result can be obtained as 82.5%. The obtained breast cancer risk assessment result cannot directly lead to the diagnosis result of the disease or the health status, and can only be used as an intermediate result.

[0039] According to the embodiments of the present disclosure, the data obtained from the remote home and follow-up service records can be input into the service platform of the centralized computing center, and the data obtained from the in-hospital examination cases can be input into the hospital electronic medical record system (Electronic Medical Record, EMR). The data in the centralized computing center service platform and the electronic medical record system can be used to build a multi-dimensional monitoring database (breast knowledge graph) for contracted high-risk and postoperative users to create a service foundation. The constructed multi-dimensional breast disease data is used to perform periodic iterative training of the evaluation model to improve the accuracy of the model. The risk assessment network and convolutional neural network are encapsulated in a multimodal data management system, and the input data is processed to obtain the physiological risk assessment probability of breast cancer. Finally, a patient-centered collaborative system of specialized institutions, data platforms and primary medical institutions is created, so that the management system will provide online physicians or offline assigned health steward services to contracted users, and provide survival rates and quality of life.

[0040] According to the embodiments of the present disclosure, the target population for the method of the present disclosure is limited, mainly including high-risk groups with family history or people who are out of hospital after receiving surgical treatment.

[0041] According to the embodiments of the present disclosure, a risk assessment network is used to process lesion segmentation images and target multi-dimensional monitoring data to obtain a physiological risk assessment probability, and the target multi-dimensional monitoring data is input into a convolutional neural network to output a psychological emotion classification probability. When the mood is stable or optimistic, the risk of recurrent breast cancer can be reduced. The psychological emotion classification probability can be used to fine-tune the breast cancer risk assessment result, and an adaptive fusion algorithm is used to process the physiological risk assessment probability and the emotion classification result, thereby obtaining a more accurate breast cancer risk assessment result.

[0042] According to an embodiment of the present disclosure, a risk assessment network includes an image classification layer and a risk assessment layer; the risk assessment network is used to process lesion segmentation images and target multi-dimensional monitoring data to obtain a physiological risk assessment probability, including: detecting the lesion morphology in the lesion segmentation image to obtain a disease recurrence probability; inputting the lesion segmentation image into the image classification layer for feature analysis, and outputting a breast classification result; based on the breast classification result, determining the lesion morphology classification probability; inputting the target multi-dimensional monitoring data into the risk assessment layer, and outputting a breast cancer staging result; based on the breast cancer staging result, determining the breast cancer staging category probability; processing the disease recurrence probability, the lesion morphology classification probability and the breast cancer staging category probability according to an adaptive fusion algorithm to obtain a physiological risk assessment probability.

[0043] According to an embodiment of the present disclosure, the risk assessment network may include an image classification layer and a risk assessment layer. The image classification layer can be used for image classification, such as a comprehensive lesion feature analysis model, a convolutional neural network model, etc., but is not limited thereto. The embodiments of the present disclosure do not limit the model used for image classification. The risk assessment layer is used to determine the breast cancer staging results. Breast tumors are divided into benign and malignant, while breast cancer is a malignant tumor. Tumors can usually be staged according to tumor size, presence or absence of lymph node metastasis, and presence or absence of distant metastasis. The staging results are generally divided into stage I, stage II, stage III, and stage IV.

[0044] According to the embodiments of the present disclosure, the lesion morphology in the lesion segmentation image can be detected to obtain the probability of disease recurrence. In the lesion morphology of the tumor, as the tumor becomes larger and its shape changes irregularly, the recurrence probability of breast cancer increases.

[0045] According to the embodiments of the present disclosure, the lesion segmentation image can be input into the image classification layer for feature analysis, and the breast classification result can be output. Based on the breast classification result, the lesion morphology classification probability can be determined. The breast classification result can be divided into benign and malignant, and the color characteristics and edge characteristics of the mass are integrated, and the density and brightness of the lesions and masses are feature analyzed to obtain the benign and malignant classification probability of the lesions and masses (i.e., the lesion morphology classification probability).

[0046] According to an embodiment of the present disclosure, target multi-dimensional monitoring data may be input into a risk assessment layer, and a breast cancer staging result may be output. Based on the breast cancer staging result, a breast cancer staging category probability may be determined.

[0047] According to the embodiments of the present disclosure, the probability of disease recurrence, the probability of lesion morphology classification and the probability of breast cancer staging category can be processed according to an adaptive fusion algorithm to obtain a physiological risk assessment probability.

[0048] According to the embodiments of the present disclosure, multimodal data is used to obtain the patient's tumor size, shape, lesion morphology classification, and breast cancer staging category, so that the patient's overall risk assessment can be performed to improve the accuracy of the assessment.

[0049] According to an embodiment of the present disclosure, a breast knowledge graph is constructed in the following manner: knowledge modeling is performed based on breast cancer disease data to obtain entity relationship information; the breast cancer disease data is preprocessed based on the entity relationship information to obtain preprocessed target data; the target data is processed based on a knowledge extraction mechanism to obtain a breast knowledge graph, wherein the knowledge extraction mechanism includes named entity recognition, relationship extraction, and entity linking.

[0050] Figure 2 A schematic diagram of the principle of constructing a breast knowledge graph according to an embodiment of the present disclosure is shown.

[0051] like Figure 2 As shown, the steps of constructing the breast knowledge graph are divided into knowledge modeling (step S210), data preprocessing (step S220), knowledge extraction (step S230) and knowledge storage (step S240). In step S210, knowledge modeling can be performed based on breast cancer disease data to obtain entity relationship information. The entity relationship information includes entity definition and relationship definition. Then step S220 can be executed to preprocess the data. The breast cancer disease data can be preprocessed based on the entity relationship information to obtain the preprocessed target data. Among them, the preprocessing can include cleaning and collating the text corpus, and the breast cancer disease data can be unstructured data, such as text and semantic information such as medical records and physical examination results of doctors and patients, imaging information of doctors and patients, emotion and expression video feature collection, etc., but it is not limited to this. The embodiments of the present disclosure do not limit the specific breast cancer disease data.

[0052] According to an embodiment of the present disclosure, step S230 can be executed based on the knowledge extraction mechanism to process the target data and obtain a breast knowledge graph. Entity extraction and relationship extraction can be completed based on the BERT-PGD-BiLSTM-CRF entity recognition model and relationship extraction technology of adversarial training, and the knowledge graph can be supplemented using an entity linking method that integrates attributes and semantic features. Finally, step S240 is executed based on the constructed breast knowledge graph, and polarity knowledge is stored and imported into a graph database management system. The knowledge extraction mechanism may include named entity recognition, relationship extraction, and entity linking, but is not limited to this. The embodiments of the present disclosure do not limit the specific algorithm of the knowledge extraction mechanism. Structured data may include breast cancer disease data in a unified standard format.

[0053] According to an embodiment of the present disclosure, obtaining breast cancer disease data includes: performing semantic segmentation on an acquired target breast scan image to obtain a lesion segmentation image; and / or performing lesion image entity extraction on an acquired breast knowledge graph to obtain a lesion segmentation image, wherein the breast knowledge graph includes lesion image entities, and the lesion image entities represent the lesion segmentation image obtained by performing semantic segmentation on the target breast scan image.

[0054] According to the embodiments of the present disclosure, semantic segmentation can be performed on the acquired target breast scan image to obtain a lesion segmentation image. Lesion image entities can also be extracted from the acquired breast knowledge graph to obtain a lesion segmentation image. The lesion segmentation image can be a segmentation image of a breast tumor at different stages. The early lesion segmentation image data can be obtained through the breast knowledge graph, and the recently acquired target breast scan image can be directly semantically segmented to obtain a lesion segmentation image.

[0055] According to the embodiments of the present disclosure, lesion segmentation images can be obtained through different channels, thereby improving the integrity and diversity of data. By acquiring lesion images at different stages, the accuracy and reliability of risk assessment are improved.

[0056] According to an embodiment of the present disclosure, the breast cancer risk assessment method also includes: obtaining an initial breast scanning image and initial multi-dimensional monitoring data; normalizing the initial breast scanning image and the initial multi-dimensional monitoring data respectively to obtain a target breast scanning image and target multi-dimensional monitoring data.

[0057] According to an embodiment of the present disclosure, an initial breast scanning image and initial multi-dimensional monitoring data are obtained, and the initial breast scanning image and the initial multi-dimensional monitoring data can be normalized respectively to obtain a target breast scanning image and target multi-dimensional monitoring data.

[0058] Figure 3 A data flow diagram of an intelligent medical diagnosis management system according to an embodiment of the present disclosure is shown.

[0059] like Figure 3As shown, the data acquisition system 310 can collect relevant multi-dimensional monitoring data, and the data acquisition system can include electrocardiogram equipment, electroencephalogram equipment, sleep equipment, blood oxygen equipment, camera equipment, electromyography equipment, in-hospital examination and medical record system, etc. The collected multimodal data is input into the multimodal data management system 320 for management, wherein the multimodal data management system can include external data, data upload, data download, data screening, data storage and data management functions. Relevant data can be obtained from the multimodal data management system 320, and the obtained data is input into the intelligent analysis model system 330 for analysis, and finally the evaluation result 340 output by the system is obtained, wherein the intelligent analysis model system 330 includes electrocardiogram analysis, electroencephalogram analysis, electromyography analysis, sleep analysis, expression analysis and behavior analysis, etc. The system output result 340 includes electrocardiogram evaluation (probability of arrhythmia, heart rate variability and premature beats, etc.), electroencephalogram evaluation (cognitive behavior and electroencephalogram model), electromyography evaluation (change in skin tissue conductivity, impedance change trend), sleep quality (mental impact, disease assessment on sleep), behavior assessment (patient expression behavior, psychological state assessment) and tissue assessment (risk of tissues and organs to previous medical history).

[0060] According to the embodiments of the present disclosure, breast cancer is taken as the entry point for disease management, a multidisciplinary integration specialist consortium cancer high-risk and postoperative population management service system is taken as the main body, and short-pipeline technical means are assisted to form an intelligent medical diagnosis and management system based on patient outcome reports. At the same time, the system has the function of multimodal risk and efficacy evaluation for the breast cancer population.

[0061] Figure 4 A flowchart of a method for training an assessment model for breast cancer risk assessment according to an embodiment of the present disclosure is shown.

[0062] like Figure 4 As shown, the method 400 includes operations S410 to S460.

[0063] In operation S410, sample breast scan images, sample multi-dimensional monitoring data, and sample breast cancer risk assessment labels are obtained.

[0064] In operation S420, the sample breast scan image and the sample multi-dimensional monitoring data are processed using a risk assessment network to obtain a sample physiological risk assessment probability.

[0065] In operation S430, the sample multi-dimensional monitoring data is input into a convolutional neural network, and the sample psychological emotion classification probability is output.

[0066] In operation S440, the sample physiological risk assessment probability and the sample psychological emotion classification probability are processed based on an adaptive fusion algorithm to obtain a sample breast cancer risk assessment result.

[0067] In operation S450, the sample breast cancer risk assessment result and the sample breast cancer risk assessment label are processed according to the loss function to obtain a target loss value.

[0068] In operation S460, the evaluation model is trained according to the target loss value to obtain a trained evaluation model.

[0069] According to the embodiments of the present disclosure, sample breast scan images, sample multi-dimensional monitoring data and sample breast cancer risk assessment labels can be obtained from authoritative medical institutions, industry associations, health management institutions and cancer center institutions. The data is cleaned, screened and classified to construct a breast knowledge graph. Knowledge discovery and mining can be performed based on the multimodal information in the breast knowledge graph to discover the invention mechanism, clinical efficacy, metastasis, metabolic markers and body characteristics of breast cancer. The sample physiological risk assessment probability can be obtained by processing the sample breast scan image and the sample multi-dimensional monitoring data using the risk assessment network. The sample multi-dimensional monitoring data is input into the convolutional neural network, and the sample psychological emotion classification probability can be output. Based on the adaptive fusion algorithm, the sample physiological risk assessment probability and the sample psychological emotion classification probability are processed to obtain the sample breast cancer risk assessment result. The sample breast cancer risk assessment result and the sample breast cancer risk assessment label are processed according to the loss function to obtain the target loss value. According to the target loss value, the evaluation model is trained to obtain the trained evaluation model. Finally, a design, development and management service system based on the computing power of the centralized computing center is formed.

[0070] Figure 5 A flow chart of a method for assessing breast cancer risk according to an embodiment of the present disclosure is shown.

[0071] like Figure 5 As shown, the present disclosure adopts a breast cancer risk assessment method to manage the high-risk postoperative population for breast cancer through three modules: "data collection module 510, assessment module 520, and management module 530".

[0072] According to the embodiment of the present disclosure, the "data collection module 510" is to form a specialized institution database 512 with data obtained from specialized institutions (cancer centers, authoritative medical institutions, industry associations, etc.) by cooperating with patient-centered specialized institutions. Cooperate with primary medical institutions to conduct preoperative and postoperative physiological monitoring and clinical examinations or add new data (collection of remote monitoring equipment, clinical data, multi-dimensional follow-up data, active entry by doctors and patients, etc.) to obtain multi-dimensional human monitoring data 511, and normalize the data obtained from the specialized institution database 512 and the multi-dimensional human monitoring data 511, and use the unified format as the data input port of the specialized intelligent centralized computing center.

[0073] According to an embodiment of the present disclosure, "Evaluation Module 520": through a specialized intelligent centralized computing center as a processing unit, a multimodal risk and efficacy evaluation system and an intelligent diagnosis and treatment system are supported. The system uses artificial intelligence and statistical principles through data input, multimodal data features as the input of a large model, and trains and generates an evaluation model. The multimodal risk and efficacy evaluation system may include a multi-level semantic segmentation large model (Segment Anything Model, SAM), a comprehensive lesion feature analysis model (image classification layer), a holographic knowledge graph and diagnosis and treatment decision model (risk assessment layer) and a 3D imaging knowledge graph (breast knowledge graph); the intelligent diagnosis and treatment system may include a physiological electrical signal analysis and evaluation model, which may include an evaluation and analysis of electrocardiogram, electroencephalogram, electromyography, sleep and behavioral expressions.

[0074] According to the embodiment of the present disclosure, "management module 530": through the output of the evaluation model, high-risk and post-operative management is performed on the contracted doctors and patients. After the doctors, patients and high-risk groups sign contracts with personal doctors and health managers, the doctors manage and track the contracted objects through the system's intelligent monitoring and early warning; the health managers conduct timely home intervention for the contracted people through the system's early warning and notification to reduce the probability of disease.

[0075] According to the embodiments of the present disclosure, the intelligent medical diagnosis and management system (i.e., the multimodal risk and efficacy assessment system and the intelligent diagnosis and treatment system) can perform risk and efficacy assessment and precise management recommendations, etc.; construct an intelligent diagnosis and treatment system to fill the gap in the management of high-risk and postoperative populations for breast cancer, enhance patients' awareness of their own diseases, establish a "patient-centered" and "integrated Chinese and Western medicine intervention" active and passive management model, extend the doctor-patient communication chain, and assist doctors in providing patients with real-time, personalized, and intelligent clinical decision-making and support.

[0076] Based on the above breast cancer risk assessment method, the present disclosure also provides a breast cancer risk assessment device. Figure 6 The device is described in detail.

[0077] Figure 6 A structural block diagram of a breast cancer risk device according to an embodiment of the present disclosure is shown.

[0078] like Figure 6 As shown, the identification device 600 of this embodiment includes an acquisition module 610, a processing module 620, a psychological emotion classification probability module 630, and a risk assessment result obtaining module 640.

[0079] The acquisition module 610 is used to acquire breast cancer disease data, which includes lesion segmentation images and target multi-dimensional monitoring data. The target multi-dimensional monitoring data includes the tester's electrocardiogram data, electroencephalogram data, electromyography data, sleep data and behavioral expression data.

[0080] The processing module 620 is used to process the lesion segmentation image and the target multi-dimensional monitoring data using the risk assessment network to obtain the physiological risk assessment probability.

[0081] The psychological emotion classification probability module 630 is used to input the target multi-dimensional monitoring data into the convolutional neural network and output the psychological emotion classification probability.

[0082] The risk assessment result obtaining module 640 is used to process the physiological risk assessment probability and the psychological emotion classification probability based on the adaptive fusion algorithm to obtain the breast cancer risk assessment result.

[0083] According to the embodiments of the present disclosure, a risk assessment network is used to process lesion segmentation images and target multi-dimensional monitoring data to obtain a physiological risk assessment probability, and the target multi-dimensional monitoring data is input into a convolutional neural network to output a psychological emotion classification probability. When the mood is stable or optimistic, the risk of recurrent breast cancer can be reduced. The psychological emotion classification probability can be used to fine-tune the breast cancer risk assessment result, and an adaptive fusion algorithm is used to process the physiological risk assessment probability and the emotion classification result, thereby obtaining a more accurate breast cancer risk assessment result.

[0084] According to an embodiment of the present disclosure, the processing module includes: a disease recurrence probability unit, a breast classification result unit, a lesion morphology classification probability unit, a staging result unit, a staging category probability unit and a physiological risk assessment probability obtaining unit.

[0085] The disease recurrence probability unit is used to detect the lesion morphology in the lesion segmentation image to obtain the disease recurrence probability.

[0086] The breast classification result unit is used to input the lesion segmentation image into the image classification layer for feature analysis and output the breast classification result.

[0087] The lesion morphology classification probability unit is used to determine the lesion morphology classification probability based on the breast classification result.

[0088] The staging result unit is used to input the target multi-dimensional monitoring data into the risk assessment layer and output the breast cancer staging results.

[0089] The staging category probability unit is used to determine the breast cancer staging category probability based on the breast cancer staging result.

[0090] The physiological risk assessment probability obtaining unit is used to process the disease recurrence probability, the lesion morphology classification probability and the breast cancer stage category probability according to the adaptive fusion algorithm to obtain the physiological risk assessment probability.

[0091] According to an embodiment of the present disclosure, the acquisition module includes: a semantic segmentation unit and an entity extraction unit.

[0092] The semantic segmentation unit is used to perform semantic segmentation on the acquired target breast scan image to obtain a lesion segmentation image.

[0093] The entity extraction unit is used to extract lesion image entities from the acquired breast knowledge graph to obtain a lesion segmentation image, wherein the breast knowledge graph includes lesion image entities, and the lesion image entity represents the lesion segmentation image obtained by semantic segmentation of the target breast scan image.

[0094] According to an embodiment of the present disclosure, a breast knowledge graph is constructed in the following manner: knowledge modeling is performed based on breast cancer disease data to obtain entity relationship information; the breast cancer disease data is preprocessed based on the entity relationship information to obtain preprocessed target data; the target data is processed based on a knowledge extraction mechanism to obtain a breast knowledge graph, wherein the knowledge extraction mechanism includes named entity recognition, relationship extraction, and entity linking.

[0095] According to an embodiment of the present disclosure, the identification device 600 of this embodiment further includes: an initial data acquisition module and a normalization module.

[0096] The initial data acquisition module is used to acquire initial breast scanning images and initial multi-dimensional monitoring data.

[0097] The normalization module is used to normalize the initial breast scan image and the initial multi-dimensional monitoring data respectively to obtain the target breast scan image and the target multi-dimensional monitoring data.

[0098] Based on the above-mentioned training method of the evaluation model for breast cancer risk assessment, the present disclosure also provides a training device for the evaluation model for breast cancer risk assessment. Figure 7 The device is described in detail.

[0099] Figure 7 A structural block diagram of a training device for an assessment model for breast cancer risk assessment according to an embodiment of the present disclosure is shown.

[0100] like Figure 7As shown, the training device 700 of the assessment model for breast cancer risk assessment of this embodiment includes a sample acquisition module 710, a sample physiological risk assessment probability module 720, a sample psychological emotion classification probability output module 730, a sample risk assessment result module 740, a target loss value acquisition module 750 and a training module 760.

[0101] The sample acquisition module 710 is used to acquire sample breast scan images, sample multi-dimensional monitoring data and sample breast cancer risk assessment labels.

[0102] The sample physiological risk assessment probability module 720 is used to process the sample breast scan image and the sample multi-dimensional monitoring data using the risk assessment network to obtain the sample physiological risk assessment probability.

[0103] The sample psychological emotion classification probability output module 730 is used to input the sample multi-dimensional monitoring data into the convolutional neural network and output the sample psychological emotion classification probability.

[0104] The sample risk assessment result module 740 is used to process the sample physiological risk assessment probability and the sample psychological emotion classification probability based on the adaptive fusion algorithm to obtain the sample breast cancer risk assessment result.

[0105] The target loss value obtaining module 750 is used to process the sample breast cancer risk assessment result and the sample breast cancer risk assessment label according to the loss function to obtain the target loss value.

[0106] The training module 760 is used to train the evaluation model according to the target loss value to obtain the trained evaluation model.

[0107] According to an embodiment of the present disclosure, any multiple modules among the modules, submodules, units, and subunits can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the modules, submodules, units, and subunits can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application-specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of them. Alternatively, at least one of the modules, submodules, units, and subunits can be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding function can be executed.

[0108] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions. It can be understood by those skilled in the art that the features recorded in the various embodiments and / or claims of the present disclosure can be combined and / or combined in a variety of ways, even if such a combination or combination is not explicitly recorded in the present disclosure. In particular, without departing from the spirit and teaching of the present disclosure, the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0109] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A method for assessing breast cancer risk, characterized in that: The evaluation method is performed by an electronic device, and the evaluation method includes: Acquire breast cancer disease data, wherein the breast cancer disease data includes lesion segmentation images and target multi-dimensional monitoring data, wherein the target multi-dimensional monitoring data includes the tester's electrocardiogram data, electroencephalogram data, electromyography data, sleep data, and behavioral expression data; Processing the lesion segmentation image and the target multi-dimensional monitoring data using a risk assessment network to obtain a physiological risk assessment probability; Inputting the target multi-dimensional monitoring data into a convolutional neural network and outputting the psychological emotion classification probability; The physiological risk assessment probability and the psychological emotion classification probability are processed based on an adaptive fusion algorithm, and the weight between the two is adjusted to obtain a breast cancer risk assessment result.

2. The method according to claim 1, characterized in that The risk assessment network includes an image classification layer and a risk assessment layer; The step of using a risk assessment network to process the lesion segmentation image and the target multi-dimensional monitoring data to obtain a physiological risk assessment probability includes: Detecting the morphology of the lesion in the lesion segmentation image to obtain the probability of disease recurrence; Inputting the lesion segmentation image into the image classification layer for feature analysis, and outputting a breast classification result; Based on the breast classification result, determining the probability of lesion morphology classification; Inputting the target multi-dimensional monitoring data into the risk assessment layer, and outputting the breast cancer staging result; Based on the breast cancer staging result, determining a breast cancer staging category probability; The disease recurrence probability, the lesion morphology classification probability and the breast cancer staging category probability are processed according to an adaptive fusion algorithm to obtain the physiological risk assessment probability.

3. The method according to claim 2, characterized in that Breast cancer disease data include: Performing semantic segmentation on the acquired target breast scan image to obtain the lesion segmentation image; and / or Lesion image entities are extracted from the acquired breast knowledge graph to obtain a lesion segmentation image, wherein the breast knowledge graph includes lesion image entities, and the lesion image entities represent the lesion segmentation image obtained by semantically segmenting the target breast scan image.

4. The method according to claim 3, characterized in that The breast knowledge graph is constructed in the following way: Perform knowledge modeling based on the breast cancer disease data to obtain entity relationship information; Preprocessing the breast cancer disease data based on the entity relationship information to obtain preprocessed target data; The target data is processed based on a knowledge extraction mechanism to obtain the breast knowledge graph, wherein the knowledge extraction mechanism includes named entity recognition, relationship extraction and entity linking.

5. The method according to claim 1, characterized in that The method further comprises: Obtaining initial breast scan images and initial multi-dimensional monitoring data; The initial breast scan image and the initial multi-dimensional monitoring data are respectively normalized to obtain the target breast scan image and the target multi-dimensional monitoring data.

6. A training method for an assessment model for breast cancer risk assessment, characterized in that: The assessment model includes a risk assessment network and a convolutional neural network; the training method includes: Obtain sample breast scan images, sample multi-dimensional monitoring data, and sample breast cancer risk assessment labels; Processing the sample breast scan image and the sample multi-dimensional monitoring data using the risk assessment network to obtain a sample physiological risk assessment probability; Inputting the sample multi-dimensional monitoring data into the convolutional neural network, and outputting the sample psychological emotion classification probability; Processing the sample physiological risk assessment probability and the sample psychological emotion classification probability based on an adaptive fusion algorithm, adjusting the weight between the sample physiological risk assessment probability and the sample psychological emotion classification probability, and obtaining a sample breast cancer risk assessment result; Processing the sample breast cancer risk assessment result and the sample breast cancer risk assessment label according to a loss function to obtain a target loss value; According to the target loss value, the evaluation model is trained to obtain a trained evaluation model.

7. A breast cancer risk assessment device, characterized in that: include: An acquisition module is used to acquire breast cancer disease data, wherein the breast cancer disease data includes lesion segmentation images and target multi-dimensional monitoring data, wherein the target multi-dimensional monitoring data includes the tester's electrocardiogram data, electroencephalogram data, electromyography data, sleep data, and behavioral expression data; A processing module, used for processing the lesion segmentation image and the target multi-dimensional monitoring data using a risk assessment network to obtain a physiological risk assessment probability; A psychological emotion classification probability module is used to input the target multi-dimensional monitoring data into a convolutional neural network and output the psychological emotion classification probability; The risk assessment result obtaining module is used to process the physiological risk assessment probability and the psychological emotion classification probability based on an adaptive fusion algorithm, adjust the weight between the two, and obtain a breast cancer risk assessment result.

8. An electronic device comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images