Female tumor comprehensive management system and method based on female tumor large model

Abstract

The invention discloses a female tumor comprehensive management system and method based on a female tumor large model, and relates to the technical field of female tumor management, and the system comprises a model training module and a personalized management suggestion generation module. The model training module is used for training a model with a double-engine architecture based on multiple groups of multi-modal data associated with the female tumor to obtain a large female tumor model; and the personalized management suggestion generation module is used for generating personalized management suggestions of the target user by using the female tumor large model when the female tumor related data of the target user is received. According to the female tumor large model, individual differences can be fully considered, so that the generated personalized management suggestions are more targeted. Moreover, the whole system architecture has good expandability and self-learning ability, can continuously optimize the female tumor large model along with continuous input of new data and new knowledge, and provides more accurate personalized management suggestions for female tumor patients.

Description

Technical Field

[0001] The present invention relates to the technical field of female tumor management, and in particular to a comprehensive female tumor management system and method based on a large female tumor model. Background Art

[0002] Female cancer diseases are characterized by high morbidity, long diagnosis and treatment pathways, and reliance on multidisciplinary collaboration. Currently, comprehensive management approaches for female cancer are deficient in many aspects. First, data integration is difficult. Data formats vary across hospitals and departments, making it difficult to centrally and effectively integrate multiple sets of multimodal data. This results in incomplete information and hinders comprehensive analysis of the condition. Second, the degree of personalization is limited. Traditional management is often based on medical guidelines, making it difficult to fully consider individual differences, and the resulting management recommendations are insufficiently targeted. Furthermore, knowledge updates lag behind, with diagnostic and treatment knowledge relying heavily on expert experience, making it difficult to update in a timely manner and unable to provide patients with the most cutting-edge management recommendations. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to address the deficiencies of the existing technology and specifically provide a comprehensive management system and method for female tumors based on a large female tumor model, specifically as follows:

[0004] 1) In the first aspect, the present invention provides a comprehensive management system for female tumors based on a large female tumor model. The specific technical solutions are as follows:

[0005] Includes model training module and personalized management suggestion generation module;

[0006] The model training module is used to train a model with a dual-engine architecture based on multiple sets of multimodal data related to female tumors to obtain a large model of female tumors. The dual-engine architecture includes an intelligent base engine and a large model inference engine. The intelligent base engine is used to convert multimodal data into structured feature vectors and, based on the structured feature vectors, generate a knowledge graph and a diagnosis and treatment pathway map, which are then transmitted to the large model inference engine. The large model inference engine is used to generate personalized management recommendations based on the knowledge graph and diagnosis and treatment pathway map.

[0007] The personalized management suggestion generation module is used to: when receiving female tumor-related data of a target user, use the female tumor large model to generate personalized management suggestions for the target user.

[0008] The beneficial effects of the comprehensive female tumor management system based on the female tumor large model provided by the present invention are as follows:

[0009] The large model for female tumors can efficiently integrate multiple sets of multimodal data. The intelligent base engine converts the data into structured feature vectors and generates knowledge graphs and treatment pathways, providing a solid foundation for subsequent reasoning and enabling deep data utilization. The large model reasoning engine generates personalized management recommendations based on these results, fully considering individual differences to make the generated personalized management recommendations more targeted. Moreover, the entire system architecture has excellent scalability and self-learning capabilities. With the continuous input of new data and new knowledge, it can continuously optimize the large model for female tumors, provide more accurate personalized management recommendations for female cancer patients, effectively make up for the shortcomings of traditional methods, and promote the development of intelligent and personalized female tumor management.

[0010] On the basis of the above scheme, the comprehensive management system for female tumors based on the large female tumor model of the present invention can also be improved as follows.

[0011] Furthermore, the intelligent base engine is also used to: standardize the multimodal data using clinical medical terminology standards before converting it into structured feature vectors, and use a privacy protection mechanism to process the standardized multimodal data, and convert the processed multimodal data into structured feature vectors.

[0012] The beneficial effects of adopting this further approach are: on the one hand, standardizing multimodal data using clinical medical terminology standards can eliminate differences in data from different sources, ensure data quality, and make the subsequently generated structured feature vectors more accurate and unified, thereby improving model training and diagnostic accuracy. On the other hand, using a privacy protection mechanism to process data effectively protects patient privacy, enhances data compliance, increases patient trust in data sharing, and promotes the integration and utilization of multi-source data. This in turn enhances the practicality and reliability of the entire model in the comprehensive management of female cancers, laying a solid foundation for providing precise and personalized diagnosis and treatment recommendations.

[0013] Furthermore, the large model inference engine is also used to: generate electronic medical records based on the knowledge graph and diagnosis and treatment pathway diagram, combined with medical knowledge related to female tumors and expert advice;

[0014] The personalized management suggestion generation module is also used to: when receiving female tumor-related data of a target user, generate an electronic medical record of the target user using a large female tumor model.

[0015] The benefits of this further approach are as follows: First, the large-scale model inference engine combines medical knowledge related to female cancers and expert advice to generate electronic medical records, making diagnosis and treatment decisions more targeted and scientific, improving both efficiency and quality. Second, the personalized management recommendation generation module utilizes the large-scale model of female cancers to generate electronic medical records, providing physicians with professional references and helping to reduce medical errors. It also facilitates the storage, query, and analysis of medical records, improving the quality of medical services and providing more precise diagnosis, treatment, and management services for female cancer patients.

[0016] Furthermore, it also includes a propaganda and education content generation module, which is used to generate personalized health knowledge propaganda and education content based on the personalized management suggestions and electronic medical records of the target users.

[0017] The beneficial effects of adopting the above-mentioned further scheme are: on the one hand, the health knowledge education content generated based on personalized management suggestions, etc., fits the actual condition and needs of the target users, provides targeted health guidance, and helps patients better understand their own conditions and management methods. On the other hand, it can promote doctor-patient communication. Personalized education content can help patients and their families understand the diagnosis and treatment process and decision-making basis more clearly, and enhance their trust and cooperation in medical plans. The third is to improve health management compliance. By acquiring accurate health knowledge, patients can more actively participate in their own health management, improve treatment effects and quality of life. Fourth, it reduces the burden of education on medical staff. The automatically generated education content can be used as auxiliary material, saving medical staff time and energy in repeated explanations.

[0018] Furthermore, it also includes a medical document automatic generation module, which is used to automatically generate the target user's medical records, discharge summary and follow-up records based on the target user's diagnosis and treatment records after the target user receives treatment.

[0019] The beneficial effects of adopting the above-mentioned further solution are as follows: ① The automatic generation module of medical documents can automatically generate medical records, discharge summaries, and follow-up records, which greatly saves medical staff time and energy in writing documents, allowing them to focus more on patient treatment and care. ② The automatically generated documents are based on standardized templates and accurate data, and the content is more standardized and complete, reducing errors and omissions caused by human negligence and improving the overall quality of medical documents. ③ These automatically generated documents can be stored as structured data to facilitate subsequent statistical analysis, quality control, and medical research, providing strong support for hospital management decisions and helping to optimize medical processes and services. ④ The latest records can be generated in real time based on the patient's treatment progress, ensuring the timeliness of medical documents, providing the medical team with the latest patient status information, and facilitating timely adjustment of treatment plans.

[0020] 2) In a second aspect, the present invention also provides a comprehensive management method for female tumors based on a large female tumor model. The specific technical solution is as follows:

[0021] Based on multiple sets of multimodal data related to female tumors, a model with a dual-engine architecture was trained to obtain a large model of female tumors. The dual-engine architecture includes an intelligent base engine and a large model inference engine. The intelligent base engine is used to convert multimodal data into structured feature vectors and, based on these structured feature vectors, generate a knowledge graph and treatment pathway map, which are then transmitted to the large model inference engine. The large model inference engine is used to generate personalized management recommendations based on the knowledge graph and treatment pathway map.

[0022] When receiving female tumor-related data of the target user, the female tumor big model is used to generate personalized management suggestions for the target user.

[0023] On the basis of the above scheme, the comprehensive management method of female tumors based on the large female tumor model of the present invention can also be improved as follows.

[0024] Furthermore, the intelligent base engine is also used to: standardize the multimodal data using clinical medical terminology standards before converting it into structured feature vectors, and use a privacy protection mechanism to process the standardized multimodal data, and convert the processed multimodal data into structured feature vectors.

[0025] Furthermore, the big model reasoning engine is also used to generate electronic medical records based on the knowledge graph and treatment pathway diagram, and in combination with medical knowledge and expert advice related to female tumors. The method also includes: when receiving female tumor-related data of the target user, using the female tumor big model to generate the electronic medical records of the target user.

[0026] Furthermore, it also includes: generating personalized health knowledge education content based on the target users' personalized management suggestions and electronic medical records.

[0027] Furthermore, it also includes: after the target user receives treatment, the target user's medical record, discharge summary and follow-up record are automatically generated according to the target user's diagnosis and treatment record.

[0028] 3) In a third aspect, the present invention further provides an electronic device, comprising a processor coupled to a memory, wherein the memory stores at least one computer program, and the at least one computer program is loaded and executed by the processor so that the electronic device implements any of the above-mentioned methods for comprehensive management of female tumors based on a large model of female tumors.

[0029] 4) In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for comprehensive management of female tumors based on a large female tumor model.

[0030] It should be noted that the beneficial effects achieved by the technical solutions of the second to fourth aspects of the present invention and the corresponding possible implementation methods can be found in the above-mentioned technical effects of the first aspect and its corresponding possible implementation methods, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments of the present invention:

[0032] Figure 1 This is a schematic structural diagram of a comprehensive female tumor management system based on a large female tumor model according to an embodiment of the present invention;

[0033] Figure 2 This is a flow chart of a method for comprehensive management of female tumors based on a large female tumor model according to an embodiment of the present invention;

[0034] Figure 3 The figure is a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The principles and features of the present invention are described below. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0036] The following describes in detail the technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems using specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The following embodiments of the present invention are described in conjunction with the accompanying drawings.

[0037] like Figure 1 As shown, an embodiment of the present invention provides a comprehensive management system for female tumors based on a large female tumor model, including a model training module and a personalized management suggestion generation module;

[0038] The model training module is used to train a model with a dual-engine architecture based on multiple sets of multimodal data related to female tumors to obtain a large model of female tumors. The dual-engine architecture includes an intelligent base engine and a large model inference engine. The intelligent base engine is used to convert multimodal data into structured feature vectors and, based on the structured feature vectors, generate a knowledge graph and a diagnosis and treatment pathway map, which are then transmitted to the large model inference engine. The large model inference engine is used to generate personalized management recommendations based on the knowledge graph and diagnosis and treatment pathway map.

[0039] Among them, the intelligent base engine includes a multimodal encoding layer, a cross-modal fusion layer, a graph neural network layer and a diagnosis and treatment path generation layer.

[0040] Among them, each female cancer patient corresponds to a set of multimodal data, which includes medical records, questionnaires, medical images (CT images, MRI images and ultrasound images), pathological images, genetic test results and laboratory test results.

[0041] The specific implementation process of the intelligent base engine converting multimodal data into structured feature vectors is as follows:

[0042] S10. Extract heterogeneous data features through the multimodal encoding layer. Specifically:

[0043] 1) Use a 3D convolutional neural network to extract spatial feature vectors of image data (medical images and pathological images), specifically:

[0044] First, medical and pathological images are preprocessed, such as through normalization and cropping, to ensure that their size and pixel value range meet the network input requirements. The processed data is then fed into a 3D convolutional neural network. The 3D convolutional layer in the 3D convolutional neural network uses a learnable 3D convolution kernel to perform convolution operations along the three-dimensional space (length, width, and channel dimensions) of the image data. The convolution kernel slides over the input data, performing a weighted sum of the pixel values ​​within the local receptive field to extract the local spatial features of the image data and generate a multi-channel feature map. These feature maps, after passing through an activation function (such as ReLU), enhance the nonlinear representation of the features. Pooling layers (such as max pooling or average pooling) then downsample the feature maps to reduce their dimensionality and computational complexity while preserving important spatial feature information. Multiple 3D convolutional and pooling layers are stacked to form the deep structure of the network. As the network depth increases, more abstract and advanced spatial features of the image data are gradually extracted. Finally, the feature map output by the last convolutional layer is flattened into a one-dimensional vector, which is the spatial feature vector of the extracted image data.

[0045] 2) A biomedical text encoder is used to extract semantic feature vectors of pathology data (medical records, questionnaires, genetic test results, and laboratory test results).

[0046] Among them, the biomedical text encoder is used to convert text information into low-dimensional dense semantic feature vectors. Specifically, it can be pre-trained language models such as BioBERT and ClinicalBERT based on the Transformer architecture. It can be pre-trained on large-scale biomedical text corpus and learn vocabulary, semantics and syntactic knowledge in the biomedical field.

[0047] The specific implementation process of using the biomedical text encoder to extract the semantic feature vector of pathological data is as follows:

[0048] First, the pathology data is preprocessed, including operations such as word segmentation and removal of irrelevant symbols. The preprocessed text sequence is then input into the biomedical text encoder. The embedding layer of the biomedical text encoder maps words into word vectors, capturing their semantic information. The multi-layer Transformer architecture of the biomedical text encoder then uses a self-attention mechanism to iteratively calculate the word vectors in the text sequence, fully considering contextual relationships and generating a context-sensitive representation of each word, thereby constructing semantic features of the text. Finally, the sequence features output by the biomedical text encoder are pooled and other operations are performed to produce a fixed-length semantic feature vector that effectively represents the semantic information of the pathology data text.

[0049] 3) A one-dimensional convolutional neural network is used to extract the SNP feature vectors of the gene sequence data in the gene test results. Specifically:

[0050] When a one-dimensional convolutional neural network is used to extract the SNP feature vector of a gene sequence, the gene sequence in the gene test results is first preprocessed and converted into a numerical form suitable for network input. Next, the processed sequence is input into the one-dimensional convolutional neural network. The one-dimensional convolution layer of the network uses a learnable one-dimensional convolution kernel to slide on the gene sequence, and through the convolution operation, it captures local patterns and features in the gene sequence, especially the variation features related to SNPs (single nucleotide polymorphisms). The result of the convolution operation generates a feature map. After the activation function (such as ReLU) enhances the nonlinear expression ability of the feature, the feature map is downsampled through the pooling layer to reduce the feature dimension while retaining key feature information. The stacking of multiple layers of convolution and pooling layers gradually extracts more advanced SNP features. Finally, the feature map of the last layer is flattened into a one-dimensional vector to form the SNP feature vector of the gene sequence data.

[0051] Among them, the SNP feature vector is a numerical vector that represents the characteristics of single nucleotide variations in a gene sequence.

[0052] S11. Through the cross-modal fusion layer, the gated attention mechanism is used to dynamically weight the spatial feature vector, semantic feature vector, and SNP feature vector to generate a unified structured feature vector. Specifically:

[0053] The gated attention unit, which implements the gated attention mechanism, receives spatial, semantic, and SNP feature vectors and calculates an attention score. The attention score is normalized and converted into a weight coefficient. The cross-modal fusion layer then weights and sums the feature vectors to generate a preliminary fused feature vector. Nonlinear transformations (such as activation functions) are then used to enhance the feature representation, enabling it to better capture the complex relationships between modalities. Finally, normalization and data dimensionality reduction are performed to obtain a unified structured feature vector.

[0054] The specific implementation process of the intelligent base engine to generate a knowledge graph based on structured feature vectors is as follows:

[0055] Based on the graph neural network layer, entity relationships are extracted from structured feature vectors to construct a knowledge graph containing disease entities, gene entities, and treatment entities;

[0056] The graph neural network layer can be specifically a GNN graph neural network (GNN graph neural network includes graph convolutional network and graph attention network, etc.), and the process of generating the knowledge graph is as follows:

[0057] First, the structured feature vectors are converted into graph-structured data. Disease entities, gene entities, and treatment entities are represented as nodes in the graph, with the node features representing the corresponding structured feature vectors. Next, edges between nodes are initialized to represent possible relationships. The initial edge weights can be determined based on prior knowledge or feature similarity. Next, the graph data is fed into a GNN (Graph Neural Network). The GNN uses a message-passing mechanism to transmit information between nodes and their neighbors. Each node updates its own features based on the features of its neighbors and the weights of its edges. Through the iterative operation of multiple layers of the GNN, node features are enriched and updated, more accurately capturing the potential relationships between different entities. Finally, based on the updated node features and edge predictions, a knowledge graph is constructed containing disease entities, gene entities, and treatment entities. Disease entities refer to specific diseases suffered by female cancer patients, including female cancers and other diseases such as breast cancer and ovarian cancer. Gene entities refer to genes that are associated with the disease or may affect the occurrence, progression, and treatment of the disease. Treatment entities refer to various treatment measures and methods used to treat the disease, including surgical treatment methods, drug treatments, and radiotherapy regimens.

[0058] The specific implementation process of generating a diagnosis and treatment pathway diagram based on the adjacent structured feature vectors is as follows:

[0059] The structured feature vector is processed through the diagnosis and treatment pathway generation layer to obtain the clinical event dependency, and a diagnosis and treatment pathway diagram with a timestamp is generated based on the clinical event dependency.

[0060] Among them, the diagnosis and treatment pathway generation layer can be specifically a temporal Transformer model, a long short-term memory network or a convolutional neural network.

[0061] Taking the temporal Transformer model as an example, the process of obtaining the dependency relationship of clinical events is explained in detail:

[0062] First, the structured feature vectors are arranged in chronological order to form a sequence input. The model's encoder uses a self-attention mechanism to capture the interdependencies between feature vectors at different time steps in the sequence. This mechanism eliminates the need for pre-specified time windows or kernel sizes, allowing for flexible attention to global temporal dependencies. The feature sequence output by the encoder undergoes nonlinear transformations, such as a multilayer perceptron, to generate a feature representation that incorporates temporal dependency information. During model training, supervised learning is used, based on labeled clinical event sequence data, to optimize model parameters so that the output features accurately reflect the dependencies between clinical events.

[0063] Clinical event dependencies refer to the temporal order and mutual influence of clinical events (such as disease diagnosis, treatment measures, and examination results). For example, symptoms appear first (event A), followed by a diagnosis through examination (event B), and then treatment (event C). Event A may trigger event B, which in turn may prompt event C. This dependency helps understand the disease progression and evaluate treatment effectiveness.

[0064] The specific implementation process of generating a diagnosis and treatment pathway diagram with timestamps based on clinical event dependencies is as follows:

[0065] First, clinical events, including disease diagnosis, treatment measures, and test results, are identified based on their dependencies. Next, timestamps are determined—the order in which these events occur on the timeline. Next, using techniques such as time series annotation, timestamps are associated with corresponding clinical events. These timestamped events are then plotted using graphical tools to form a treatment pathway diagram. Each node in the pathway diagram represents a clinical event, while edges indicate dependencies between events. The timestamps on the edges illustrate the order in which these events occur. The resulting treatment pathway diagram intuitively illustrates the patient's entire journey from diagnosis to treatment.

[0066] The large-model inference engine includes a knowledge embedding layer, a knowledge embedding layer, a pathway modeling layer, and a multi-source inference layer. The knowledge embedding layer includes knowledge graph embedding models (TransR model, DistMult model, and ComplEx model, etc.), the pathway modeling layer includes a temporal coding network, and the decision output layer includes a large language model. The specific implementation process of the large-model inference engine generating personalized management recommendations based on the knowledge graph and diagnosis and treatment pathway diagram is as follows:

[0067] S12. Use the knowledge graph embedding model to map the entity relationships of the knowledge graph to a low-dimensional vector space to generate a set of knowledge vectors. Specifically:

[0068] First, each entity and relationship in the knowledge graph is encoded. Entity encoding can be based on unique identifiers or attribute features, while relationship encoding is based on its type and semantics. Next, the encoded entities and relationships are input into the knowledge graph embedding model. The knowledge graph embedding model converts high-dimensional entity and relationship features into low-dimensional vectors through a mapping function. During the conversion process, the knowledge graph embedding model retains the semantic information of entities and relationships and their structural relationships in the knowledge graph. For example, entities with similar attributes or in similar contexts will be close to each other in the vector space. At the same time, the knowledge graph embedding model also ensures that the relationship vectors can reflect the reasonable associations and semantic constraints between entities. Ultimately, the resulting set of knowledge vectors contains low-dimensional representations of all entities and relationships.

[0069] S13. Use a time series coding network to perform state coding on the time series events in the diagnosis and treatment pathway diagram and output a path time series vector. Specifically:

[0070] First, the time series events in the treatment pathway diagram are arranged in chronological order, and each event is assigned a timestamp. Then, the timestamped event sequence is input into the temporal encoding network. The network performs state encoding on each event, extracts its features, and generates corresponding state vectors. These state vectors contain both semantic and temporal information about the event. Finally, all state vectors are combined in chronological order to form a path time series vector, which fully represents the dynamic changes in the treatment pathway.

[0071] S14. Through the multi-source reasoning layer, the knowledge vector set and the path time series vector are directly spliced ​​by dimension to obtain the joint reasoning input data;

[0072] S15. Input the joint reasoning input data into the large language model to generate personalized management suggestions, which include at least one of dietary management suggestions, psychological management suggestions, sleep management suggestions and treatment plans (also known as customized decision paths).

[0073] The large language model can be a medically enhanced large language model. This is a deep learning model designed specifically for the medical field, with advanced reasoning and semantic understanding capabilities. It is trained on large-scale medical text data to learn medical knowledge, terminology, and grammatical rules. This model is enhanced for processing medical information and can accurately understand the semantics and context of medical text. It can generate medical reports, analyze medical records, assist in medical decision-making, and provide services such as medical information retrieval, helping to improve medical efficiency and quality.

[0074] Based on multiple sets of multimodal data related to female tumors, a model with a dual-engine architecture was trained to obtain a large model of female tumors. The specific implementation process is as follows:

[0075] Multiple sets of multimodal data related to female cancer are collected, including medical records, questionnaires, medical imaging (CT, MRI, ultrasound), pathological images, genetic testing, and laboratory tests. This data is then annotated, covering psychological management recommendations, dietary management recommendations, sleep management recommendations, and treatment plans. The annotation process, conducted by professional physicians or medical experts, ensures accuracy and consistency, based on medical guidelines, expert consensus, and clinical practice. The annotated data is then fed into a model with a dual-engine architecture for training. One engine focuses on understanding and extracting features from multimodal data, processing heterogeneous data and generating structured feature vectors. The other engine utilizes the annotated information for decision-making and reasoning, learning how to generate appropriate management recommendations and treatment plans based on the feature vectors. During training, the model continuously optimizes parameters to improve the accuracy and consistency of predictions based on the annotated data. Through supervised learning using a large amount of annotated data, the model gradually learns relevant medical knowledge and patterns related to female cancer, enabling accurate analysis and decision-making on new multimodal data. Ultimately, the fully trained models are integrated into a comprehensive female cancer model, enabling personalized recommendations and decision-making support for the diagnosis, treatment, and management of female cancers.

[0076] In another implementation method, the model with a dual-engine architecture is the "dual-engine + multimodal" architecture of the female tumor AI large model "Mulan", and the trained female tumor large model is the female tumor AI large model "Mulan".

[0077] In another implementation, a multimodal data processing module can be set up separately to convert multimodal data into structured feature vectors, and transmit the structured feature vectors to the intelligent base engine. The intelligent base engine generates a knowledge graph and a diagnosis and treatment pathway map based on the structured feature vectors.

[0078] The personalized management suggestion generation module is used to: when receiving female tumor-related data of a target user, use the female tumor large model to generate personalized management suggestions for the target user.

[0079] Among them, the target users are female cancer patients. The target users can directly input female cancer-related data, or ask others to help the target users input female cancer-related data. The target users' female cancer-related data include: medical records, questionnaires, medical images (CT images, MRI images and ultrasound images), pathological images, genetic test results and at least one of laboratory test results.

[0080] Optionally, in the above technical solution, the intelligent base engine is also used to: standardize the multimodal data using clinical medical terminology standards before converting the multimodal data into structured feature vectors, and use a privacy protection mechanism to process the standardized multimodal data, and convert the processed multimodal data into structured feature vectors.

[0081] The specific implementation process of standardizing multimodal data using clinical medical terminology standards is as follows:

[0082] First, according to clinical medical terminology standards, medical terms in multimodal data are identified and annotated, covering disease names, test results, treatment plans, etc., to ensure uniform terminology. Then, entity recognition and standardized conversion are performed on unstructured text data (such as medical records and questionnaires) to align them with standard terminology. For non-text data such as medical images and pathological images, the associated medical terms are annotated and the description format is unified. Data such as genetic testing and laboratory tests are named and expressed in a standardized manner according to standard terminology.

[0083] The SNOMED CT / ICD-10 clinical medical terminology standards can be used as clinical medical terminology standards. Privacy protection mechanisms can include federated learning technology and desensitizing encryption technology.

[0084] Before converting multimodal data into structured feature vectors, the intelligent base engine first standardizes the multimodal data using clinical medical terminology standards. This step ensures the consistency and standardization of all medical terminology in the dataset, facilitating subsequent analysis and processing. Subsequently, a privacy protection mechanism is used to process the standardized multimodal data to protect patient privacy and data security. Specifically, technologies such as data desensitization and encryption can be used to ensure that sensitive information is not leaked during data conversion and use. Finally, the processed multimodal data is converted into structured feature vectors through feature extraction and encoding techniques to facilitate subsequent analysis, reasoning, and decision support. These structured feature vectors integrate key information from the multimodal data.

[0085] It should be noted that: ① During the privacy protection process, the use of technologies such as data desensitization can effectively remove sensitive information from the data while retaining the data's characteristic information and ensuring data availability. ② When converting data into structured feature vectors, feature information related to model training is extracted rather than the original data itself, which helps to further reduce the risk of privacy leakage. ③ The standardized processing and privacy protection mechanism are designed to maintain data consistency, ensuring that the processed data can still accurately reflect the characteristics and distribution of the original data. ④ The privacy protection mechanism is designed with the performance requirements of model training in mind, ensuring that while protecting privacy, the training effect of the model will not be significantly reduced. ⑤ Privacy protection processing will not undermine the integrity and availability of the data. The model can still learn effective features from the processed data for training and prediction.

[0086] Optionally, in the above technical solution, the large model inference engine is further used to: generate electronic medical records based on the knowledge graph and the diagnosis and treatment pathway diagram, combined with medical knowledge related to female tumors and expert advice;

[0087] The large-scale inference engine utilizes a large language model, taking the knowledge graph and treatment pathway map as prior knowledge input. Combined with relevant medical knowledge about female cancers and expert advice, it performs a comprehensive analysis of a patient's multimodal data. It first understands key information in the patient data, such as disease characteristics and genetic status. Then, through deep interaction with the knowledge graph and treatment pathway map, it queries diagnostic and treatment patterns for similar cases. Integrating expert advice, it generates electronic medical records, ensuring they contain key diagnostic and treatment information.

[0088] The personalized management suggestion generation module is also used to: when receiving female tumor-related data of a target user, generate an electronic medical record of the target user using a large female tumor model.

[0089] Optionally, the above technical solution further includes a propaganda content generation module, which is used to generate personalized health knowledge propaganda content based on the personalized management suggestions and electronic medical records of the target user, specifically:

[0090] Obtain personalized management recommendations and electronic medical records for target users. These data contain key information such as the user's disease status, treatment plan, lifestyle, etc. Then, using natural language processing technology, extract topics and key points related to health knowledge education from these data, such as dietary recommendations, exercise guidance, and precautions for taking medication. Next, the system retrieves health knowledge content that matches the extracted topics from the medical knowledge base and expert advice to ensure the accuracy and scientific nature of the information. Based on the characteristics of the target users (such as age, gender, educational level, etc.), the retrieved knowledge content is personalized and adjusted using easy-to-understand language and expressions. Finally, the adjusted content is integrated into structured health education materials and pushed to users through appropriate channels to help them better manage their health.

[0091] Optionally, the above technical solution further includes a medical document automatic generation module, which is used to automatically generate the target user's medical records, discharge summary and follow-up records based on the target user's diagnosis and treatment records after the target user receives treatment. Specifically:

[0092] Collect all kinds of diagnosis and treatment records of target users after treatment, including examination reports, treatment execution details and medical advice. The natural language processing module extracts key information, such as treatment effects, changes in physical indicators and recovery status. The medical record module organizes this information into a timeline medical record based on preset templates and logic, highlighting changes in the condition and treatment response. Filter key discharge-related content, such as the final diagnosis, treatment summary and discharge instructions, and generate a discharge summary in accordance with the standard format of medical documents. Create a follow-up plan template based on the patient's recovery progress and medical advice requirements to record the specific circumstances of each follow-up, i.e., follow-up records, such as review results and rehabilitation guidance.

[0093] Optionally, the above technical solution further includes an intelligent consultation and early screening module, which is used to actively collect symptoms and medical history of female tumors (i.e., female tumor-related data) based on medical guidelines and risk assessment models, and generate early screening conclusions and recommendations. Specifically:

[0094] First, the system identifies the typical symptoms and key medical history information of female tumors based on authoritative medical guidelines. Through multiple channels, including medical record systems, questionnaires, and medical databases, the system proactively collects information on patients' symptoms, such as abnormal bleeding and lumps, as well as medical history information such as family genetic disease history and previous medical history. Then, combined with a risk assessment model, the collected information is quantitatively analyzed to calculate the patient's risk of developing female tumors. Based on the risk assessment results, the system compares the early screening standards in medical guidelines to generate personalized early screening conclusions for the patient, such as recommending specific imaging examinations or tumor marker testing, and providing corresponding recommendations such as examination frequency and precautions to assist doctors in early diagnosis and intervention. Specifically, the risk assessment model can be obtained by training a neural network.

[0095] Optionally, the above technical solution further includes a diagnosis and treatment decision support module, which is used to automatically generate classification, staging and treatment plan recommendations that comply with medical guidelines based on female tumor-related data, specifically:

[0096] The data related to female tumors are standardized and feature extracted to ensure the accuracy and consistency of the data. The processed data is input into a trained neural network model, which is trained based on a large amount of medical guidelines and clinical data and can understand the classification and staging standards of tumors and the corresponding treatment plans. The trained neural network model automatically matches the classification and staging standards in the medical guidelines based on the characteristics of the input data to generate preliminary classification and staging results. At the same time, combined with the individual characteristics of the patient (such as age, health status, etc.), the model extracts appropriate treatment plan recommendations from medical guidelines and clinical best practices, such as surgical methods, chemotherapy drug selection, radiotherapy dosage, etc.

[0097] Optionally, the above technical solution further includes a medical quality control module, which supports the question-and-answer quality control of the doctor's behavior process and the quality control of medical records. Specifically:

[0098] A question-and-answer system based on natural language processing was built to collect conversations between doctors and patients and identify the type of question (e.g., symptom consultation, treatment plan inquiry). Using a pre-set medical knowledge graph and rules, the system analyzed the completeness, accuracy, and standardization of the doctor's answers. If a doctor's answer lacked key information or did not comply with guidelines, the system would flag it and prompt the doctor to provide additional corrections.

[0099] Use document parsing technology to extract key medical record information, such as diagnosis results, treatment plans, and medical history. Set quality control rules, such as data integrity and logical consistency. The system compares medical record content with these rules, flags non-compliant records, and provides feedback to physicians for correction.

[0100] The female tumor comprehensive management system based on the female tumor large model of the present invention supports local deployment and cloud integration, can be connected to the hospital HIS / PACS / LIS system, and has an interface structure, which is convenient for expansion and adaptation to other subsystems.

[0101] The beneficial effects of the female tumor comprehensive management system based on the female tumor large model of the present invention are as follows:

[0102] 1) Full-process support: covering prevention, early screening, diagnosis, personalized treatment, and follow-up management; high adaptability and credibility: optimized training for Chinese women, in line with Chinese guidelines; multimodal fusion and high interpretability: heterogeneous data fusion + explainable prediction mechanism; high efficiency and low threshold for use: supporting question-and-answer interaction among primary care physicians, automatic document generation, and quality control assistance; data privacy protection and standardization: equipped with international standard access and secure data processing mechanisms, and supporting multiple rounds of question-and-answer consultations, risk prediction, disease diagnosis, and treatment recommendation output.

[0103] 2) The first "Women's Tumor Specialty Model" system architecture that combines a multimodal intelligent base with a medical big language model; systematically integrates intelligent medical consultation, classification and staging prediction, personalized treatment recommendations, and quality control document generation; multimodal semantic unified modeling and medical knowledge graph fusion mechanism; dual-engine collaborative working mechanism achieves a 1+1>2 effect; the first domestic dual-engine medical reasoning platform designed and implemented specifically for women's tumor scenarios; fully compatible with international standards such as SNOMED CT and has strong data privacy protection capabilities.

[0104] The application prospects of the comprehensive female tumor management system based on the large-scale female tumor model of this invention are as follows: 1. Early screening scenario: After patients at primary care hospitals enter basic information, the system completes a risk assessment through intelligent question-and-answer, identifies high-risk groups, and recommends further examinations. 2. Diagnosis scenario: Based on the patient's CT images, pathology reports, and genetic test results, the system provides a staging diagnosis and treatment plan recommendations. 3. Quality control scenario: Automatically analyze surgical records, check the integrity of key elements against medical guidelines, and generate a quality control score report.

[0105] In the above embodiments, although the steps are numbered S1, S2, etc., these are only specific embodiments given by the present invention. Those skilled in the art may adjust the execution order of S1, S2, etc. according to actual conditions, which is also within the scope of protection of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0106] like Figure 2 As shown, a comprehensive management method for female tumors based on a large female tumor model according to an embodiment of the present invention includes the following steps:

[0107] S1. Based on multiple sets of multimodal data related to female tumors, a model with a dual-engine architecture is trained to obtain a large model of female tumors. The dual-engine architecture includes an intelligent base engine and a large model inference engine. The intelligent base engine is used to convert multimodal data into structured feature vectors, and based on the structured feature vectors, generate a knowledge graph and a diagnosis and treatment pathway map, and transmit them to the large model inference engine. The large model inference engine is used to generate personalized management recommendations based on the knowledge graph and diagnosis and treatment pathway map.

[0108] S2. When receiving the target user's female tumor-related data, use the female tumor large model to generate personalized management suggestions for the target user.

[0109] Optionally, in the above technical solution, the intelligent base engine is also used to: standardize the multimodal data using clinical medical terminology standards before converting the multimodal data into structured feature vectors, and use a privacy protection mechanism to process the standardized multimodal data, and convert the processed multimodal data into structured feature vectors.

[0110] Optionally, in the above technical solution, the big model inference engine is also used to: generate electronic medical records based on the knowledge graph and treatment pathway diagram, and in combination with medical knowledge and expert advice related to female tumors. The method also includes: when receiving female tumor-related data of the target user, using the female tumor big model to generate an electronic medical record for the target user.

[0111] Optionally, the above technical solution also includes: generating personalized health knowledge education content based on the personalized management suggestions and electronic medical records of the target users.

[0112] Optionally, the above technical solution further includes: after the target user receives treatment, the target user's medical record, discharge summary and follow-up record are automatically generated according to the target user's diagnosis and treatment record.

[0113] It should be noted that the beneficial effects of the female tumor comprehensive management method based on the large female tumor model provided by the above embodiment are the same as the beneficial effects of the female tumor comprehensive management system based on the large female tumor model, which will not be repeated here. In addition, when the system provided by the above embodiment realizes its functions, it only uses the division of the above functional modules as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to actual conditions to complete all or part of the functions described above. In addition, the system and method embodiments provided by the above embodiment belong to the same concept. The specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0114] Among them, the comprehensive management system for female tumors based on the large model of female tumors of the present invention can be a computer program (including program code) running on a computer device. For example, the comprehensive management system for female tumors based on the large model of female tumors of the present invention is an application software that can be used to execute the corresponding steps in the comprehensive management method for female tumors based on the large model of female tumors of the present invention.

[0115] In some embodiments, the comprehensive management system for female tumors based on the large model of female tumors of the present invention can be implemented by a combination of software and hardware. As an example, the comprehensive management system for female tumors based on the large model of female tumors of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the comprehensive management method for female tumors based on the large model of female tumors of the present invention. For example, the processor in the form of a hardware decoding processor can adopt one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0116] The modules described in the embodiments of the present invention may be implemented in software or hardware, and the name of a module does not necessarily limit the module itself.

[0117] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any one of the above-mentioned methods for comprehensive management of female tumors based on the large model of female tumors is implemented. That is, an electronic device according to an embodiment of the present invention may include but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the method for comprehensive management of female tumors based on the large model of female tumors shown in any embodiment of the present invention by calling the computer program.

[0118] In an alternative embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3The electronic device 4000 shown includes: a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data exchange between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the number of transceivers 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0119] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the present disclosure. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

[0120] Bus 4002 may include a path for transmitting information between the above components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. Bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 In the figure, only one thick line is used to represent the bus 4002, but this does not mean that there is only one bus or one type of bus.

[0121] The memory 4003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0122] The memory 4003 is used to store application code (computer program) for executing the solution of the present invention, and is controlled by the processor 4001. The processor 4001 is used to execute the application code stored in the memory 4003 to implement the content shown in the above method embodiment.

[0123] Among them, the electronic device can also be a terminal device, and the terminal device can be any device that can install applications, including at least one of a smartphone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a smart TV, and a smart car device.

[0124] It should be noted that Figure 3 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0125] A computer-readable storage medium according to an embodiment of the present invention stores a computer program, which, when executed by a processor, implements any of the above-mentioned methods for comprehensive management of female tumors based on a large female tumor model.

[0126] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, or the like.

[0127] In an exemplary embodiment, a computer program product or computer program is also provided, comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform any of the aforementioned methods for comprehensive management of female tumors based on a large female tumor model.

[0128] Computer program code for performing the operations of the present invention may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0129] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of 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 can also occur in a different order than the order 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 and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0130] The computer-readable storage medium provided in the embodiment of the present invention may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EEPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or component.

[0131] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

[0132] The above description is merely a preferred embodiment of the present invention and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of disclosure involved in the present invention is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also includes other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned disclosed concepts. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in the present invention.

[0133] It should be noted that the terms "first," "second," and the like in the specification and claims of this application are used to distinguish similar objects and to define a specific order or precedence. Where appropriate, the order used for similar objects may be interchanged, such that the embodiments of the present application described herein can be implemented in an order other than the order shown or described.

[0134] Those skilled in the art will appreciate that the present invention may be implemented as a system, method, or computer program product. Therefore, the present invention may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, the present invention may be implemented in the form of a computer program product embodied in one or more computer-readable media containing computer-readable program code.

[0135] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A comprehensive management system for female tumors based on a large female tumor model, characterized by: Includes model training module and personalized management suggestion generation module; The model training module is used to train a model with a dual-engine architecture based on multiple sets of multimodal data associated with female tumors to obtain a large model of female tumors; wherein the dual-engine architecture includes an intelligent base engine and a large model inference engine; the intelligent base engine is used to convert the multimodal data into structured feature vectors, and based on the structured feature vectors, generate a knowledge graph and a diagnosis and treatment pathway map, and transmit them to the large model inference engine; the large model inference engine is used to generate personalized management recommendations based on the knowledge graph and the diagnosis and treatment pathway map; The personalized management suggestion generating module is used to: when receiving female tumor related data of a target user, generate personalized management suggestions for the target user by using the female tumor macro model.

2. A comprehensive female tumor management system based on a large female tumor model according to claim 1, characterized in that: The intelligent base engine is also used to: before converting the multimodal data into the structured feature vector, standardize the multimodal data using clinical medical terminology standards, and use a privacy protection mechanism to process the standardized multimodal data, and convert the processed multimodal data into the structured feature vector.

3. The female tumor comprehensive management system based on the female tumor large model according to claim 2 is characterized in that: The large model inference engine is further used to: generate an electronic medical record based on the knowledge graph and the diagnosis and treatment pathway diagram, combined with medical knowledge and expert advice related to female tumors; The personalized management suggestion generation module is further configured to: upon receiving female tumor-related data of a target user, generate an electronic medical record of the target user using the female tumor macro model.

4. The female tumor comprehensive management system based on the female tumor large model according to claim 3 is characterized in that: It also includes a propaganda and education content generation module, which is used to generate personalized health knowledge propaganda and education content based on the personalized management suggestions and electronic medical records of the target user.

5. The female tumor comprehensive management system based on the female tumor large model according to claim 4 is characterized in that: It also includes a medical document automatic generation module, which is used to automatically generate the target user's medical record, discharge summary and follow-up record based on the target user's diagnosis and treatment record after the target user receives treatment.

6. A comprehensive management method for female tumors based on a large female tumor model, characterized in that: include: Based on multiple sets of multimodal data associated with female tumors, a model with a dual-engine architecture is trained to obtain a large model of female tumors. The dual-engine architecture includes an intelligent base engine and a large model inference engine. The intelligent base engine is used to convert the multimodal data into structured feature vectors, and based on the structured feature vectors, generate a knowledge graph and a diagnosis and treatment pathway map, and transmit them to the large model inference engine. The large model inference engine is used to generate personalized management recommendations based on the knowledge graph and the diagnosis and treatment pathway map. When receiving the female tumor-related data of the target user, the female tumor macro model is used to generate personalized management suggestions for the target user.

7. A comprehensive management method for female tumors based on a large female tumor model according to claim 6, characterized in that: The intelligent base engine is also used to: before converting the multimodal data into the structured feature vector, standardize the multimodal data using clinical medical terminology standards, and use a privacy protection mechanism to process the standardized multimodal data, and convert the processed multimodal data into the structured feature vector.

8. A comprehensive management method for female tumors based on a large female tumor model according to claim 7, characterized in that: The large model inference engine is also used to generate electronic medical records based on the knowledge graph and the diagnosis and treatment pathway map, combined with medical knowledge and expert advice related to female tumors. The method also includes: when receiving female tumor-related data of a target user, using the female tumor large model to generate an electronic medical record for the target user.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for comprehensive management of female tumors based on a large female tumor model as claimed in any one of claims 6 to 8 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, it implements the comprehensive management method of female tumors based on a large model of female tumors as described in any one of claims 6 to 8.

Images