An information recording and storage method and system for chronic disease management

Abstract

The present invention relates to the technical field of data storage and processing, and particularly to a method and system for information recording and storage for chronic disease management. The method includes the following steps: obtaining a chronic disease medical record image of a patient; identifying the text area of the chronic disease medical record image of the patient, performing a detection on the shaping degree of chronic disease vocabulary in the text area to obtain the shaping degree of chronic disease vocabulary; performing feature fusion of chronic disease data on the text area based on the shaping degree of chronic disease vocabulary to obtain multi-modal chronic disease feature data; obtaining the medical images of the patient's chronic diseases and performing marking of the patient's physiological monitoring parameters on the multi-modal chronic disease feature data. Through data processing technology, pattern recognition technology, and deep learning technology, the present invention realizes multi-modal chronic disease feature recognition for chronic disease information and performs multi-modal association; realizes dynamic adjustment of the chronic disease data storage architecture for the real-time status data of chronic disease storage, thereby improving the performance of chronic disease information recording and storage.

Description

Technical Field

[0001] The present invention relates to the technical field of data storage and processing, and in particular to a method and system for recording and storing information for chronic disease management. Background Art

[0002] Early chronic disease management mainly relied on a combination of patient self-management and regular doctor management. Since the chronic disease information between patients and doctors could not be communicated in real time, the efficiency of chronic disease management was low. With the application of digital technology in the field of chronic disease management, researchers have constructed a digital chronic disease information collaborative management model, which specifically stores and manages general chronic disease information, and extracts common task feature information of common chronic diseases through analysis and induction of common chronic disease types, thereby constructing a digital chronic disease information collaborative management model. However, chronic disease information is characterized by large quantity and complex types. The existing chronic disease information recording and storage methods make it difficult to identify multimodal chronic disease features of chronic disease information and to correlate multimodal chronic disease feature information. In addition, the chronic disease data storage architecture fails to be dynamically adjusted according to the real-time status data of chronic disease storage, resulting in low performance of chronic disease information recording and storage. Summary of the invention

[0003] Based on this, it is necessary to provide an information recording and storage method and system for chronic disease management to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a method for recording and storing information for chronic disease management is provided, the method comprising the following steps:

[0005] Step S1: Acquire a patient's chronic disease medical record image; identify the text area of ​​the patient's chronic disease medical record image, perform a chronic disease vocabulary formation test on the text area, and obtain the chronic disease vocabulary formation; perform chronic disease data feature fusion on the text area based on the chronic disease vocabulary formation to obtain multimodal chronic disease feature data;

[0006] Step S2: Obtain the patient's chronic disease medical images and mark the multimodal chronic disease feature data with the patient's physiological monitoring parameters; extract the patient's clinical manifestation data from the multimodal chronic disease feature data; perform implicit feature mining based on the patient's physiological monitoring parameters and the patient's clinical manifestation data, and construct a clinical reliability feature network;

[0007] Step S3: Perform structural attribute analysis on the clinical credibility feature network to obtain credibility structure attributes; divide the multimodal chronic disease feature data into chronic disease unstructured data and chronic disease structured data based on the credibility structure attributes; reconstruct the chronic disease unstructured data and chronic disease structured data into chronic disease compressed data; establish a storage layout for the chronic disease dynamic compressed data and generate chronic disease storage layout information; construct a preliminary architecture for chronic disease data storage based on the chronic disease storage layout information;

[0008] Step S4: Input the chronic disease data of the patient into the preliminary architecture of chronic disease data storage and monitor the real-time status data of chronic disease storage; adjust the preliminary architecture of chronic disease data storage according to the real-time status data of chronic disease storage to obtain the chronic disease data storage architecture; record and store the chronic disease data of the patient based on the chronic disease data storage architecture to generate a chronic disease information storage report.

[0009] The present invention performs text region recognition on the chronic disease medical record image of the patient and conducts a detection of the shaping degree of chronic disease vocabulary, and can accurately extract the text information related to chronic diseases. The text region chronic disease data feature fusion based on the shaping degree of chronic disease vocabulary generates multi-modal chronic disease feature data. This process effectively integrates the key information in the medical record image, provides a high-quality data basis for subsequent analysis and processing, and ensures the integrity and accuracy of chronic disease data. Mark the patient's physiological monitoring parameters for the multi-modal chronic disease feature data, and extract the patient's clinical manifestation data. Based on these data, implicit feature mining is performed to construct a clinical reliability feature network. This process not only covers the quantitative data of the patient's physiological monitoring, but also integrates the qualitative information of the clinical manifestations. Through the construction of the clinical reliability feature network, the chronic disease state of the patient can be more comprehensively reflected, providing a scientific basis for subsequent data classification and compression, and improving the depth and breadth of chronic disease data. Analyze the structural attributes of the clinical reliability feature network to obtain the reliability structural attributes. Based on this, the multi-modal chronic disease feature data is divided into chronic disease unstructured data and chronic disease structured data, and further reconstructed into chronic disease compressed data. Through this classification and compression method, the requirement for data storage space is effectively reduced, while the key information is retained, improving the efficiency and manageability of data storage. In addition, a storage layout is established for the chronic disease dynamic compressed data and chronic disease storage layout information is generated, further optimizing the data storage structure and providing clear guidance for constructing the preliminary architecture of chronic disease data storage. Input the chronic disease data of the patient into the preliminary architecture of chronic disease data storage, and monitor the real-time status data of chronic disease storage. Adjust the preliminary architecture of chronic disease data storage according to the real-time status data, and finally obtain an optimized chronic disease data storage architecture. This dynamic adjustment process can flexibly optimize the storage architecture according to the actual data storage situation, ensuring the stability and reliability of data storage. Record and store the chronic disease data of the patient based on the optimized chronic disease data storage architecture to generate a chronic disease information storage report. Therefore, the present invention realizes multi-modal chronic disease feature recognition and multi-modal association of chronic disease information through data processing technology, pattern recognition technology and deep learning technology; realizes dynamic adjustment of the chronic disease data storage architecture for the real-time status data of chronic disease storage, thereby improving the performance of chronic disease information recording and storage.

[0010] Preferably, step S1 includes the following steps:

[0011] Step S11: Obtain the chronic disease medical record image of the patient; perform image color enhancement on the chronic disease medical record image of the patient and remove noise to obtain a standardized medical record image;

[0012] Step S12: Mark the text features of the standardized medical record image, and identify the text area of the standardized medical record image according to the text features;

[0013] Step S13: Perform handwritten font contour recognition on the text area to obtain handwritten font contour features;

[0014] Step S14: Extract the stroke connection points of the handwritten font contour features, and detect the font connection complexity according to the stroke connection points to generate the font connection complexity;

[0015] Step S15: Identify the font aspect ratio of the handwritten font contour features, and detect the shape presentation degree according to the font aspect ratio to generate the font shape presentation degree;

[0016] Step S16: Perform chronic disease vocabulary shaping degree mapping on the font connection complexity and the font shape presentation degree to obtain the chronic disease vocabulary shaping degree;

[0017] Step S17: Optimize the font presentation shape of the handwritten font contour according to the chronic disease vocabulary shaping degree to obtain chronic disease vocabulary optimization data;

[0018] Step S18: Based on the chronic disease vocabulary optimization data, perform chronic disease data feature fusion on the text area to obtain multi-modal chronic disease feature data.

[0019] The present invention performs image color enhancement and noise removal on the chronic disease medical record image of the patient to obtain a standardized medical record image. This process effectively improves the quality and consistency of the image, providing a clear and accurate image basis for subsequent text feature extraction and recognition; by marking the text features of the standardized medical record image and identifying the text area, the text content in the medical record image can be accurately located. This accurate text area marking ensures the pertinence and accuracy of subsequent processing; the recognition of handwritten font contour features, the extraction of stroke connection points, and the detection of font connection complexity and shape presentation degree realize the refined analysis of handwritten fonts. Further, through chronic disease vocabulary shaping degree mapping, the handwritten font contour is optimized to generate chronic disease vocabulary optimization data. This series of operations improves the readability and accuracy of handwritten fonts, providing high-quality data for subsequent feature fusion; based on the chronic disease vocabulary optimization data, chronic disease data feature fusion is performed on the text area to generate multi-modal chronic disease feature data. This process integrates the text information and other relevant features in the medical record image to form comprehensive multi-modal data, providing comprehensive and accurate data support for chronic disease management.

[0020] Preferably, step S2 includes the following steps:

[0021] Step S21: Obtain the chronic disease medical images of the patient;

[0022] Step S22: Locate the diseased part of the patient in the chronic disease medical images of the patient to obtain diseased part data; Extract the patient's physiological monitoring parameters from the multi-modal chronic disease feature data according to the diseased part data, and generate the patient's physiological monitoring parameters;

[0023] Step S23: Mark the patient's chronic disease clinical features in the multi-modal chronic disease feature data to obtain the patient's chronic disease clinical feature marking information; Extract the patient's chronic disease clinical data from the multi-modal chronic disease feature data based on the patient's chronic disease clinical feature marking information to obtain the patient's clinical manifestation data;

[0024] Step S24: Conduct implicit feature mining based on the patient's physiological monitoring parameters and the patient's clinical manifestation data, and construct a clinical reliability feature network.

[0025] The present invention conducts patient chronic disease medical image recognition on multi-modal chronic disease feature data, can obtain patient chronic disease medical image information, and provides basic data for medical image analysis; The application of medical image enhancement technology improves the quality and readability of medical images, making the patient's imaging medical data more accurate; The location of the diseased part of the patient's imaging medical data provides a basis for accurately identifying the diseased area; Combining the extraction of the patient's physiological monitoring parameters and correlating the patient's physiological monitoring parameters with the corresponding imaging data enhances the accuracy and practicality of the imaging-physiological correlation data; The recognition of the patient's chronic disease clinical feature marking provides standardized information for the classification and marking of clinical data; Based on the marking information, the extraction of the patient's clinical manifestation data is more accurate, provides structured data support for clinical data analysis, and enhances the comprehensiveness and depth of the data.

[0026] Preferably, step S24 includes the following steps:

[0027] Step S241: Analyze the patient's physiological monitoring parameters layer by layer, and extract the physiological signal fluctuation pattern, abnormal peak value, and periodic change;

[0028] Step S242: Decompose the patient's clinical manifestation data item by item, and extract the clinical manifestation frequency, intensity, and duration;

[0029] Step S243: Conduct time series analysis on the physiological signal fluctuation pattern, determine the signal frequency distribution characteristics, and non-linearly correlate with the clinical manifestation frequency to obtain frequency implicit correlation data;

[0030] Step S244: Record the peak points of the physiological signal abnormal peak value, detect the duration of the peak signal amplitude, and perform a correlation function correspondence with the clinical manifestation intensity to obtain intensity implicit correlation data;

[0031] Step S245: Extract the cycle length of the periodic change of the physiological signal, identify the change of the cycle amplitude, and perform a time series comparison with the duration of the clinical manifestation to obtain cycle implicit association data;

[0032] Step S246: Construct a clinical reliability feature network based on the frequency implicit association data, intensity implicit association data, and cycle implicit association data.

[0033] The present invention analyzes the patient's physiological monitoring parameters layer by layer, extracts the physiological signal fluctuation pattern, abnormal peak, and periodic change, and can accurately capture the key features of the physiological signal, providing high-precision quantitative data for subsequent correlation analysis; disassembles the patient's clinical manifestation data item by item, and extracts the frequency, intensity, and duration of the clinical manifestation. This detailed data disassembly method can comprehensively reflect the clinical symptom characteristics of the patient; performs a time series analysis on the physiological signal fluctuation pattern, determines the signal frequency distribution characteristics, and performs a non-linear correlation with the clinical manifestation frequency to obtain frequency implicit association data. This correlation analysis can reveal the complex relationship between the physiological signal frequency and the clinical manifestation frequency, providing important frequency dimension data for the construction of the clinical reliability feature network; records the peak points of the abnormal peaks of the physiological signal, detects the duration of the peak signal amplitude, and performs a correlation function correspondence with the clinical manifestation intensity to obtain intensity implicit association data. This process can accurately associate the abnormal peak characteristics of the physiological signal with the intensity characteristics of the clinical manifestation, further enriching the information of the clinical reliability feature network; extracts the cycle length of the periodic change of the physiological signal, identifies the change of the cycle amplitude, and performs a time series comparison with the duration of the clinical manifestation to obtain cycle implicit association data. This comparative analysis can effectively associate the periodic characteristics of the physiological signal with the duration characteristics of the clinical manifestation, providing data support for the cycle dimension for the construction of the clinical reliability feature network; constructs a clinical reliability feature network based on the frequency implicit association data, intensity implicit association data, and cycle implicit association data, which can comprehensively reflect the multi-dimensional association relationship between the patient's physiological monitoring parameters and clinical manifestation data.

[0034] In this specification, an information recording and storage system for chronic disease management is provided, which is used to execute the above-mentioned information recording and storage method for chronic disease management. The information recording and storage system for chronic disease management includes:

[0035] A chronic disease data acquisition module, which is used to obtain the patient's chronic disease medical record image; identify the text area of the patient's chronic disease medical record image, perform a detection on the shaping degree of chronic disease vocabulary in the text area to obtain the shaping degree of chronic disease vocabulary; perform a fusion of chronic disease data features on the text area based on the shaping degree of chronic disease vocabulary to obtain multi-modal chronic disease feature data;

[0036] The chronic disease information association module is used to obtain the medical images of patients with chronic diseases and mark the patient's physiological monitoring parameters for multi-modal chronic disease feature data; extract the clinical manifestation data of patients from the multi-modal chronic disease feature data; perform implicit feature mining based on the patient's physiological monitoring parameters and the patient's clinical manifestation data, and construct a clinical reliability feature network;

[0037] The chronic disease information storage architecture module is used to perform structural attribute analysis on the clinical reliability feature network to obtain the reliability structural attributes; divide the multi-modal chronic disease feature data into chronic disease unstructured data and chronic disease structured data based on the reliability structural attributes; reconstruct the chronic disease unstructured data and the chronic disease structured data into chronic disease compressed data; establish a storage layout for the chronic disease dynamic compressed data to generate chronic disease storage layout information; construct a preliminary architecture for chronic disease data storage based on the chronic disease storage layout information;

[0038] The chronic disease information storage architecture adjustment module is used to input the patient's chronic disease data into the preliminary architecture for chronic disease data storage and monitor the real-time status data of chronic disease storage; adjust the architecture of the preliminary architecture for chronic disease data storage according to the real-time status data of chronic disease storage to obtain the chronic disease data storage architecture; record and store the patient's chronic disease data based on the chronic disease data storage architecture to generate a chronic disease information storage report.

[0039] Through the integration of multiple functional modules, the present invention realizes the full-process automatic processing from chronic disease data collection, information association, storage architecture construction to architecture adjustment, and has significant beneficial effects. The system can efficiently obtain and process the medical record images of patients with chronic diseases, accurately identify the text areas and perform the fusion of chronic disease data features to generate multi-modal chronic disease feature data. By performing implicit feature mining to construct a clinical reliability feature network, further perform structural attribute analysis on the data, realize the classification and compression of the data, and optimize the storage layout. In addition, the system can dynamically monitor the storage status and adjust the architecture in real time to ensure the high efficiency and stability of data storage, and finally generate a detailed chronic disease information storage report, providing comprehensive, accurate and efficient data support for chronic disease management, and improving the scientificity and convenience of chronic disease management. Brief Description of the Drawings

[0040] Figure 1 It is a schematic diagram of the step flow of a method for recording and storing information for chronic disease management;

[0041] Figure 2 For Figure 1 It is a schematic diagram of the detailed implementation step flow of step S2 in

[0042] Figure 3 For Figure 2 It is a schematic diagram of the detailed implementation step flow of step S24 in

[0043] The realization, functional features and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the accompanying drawings. Specific embodiments

[0044] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative work belong to the scope of protection of the present invention.

[0045] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0046] It should be understood that although terms such as "first" and "second" may be used here to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit can be called the second unit, and similarly the second unit can be called the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed related items.

[0047] To achieve the above object, please refer to Figures 1 to 3 , an information recording and storage method for chronic disease management, the method comprising the following steps:

[0048] Step S1: Obtain the chronic disease medical record image of the patient; identify the text area of the chronic disease medical record image of the patient, perform a detection on the shaping degree of chronic disease vocabulary on the text area to obtain the shaping degree of chronic disease vocabulary; based on the shaping degree of chronic disease vocabulary, perform a fusion of chronic disease data features on the text area to obtain multi-modal chronic disease feature data;

[0049] Step S2: Obtain the chronic disease medical image of the patient and mark the patient's physiological monitoring parameters; extract the clinical manifestation data of the patient of the multi-modal chronic disease feature data; perform implicit feature mining based on the patient's physiological monitoring parameters and the patient's clinical manifestation data, and construct a clinical reliability feature network;

[0050] Step S3: Conduct a structural attribute analysis on the clinical reliability feature network to obtain the reliability structural attributes; divide the multi-modal chronic disease feature data into chronic disease unstructured data and chronic disease structured data based on the reliability structural attributes; reconstruct the chronic disease unstructured data and chronic disease structured data into chronic disease compressed data; establish a storage layout for the chronic disease dynamic compressed data to generate chronic disease storage layout information; construct a preliminary architecture for chronic disease data storage based on the chronic disease storage layout information;

[0051] Step S4: Input the patient's chronic disease data into the preliminary architecture for chronic disease data storage and monitor the real-time status data of chronic disease storage; adjust the architecture of the preliminary architecture for chronic disease data storage according to the real-time status data of chronic disease storage to obtain the chronic disease data storage architecture; record and store the patient's chronic disease data based on the chronic disease data storage architecture to generate a chronic disease information storage report.

[0052] The present invention performs text area recognition on the chronic disease medical record images of patients and conducts detection on the shaping degree of chronic disease vocabulary, capable of accurately extracting text information related to chronic diseases. Through the integration of text area chronic disease data features based on the shaping degree of chronic disease vocabulary, multi-modal chronic disease feature data is generated. This process effectively integrates the key information in the medical record images, providing a high-quality data foundation for subsequent analysis and processing, and ensuring the integrity and accuracy of chronic disease data. Mark the physiological monitoring parameters of patients for the multi-modal chronic disease feature data, and extract the clinical manifestation data of patients. Based on these data, implicit feature mining is carried out to construct a clinical reliability feature network. This process not only covers the quantitative data of patients' physiological monitoring but also integrates the qualitative information of clinical manifestations. Through the construction of the clinical reliability feature network, the chronic disease status of patients can be more comprehensively reflected, providing a scientific basis for subsequent data classification and compression, and enhancing the depth and breadth of chronic disease data. Conduct structural attribute analysis on the clinical reliability feature network to obtain the reliability structural attributes. Based on this, the multi-modal chronic disease feature data is divided into chronic disease unstructured data and chronic disease structured data, and further reconstructed into chronic disease compressed data. Through this classification and compression method, the demand for data storage space is effectively reduced, while the key information is retained, improving the efficiency and manageability of data storage. In addition, a storage layout is established for the dynamically compressed chronic disease data and chronic disease storage layout information is generated, further optimizing the data storage structure and providing clear guidance for constructing the preliminary architecture of chronic disease data storage. Input the chronic disease data of patients into the preliminary architecture of chronic disease data storage, and monitor the real-time status data of chronic disease storage. According to the real-time status data, adjust the architecture of the preliminary architecture of chronic disease data storage, and finally obtain an optimized chronic disease data storage architecture. This dynamic adjustment process can flexibly optimize the storage architecture according to the actual data storage situation, ensuring the stability and reliability of data storage. Record and store the chronic disease data of patients based on the optimized chronic disease data storage architecture to generate a chronic disease information storage report. Therefore, the present invention realizes multi-modal chronic disease feature recognition and multi-modal association of chronic disease information through data processing technology, pattern recognition technology, and deep learning technology; realizes dynamic adjustment of the chronic disease data storage architecture for the real-time status data of chronic disease storage, thereby improving the performance of chronic disease information recording and storage.

[0053] In an embodiment of the present invention, with reference to Figure 1 as shown, it is a schematic flowchart of the steps of a method for recording and storing information for chronic disease management according to the present invention. In this example, the method for recording and storing information for chronic disease management includes the following steps:

[0054] Step S1: Obtain the chronic disease medical record images of patients; recognize the text areas of the chronic disease medical record images of patients, conduct detection on the shaping degree of chronic disease vocabulary for the text areas to obtain the shaping degree of chronic disease vocabulary; perform chronic disease data feature fusion on the text areas based on the shaping degree of chronic disease vocabulary to obtain multi-modal chronic disease feature data;

[0055] In the embodiments of the present invention, the optical character recognition (OCR) technology is used to process the chronic disease medical record images of patients and identify the text regions therein. The OCR technology converts the text content in the medical record images into an editable text format through image scanning and text recognition algorithms. Subsequently, a text analysis method based on deep learning is used to detect the shaping degree of chronic disease vocabulary in the identified text regions. Specifically, a pre-trained BERT model is used to tokenize and embed the text, and the shaping degree of each vocabulary is evaluated by calculating the semantic similarity between each vocabulary and chronic disease-related vocabulary. The cosine similarity formula is used to calculate the semantic similarity, that is, the dot product of two vectors divided by the product of their norms. After obtaining the shaping degree of chronic disease vocabulary, the chronic disease data features of the text regions are fused based on this shaping degree. By constructing a multi-modal feature fusion framework, the text features are fused with other modal data (such as medical image features) in the medical record images. The medical image features are extracted using a convolutional neural network (CNN) to extract the spatial features in the images. Then, a bidirectional multi-modal attention mechanism is used to weight-fuse the text features and the image features, and the weights are dynamically adjusted according to the shaping degree of chronic disease vocabulary, and finally multi-modal chronic disease feature data are obtained.

[0056] Step S2: Obtain the chronic disease medical images of the patient and mark the patient's physiological monitoring parameters for the multi-modal chronic disease feature data; extract the clinical manifestation data of the patient from the multi-modal chronic disease feature data; perform implicit feature mining based on the patient's physiological monitoring parameters and the patient's clinical manifestation data, and construct a clinical reliability feature network;

[0057] In the embodiments of the present invention, through the PACS (Picture Archiving and Communication System) system of the hospital, chronic disease medical image data of patients is obtained from medical imaging devices (such as CT, MRI, X-ray, etc.). The image data is stored in DICOM format and contains the basic information of the patient (such as name, age, gender) and the metadata of the image (such as imaging time, device type, imaging parameters, etc.); the resolution of the image data is usually 512×512 pixels, the bit depth is 16 bits, and the storage format is lossless compressed JPEG2000 format; wearable physiological parameter monitoring devices (such as smart bracelets, ambulatory blood pressure monitors, continuous glucose monitors, etc.) are used to collect the physiological monitoring parameters of patients, including heart rate, blood pressure, blood glucose, blood oxygen saturation, etc. These devices transmit the data to the local server or cloud database through Bluetooth or Wi-Fi, and the data is stored in time series format with a time stamp accuracy of the second level; the obtained medical image data is associated with the physiological monitoring parameters. The image data and the physiological monitoring data are matched through the unique identifier of the patient (such as the medical record number). In the multi-modal chronic disease feature data, corresponding physiological monitoring parameter marks are added to each image data. For example, for a CT image at a certain time point, the average heart rate, blood pressure, and blood glucose values within 1 hour before and after that time point are marked. The marking process is completed through an automated script to ensure the accuracy and consistency of the data. The text part in the multi-modal chronic disease feature data is processed to extract the clinical manifestation data of the patient. Using natural language processing (NLP) technology, preprocessing operations such as word segmentation, part-of-speech tagging, and syntactic analysis are performed on the text data to extract keywords and semantic information related to the clinical manifestations. The medical image data is preprocessed, including operations such as denoising, contrast enhancement, and segmentation, to improve the quality of the image data. Then, features in the image are extracted through image processing technology, and these features are closely related to the clinical manifestations of the patient. The extracted physiological monitoring parameters and clinical manifestation data are fused, and an association rule mining algorithm (such as the Apriori algorithm) is used, with a minimum support of 0.3 and a minimum confidence of 0.7, to mine the potential association rules between the two. Through these rules, implicit features of great significance for chronic disease management are identified. The mined implicit features are used to construct a clinical confidence feature network. Using graph neural network technology, the implicit features are used as nodes, and the association relationships between the features are used as edges to construct a feature network. The feature vectors of the nodes are encoded through a multi-layer perceptron (MLP), and the weights of the edges are dynamically adjusted according to the confidence of the association rules. In this way, a network structure that can reflect the clinical feature confidence and relevance of the patient is formed.

[0058] Step S3: Perform a structural attribute analysis on the clinical reliability feature network to obtain the reliability structural attributes; divide the multi-modal chronic disease feature data into chronic disease unstructured data and chronic disease structured data based on the reliability structural attributes; reconstruct the chronic disease unstructured data and chronic disease structured data into chronic disease compressed data; establish a storage layout for the chronic disease dynamic compressed data to generate chronic disease storage layout information; construct a preliminary architecture for chronic disease data storage based on the chronic disease storage layout information;

[0059] In the embodiments of the present invention, a graph theory analysis method is used to perform a structural attribute analysis on the clinical reliability feature network. Calculate the global attributes of the network, including global efficiency (Eg), local efficiency (Eloc), clustering coefficient (Cp), characteristic path length (Lp), normalized clustering coefficient (Gamma), normalized characteristic path length (Lambda), and small-world attribute (Sigma). At the same time, calculate the node attributes, such as node betweenness (Bc), node degree (Dc), and node efficiency (Ne). Analyze the overall topological structure of the network and the importance of nodes through these parameters; according to the results of the structural attribute analysis, divide the multi-modal chronic disease feature data into chronic disease unstructured data and chronic disease structured data. Chronic disease structured data includes data with clear numerical values and formats, such as physiological monitoring parameters (heart rate, blood pressure, blood sugar, etc.) and imaging features (lesion area size, texture, etc.). Chronic disease unstructured data includes text data (medical record descriptions, imaging reports, etc.) and imaging data that is difficult to directly quantify; use data compression technology to reconstruct the chronic disease unstructured data and chronic disease structured data. For structured data, use dimensionality reduction methods such as principal component analysis (PCA) to extract key features and reduce the data dimension. For unstructured data, use text summarization technology to extract the core information of the text and perform lossy compression on the imaging data to retain the details of the key regions. In this way, the multi-modal data is reconstructed into chronic disease compressed data to reduce the storage space and improve the data processing efficiency. According to the characteristics of the chronic disease compressed data, establish a storage layout. Adopt a hierarchical storage strategy, store the frequently accessed data (such as recent physiological monitoring parameters) in a fast-access storage medium (such as SSD), and store the infrequently accessed data (such as historical imaging data) in a storage medium with a larger capacity (such as HDD). At the same time, allocate independent storage areas for each data type and establish an index for quick retrieval. The generated chronic disease storage layout information includes data storage locations, storage medium types, and index structures. According to the chronic disease storage layout information, construct a preliminary architecture for chronic disease data storage. This architecture includes a data storage layer, an index management layer, and a data access layer. The data storage layer is responsible for storing the compressed chronic disease data; the index management layer is responsible for managing and maintaining the data index for quick location and retrieval of data; the data access layer provides a data access interface to support data read and write operations. Through this hierarchical architecture, efficient storage and management of chronic disease data are achieved.

[0060] Step S4: Input the patient's chronic disease data into the preliminary architecture of chronic disease data storage and monitor the real-time status data of chronic disease storage; adjust the preliminary architecture of chronic disease data storage according to the real-time status data of chronic disease storage to obtain the chronic disease data storage architecture; record and store the patient's chronic disease data based on the chronic disease data storage architecture to generate a chronic disease information storage report.

[0061] In the embodiments of the present invention, the chronic disease data of patients is stored in two forms: structured and unstructured. Structured data (such as physiological monitoring parameters, examination results, etc.) is stored in a relational database in tabular form, and unstructured data (such as imaging data, medical record texts, etc.) is stored in a distributed file system in file form. The data input process is implemented through a data import tool, which supports batch import and real-time data stream transmission. For structured data, SQL statements are used for batch insertion operations; for unstructured data, the File Transfer Protocol (FTP) or the API of a distributed file system (such as HDFS) is used for data upload. The configuration parameters of the data import tool include: the transaction size of batch insertion (such as 1000 records per time), the buffer size of file transfer (such as 1MB), and the number of concurrent threads for data import (such as 5 threads). A storage monitoring system is deployed, and an open-source monitoring tool Prometheus combined with Grafana is used for real-time status monitoring. Prometheus collects the performance metrics of the storage system, including disk utilization rate, I / O latency, network bandwidth, etc. The configuration parameters of the monitoring system include: data collection interval (such as 10 seconds), alarm threshold (such as an alarm is triggered when the disk utilization rate exceeds 80%), and data retention period (such as 30 days). The monitoring system interacts with the storage system through an API interface to obtain real-time status data and stores the data in a time series database. When the monitoring system detects that the utilization rate of the storage medium exceeds the preset threshold (such as 80%), a storage capacity expansion operation is automatically triggered. For a distributed file system, the capacity is expanded by adding storage nodes; for a relational database, the storage space is expanded by adding storage volumes or adjusting the table partitioning strategy. The storage capacity expansion operation includes: configuring parameters such as the IP address of a new node and the node storage capacity in the distributed file system; adjusting the size and number of table partitions in the relational database. According to the analysis results of data access frequency, frequently accessed data is migrated to high-performance storage media (such as SSDs), and infrequently accessed data is migrated to low-performance storage media (such as HDDs). The data migration operation is implemented through a storage management tool, and the tool supports hot migration of data to ensure that data access is not affected during the data migration process. The configuration parameters of the data migration tool include: the size of the migrated data block (such as 64KB), the priority of the migration operation (such as low priority is executed at night), and the number of concurrent migrations (such as 3 concurrent migration tasks). For a distributed storage system, the availability and fault tolerance of data are improved by adjusting the number of replicas and the distribution of storage nodes. The number of replicas is dynamically adjusted according to the importance and access frequency of the data, and the distribution of storage nodes is optimized according to the network topology and data access patterns. The configuration parameters of the distributed storage system include: the number of replicas (such as 3 replicas), the data block distribution strategy (such as random distribution or distribution according to access frequency), and the reserved network bandwidth (such as reserving 20% of the bandwidth for data synchronization). The chronic disease data of patients is stored according to the optimized storage architecture.Structured data is stored in a relational database, and the table structure includes information such as field names, data types, indexes, etc.; unstructured data is stored in a distributed file system, and file metadata includes information such as file names, file sizes, storage paths, etc. The data storage process is implemented through automated scripts, and the scripts support data verification and integrity checks. For structured data, verify the non-emptiness and data types of fields; for unstructured data, verify the integrity of the file (such as through MD5 verification). The data storage script configuration parameters include: data verification rules (such as lists of non-empty fields, data type lists), integrity verification algorithms (such as MD5), and the number of retries for data storage (such as 3 times). The chronic disease information storage report is generated through a report generation tool, and the report content includes information such as data storage location, storage time, data type, data integrity verification results, etc. The report generation tool supports customizing report templates, and the templates include parts such as report titles, data storage summaries, and data integrity verification results.

[0062] Preferably, step S1 includes the following steps:

[0063] Step S11: Obtain the chronic disease medical record images of the patient; perform image color enhancement on the chronic disease medical record images of the patient, and remove noise to obtain standardized medical record images;

[0064] Step S12: Mark the text features of the standardized medical record images, and identify the text areas of the standardized medical record images according to the text features;

[0065] Step S13: Perform handwritten font contour recognition on the text areas to obtain handwritten font contour features;

[0066] Step S14: Extract the stroke connection points of the handwritten font contour features, and detect the font connection complexity according to the stroke connection points to generate the font connection complexity;

[0067] Step S18: Map the font connection complexity and the font shape presentation degree to the chronic disease vocabulary shaping degree to obtain the chronic disease vocabulary shaping degree;

[0068] Step S16: Map the font connection complexity and the font shape presentation degree to the chronic disease vocabulary shaping degree to obtain the chronic disease vocabulary shaping degree;

[0069] Step S17: Optimize the font presentation shape of the handwritten font contour according to the chronic disease vocabulary shaping degree to obtain chronic disease vocabulary optimization data;

[0070] Step S18: Based on the chronic disease vocabulary optimization data, perform chronic disease data feature fusion on the text areas to obtain multi-modal chronic disease feature data.

[0071] In the embodiments of the present invention, through the medical record image acquisition system of the hospital, paper medical records are converted into digital images by a high-resolution scanner. The resolution of the scanner is set to 300 dpi to ensure the clarity and details of the images. The scanned images are stored in TIFF format. Image processing software is used to enhance the color of the medical record images. By adjusting the contrast and brightness of the images, the text in the images becomes clearer. The specific parameter settings are: the contrast enhancement factor is 1.5, and the brightness adjustment value is 20. At the same time, the median filtering algorithm is adopted to remove the noise in the images, and the filter window size is set to 3×3 pixels. The processed images are the standardized medical record images. Optical Character Recognition (OCR) technology is used to mark the text features of the standardized medical record images. The OCR system identifies the outlines and positions of the text by analyzing the pixel distribution in the images. By setting the confidence threshold of text recognition to 0.8, the accuracy of recognition is ensured. The recognized text areas are marked on the images in the form of coordinate frames. For the handwritten fonts in the text areas, a contour recognition algorithm based on deep learning is used. This algorithm extracts the contour features of the handwritten fonts through a Convolutional Neural Network (CNN). The input of the network is the image patch of the text area, and the output is the coordinate points of the contour of the handwritten font. By setting the number of convolutional layers of the network to 5 layers, the size of the convolutional kernel of each layer to 3×3, and the stride to 1, the contour features of the handwritten fonts can be effectively extracted. The stroke connection points are extracted from the handwritten font contour features. By analyzing the geometric relationship of the contour points, the starting points and ending points of the strokes are identified. For each stroke, its connection complexity is calculated, that is, the number of intersections and the degree of curvature between the strokes. The calculation formula of the connection complexity is: complexity = α×number of intersections + β×degree of curvature, where α and β are weight coefficients, which are set to 0.6 and 0.4 respectively. The aspect ratio of the handwritten font contour is calculated, that is, the ratio of the height to the width of the font. By setting the threshold range of the aspect ratio to 0.5 to 2.0, the font shapes that meet the standards are identified. For each font, its shape representation degree is calculated, that is, the similarity between the font contour and the standard font shape. The calculation formula of the shape representation degree is: representation degree = 1 - |actual aspect ratio - standard aspect ratio| / standard aspect ratio. The font connection complexity and shape representation degree are comprehensively evaluated to generate the shaping degree of chronic disease vocabulary. By setting the weight coefficients, the connection complexity and shape representation degree are multiplied by the weights and then added together to obtain the final shaping degree value. The weight coefficients are 0.4 and 0.6 respectively, and the calculation formula of the shaping degree is: shaping degree = 0.4×connection complexity + 0.6×shape representation degree. According to the shaping degree of chronic disease vocabulary, the handwritten font contour is optimized. For the fonts with a lower shaping degree, by adjusting the connection points and shapes of the strokes, they are made closer to the standard fonts. The optimized font contour data is used as the optimized data of chronic disease vocabulary for subsequent multi-modal data fusion. The optimized chronic disease vocabulary data is fused with other features in the medical record images (such as imaging features, physiological monitoring parameters, etc.).Through multi-modal data fusion technology, data of different modalities are uniformly represented to generate multi-modal chronic disease feature data. During the fusion process, a weighted average method is adopted, and corresponding weights are set according to the importance and relevance of each modality data to complete the data fusion.

[0072] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0073] Step S21: Obtain the medical images of the patient's chronic diseases;

[0074] Step S22: Locate the patient's affected part in the medical images of the patient's chronic diseases to obtain the data of the affected part; extract the patient's physiological monitoring parameters from the multi-modal chronic disease feature data according to the data of the affected part to generate the patient's physiological monitoring parameters;

[0075] Step S23: Mark the clinical features of the patient's chronic diseases in the multi-modal chronic disease feature data to obtain the clinical feature marking information of the patient's chronic diseases; extract the clinical data of the patient's chronic diseases from the multi-modal chronic disease feature data based on the clinical feature marking information of the patient's chronic diseases to obtain the clinical manifestation data of the patient;

[0076] Step S24: Perform implicit feature mining based on the patient's physiological monitoring parameters and the patient's clinical manifestation data, and construct a clinical reliability feature network.

[0077] In the embodiments of the present invention, through the PACS (Picture Archiving and Communication System) system of the hospital, the chronic disease medical image data of the patient is obtained from medical imaging devices (such as CT, MRI, X-ray, etc.). The image data is stored in DICOM format and contains the patient's basic information (such as name, age, gender) and the metadata of the image (such as imaging time, device type, imaging parameters, etc.). The resolution of the image data is usually 512×512 pixels, the bit depth is 16 bits, and the storage format is lossless compressed JPEG2000 format. Using medical image processing technology, the diseased part of the obtained medical image is located. An image segmentation algorithm (such as threshold-based segmentation or deep learning segmentation model) is used to process the image to identify the contour and position of the diseased part. The positioning result is stored in the form of a coordinate box. According to the data of the diseased part, the physiological monitoring parameters of the patient are extracted from the multi-modal chronic disease characteristic data. Wearable devices (such as smart bracelets, ambulatory blood pressure monitors, continuous glucose monitors, etc.) are used to collect the physiological monitoring parameters of the patient, including heart rate, blood pressure, blood glucose, blood oxygen saturation, etc. These devices collect data through sensors (such as PPG, ECG, IMU, etc.), and their technical parameters include signal frequency, acquisition frequency, signal amplitude, and resolution, etc. For example, the acquisition frequency of the PPG sensor is 5 - 500 Hz, the signal amplitude is -10 - 10 mV, and the resolution is 16 bits. The collected physiological monitoring parameters are stored in time series format, and the time stamp accuracy is at the second level. Using medical image processing technology, the patient's chronic disease clinical characteristics are marked on the multi-modal chronic disease characteristic data. Through an image recognition algorithm (such as the convolutional neural network CNN), the lesion characteristics in the image data are marked, and at the same time, natural language processing technology (NLP) is combined to mark the clinical manifestations in the medical record text. The marked information includes the type, location, size, etc. of the lesion, as well as the symptom description and disease name in the medical record text. According to the marked information, the clinical manifestation data of the patient are extracted from the multi-modal chronic disease characteristic data. The extracted data includes the lesion characteristics in the image, the symptom description in the medical record text, the examination results, etc. These data are integrated into structured clinical manifestation data. The extracted physiological monitoring parameters and clinical manifestation data are fused, and an association rule mining algorithm (such as the Apriori algorithm) is used, with the minimum support set to 0.3 and the minimum confidence set to 0.7, to mine the potential association rules between the two. Through these rules, the implicit features of great significance for chronic disease management are identified. The mined implicit features are used to construct a clinical confidence feature network. Using graph neural network technology, the implicit features are used as nodes, and the association relationships between the features are used as edges to construct a feature network. The feature vectors of the nodes are encoded through a multi-layer perceptron (MLP), and the weights of the edges are dynamically adjusted according to the confidence of the association rules. In this way, a network structure that can reflect the clinical feature confidence and relevance of the patient is formed.

[0078] As an example of the present invention, refer to Figure 3 As shown, in this example, step S24 includes:

[0079] Step S241: Layer-by-layer analyze the patient's physiological monitoring parameters, and extract the physiological signal fluctuation pattern, abnormal peaks, and periodic changes;

[0080] Step S242: Item-by-item disassemble the patient's clinical manifestation data, and extract the clinical manifestation frequency, intensity, and duration;

[0081] Step S243: Perform a time series analysis on the physiological signal fluctuation pattern, determine the signal frequency distribution characteristics, and non-linearly correlate with the clinical manifestation frequency to obtain frequency implicit correlation data;

[0082] Step S244: Record the peak points of the physiological signal abnormal peaks, detect the duration of the peak signal amplitude, and correspond to the clinical manifestation intensity with a correlation function to obtain intensity implicit correlation data;

[0083] Step S245: Extract the cycle length of the physiological signal periodic change, identify the cycle amplitude change, and compare it with the clinical manifestation duration in a time series to obtain cycle implicit correlation data;

[0084] Step S246: Construct a clinical reliability feature network based on the frequency implicit correlation data, intensity implicit correlation data, and cycle implicit correlation data.

[0085] In the embodiments of the present invention, time-frequency analysis is performed on physiological monitoring parameters (such as heart rate, blood pressure, etc.) using the wavelet transform method. By selecting an appropriate wavelet basis function, the signal is decomposed into different frequency components, and the fluctuation pattern of the signal is extracted. For example, for the heart rate signal, the wavelet transform can be used to extract its change characteristics at different time scales; the empirical mode decomposition (EMD) method is used to decompose the physiological signal. The signal is decomposed into several intrinsic mode functions (IMFs), and by analyzing the amplitude and frequency characteristics of the IMFs, abnormal peaks are identified. For example, in the blood pressure signal, the IMFs after EMD decomposition can clearly show the abnormal fluctuations of blood pressure; the singular spectrum analysis (SSA) method is used to perform periodic analysis on the physiological signal. By constructing a trajectory matrix and performing singular value decomposition, the periodic components of the signal are extracted. For example, for the respiratory signal, SSA can effectively identify the length and amplitude changes of the respiratory cycle; a medical entity data extraction framework based on a large language model (LLM) (such as GAMedX) is used to process the clinical manifestation data. Through named entity recognition (NER) technology, entities such as symptoms and disease names are extracted from the medical record text, and the frequency of their occurrence is counted. For example, the number of occurrences of the symptom "dizziness" is extracted from the medical record, and its frequency is calculated; combined with text mining technology, the intensity of the clinical manifestation is quantified. For example, by analyzing the description of the symptom in the medical record (such as "mild dizziness", "severe dizziness"), the symptom intensity is classified and extracted using a predefined intensity level; regular expressions and time series analysis techniques are used to extract the duration of the symptom from the medical record text. For example, by identifying descriptions such as "lasting for 2 weeks", "lasting for 3 days", the duration of the clinical manifestation is extracted in a standardized time format; time series analysis is performed on the extracted physiological signal fluctuation pattern to calculate the frequency distribution characteristics of the signal. For example, the frequency components of the heart rate signal are analyzed by the fast Fourier transform (FFT) to determine its main frequency distribution; a non-linear regression model (such as polynomial regression) is used to perform correlation analysis between the frequency distribution characteristics of the physiological signal and the clinical manifestation frequency. By calculating the correlation between the two, frequency implicit association data is obtained. The occurrence time point and amplitude of the abnormal peak in the physiological signal are recorded. For example, for the abnormal peak in the blood pressure signal, the time stamp of the peak point and the peak amplitude are recorded; through time series analysis, the amplitude duration of the abnormal peak signal is calculated. For example, the time length from the appearance of the blood pressure abnormal peak to its return to normal is detected; the correlation analysis method (such as the Pearson correlation coefficient) is used to perform corresponding analysis between the peak signal amplitude duration and the clinical manifestation intensity to obtain intensity implicit association data. The cycle length of the physiological signal is extracted by singular spectrum analysis (SSA). For example, for the respiratory signal, the length of its respiratory cycle is extracted; the amplitude change of the physiological signal cycle is analyzed to identify the amplitude difference in different cycles; time series comparison analysis is performed between the cycle length and amplitude change of the physiological signal and the duration of the clinical manifestation.By calculating the time series similarity between the two, periodic implicit association data is obtained. Using the frequency implicit association data, intensity implicit association data, and periodic implicit association data as node features, a clinical reliability feature network is constructed. Utilizing graph neural network (GNN) technology, the association relationships between nodes are used as edges to construct a graph structure; by adjusting the node weights and edge weights, the structure of the clinical reliability feature network is optimized. The node weights are adjusted according to the confidence of the implicit association data, and the edge weights are dynamically adjusted according to the association strength.

[0086] The structural attribute analysis of the clinical reliability feature network described in step S3 includes:

[0087] Assign a unique identifier to each node in the clinical reliability feature network to form a node index sequence;

[0088] Determine the connection feature parameters according to the node index sequence, and construct a connection feature matrix based on the connection feature parameters;

[0089] Perform a block processing on the connection feature matrix, divide the connection feature matrix into several sub-regions, and each sub-region contains a certain number of nodes and connections;

[0090] Identify the local connection patterns and local node distribution positions within each sub-region;

[0091] Map the topological relationships of the local connection patterns and local node distribution positions to generate a network topological structure;

[0092] Evaluate the node weights of the network topological structure, detect the proportion of nodes in the network, and obtain node weight distribution data;

[0093] Detect the node association quantity of the network topological structure, determine the node attribute features according to the node association quantity, and set the initial node weights based on the node attribute features;

[0094] Continuously monitor the dynamic interaction behavior of the nodes, and update the initial node weights in real time according to the dynamic interaction behavior to obtain node weight distribution data;

[0095] Evaluate the connection strength of the network topological structure according to the node weight distribution data to obtain node connection strength data;

[0096] Calculate the initial connection strength value of the nodes in the network topological structure based on the node weight distribution data;

[0097] Continuously monitor the dynamic connection situation of the nodes, and update the connection strength of the nodes in real time according to the dynamic connection situation to obtain node connection strength data.

[0098] In the embodiments of the present invention, in the clinical reliability feature network, a unique identifier is assigned to each node to form a node index sequence. The specific operation is as follows: Combine the feature information of each node, such as node name, type, data source, etc., into a string, and then process this string through a hash algorithm (such as SHA-256) to generate a unique hash value as the node identifier. Arrange the hash values of all nodes in ascending order to form a complete node index sequence. According to the node index sequence, determine the connection feature parameters. The connection feature parameters include the connection strength, connection direction, connection type, etc. between nodes. The connection strength can be calculated through the data correlation between nodes. For example, using the Pearson correlation coefficient method, perform a correlation analysis on the data values of two nodes to obtain a value between -1 and 1 as the connection strength. The connection direction is determined according to the data flow direction, and the connection type is divided according to the nature of the data (such as causal relationship, association relationship, etc.). Based on these connection feature parameters, construct a connection feature matrix. The rows and columns of the matrix respectively correspond to the nodes in the node index sequence, and the elements in the matrix represent the connection feature parameters between nodes. Perform a block processing on the connection feature matrix, and divide the matrix into several sub-regions. The specific operation is as follows: According to the type and function of the nodes, divide the nodes into different categories, such as clinical data nodes, patient information nodes, treatment plan nodes, etc. Divide the rows and columns of the matrix according to the categories, so that each sub-region contains nodes of the same category and their connections. For example, divide the clinical data nodes into one sub-region, and the patient information nodes into another sub-region. Each sub-region contains a certain number of nodes and connections, and the number of nodes is determined according to actual needs. For example, each sub-region contains 10 to 50 nodes. In each sub-region, identify the local connection pattern and the node distribution position. The local connection pattern can be identified through a clustering algorithm (such as K-Means). Divide the nodes into different clusters according to the connection feature parameters, and each cluster represents a local connection pattern. The node distribution position can be determined through a spatial analysis method. For example, use multidimensional scaling analysis (MDS) to map the connection features of the nodes into a two-dimensional or three-dimensional space to obtain the spatial distribution position of the nodes. For example, in the sub-region of clinical data nodes, divide the nodes into 3 clusters through the K-Means algorithm, and each cluster represents a local connection pattern. For example, cluster 1 represents a strong association between blood glucose monitoring nodes, cluster 2 represents a weak association between blood pressure monitoring nodes, and cluster 3 represents a medium association between blood glucose and blood pressure monitoring nodes. Through MDS analysis, map the nodes into a two-dimensional space to obtain the node distribution position map. Map the topological relationship between the local connection pattern and the node distribution position to generate a network topology structure. The specific operation is as follows: According to the local connection pattern and the node distribution position, construct a connection relationship graph between nodes. The connection relationship graph is represented by an adjacency list or adjacency matrix in graph theory. Each node corresponds to a vertex, and each connection corresponds to an edge. The weight of the edge is determined according to the connection feature parameters, such as the connection strength.For example, according to the local connection pattern and the node distribution positions, a directed graph is generated. There is a directed edge between node A and node B in the graph, and the weight of the edge is 0.8, indicating that the connection strength from A to B is 0.8. The node weights of the network topology are evaluated to detect the proportion degree of nodes in the network. The specific operation is as follows: The node centrality analysis method (such as degree centrality, closeness centrality, betweenness centrality) is used to calculate the weights of the nodes. Degree centrality represents the number of connections of a node, closeness centrality represents the average distance from a node to other nodes, and betweenness centrality represents the frequency of a node acting as a mediator in the network. For example, for node A, its degree centrality is calculated to be 5, closeness centrality is 0.3, and betweenness centrality is 0.2. These centrality indicators are weighted and summed to obtain the comprehensive weight of the node. For example, the comprehensive weight calculation method is: degree centrality accounts for 40%, closeness centrality accounts for 30%, and betweenness centrality accounts for 30%, and the comprehensive weight of node A is obtained as 2.5. The node association quantity of the network topology is detected, and the node attribute characteristics are determined according to the node association quantity, and the initial weights of the nodes are set. The specific operation is as follows: The association degree between nodes is calculated. For example, the Jaccard similarity method is used to compare the attribute sets of two nodes, and the ratio of their intersection to union is calculated as the association degree. The attribute characteristics of the nodes are determined according to the association degree. For example, a high association degree indicates that the nodes have similar attribute characteristics. The initial weights of the nodes are set based on the node attribute characteristics. For example, for nodes with a high association degree, the initial weight is set to a higher value, such as 0.8; for nodes with a lower association degree, the initial weight is set to a lower value, such as 0.2. The dynamic interaction behaviors of the nodes are continuously monitored, and the initial weights of the nodes are updated in real time according to the dynamic interaction behaviors. The specific operation is as follows: The data stream monitoring technology (such as the sliding window algorithm) is used to monitor the data interaction frequency and data volume between nodes in real time. For example, the sliding window size is set to 10 minutes, and the data interaction situation between nodes is counted every 10 minutes. The weights of the nodes are adjusted according to the changes in the data interaction frequency and data volume. For example, if the data interaction frequency between node A and node B increases by 50% within 10 minutes, the weight of node A is increased by 0.1, and the weight of node B is increased by 0.1. The connection strength of the network topology is evaluated according to the node weight distribution data to obtain the node connection strength data. The initial connection strength value of the nodes in the network topology is calculated based on the node weight distribution data. The specific operation is as follows: According to the initial weights of the nodes and the connection characteristic parameters, the initial connection strength value is calculated. For example, for the connection between node A and node B, the initial weights of node A and node B are multiplied, and then multiplied by the connection characteristic parameters to obtain the initial connection strength value. The dynamic connection situation of the nodes is continuously monitored, and the connection strength of the nodes is updated in real time according to the dynamic connection situation. The specific operation is as follows: The dynamic graph update algorithm (such as the incremental update algorithm) is used to monitor the connection changes between nodes in real time.

[0099] The structural attribute analysis of the clinical reliability feature network described in step S3 further includes:

[0100] Calculating the topological path strength value for the network topology based on the node connection strength data; comparing the topological path strength value with a preset topological path strength threshold, and classifying the topological paths higher than the threshold as strong paths and the rest as weak paths to obtain path division data;

[0101] Determining the core area and the global area of the network according to the path division data. The core area consists of strong paths and the nodes they connect, and the global area contains all other nodes and weak paths in the network;

[0102] Recalculating the node attribution for the core area and the global area, excluding the node attribution quantity of the global area, and retaining the node attribution quantity of the core area;

[0103] Calculating the proportion of the node attribution quantity of the core area in the network topology, including the proportion of the number of nodes, the proportion of the number of connections, and the proportion of data traffic, to generate network core proportion data;

[0104] Sampling the node time series of the clinical reliability feature network based on the network core proportion data, and identifying the sequence point intensity period of the node time series to generate node stability features;

[0105] Statistically analyzing the connection strength values of the clinical reliability feature network according to the node time series, and determining the connection strength fluctuation range to generate connection reliability features;

[0106] Merging the node stability features and the connection reliability features for core attribute to obtain reliability structural attributes.

[0107] In the embodiments of the present invention, the topological path strength value is calculated for the network topology structure according to the node connection strength data. The specific operation is as follows: First, all possible paths are extracted from the network topology structure. A path refers to a sequence starting from a node, passing through several connections, and reaching another node. For each path, its path strength value is calculated by multiplying the connection strength values of all the connections in the path. For example, if path P contains connections C1, C2, and C3, and their connection strength values are 0.8, 0.7, and 0.9 respectively, then the strength value of path P is 0.8×0.7×0.9 = 0.504. In this way, the strength values of all paths in the network are calculated. The topological path strength value is compared with a preset topological path strength threshold. The preset topological path strength threshold is set according to actual needs, for example, set to 0.6. The topological paths higher than this threshold are classified as strong paths, and the rest are weak paths to obtain path classification data. For example, the strength value of path P is 0.504, which is lower than the threshold 0.6, so path P is classified as a weak path; while the strength value of path Q is 0.72, which is higher than the threshold 0.6, so path Q is classified as a strong path. In this way, path classification data is obtained, including the strength value of each path and its corresponding strong / weak classification. The core area and the global area of the network are determined according to the path classification data. The core area consists of strong paths and the nodes they connect, and the global area contains all other nodes and weak paths in the network. The specific operation is as follows: Traverse the path classification data, extract all strong paths and the nodes they connect to form the core area. The remaining nodes and weak paths in the network are classified as the global area. For example, strong path Q connects nodes A, B, and C, and these nodes and path Q together form the core area; while the other nodes and weak paths not connected by strong paths, such as path P and the nodes it connects, form the global area. The node attribution of the core area and the global area is recalculated, and the node attribution quantity of the global area is removed, and the node attribution quantity of the core area is retained. The specific operation is as follows: Count the attribution quantity of each node in the core area and the global area, that is, how many paths each node belongs to. Then, remove the attribution quantity of the nodes in the global area, and only retain the attribution quantity of the nodes in the core area. For example, node A belongs to 3 strong paths in the core area and 2 weak paths in the global area. After recalculation, only the attribution quantity of node A in the core area, which is 3, is retained. Calculate the proportion of the node attribution quantity in the core area in the network topology structure, including the proportion of the number of nodes, the proportion of the number of connections, and the proportion of data traffic, to generate network core proportion data. The specific operation is as follows: First, calculate the number of nodes, the number of connections, and the data traffic in the core area. For example, the core area contains 10 nodes, 15 connections, and the data traffic is 1000 units. Then, calculate the number of nodes, the number of connections, and the data traffic of the entire network. The node time series of the clinical reliability feature network is sampled according to the network core proportion data, and the sequence point intensity period of the node time series is identified to generate node stability features.The specific operation is as follows: According to the network core ratio data, key nodes in the core area are selected for time series sampling. Based on the node time series, the connection strength values of the clinical reliability feature network are statistically segmented, and the connection strength fluctuation range is determined to generate the connection reliability feature. The specific operation is as follows: According to the time period division of the node time series, the connection strength values are statistically segmented. The node stability feature and the connection reliability feature are merged for core attributes to obtain the reliability structure attribute. The specific operation is as follows: The stability features (such as period and period strength) of each node are merged with the reliability features of the connection (such as fluctuation range and average strength value).

[0108] Preferably, in step S3, dividing the multi-modal chronic disease feature data into chronic disease unstructured data and chronic disease structured data based on the reliability structure attribute includes:

[0109] Determine the feature dimension of the multi-modal chronic disease feature data according to the reliability structure attribute, and dynamically adjust the mapping dimension range based on the feature dimension to obtain the feature mapping result;

[0110] Based on the feature mapping result and using the clustering algorithm to automatically identify the inherent type features in the multi-modal chronic disease feature data, and divide the inherent type features into multiple subsets, where each subset represents a chronic disease data type feature to obtain the clustering feature data;

[0111] Extract the chronic disease structure features of the clustering feature data, and divide the multi-modal chronic disease feature data into chronic disease unstructured data and chronic disease structured data according to the chronic disease structure features.

[0112] In the embodiments of the present invention, by analyzing the reliability structure attributes of the clinical reliability feature network, the feature dimensions of multi-modal chronic disease feature data are determined. The reliability structure attributes include node weight distribution, connection strength, network topology structure, etc. According to these attributes, dimensions related to chronic disease features are selected, such as node betweenness, node degree, edge weight, etc.; based on the determined feature dimensions, the feature data is normalized to map the data to a unified dimension range. A dynamic adjustment strategy is adopted to adjust the mapping range in real time according to the distribution and changes of the data. For example, for the node weight dimension, the normalization range is dynamically adjusted according to the maximum and minimum values of the weight distribution; after the mapping dimension range is adjusted, a feature mapping result is obtained. Each feature data point is represented as a set of normalized values within the new dimension range, facilitating subsequent analysis and processing. A clustering algorithm suitable for multi-modal data is selected, such as K-Means or DBSCAN. For the K-Means algorithm, the number of clustering centers is set (such as 3 clustering centers), and the random seed is 42 to ensure the repeatability of the results. For the DBSCAN algorithm, the neighborhood radius (eps) is set to 0.3, and the minimum number of samples (min_samples) is set to 10; the feature mapping result is input into the clustering algorithm, and the algorithm automatically identifies the inherent type features in the data. The K-Means algorithm iteratively optimizes the clustering centers and assigns data points to the nearest clustering centers; the DBSCAN algorithm identifies clusters through density connectivity; after clustering is completed, multiple subsets are obtained, and each subset represents a chronic disease data type feature. For example, the K-Means algorithm may divide the data into three subsets: high risk, medium risk, and low risk. Chronic disease structure features are extracted from each clustering subset, including the distribution, correlation, change trend, etc. of the features. For example, for the high-risk clustering subset, the distribution range and correlation of its feature values are analyzed; according to the extracted chronic disease structure features, a division standard for multi-modal chronic disease feature data is formulated. Structured data has a clear format and type, such as physiological monitoring parameters, examination results, etc.; unstructured data includes image data, medical record texts, etc.; according to the formulated standard, multi-modal chronic disease feature data is divided into chronic disease unstructured data and chronic disease structured data. The structured data is stored in a relational database, and the unstructured data is stored in a distributed file system.

[0113] Preferably, in step S3, the reconstruction of the chronic disease unstructured data and the chronic disease structured data into chronic disease compressed data includes:

[0114] Perform unstructured information semantic analysis on the chronic disease unstructured data to obtain chronic disease unstructured semantic data; perform structured information relationship enhancement on the chronic disease structured data to obtain chronic disease structured enhanced data;

[0115] Detect the information usage frequency of non-structured semantic data for chronic diseases to generate non-structured information usage frequency data; perform data chunking on the non-structured data for chronic diseases according to the non-structured information usage frequency data to obtain non-structured data chunking information; perform chunk data compression on the non-structured data chunking information to generate non-structured chunk compressed data;

[0116] Identify the access patterns of structured enhanced data for chronic diseases to generate structured information access pattern data; detect the sensitivity of structured data based on the structured information access pattern data to obtain the sensitivity of structured data;

[0117] Eliminate data redundancy from the structured data for chronic diseases based on the sensitivity of structured data to obtain structured redundancy elimination information; perform lightweight data compression on the structured redundancy elimination information to generate structured lightweight compressed data;

[0118] Dynamically compress and merge the non-structured chunk compressed data and the structured lightweight compressed data to obtain the compressed data for chronic diseases.

[0119] In the embodiments of the present invention, natural language processing (NLP) technology is used to perform semantic analysis on unstructured data of chronic diseases (such as medical record texts, doctor's orders, etc.). Through techniques such as word segmentation, part-of-speech tagging, named entity recognition (NER), and dependency syntax analysis, key information in the text is extracted, such as symptoms, disease names, treatment measures, etc. For example, the TF-IDF algorithm is used to extract keywords to form a keyword index. At the same time, topic models (such as LDA) are used to extract latent topics from the text corpus to achieve topic indexing of the text. Through these techniques, unstructured data is converted into structured data with semantic information to obtain unstructured semantic data of chronic diseases. Through association analysis and data mining techniques, the relationships in the structured data of chronic diseases (such as physiological monitoring parameters, examination results, etc.) are strengthened. For example, using association queries in the database, the physiological monitoring data at different time points is associated with the corresponding examination results to strengthen the internal relationships between the data. In this way, potential associations between the data can be discovered to obtain structured enhanced data of chronic diseases. Using data statistics and analysis techniques, the usage frequencies of keywords and topics in the unstructured semantic data of chronic diseases are detected. By counting the number of times each keyword and topic appears in the text, unstructured information usage frequency data is generated. For example, the inverted index technique is used to record the position information of each term in the document, and then its usage frequency is counted. According to the unstructured information usage frequency data, the unstructured data of chronic diseases is divided into different data blocks. Data blocks with higher usage frequencies are preferentially stored and processed to improve data access efficiency. For example, text paragraphs containing high-frequency keywords are divided into one data block, and text paragraphs with low-frequency keywords are divided into another data block. Data compression techniques are used to compress the information of the unstructured data blocks. For example, the Huffman coding or LZ77 algorithm is used to compress each data block to reduce the data storage space. In this way, unstructured block compressed data is generated. Using data mining and machine learning techniques, the access patterns of the structured enhanced data of chronic diseases are identified. By analyzing information such as the access time, access frequency, and access path of the data, structured information access pattern data is generated. For example, clustering algorithms are used to identify the access patterns of different users to the data, and data with similar access patterns are divided into one category. According to the structured information access pattern data, the sensitivity of the structured data is detected. By analyzing information such as the access frequency, access user permissions, and data update frequency, the sensitivity of the data is evaluated. For example, data with a high access frequency and a low update frequency is considered high-sensitivity data. According to the sensitivity of the structured data, redundancy elimination is performed on the structured data of chronic diseases. For high-sensitivity data, more stringent data deduplication and cleaning strategies are adopted to ensure the accuracy and consistency of the data. For example, by comparing the values of data fields, duplicate data records are deleted. A lightweight data compression technique is used to compress the structured redundancy elimination information.For example, data is compressed using differential coding or dictionary coding to reduce the data storage space. In this way, structured lightweight compressed data is generated. The unstructured block compressed data and the structured lightweight compressed data are merged. According to the access frequency and importance of the data, the storage location and compression method of the data are dynamically adjusted. For example, the data with high access frequency is stored in a fast storage device and a lower compression ratio is adopted to improve the data access speed. In this way, the chronic disease compressed data is obtained.

[0120] Of particular importance is that the establishment of the storage layout for the dynamically compressed chronic disease data in step S3 includes:

[0121] Performing block identifier marking on the unstructured block compressed data to obtain block identifier information; extracting structured metadata from the structured lightweight compressed data to obtain structured metadata information;

[0122] Performing chronic disease information storage mapping on the block identifier information and the structured metadata information to generate chronic disease information storage mapping data; positioning the access location of the chronic disease information for the dynamically compressed chronic disease data based on the chronic disease information storage mapping data to obtain the chronic disease information access location points;

[0123] Identifying the chronic disease information query path for the chronic disease information access location points to obtain chronic disease information query path data; determining the storage retrieval type according to the chronic disease information query path data to obtain chronic disease storage retrieval type data;

[0124] Establishing storage retrieval for the dynamically compressed chronic disease data based on the chronic disease storage retrieval type data to generate storage retrieval data; encapsulating the storage retrieval interface for the storage retrieval data to generate chronic disease storage layout information.

[0125] In the embodiments of the present invention, the data marking technology is adopted to mark the identifiers of the unstructured block-compressed data. After the unstructured data is block-compressed, a unique identifier (such as UUID) is assigned to each data block and recorded in the metadata to obtain the block identifier information. The data extraction technology is used to scan the structured lightweight compressed data and extract the metadata of each data record, including the field name, data type, and record identifier. The data mapping technology is applied to associate the identifier of the unstructured block-compressed data with the structured metadata information, create a mapping table for quickly locating the data storage location, and generate the chronic disease information storage mapping data. The data positioning technology is used to quickly locate the physical location of the specific chronic disease information on the storage medium by querying the mapping table to obtain the chronic disease information access location point. The path recognition technology is adopted to perform query path recognition on the chronic disease information access location point, specifically analyze the chronic disease information access mode, and determine the optimal data retrieval path. The data type recognition technology is used to determine the retrieval type of the data according to the query path data, such as full-text retrieval and field retrieval. The data index technology is applied to establish the corresponding data index according to the chronic disease storage retrieval type data to improve the data retrieval efficiency. The interface encapsulation technology is used to encapsulate the storage retrieval data into an API interface and provide it to the application program for data retrieval to obtain the chronic disease storage layout information.

[0126] In the embodiments of the present invention, data monitoring technology is used to monitor the storage capacity of the dynamically compressed data of chronic diseases in real time to obtain storage capacity requirement information; specifically, through a data monitoring system, the remaining space in the data storage system is regularly checked, and the future data growth trend is predicted to obtain the storage capacity requirement information; data management technology is applied to divide the storage resources into different data storage areas according to the storage capacity requirement, and each area is optimized and configured according to the data type and access frequency; permission management technology is used to analyze the chronic disease storage retrieval information, and through a permission management system, access permissions are assigned to users of different roles and recorded in the permission list to obtain chronic disease access permission information; security monitoring technology is used to monitor the access behavior of the dynamically compressed data of chronic diseases through the chronic disease access permission information, specifically analyze the access mode, and identify unauthorized access or abnormal access behavior; dynamic permission adjustment technology is applied to adjust the chronic disease access permission information according to the chronic disease abnormal access status data. Once abnormal access is found, the access permissions of the relevant users are immediately reduced or the accounts are temporarily frozen to protect data security; system performance evaluation technology is used to evaluate the performance indicators of the storage system, such as I / O throughput and response time, and predict the performance of the system when the data volume increases to obtain storage data scalability information; data classification technology is used to analyze the storage data scalability information; specifically, according to the storage data scalability information, it is identified whether horizontal expansion (adding more storage nodes) or vertical expansion (enhancing the performance of existing nodes) is required; system configuration technology is applied to configure the storage system to add new storage nodes or upgrade the hardware resources of existing nodes according to the storage expansion type data; comprehensive data management technology is used to construct a chronic disease data storage architecture according to the data storage capacity division data, access permission setting data, and expansion content setting data; specifically, considering storage capacity, security, and scalability comprehensively, a chronic disease data storage architecture is designed, including data distribution, backup, and disaster recovery plans, to obtain a preliminary chronic disease data storage architecture.

[0127] Especially importantly, the construction of the chronic disease data storage architecture based on the chronic disease storage retrieval information in step S3 includes:

[0128] Monitoring the storage capacity requirement of the dynamically compressed data of chronic diseases based on the chronic disease storage retrieval information to obtain the storage capacity requirement information; dividing the data storage capacity according to the storage capacity requirement information to generate data storage capacity division data;

[0129] Identifying the data access permissions of the chronic disease storage retrieval information to obtain the chronic disease access permission information; detecting the abnormal access status of the dynamically compressed data of chronic diseases through the chronic disease access permission information to generate chronic disease abnormal access status data; reducing the access permissions of the chronic disease access permission information according to the chronic disease abnormal access status data to obtain access permission setting data;

[0130] Perform storage data scalability detection on the chronic disease dynamic compressed data based on the retrieved information of chronic disease storage, and obtain storage data scalability information; perform expansion type identification on the storage data scalability information to generate storage expansion type data; perform expansion content setting on the chronic disease dynamic compressed data based on the storage expansion type data to obtain expansion content setting data;

[0131] Construct a chronic disease data storage architecture according to the data storage capacity division data, access permission setting data, and expansion content setting data, and obtain a preliminary chronic disease data storage architecture.

[0132] In the embodiments of the present invention, by deploying a storage monitoring system, the storage capacity requirements of the dynamically compressed data of chronic diseases are monitored in real time. The monitoring system uses storage capacity monitoring tools (such as Prometheus combined with Grafana) to collect and analyze the usage of storage devices in real time. The monitoring metrics include the used capacity, remaining capacity, read and write speeds of the storage devices, etc. The monitoring system automatically generates storage capacity requirement information according to preset thresholds (such as the remaining capacity is lower than 20%), records the current capacity status of the storage device and the prediction of future capacity requirements. According to the storage capacity requirement information, the storage space is dynamically partitioned through storage management software (such as Ceph or HDFS). The storage management software divides the storage space into different regions according to the access frequency and importance of the data, such as a hot storage area (for frequently accessed data) and a cold storage area (for infrequently accessed data). The partitioning process is implemented through a configuration file, and the size, location, and access permissions of each storage area are defined in the configuration file. The storage management software generates data storage capacity partitioning data according to the configuration file and records the detailed information of each storage area. The data access permissions of the chronic disease storage retrieval information are identified through an access control management system (such as RBAC or ABAC). The access control management system identifies the access permissions of users to data according to the roles, attributes, and policies of the users. For example, a doctor role may have read and write permissions to patient medical records, while ordinary users may only have read permissions. The access control management system generates chronic disease access permission information and records the access permission levels and scopes of each user or user group. An intrusion detection system (IDS) or a security information and event management system (SIEM) is used to detect the abnormal access status of the dynamically compressed data of chronic diseases in combination with the chronic disease access permission information. The IDS or SIEM system identifies abnormal access behaviors, such as frequently accessing sensitive data, unauthorized access, etc., by analyzing access logs and network traffic. The system generates chronic disease abnormal access status data according to preset abnormal detection rules (such as the access frequency exceeds the threshold or the access time is abnormal), and records the specific information of the abnormal access, including the time, user, access path, etc. According to the chronic disease abnormal access status data, the access control management system adjusts the chronic disease access permission information. For users or user groups detected with abnormal access behaviors, the system automatically reduces their access permissions, such as changing the read and write permissions to read-only permissions. The access control management system generates access permission setting data according to the adjusted permission information and records the new access permission levels and scopes of each user or user group. The storage data scalability of the dynamically compressed data of chronic diseases is detected through the storage management software. The storage management software analyzes the expansion capabilities of the storage devices, including the scalability of storage capacity, the scalability of storage nodes, etc. The detection results are recorded in the form of storage data scalability information, including the current expansion status of the storage devices. According to the storage data scalability information, the storage management software identifies the expansion type of the storage devices.The expansion types include horizontal expansion (adding storage nodes) and vertical expansion (increasing storage capacity). The storage management software generates storage expansion type data according to the expansion type, records the applicable scenarios and operation steps of each expansion type; according to the storage expansion type data, the storage management software sets the expansion content for the dynamic compression data of chronic diseases. For horizontal expansion, the storage management software configures new storage nodes and adds them to the storage cluster; for vertical expansion, the storage management software increases the capacity of the storage device. The expansion content setting data records the specific parameters of the expansion operation, such as the IP address of the new node, the size of the newly added capacity, etc.; combining the data storage capacity division data, access permission setting data and expansion content setting data, the storage management software constructs a preliminary architecture for chronic disease data storage. The storage management software generates a configuration file based on these data, defining each component of the storage architecture and its parameters. The preliminary architecture includes the layout of the storage device, the allocation of access permissions, and the planning of expansion operations. The storage management software deploys the storage architecture according to the configuration file to complete the construction of the preliminary architecture for chronic disease data storage.

[0133] Preferably, the architecture adjustment of the preliminary architecture for chronic disease data storage according to the real-time status data of chronic disease storage in step S4 includes:

[0134] Detect abnormal storage response time for the real-time status information of chronic disease storage, and generate abnormal response time information;

[0135] Locate the abnormal storage block for the abnormal response time information to obtain the abnormal response storage block data; based on the abnormal response storage block data, allocate block resources for the dynamic compression data of chronic diseases to obtain abnormal block resource allocation measures;

[0136] Detect abnormal writing speed for the real-time status information of chronic disease storage, and generate abnormal writing speed information;

[0137] Set the write buffer for the abnormal writing speed information to generate write buffer information; according to the write buffer information, perform data parallel writing on the dynamic compression data of chronic diseases to obtain data parallel writing measures;

[0138] Merge the abnormal block resource allocation measures and the data parallel writing measures for architecture adjustment to generate the adjustment data for chronic disease data storage; based on the adjustment data for chronic disease data storage, perform architecture adjustment on the preliminary architecture for chronic disease data storage to obtain the architecture for chronic disease data storage.

[0139] In the embodiments of the present invention, the time series anomaly detection technology is used to monitor the storage response time in the real-time status information of chronic disease storage; specifically, through an algorithm process, such as sample sorting, calculating mean and std.dev, and finding the difference between min / max and mean, it is judged whether the storage response time exceeds the normal range; the distance-based method or the proximity-based method is applied to analyze the response time anomaly information; and a non-linear method, such as a neural network, is used to deeply analyze the abnormal response time to locate the abnormal storage block; the system resource management technology is used to allocate resources for the dynamic compression data of chronic diseases based on the data of the abnormal response storage block, specifically, reallocate resources for the located abnormal storage block, such as increasing CPU or memory resources, to optimize performance and obtain the resource allocation measures for the abnormal block; the deep learning time series anomaly detection method is used to monitor the writing speed in the real-time status information of chronic disease storage to identify anomalies; specifically, through a recurrent neural network (RNN) model, the writing speed is predicted and compared with the actual value to generate the writing speed anomaly information; the data buffering technology is applied to process the writing speed anomaly information and set a suitable writing buffer; specifically, according to the writing speed anomaly information, the size of the writing buffer of the database is adjusted to optimize the data writing process and generate the writing buffer information; the parallel processing technology is used to perform parallel writing on the dynamic compression data of chronic diseases according to the writing buffer information; specifically, through the parallel writing technology, the data writing efficiency is improved and the writing delay is reduced to obtain the data parallel writing measures; the system architecture optimization technology is applied to merge the resource allocation measures for the abnormal block and the data parallel writing measures to generate the adjustment data for chronic disease data storage; the cloud storage scalability technology is used to adjust the cloud storage architecture according to the adjustment data for chronic disease data storage, such as adding storage nodes or optimizing data distribution, to improve the scalability and performance of the system and obtain the chronic disease data storage architecture.

[0140] In this specification, an information recording and storage system for chronic disease management is provided, which is used to execute the above-mentioned information recording and storage method for chronic disease management. The information recording and storage system for chronic disease management includes:

[0141] A chronic disease data acquisition module, which is used to obtain the chronic disease medical record images of patients; identify the text area of the chronic disease medical record images of patients, perform the shaping degree detection of chronic disease vocabulary on the text area to obtain the shaping degree of chronic disease vocabulary; and perform chronic disease data feature fusion on the text area based on the shaping degree of chronic disease vocabulary to obtain multi-modal chronic disease feature data.

[0142] A chronic disease information association module, which is used to obtain the chronic disease medical images of patients and mark the patient's physiological monitoring parameters for the multi-modal chronic disease feature data; extract the patient's clinical manifestation data of the multi-modal chronic disease feature data; perform implicit feature mining based on the patient's physiological monitoring parameters and the patient's clinical manifestation data, and construct a clinical reliability feature network.

[0143] The chronic disease information storage architecture module is used to perform structural attribute analysis on the clinical reliability feature network to obtain reliability structural attributes; divide the multi-modal chronic disease feature data into chronic disease unstructured data and chronic disease structured data based on the reliability structural attributes; reconstruct the chronic disease unstructured data and chronic disease structured data into chronic disease compressed data; establish a storage layout for the chronic disease dynamic compressed data to generate chronic disease storage layout information; construct a preliminary chronic disease data storage architecture based on the chronic disease storage layout information;

[0144] The chronic disease information storage architecture adjustment module is used to input the patient's chronic disease data into the preliminary chronic disease data storage architecture and monitor the real-time status data of chronic disease storage; adjust the preliminary chronic disease data storage architecture according to the real-time status data of chronic disease storage to obtain the chronic disease data storage architecture; record and store the patient's chronic disease data based on the chronic disease data storage architecture to generate a chronic disease information storage report.

[0145] Through the integration of multiple functional modules, the present invention realizes the full-process automated processing from chronic disease data collection, information association, storage architecture construction to architecture adjustment, and has significant beneficial effects. The system can efficiently acquire and process the patient's chronic disease medical record images, accurately identify the text area and perform chronic disease data feature fusion to generate multi-modal chronic disease feature data. By implicitly mining features to construct a clinical reliability feature network, further perform structural attribute analysis on the data, realize data classification and compression, and optimize the storage layout. In addition, the system can dynamically monitor the storage status and adjust the architecture in real time to ensure the efficiency and stability of data storage, and finally generate a detailed chronic disease information storage report, providing comprehensive, accurate and efficient data support for chronic disease management, and improving the scientificity and convenience of chronic disease management.

[0146] Therefore, in any regard, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to embrace all changes within the meaning and scope of the equivalent elements of the application documents in the present invention.

[0147] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An information recording and storage method for chronic disease management, characterized in that, Including the following steps: Step S1: Obtain the chronic disease medical record image of the patient; identify the text area of the chronic disease medical record image of the patient, perform a detection on the shaping degree of chronic disease vocabulary for the text area to obtain the shaping degree of chronic disease vocabulary; based on the shaping degree of chronic disease vocabulary, perform a fusion of chronic disease data features on the text area to obtain multi-modal chronic disease feature data; Step S2: Obtain the chronic disease medical image of the patient and perform a marking of the patient's physiological monitoring parameters on the multi-modal chronic disease feature data; extract the clinical manifestation data of the patient from the multi-modal chronic disease feature data; perform an implicit feature mining based on the patient's physiological monitoring parameters and the patient's clinical manifestation data, and construct a clinical reliability feature network; Among them, the performing an implicit feature mining based on the patient's physiological monitoring parameters and the patient's clinical manifestation data, and constructing a clinical reliability feature network includes: Perform a layer-by-layer analysis on the patient's physiological monitoring parameters to extract the physiological signal fluctuation pattern, abnormal peak value, and periodic change; Perform an item-by-item decomposition on the patient's clinical manifestation data to extract the clinical manifestation frequency, intensity, and duration; Perform a time series analysis on the physiological signal fluctuation pattern to determine the signal frequency distribution feature, and perform a non-linear association with the clinical manifestation frequency to obtain frequency implicit association data; Record the peak points of the physiological signal abnormal peak value, detect the duration of the peak signal amplitude, and perform a corresponding relationship with the clinical manifestation intensity through a correlation function to obtain intensity implicit association data; Extract the cycle length of the physiological signal periodic change, identify the cycle amplitude change, and perform a time series comparison with the clinical manifestation duration to obtain cycle implicit association data; Construct a clinical reliability feature network based on the frequency implicit association data, intensity implicit association data, and cycle implicit association data; Step S3: Perform a structural attribute analysis on the clinical reliability feature network to obtain the reliability structural attribute; based on the reliability structural attribute, divide the multi-modal chronic disease feature data into chronic disease unstructured data and chronic disease structured data; reconstruct the chronic disease unstructured data and chronic disease structured data into chronic disease compressed data; establish a storage layout for the chronic disease dynamic compressed data to generate chronic disease storage layout information; construct a preliminary architecture for chronic disease data storage based on the chronic disease storage layout information; Step S4: Input the patient's chronic disease data into the preliminary architecture for chronic disease data storage and monitor the real-time status data of chronic disease storage; perform an architecture adjustment on the preliminary architecture for chronic disease data storage according to the real-time status data of chronic disease storage to obtain the chronic disease data storage architecture; perform an information recording and storage on the patient's chronic disease data based on the chronic disease data storage architecture to generate a chronic disease information storage report.

2. The information recording and storage method for chronic disease management according to claim 1, wherein Step S1 includes the following steps: Step S11: Obtain the chronic disease medical record image of the patient; perform an image color enhancement on the chronic disease medical record image of the patient, and perform noise removal to obtain a standardized medical record image; Step S12: Mark the text features of the standardized medical record image, and identify the text area of the standardized medical record image according to the text features; Step S13: Perform a recognition on the handwritten font outline of the text area to obtain the handwritten font outline features; Step S14: Extract the stroke connection points of the handwritten font outline features, and detect the font connection complexity according to the stroke connection points to generate the font connection complexity; Step S15: Identify the aspect ratio of the handwritten font outline features, detect the shape presentation degree according to the aspect ratio of the font, and generate the font shape presentation degree; Step S16: Map the font connection complexity and the font shape presentation degree to the chronic disease vocabulary shaping degree to obtain the chronic disease vocabulary shaping degree; Step S17: Optimize the font presentation shape of the handwritten font outline according to the chronic disease vocabulary shaping degree to obtain the chronic disease vocabulary optimization data; Step S18: Based on the chronic disease vocabulary optimization data, fuse the chronic disease data features of the text area to obtain the multi-modal chronic disease feature data.

3. The information recording and storage method for chronic disease management according to claim 1, characterized in that Step S2 includes the following steps: Step S21: Obtain the medical images of the patient's chronic diseases; Step S22: Locate the patient's diseased part in the medical images of the patient's chronic diseases to obtain the diseased part data; Extract the patient's physiological monitoring parameters from the multi-modal chronic disease feature data according to the diseased part data to generate the patient's physiological monitoring parameters; Step S23: Mark the patient's chronic disease clinical features for the multi-modal chronic disease feature data to obtain the patient's chronic disease clinical feature marking information; Extract the patient's chronic disease clinical data from the multi-modal chronic disease feature data based on the patient's chronic disease clinical feature marking information to obtain the patient's clinical manifestation data; Step S24: Mine implicit features based on the patient's physiological monitoring parameters and the patient's clinical manifestation data, and construct a clinical reliability feature network.

4. The information recording and storage method for chronic disease management according to claim 1, wherein, In Step S3, the structural attribute analysis of the clinical reliability feature network includes: Assign a unique identifier to each node in the clinical reliability feature network to form a node index sequence; Determine the connection feature parameters according to the node index sequence, and construct a connection feature matrix based on the connection feature parameters; Perform block processing on the connection feature matrix, divide the connection feature matrix into several sub-regions, and each sub-region contains a certain number of nodes and connections; Identify the local connection pattern and the local node distribution position within each sub-region; Map the local connection pattern and the local node distribution position to generate a network topology structure; Evaluate the node weights of the network topology structure, detect the proportion of nodes in the network, and obtain the node weight distribution data; Detect the node association quantity of the network topology structure, determine the node attribute features according to the node association quantity, and set the initial node weights based on the node attribute features; Continuously monitor the dynamic interaction behavior of the nodes, and update the initial node weights in real time according to the dynamic interaction behavior to obtain the node weight distribution data; Evaluate the connection strength of the network topology structure according to the node weight distribution data to obtain the node connection strength data; Calculate the initial connection strength value of the nodes of the network topology structure based on the node weight distribution data; Continuously monitor the dynamic connection situation of the nodes, and update the connection strength of the nodes in real time according to the dynamic connection situation to obtain the node connection strength data.

5. The information recording and storage method for chronic disease management according to claim 4, wherein, In Step S3, the structural attribute analysis of the clinical reliability feature network also includes: Calculate the topological path strength value for the network topology structure according to the node connection strength data; Compare the topological path strength value with the preset topological path strength threshold, and divide the topological paths higher than the threshold into strong paths, and the rest are weak paths to obtain the path division data; Determine the core area and the global area of the network according to the path-partitioned data. The core area consists of strong paths and the nodes they connect, and the global area contains all other nodes and weak paths in the network; Recalculate the node attribution for the core area and the global area, excluding the node attribution quantity of the global area and retaining the node attribution quantity of the core area; Calculate the proportion of the node attribution quantity of the core area in the network topology structure, including the proportion of the number of nodes, the proportion of the number of connections, and the proportion of data traffic, to generate network core proportion data; Perform node time-series sampling on the clinical reliability feature network based on the network core proportion data, and identify the sequence point intensity period of the node time series to generate node stability features; Statistically segment the connection strength values of the clinical reliability feature network according to the node time series, and determine the connection strength fluctuation range to generate connection reliability features; Perform core attribute merging on the node stability features and the connection reliability features to obtain the reliability structure attributes.

6. The information recording and storage method for chronic disease management according to claim 1, wherein The division of multi-modal chronic disease feature data into chronic disease unstructured data and chronic disease structured data described in step S3 includes: Determine the feature dimensions of the multi-modal chronic disease feature data according to the reliability structure attributes, and dynamically adjust the mapping dimension range based on the feature dimensions to obtain the feature mapping results; Based on the feature mapping results and using a clustering algorithm, automatically identify the inherent type features in the multi-modal chronic disease feature data, and divide the inherent type features into multiple subsets, where each subset represents a chronic disease data type feature, to obtain clustering feature data; Extract the chronic disease structure features of the clustering feature data, and divide the multi-modal chronic disease feature data into chronic disease unstructured data and chronic disease structured data according to the chronic disease structure features.

7. The information recording and storage method for chronic disease management according to claim 6, characterized in that The reconstruction of the chronic disease unstructured data and the chronic disease structured data into chronic disease compressed data described in step S3 includes: Perform semantic analysis of the unstructured information of the chronic disease unstructured data to obtain chronic disease unstructured semantic data; perform strengthening of the structured information relationship of the chronic disease structured data to obtain chronic disease structured strengthened data; Detect the information usage frequency of the chronic disease unstructured semantic data to generate unstructured information usage frequency data; perform data chunking on the chronic disease unstructured data according to the unstructured information usage frequency data to obtain unstructured data chunking information; perform chunked data compression on the unstructured data chunking information to generate unstructured chunked compressed data; Identify the access mode of the chronic disease structured strengthened data to generate structured information access mode data; detect the sensitivity of the structured data for the structured information access mode data to obtain the structured data sensitivity; Eliminate data redundancy for the chronic disease structured data based on the structured data sensitivity to obtain structured redundancy elimination information; perform lightweight data compression on the structured redundancy elimination information to generate structured lightweight compressed data; Perform dynamic data compression merging on the unstructured chunked compressed data and the structured lightweight compressed data to obtain chronic disease compressed data.

8. The information recording and storage method for chronic disease management according to claim 1, wherein The architecture adjustment of the preliminary architecture for chronic disease data storage according to the real-time status data of chronic disease storage described in step S4 includes: Perform abnormal detection on the storage response time of the real-time status information of chronic disease storage, and generate abnormal response time information; Locate the abnormal storage block for the abnormal response time information to obtain the abnormal response storage block data; Based on the abnormal response storage block data, allocate block resources for the dynamically compressed data of chronic diseases to obtain abnormal block resource allocation measures; Perform abnormal detection on the writing speed of the real-time status information of chronic disease storage, and generate abnormal writing speed information; Set the write buffer for the abnormal writing speed information to generate write buffer information; According to the write buffer information, perform parallel data writing on the dynamically compressed data of chronic diseases to obtain parallel data writing measures; Adjust and merge the abnormal block resource allocation measures and the parallel data writing measures in the architecture to generate adjusted data for chronic disease data storage; Based on the adjusted data for chronic disease data storage, adjust the initial architecture of chronic disease data storage to obtain the chronic disease data storage architecture.

9. An information recording and storage system for chronic disease management, characterized in that, For implementing the information recording and storage method for chronic disease management as described in claim 1, the information recording and storage system for chronic disease management includes: A chronic disease data acquisition module, which is used to obtain the chronic disease medical record images of patients; Identify the text areas of the chronic disease medical record images of patients, perform detection on the shaping degree of chronic disease vocabulary in the text areas to obtain the shaping degree of chronic disease vocabulary; Based on the shaping degree of chronic disease vocabulary, perform feature fusion of chronic disease data on the text areas to obtain multi-modal chronic disease feature data; A chronic disease information association module, which is used to obtain the chronic disease medical images of patients and mark the patient's physiological monitoring parameters for the multi-modal chronic disease feature data; Extract the patient's clinical manifestation data from the multi-modal chronic disease feature data; Perform implicit feature mining based on the patient's physiological monitoring parameters and the patient's clinical manifestation data, and construct a clinical reliability feature network; A chronic disease information storage architecture module, which is used to analyze the structural attributes of the clinical reliability feature network to obtain reliability structural attributes; Based on the reliability structural attributes, divide the multi-modal chronic disease feature data into chronic disease unstructured data and chronic disease structured data; Reconstruct the chronic disease unstructured data and the chronic disease structured data into compressed chronic disease data; Establish a storage layout for the dynamically compressed chronic disease data to generate chronic disease storage layout information; Based on the chronic disease storage layout information, construct an initial architecture for chronic disease data storage; A chronic disease information storage architecture adjustment module, which is used to input the patient's chronic disease data into the initial architecture of chronic disease data storage and monitor the real-time status data of chronic disease storage; According to the real-time status data of chronic disease storage, adjust the initial architecture of chronic disease data storage to obtain the chronic disease data storage architecture; Based on the chronic disease data storage architecture, perform information recording and storage on the patient's chronic disease data to generate a chronic disease information storage report.

Images