Medical data classified storage method for blood sampling inspection

Abstract

The invention discloses a medical data classified storage method for blood sampling inspection, and belongs to the technical field of data processing, and the method specifically comprises the following steps: collecting identity information and blood sampling inspection data of a patient; constructing an identity data link, and generating an encrypted identity data link and an encrypted inspection data block; performing logical association on the encrypted identity data link and the encrypted inspection data block to obtain an encrypted patient data unit, packaging the encrypted patient data unit and the screening data link into a data task unit, and routing the data task unit to a cache database corresponding to the initial screening department; after affiliation confirmation feedback is received, a definite diagnosis department label is generated and stored in a central data storage library in an associated mode; and performing logic classification on the data based on the definite diagnosis department label and establishing an index structure. According to the invention, intelligent distribution, active routing and structured storage of inspection data are realized, the clinical data scheduling efficiency and diagnosis support capability are effectively improved, and a closed-loop optimized medical data management system is formed.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and specifically to a method for classifying and storing medical data for blood collection and testing. Background Technology

[0002] In modern medical diagnosis, blood tests are a crucial means of obtaining patients' physiological and biochemical indicators to aid in disease diagnosis and treatment. Various automated testing devices can rapidly generate massive amounts of test data containing multiple indicators. How to efficiently classify, store, and manage this data is of great significance for improving clinical diagnostic efficiency and data reuse value.

[0003] Currently, most common medical data storage methods employ a centralized architecture, storing laboratory data from various sources uniformly in data centers or cloud storage systems. These systems typically rely on manual entry of diagnostic results or labeling and archiving data based on simple rules. While they can achieve basic data management and retrieval, they still have significant limitations in practical applications. On the one hand, laboratory data itself lacks effective classification labels for clinical departments, leading to low data retrieval and scheduling efficiency; doctors often have to spend a lot of time sifting through irrelevant data to find key information. On the other hand, because laboratory data and clinical diagnostic conclusions are separate and do not form a closed-loop relationship, the data is difficult to use efficiently for model training and system self-optimization.

[0004] Furthermore, while some technical solutions have attempted to use artificial intelligence models for preliminary triage of test data, their outputs are often merely suggestions and are not organically integrated with the physical or logical storage structure of the data, multi-departmental collaboration mechanisms, or continuous model optimization processes. As a result, the model and data management are often disconnected, making it impossible to achieve dynamic data classification and self-correction in actual clinical workflows, thus limiting the overall intelligence and practical value of the system.

[0005] Therefore, there is an urgent need in this field for a novel classification and storage method that can deeply integrate intelligent triage, dynamic routing, clinical feedback and data storage, so as to achieve high-precision, automated and continuously optimized management of blood collection and testing data. Summary of the Invention

[0006] The purpose of this invention is to provide a method for classifying and storing medical data for blood collection and testing, thereby solving the following technical problems: There is an urgent need in this field for a novel classification and storage method that can deeply integrate intelligent triage, dynamic routing, clinical feedback and data storage, so as to achieve high-precision, automated and continuously optimized management of blood collection and testing data.

[0007] The objective of this invention can be achieved through the following technical solutions: A method for classifying and storing medical data for blood collection and testing, comprising the following steps: S1, collect the patient's identity information and blood test text data, determine the screening data chain based on the blood test text data, and mark the first screening department in the screening data chain as the initial screening department; S2, construct an identity data chain based on the identity information, and use an encryption algorithm to encrypt the identity data chain and blood collection test text data to generate an encrypted identity data chain and an encrypted test data block. S3, logically associate the encrypted identity data chain with the encrypted test data block to obtain an encrypted patient data unit, encapsulate the encrypted patient data unit and the screening data chain into a data task unit, and route it to the cache database corresponding to the initial screening department; S4, Receive the treatment feedback returned by the initial screening department. If the treatment feedback is a negative attribution, route the data task unit to the cache database corresponding to the next screening department and update the routing status information in the data task unit. S5. When a confirmed attribution is received from any screening department, a confirmed department label is generated and associated with the encrypted patient data unit by signing. S6. Store the encrypted patient data unit after signature association to the central data repository, construct a first key-value index based on the confirmed department label, and construct a second feature index based on the encrypted test data block to generate a dual-index encrypted storage structure.

[0008] As a further aspect of the present invention: in step S1, the specific process of acquiring the screening data chain is as follows: The test item identifiers and measurement values ​​in the blood collection test data are combined to form a standardized input vector; The standardized input vector is input into the neural network classifier of the preset department prediction model. The neural network classifier outputs a set of category codes representing different departments and a confidence value corresponding to each category code. The category codes are sorted in descending order of confidence scores, and the sorted category code sequence is mapped to the corresponding department identifier sequence to generate a screening data chain.

[0009] As a further aspect of the present invention: in step S2, the specific process of generating the encrypted identity data chain and the encrypted verification data block is as follows: Extract identity elements from the identity information, including patient identifier, name code, gender code and age data, and connect the identity elements in a predetermined order with priority given to the patient identifier to form an identity data chain. The identity data chain is encrypted using a first encryption algorithm based on asymmetric encryption to generate an encrypted identity data chain. The blood test text data is divided into data blocks according to the test items. Each data block containing the test item identifier and measurement value is encrypted using a second encryption algorithm based on a symmetric encryption algorithm to generate a corresponding encrypted data block. The encrypted data blocks are then combined in the order of the test items to form a complete encrypted test data block.

[0010] As a further aspect of the present invention: in S3, the specific generation process of the data task unit is as follows: The encrypted identity data chain and the encrypted verification data block are used to generate a unique association identifier through a cryptographic hash function. A logical mapping relationship is established between the two based on the association identifier to form an association data group. The association data group is then encapsulated with the screening data chain containing the screening department sequence. During the encapsulation process, a timestamp and task status identifier are added to the data task unit to generate a data task unit.

[0011] As a further aspect of the present invention: in step S4, the specific process of routing the data task unit to the first cache database corresponding to the next screening department is as follows: Obtain the real-time data task unit inventory value and its confidence level for each screening department; measure the physical path distance between the initial department and any screening department to obtain the measured physical path distance value; By combining the existing data task unit values, confidence scores, and measured physical path distances, and performing weighted calculations, the route optimization score for each selected department is obtained. By comparing the routing optimization scores of each screening department, the screening department with the highest score is selected as the next initial department, and the data task unit is routed to the cache database corresponding to the selected next initial department.

[0012] As a further aspect of the present invention: if there are two or more screening departments with equal and maximum routing scores, then the confidence level of the screening department is obtained, and the screening department with the highest confidence level is selected as the next initial department.

[0013] As a further aspect of the present invention: In step S6, the specific process of storing the encrypted patient data unit after signature association to the central data repository is as follows: Extract the confirmed department code and corresponding disease classification code from the treatment feedback, convert the department code into a standard department identifier, and convert the disease classification code into a standard disease identifier; combine the standard department identifier and the standard disease identifier according to a predetermined label format to generate a structured confirmed department label; A digital signature algorithm is used to sign the confirmed department label and the encrypted patient data unit to generate a digital signature identifier; a logical binding relationship is established between the digital signature identifier and the encrypted patient data unit; the logically bound encrypted patient data unit and the corresponding digital signature identifier are stored together in a designated data partition of the central data repository.

[0014] As a further aspect of the present invention: the specific process of generating the dual-index encrypted storage structure in step S6 is as follows: The department identifier and disease identifier contained in the confirmed department label are parsed, and the parsed identifier combination is used as the first key value. The mapping relationship between the first key value and the storage address of the encrypted patient data unit in the central data repository is established to form the first key value index. Extract the test item identifier and measurement value from the original blood collection test text data corresponding to the encrypted test data block, generate a feature hash value based on the extracted test features, use the feature hash value as the second key value, establish a mapping relationship between the second key value and the storage address of the encrypted patient data unit, and form a second feature index; associate the first key value index and the second feature index together with the same encrypted patient data unit, and establish a complete dual-index encrypted storage structure in the central data repository.

[0015] As a further aspect of the present invention: S6 further includes transmitting the data task unit, screening data chain, routing status information and confirmed department label as feedback data back to the preset department prediction model; The measured values ​​and test item identifiers of the blood collection and testing data are extracted from the feedback data and converted into numerical feature vectors as input features. At the same time, the diagnostic department label is converted into a unique hot code form as the target output. The predicted output of the preset department prediction model is obtained through forward propagation calculation, and the error value between the predicted output and the target output is calculated using the cross-entropy loss function. The error value is propagated back along the network structure to calculate the error signal of each hidden layer neuron node. The gradient value of each connection weight in the network is calculated based on the error signal. The value of each connection weight is updated using the stochastic gradient descent algorithm. The updated connection weight parameters are deployed to the preset department prediction model to complete the correction of the preset department prediction model.

[0016] The beneficial effects of this invention are: 1) This invention automatically generates confirmed department and disease tags after attribution confirmation, and archives them after strongly associating them with the original data, ensuring the integrity and high value of the data in the database. Subsequently, a logical index structure built based on these tags enables massive amounts of data to be organized in an orderly manner according to disease and department dimensions, fundamentally improving the reuse value and retrieval efficiency of medical data, laying a high-quality data foundation for precision medicine and scientific research analysis. This makes it possible to subsequently perform precise data retrieval based on disease type and batch retrieval for clinical research or statistical analysis, solving the core contradiction of "large data volume but low usability".

[0017] 2) This invention forms a complete intelligent closed loop that runs through data classification, storage and model self-optimization. This invention forms a structured knowledge base that can be quickly retrieved by encrypting and associating diagnostic results with test data, based on departments and diseases. Each diagnostic label is fed back to the prediction model as a training sample. The closed-loop learning mechanism is used to continuously optimize the model parameters. Finally, a medical data management system integrating intelligent pre-triage, dynamic routing, secure storage and autonomous evolution is built.

[0018] 3) This invention significantly improves the processing efficiency of medical test data by constructing data task units and realizing their dynamic intelligent routing among multiple departments. This invention analyzes blood test data through a department prediction model, automatically generates a sequence of screened departments sorted by confidence level, and actively routes the data to the most likely initial department, directly presenting it in the pending tasks on the physician's terminal. This transforms "people looking for data" into "data looking for people." After receiving negative feedback from a department, it can dynamically calculate the optimal path and reroute the data in real time by combining the data load, physical distance, and model confidence level of each screened department, forming a flexible and intelligent task redistribution system, thereby improving the processing efficiency of medical test data in diverse clinical scenarios. Attached Figure Description

[0019] The invention will now be further described with reference to the accompanying drawings.

[0020] Figure 1 This is a schematic diagram of a medical data classification and storage method for blood collection and testing according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1As shown, this invention is a method for classifying and storing medical data for blood collection and testing, comprising the following steps: S1, collect the patient's identity information and blood test text data, determine the screening data chain based on the blood test text data, and mark the first screening department in the screening data chain as the initial screening department; S2, construct an identity data chain based on the identity information, and use an encryption algorithm to encrypt the identity data chain and blood collection test text data to generate an encrypted identity data chain and an encrypted test data block. S3, logically associate the encrypted identity data chain with the encrypted test data block to obtain an encrypted patient data unit, encapsulate the encrypted patient data unit and the screening data chain into a data task unit, and route it to the cache database corresponding to the initial screening department; S4, Receive the treatment feedback returned by the initial screening department. If the treatment feedback is a negative attribution, route the data task unit to the cache database corresponding to the next screening department and update the routing status information in the data task unit. S5. When a confirmed attribution is received from any screening department, a confirmed department label is generated and associated with the encrypted patient data unit by signing. S6. Store the encrypted patient data unit after signature association to the central data repository, construct a first key-value index based on the confirmed department label, and construct a second feature index based on the encrypted test data block to generate a dual-index encrypted storage structure.

[0023] In a preferred embodiment, the specific process of acquiring the screening data chain in S1 is as follows: The test item identifiers and measurement values ​​in the blood collection test data are combined to form a standardized input vector; The standardized input vector is input into the neural network classifier of the preset department prediction model. The neural network classifier outputs a set of category codes representing different departments and a confidence value corresponding to each category code. The category codes are sorted in descending order of confidence scores, and the sorted category code sequence is mapped to the corresponding department identifier sequence to generate a screening data chain.

[0024] First, the various test item identifiers in the blood test data are combined and paired with their corresponding measurement values. For example, "GLU" (blood glucose identifier) ​​is combined with the measured value "8.5" to form a standardized input vector that is machine-recognizable and has a uniform format. The purpose of this step is to transform the originally heterogeneous and multi-source medical data into a numerical feature expression suitable for artificial intelligence models to process. Next, this standardized input vector is input into the neural network classifier of a pre-trained department prediction model. This classifier uses its multi-layer nonlinear mapping capability to perform high-level abstraction and pattern matching on the input features, and finally outputs a set of category codes representing different departments and a confidence value corresponding to each code. The confidence value reflects the probability that the model determines that the current test data belongs to a certain department. Then, these category codes are sorted in descending order of confidence value, with the aim of prioritizing the departments with the highest probability, forming a decision sequence arranged according to urgency and probability. Finally, the sorted category code sequence is converted into a readable department identifier sequence through predefined mapping rules. For example, the code "CARD" is mapped to "cardiology", thereby generating the final screening department chain used for subsequent routing decisions.

[0025] This system transforms raw, isolated test data into a logically ordered sequence of triage suggestions with clear clinical relevance, providing crucial decision-making support for intelligent routing of subsequent data task units. By simulating the reasoning process of experienced triage physicians using artificial intelligence technology, it can automatically and rapidly extract key features from complex, multi-indicator test data and correlate them with the most likely clinical departments. This significantly reduces the subjectivity and risk of missed diagnoses inherent in manual triage and lays a solid foundation for automated and intelligent data classification and flow throughout the system. This enables medical data to be efficiently and accurately allocated to the most appropriate diagnostic departments, thereby comprehensively improving data utilization efficiency and the reliability of collaborative clinical diagnosis.

[0026] In another preferred embodiment, the specific process of generating the encrypted identity data chain and the encrypted verification data block in step S2 is as follows: Extract identity elements from the identity information, including patient identifier, name code, gender code and age data, and connect the identity elements in a predetermined order with priority given to the patient identifier to form an identity data chain. The identity data chain is encrypted using a first encryption algorithm based on asymmetric encryption to generate an encrypted identity data chain. The blood test text data is divided into data blocks according to the test items. Each data block containing the test item identifier and measurement value is encrypted using a second encryption algorithm based on a symmetric encryption algorithm to generate a corresponding encrypted data block. The encrypted data blocks are then combined in the order of the test items to form a complete encrypted test data block.

[0027] First, key identity elements are extracted from the patient's identity information, including basic identity information such as patient identifier, name code, gender code, and age data. These elements are arranged and connected in a predetermined order prior to the patient identifier to form a complete identity data chain. The purpose of this is to establish a standardized identity information structure for subsequent processing. Next, the identity data chain is encrypted using a first encryption algorithm based on asymmetric encryption. The encrypted identity data chain is generated through the public-key encryption and private-key decryption mechanism unique to asymmetric encryption. This encryption method can effectively protect the patient's sensitive identity information. At the same time, the blood collection test text data is divided into multiple independent data blocks according to the test items. Each data block contains a specific test item identifier and corresponding measurement value. Each data block is encrypted using a second encryption algorithm based on symmetric encryption to generate a corresponding encrypted data block. The application of symmetric encryption here ensures the efficiency of encryption and decryption of a large amount of test data. Finally, the encrypted data blocks are recombined according to the original test item order to form a complete encrypted test data block, ensuring the integrity and orderliness of the test data.

[0028] By employing differentiated encryption strategies, layered security protection for sensitive identity information and test data is achieved. This ensures both a high level of security for patient privacy data and efficient processing of test data encryption and decryption, providing reliable security for subsequent data transmission and storage. Meanwhile, standardized data processing procedures lay a solid foundation for data exchange and sharing within the system, enabling the entire medical data management system to operate efficiently while ensuring security. Ultimately, this provides key technical support for building a secure and reliable medical data storage system.

[0029] In another preferred embodiment, the specific generation process of the data task unit in step S3 is as follows: The encrypted identity data chain and the encrypted verification data block are used to generate a unique association identifier through a cryptographic hash function. A logical mapping relationship is established between the two based on the association identifier to form an association data group. The association data group is then encapsulated with the screening data chain containing the screening department sequence. During the encapsulation process, a timestamp and task status identifier are added to the data task unit to generate a data task unit.

[0030] First, the encrypted identity data chain and encrypted verification data block, after being encrypted, are used as input. A unique association identifier is generated through cryptographic hash function calculation. This identifier is similar to a digital fingerprint of the data, uniquely representing the association between the two sets of encrypted data. Based on this association identifier, a logical mapping relationship is established between the encrypted identity data chain and the encrypted verification data block, forming an inherently related data group. This ensures that although the two are physically stored independently, they are logically closely linked. Then, the related data group is encapsulated with a screening data chain containing the screening department sequence. During the encapsulation process, a timestamp accurate to milliseconds and a task status identifier indicating the current processing status are added to the data task unit. The timestamp records the creation time of the data task unit, and the task status identifier reflects the stage of the unit in the triage process. Finally, a data task unit with complete information is generated.

[0031] Cryptographic hash functions ensure the immutable correlation between different encrypted data blocks, timestamps and status identifiers enable full traceability of data task units, and standardized encapsulation formats guarantee the consistency of data exchange between system modules. This not only prevents data loss or confusion during transmission but also provides a complete information foundation for subsequent data routing and status tracking, enabling the entire medical data classification and storage system to operate in an orderly and reliable manner. Ultimately, it provides crucial intermediate data carrier support for achieving accurate and efficient medical data management.

[0032] In another preferred embodiment, the specific process of routing the data task unit to the first cache database corresponding to the next screening department in step S4 is as follows: Obtain the real-time data task unit inventory value and its confidence level for each screening department; measure the physical path distance between the initial department and any screening department to obtain the measured physical path distance value; By combining the existing data task unit values, confidence scores, and measured physical path distances, and performing weighted calculations, the route optimization score for each selected department is obtained. By comparing the routing optimization scores of each screening department, the screening department with the highest score is selected as the next initial department, and the data task unit is routed to the cache database corresponding to the selected next initial department.

[0033] First, the system queries the cache database corresponding to each selected department in real time to obtain the number of currently stored data task units, i.e., the inventory value. This value reflects the workload to be processed in each department. Simultaneously, the system reads the confidence scores previously predicted for these departments by the triage analysis model. Next, by retrieving indoor map data of the hospital building, the system calculates the shortest feasible path length from the patient's current initial department location to the consultation rooms of each selected department, obtaining the measured physical path distance, which is the actual spatial length the patient needs to move. Then, the system combines the inventory value of data task units, the confidence scores, and the physical path distances. The measured values, along with these three parameters, are weighted and fused together. The principle is that the smaller the stored value, the lighter the workload of the initial department and the faster the processing speed; the higher the confidence score, the greater the probability that the department is the correct one; and the shorter the physical path distance, the lower the patient's travel cost. By assigning weights to each parameter, a route optimization score is calculated. Finally, the route optimization scores of all selected departments are compared, and the department with the highest score is selected as the next initial department. The data task unit is then transferred from its current cache database to the cache database corresponding to the selected department, and the patient is notified to go to the new department for treatment.

[0034] Based on respect for medical professional judgment, the goal is to optimize the patient experience and overall treatment efficiency to the greatest extent possible. This is because by comprehensively considering departmental workload, diagnostic possibilities, and patient travel costs, the best balance is achieved between the utilization of medical resources and the physical exertion of patients. This allows each patient to receive the most efficient sequential medical services within the shortest possible physical travel range, thereby reducing the patient's stay in the hospital and physical exertion in the overall process.

[0035] In another preferred embodiment, if two or more screening departments have the same route selection score and both are the maximum value, then the confidence level of the screening department is obtained, and the screening department with the highest confidence level is selected as the next initial department.

[0036] In another preferred embodiment, the specific process of storing the signed and associated encrypted patient data unit to the central data repository in step S6 is as follows: Extract the confirmed department code and corresponding disease classification code from the treatment feedback, convert the department code into a standard department identifier, and convert the disease classification code into a standard disease identifier; combine the standard department identifier and the standard disease identifier according to a predetermined label format to generate a structured confirmed department label; A digital signature algorithm is used to sign the confirmed department label and the encrypted patient data unit to generate a digital signature identifier; a logical binding relationship is established between the digital signature identifier and the encrypted patient data unit; the logically bound encrypted patient data unit and the corresponding digital signature identifier are stored together in a designated data partition of the central data repository.

[0037] The department code and disease classification code are obtained from the treatment feedback, and the corresponding department identifier and disease identifier are extracted. The department identifier and disease identifier are combined according to a predetermined format to generate a confirmed department label. The confirmed department label is logically associated with the corresponding blood test data, and the logically associated blood test data and confirmed department label are stored together in the central data repository.

[0038] When the system receives confirmation of attribution and treatment feedback from the physician's terminal, it first parses the department code and disease classification code contained in the feedback information; then, according to the preset mapping rules, these medical standard codes are converted into identifiers used internally by the system; next, these identifiers are combined according to the established combination rules to generate structured confirmed department labels; then, a unique identification code is assigned to the original blood test data, and a mapping relationship is established between the identification code and the generated confirmed department labels to realize the logical association between data and labels; finally, the blood test data and confirmed department labels that have completed the logical association are stored as a complete data record in the designated storage area of ​​the central data repository.

[0039] By organically integrating the professional diagnostic conclusions of clinicians with raw test data, and generating structured labels with clear clinical semantics that are inextricably linked to the raw data, the originally singular test values ​​are transformed into high-value medical data records containing complete diagnostic information. This not only provides accurate classification criteria for subsequent data retrieval and access, but more importantly, it forms high-quality data assets that can be directly used for medical analysis and scientific research. This ensures that every piece of data in the central data repository has clear clinical significance and application value, providing a solid data foundation for medical big data analysis, clinical research, and intelligent diagnostic assistance. Ultimately, this achieves a major transformation of medical data from simple information storage to knowledge-based management.

[0040] In another preferred embodiment, the specific process of generating the dual-index encrypted storage structure in step S6 is as follows: The department identifier and disease identifier contained in the confirmed department label are parsed, and the parsed identifier combination is used as the first key value. The mapping relationship between the first key value and the storage address of the encrypted patient data unit in the central data repository is established to form the first key value index. Extract the test item identifier and measurement value from the original blood collection test text data corresponding to the encrypted test data block, generate a feature hash value based on the extracted test features, use the feature hash value as the second key value, establish a mapping relationship between the second key value and the storage address of the encrypted patient data unit, and form a second feature index; associate the first key value index and the second feature index together with the same encrypted patient data unit, and establish a complete dual-index encrypted storage structure in the central data repository.

[0041] Based on the parsing of the departmental and disease identifiers contained in the confirmed departmental labels, the specific identifiers representing the department and disease are separated by analyzing the structural composition of the labels. Then, corresponding metadata entries are generated based on the parsing results. These entries contain the key identifier information and related attributes obtained from the parsing. Next, the blood test data is encrypted using cryptographic algorithms. A specific encryption transformation method converts the original data into encrypted data blocks that cannot be directly read. Simultaneously, this encrypted data block is logically associated with the previously generated metadata entries using a cryptographic association identifier. This association is based on a unique identifier generated by a cryptographic hash function to achieve a reliable correspondence between the data block and the metadata. The associated encrypted data block and its metadata entries are then stored in a specially designated area of ​​the central data repository. This area has specific access control and security protection mechanisms. Finally, using the identifiers contained in the confirmed departmental labels as keys, an index structure based on key-value mapping is constructed. By establishing the correspondence between key values ​​and storage locations, the system can quickly locate the corresponding encrypted data blocks based on the department and disease identifiers, thus forming an ordered set of blood test data logically organized by department and disease identifiers.

[0042] This approach enables intelligent organization and management of medical data while ensuring data security. Cryptographic encryption technology safeguards patient privacy data and prevents the leakage of sensitive medical information. Meanwhile, the association mechanism between metadata and encrypted data maintains data availability and integrity. The key-value mapping-based index structure greatly improves the efficiency of data retrieval and access, allowing massive amounts of medical data to be organized and managed in an orderly manner according to clinical diagnostic characteristics. This provides a high-quality data foundation for subsequent data analysis and knowledge discovery, ultimately maximizing the value of medical data in a secure and controlled environment.

[0043] In another preferred embodiment, step S6 further includes sending the data task unit, the corresponding screening department sequence, the routing status information, and the confirmed department label back to the preset department prediction model as feedback data; The measured values ​​and test item identifiers of the blood collection and testing data are extracted from the feedback data and converted into numerical feature vectors as input features. At the same time, the diagnostic department label is converted into a unique hot code form as the target output. The predicted output of the preset department prediction model is obtained through forward propagation calculation, and the error value between the predicted output and the target output is calculated using the cross-entropy loss function. The error value is propagated back along the network structure to calculate the error signal of each hidden layer neuron node. The gradient value of each connection weight in the network is calculated based on the error signal. The value of each connection weight is updated using the stochastic gradient descent algorithm. The updated connection weight parameters are deployed to the preset department prediction model to complete the correction of the preset department prediction model.

[0044] The complete data task unit, the corresponding screening department sequence, routing status information, and the finally determined confirmed department label are fed back as a feedback dataset to the preset department prediction model. First, the measurement values ​​and test item identifiers of the blood sampling test data are extracted from the feedback data. Through a numerical conversion process, different types of test data are transformed into a unified numerical feature representation, forming a machine-processable input feature vector. Simultaneously, the confirmed department label is converted into a discretized target output representation using one-hot encoding. Then, through a forward propagation calculation process, the input feature vector is passed and transformed layer by layer in the neural network of the department prediction model, ultimately obtaining... The model predicts the output of each department; then, the cross-entropy loss function is used to calculate the difference between the model's predicted output and the actual target output, obtaining a quantified error value; subsequently, this error value is propagated back along the network structure, and the error contribution of each hidden layer neuron node is calculated sequentially. Based on these error signals, the adjustment direction and magnitude of each connection weight in the network are calculated, generating the corresponding gradient value; finally, the stochastic gradient descent algorithm is used to iteratively update each connection weight in the network according to the calculated gradient value, and the updated weight parameters are redeployed into the department prediction model, thereby completing the fine adjustment and optimization of the model parameters.

[0045] By feeding the complete decision data generated during actual diagnosis and treatment into the prediction model as training samples, the model can continuously learn new knowledge patterns and empirical rules from real medical scenarios, thereby gradually improving the accuracy and reliability of its triage prediction. This closed-loop learning mechanism not only reduces the model's dependence on the initial training data, but also provides continuous optimization technical support for achieving precision medicine and intelligent triage, laying a solid foundation for building a smart medical system with self-learning capabilities.

[0046] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A medical data classification storage method for blood sampling tests, characterized by, The method comprises the following steps: S1, collecting the identity information of the patient and the blood test text data, determining a screening data chain according to the blood test text data, and marking the first screening department in the screening data chain as an initial screening department; S2, constructing an identity data chain according to the identity information, and performing encryption processing on the identity data chain and the blood test text data by using an encryption algorithm to generate an encrypted identity data chain and an encrypted test data block; S3, logically associating the encrypted identity data chain with the encrypted test data block to obtain an encrypted patient data unit, encapsulating the encrypted patient data unit and the screening data chain into a data task unit, and routing the data task unit to a cache database corresponding to the initial screening department; S4, receiving the treatment feedback returned by the initial screening department, and if the treatment feedback is a negative attribution, routing the data task unit to a cache database corresponding to a next screening department and updating the routing state information in the data task unit; S5, when obtaining a treatment feedback of confirmed attribution from any screening department, generating a diagnosis department label and signing the encrypted patient data unit; S6, storing the signed and associated encrypted patient data unit to a central data storage, constructing a first key-value index based on the diagnosis department label, and constructing a second feature index based on the encrypted test data block to generate a double-index encrypted storage structure.

2. The medical data classified storage method for blood sampling test according to claim 1, wherein, In S1, the specific acquisition process of the screening data chain is as follows: combining the test item identifier and the measurement value in the blood test data to form a standardized input vector; inputting the standardized input vector into a neural network classifier of a preset department prediction model, the neural network classifier outputting a set of class codes representing different departments and a confidence value corresponding to each class code; sorting the class codes in descending order of the confidence values, mapping the sorted class code sequence to a corresponding department identifier sequence to generate a screening data chain.

3. The medical data classified storage method for blood sampling test according to claim 1, wherein, In S2, the specific process of generating the encrypted identity data chain and the encrypted test data block is as follows: extracting identity elements including a patient identifier, a name code, a gender code and age data from the identity information, connecting the identity elements in a predetermined order with the patient identifier as the priority to form an identity data chain; performing encryption processing on the identity data chain by using a first encryption algorithm based on an asymmetric encryption algorithm to generate an encrypted identity data chain; dividing the blood test text data into data blocks according to the test items, and performing encryption on each data block containing the test item identifier and the measurement value by using a second encryption algorithm based on a symmetric encryption algorithm to generate a corresponding encrypted data block; combining the encrypted data blocks in the order of the test items to form a complete encrypted test data block.

4. The medical data classified storage method for blood sampling test according to claim 1, wherein, In S3, the specific generation process of the data task unit is as follows: The encrypted identity data chain and the encrypted test data block are generated with a unique association identifier by a cryptographic hash function, a logical mapping relationship between the two is established based on the association identifier, an association data group is formed, the association data group is data-encapsulated with a screening data chain containing a screening department sequence, and a timestamp and a task status identifier are added to the data task unit during the encapsulation process to generate a data task unit.

5. The medical data classified storage method for blood sampling test according to claim 1, wherein, In the S4, the specific process of routing the data task unit to a specific cache database corresponding to the next screening department is as follows: An inventory value of the real-time data task unit corresponding to each screening department and a confidence score thereof are obtained; a physical path distance value between the initial department and any screening department is measured to obtain a physical path distance measured value; The data task unit inventory value, the confidence score and the physical path distance measured value are combined and weighted to obtain a routing optimization score of each screening department; The routing optimization scores of the screening departments are compared, and the screening department with the maximum score is selected as the next initial department, and the data task unit is routed to the cache database corresponding to the selected next initial department.

6. The medical data classified storage method for blood sampling test according to claim 5, wherein, If the routing optimization scores of two or more screening departments are equal and are the maximum, the confidence scores of the screening departments are obtained, and the screening department with the maximum confidence score is selected as the next initial department.

7. The medical data classified storage method for blood sampling test according to claim 1, wherein, In the S6, the specific process of storing the encrypted patient data unit associated with the signature to the central data storage is as follows: The confirmed department code and the corresponding disease classification code are extracted from the treatment feedback, the department code is converted into a standard department identifier, and the disease classification code is converted into a standard disease identifier; the standard department identifier and the standard disease identifier are combined according to a predetermined label format to generate a structured diagnosis department label; A digital signature algorithm is used to sign the diagnosis department label and the encrypted patient data unit to generate a digital signature identifier; The digital signature identifier and the encrypted patient data unit are logically bound; The encrypted patient data unit and the corresponding digital signature identifier that have completed logical binding are stored together in a specified data partition of the central data storage.

8. The medical data classified storage method for blood sampling test according to claim 7, wherein, In the S6, the specific process of generating a double-index encrypted storage structure is as follows: The department identifier and the disease identifier contained in the diagnosis department label are parsed, and the obtained identifier combination is used as a first key value to establish a mapping relationship between the first key value and the storage address of the encrypted patient data unit in the central data storage, forming a first key value index; The test item identifier and the measurement value are extracted from the original blood sampling test text data corresponding to the encrypted test data block, a feature hash value is generated based on the extracted test features, the feature hash value is used as a second key value, a mapping relationship between the second key value and the storage address of the encrypted patient data unit is established, forming a second feature index; the first key value index and the second feature index are associated to the same encrypted patient data unit, and a complete double-index encrypted storage structure is established in the central data storage.

9. The medical data classified storage method for blood sampling test according to claim 2, wherein, The S6 further comprises feeding back the data task unit, screening data chain, routing state information and diagnosis department label to the preset department prediction model as feedback data; The measurement value and test item identification of blood sampling test data are extracted from the feedback data, converted into a numerical feature vector as input features, and the diagnosis department label is converted into a one-hot encoding form as a target output; The prediction output of the preset department prediction model is obtained by forward propagation calculation, and the error value between the prediction output and the target output is calculated by using the cross-entropy loss function; The error value is back-propagated along the network structure, the error signals of each hidden layer neuron node are calculated, the gradient values of each connection weight in the network are calculated according to the error signals, the random gradient descent algorithm is used to update the numerical values of each connection weight, the updated connection weight parameters are deployed to the preset department prediction model, and the correction of the preset department prediction model is completed.

Images