Intelligent anesthesia record sheet automatic generation system

Abstract

The invention relates to the technical field of medical report generation, in particular to an intelligent anesthesia record sheet automatic generation system. According to the system, patient information is automatically recognized through the OCR unit, and real-time collection and structured storage of physical sign data, medication records and surgical operation are achieved in combination with a medical equipment interface and a voice recognition technology; time sequence data analysis and multi-source data fusion modeling are adopted to construct a medical event causal relationship graph so as to improve the accuracy of anomaly recognition and patient sign evaluation; based on a time axis construction unit and a structured template engine, an anesthesia record list is automatically analyzed and filled, and the manual input workload is greatly reduced; and finally, in combination with block chain evidence storage and a multi-level encryption technology, the security and non-tampering property of the anesthesia record list are ensured, the reliability of medical data management is improved, the recording, monitoring and storage efficiency of the anesthesia process is comprehensively optimized, and the medical quality and the patient safety are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical report generation, and particularly to an intelligent anesthesia record sheet automatic generation system. Background Art

[0002] With the rapid development of medical informatization, digital recording of the surgical anesthesia process has become an important means to improve medical quality and patient safety. Traditional anesthesia record sheets mainly rely on manual filling. Medical staff need to manually enter the patient's vital signs data, medication conditions, and key surgical nodes, which not only increases the work burden but also easily leads to data delay, omission, or error, affecting the accurate assessment of anesthesia quality.

[0003] Although existing electronic anesthesia record systems can achieve partial automation, there are still problems such as incomplete data collection, low information integration, and untimely anomaly recognition. For example, some systems can only collect vital signs data through monitoring devices and cannot effectively integrate multi-source data such as surgical information and voice input, resulting in insufficient integrity and real-time performance of the record sheet. In addition, existing systems lack effective protection in terms of data security and may face the risk of tampering, which is not conducive to postoperative data traceability and the handling of medical disputes.

[0004] Therefore, an intelligent anesthesia record sheet automatic generation system is proposed. Summary of the Invention

[0005] The present invention provides an intelligent anesthesia record sheet automatic generation system. By using OCR to identify patient information, automatically collect vital signs data, medication records, and surgical operation information, and combining time series analysis and causal relationship modeling, accurate anomaly detection and patient vital sign assessment are realized; using time axis construction and template engine to automatically fill the record sheet, reducing manual operation and improving efficiency; combining blockchain evidence storage and multi-level encryption to ensure data security and immutability.

[0006] To achieve the above object, the present invention provides the following technical solutions: An intelligent anesthesia record sheet automatic generation system, comprising: An information acquisition module, configured to identify the outpatient number of the patient through the OCR unit and obtain the patient's basic information through outpatient number indexing; A data collection module, configured to obtain the patient's vital signs data through the medical device interface, synchronize the medication record and surgical operation node data through the surgical information system interface, and extract the voice input of medical staff through the voice command recognition unit and convert it into structured text data; An intelligent processing module, which is used to extract features from the patient's vital sign data through a time series data analysis unit, perform multi-source data fusion modeling through a medical event extraction unit, construct a causal relationship map of medical events to identify medical events, and then monitor the patient's vital sign data through an anomaly detection unit to evaluate the patient's vital sign status in real time; An automatic generation module, which is used to establish an event timeline according to the time series through a timeline construction unit, and parse the event timeline through a structured template engine, and automatically fill the parsing result into the anesthesia record sheet; A secure storage module, which is used to record the hash value of the anesthesia record sheet through a blockchain certification unit, encrypt the anesthesia record sheet in combination with a multi-level encryption protocol, and store the encrypted anesthesia record sheet in a medical database.

[0007] Further, the OCR unit is implemented by a convolutional neural network, and the obtained patient basic information includes: Identify the outpatient number of the patient through the OCR unit, retrieve the hospital database with the outpatient number as the index, extract the patient's basic information and associate the relevant medical record data.

[0008] Further, the time series data analysis unit is implemented by an LSTM network, and the feature extraction of the patient's vital sign data includes: Obtain historical vital sign data from the medical device interface and preprocess the data; Construct a sliding window according to a preset time step to obtain time series data; Input the time series data into the LSTM network, and extract the patient's time series feature data through the forget gate, input gate and output gate.

[0009] Further, the medical event extraction unit is implemented by a GNN, and the construction of the causal relationship map of medical events includes: Obtain time series feature data, medication records, surgical operation node data and structured text data, and align the data based on the time stamp; Perform medical event modeling through a GNN, define the nodes and edges of the graph, and calculate the node features of each event node through a neighborhood aggregation mechanism; Based on the node features calculated by the GNN, adopt a causal reasoning method to extract the causal relationship between medical events, and form a causal relationship map of medical events.

[0010] Further, the evaluation method of the patient's vital sign status includes: Obtain historical vital sign data from the medical device interface and preprocess the data; Calculate the mean and standard deviation based on the historical vital sign data as the statistical features of the baseline model; Calculate the sign deviation based on the real-time patient's vital sign data, and evaluate the patient's vital sign status according to the deviation.

[0011] Further, the calculation formula for the physical sign deviation degree is as follows: ; wherein, represents the physical sign deviation degree, represents the th physiological parameter of the patient's physical sign data, represents the mean value of the th physiological parameter of the baseline model, represents the standard deviation of the th physiological parameter of the baseline model, represents the weight of the th physiological parameter of the baseline model, represents the total number of items of the physiological parameters of the patient's physical sign data.

[0012] Further, the establishment of the event time axis includes: Obtaining event nodes and causal relationships from the medical event causal relationship graph, and attaching time stamps to each event node; Sorting all event nodes according to the time stamps to form an initial time series; Establishing a structured time axis including time stamps, event nodes, and causal relationships, and storing it in a time series database.

[0013] Further, the parsing process of the event time axis includes: Loading the template corresponding to the surgical type from the template library according to the predefined anesthesia record sheet template; Establishing a field mapping relationship according to the event nodes and parsing the causal relationships; Filling the parsed physical sign data, medication records, and surgical operation nodes into the anesthesia record sheet template item by item in chronological order.

[0014] Further, the encryption process of the anesthesia record sheet includes: Performing a hash calculation on all data of the anesthesia record sheet through an encryption hash algorithm to generate a unique hash value; Encrypting the data of the anesthesia record sheet by using a multi-level encryption protocol, wherein the first layer encrypts the data by using a symmetric encryption algorithm, and the second layer encrypts the encryption key by using an asymmetric encryption algorithm; Storing the generated hash value through the blockchain evidence storage unit, and storing the encrypted anesthesia record sheet data in the hospital database.

[0015] The beneficial effects of the present invention are: 1. The construction of a medical event causal relationship graph through multi-source data fusion modeling can integrate multi-dimensional information such as patient vital sign data, medication records, surgical operation node data, and medical staff voice input, improving the accuracy and timeliness of medical event recognition. Through time-series data analysis and causal reasoning, the system can automatically identify key medical events and their correlations, enabling accurate event tracing and decision support, thereby enhancing the monitoring accuracy during anesthesia, reducing human errors, optimizing the efficiency of anesthesia form generation, and providing a scientific basis for rapid response to intraoperative abnormalities.

[0016] 2. Evaluating the patient's vital sign status based on the deviation degree can achieve real-time and accurate monitoring of the patient's health condition. By calculating the deviation degree of the patient's vital sign data relative to the baseline model, the system can quickly identify abnormal changes. This method reduces the reliance on manual judgment, improves the sensitivity and accuracy of abnormal recognition, thereby optimizing the intraoperative monitoring efficiency, reducing medical risks, and ensuring patient safety.

[0017] 3. Automatically identify the patient's outpatient number through OCR technology, and synchronously obtain the patient's vital signs, medication, and surgical records by combining medical device data, voice recognition, and surgical information to comprehensively collect multi-source data. Through time-series data analysis and the construction of a medical event causal relationship graph, the system can monitor the changes in the patient's vital signs in real time and evaluate their health status. At the same time, an anesthesia record sheet is automatically generated and the security and immutability of the data are ensured through blockchain and encryption technologies. This system effectively reduces errors and delays in manual operations and improves the intelligent level of anesthesia form generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the drawings: Figure 1 is a schematic structural diagram of an intelligent anesthesia record sheet automatic generation system provided by the present invention; Figure 2 is a flowchart of the implementation of an intelligent anesthesia record sheet automatic generation system provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The following describes the preferred embodiments of the present invention with reference to the drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention. Embodiment 1

[0020] An intelligent anesthesia record sheet automatic generation system, as Figure 1 shown, includes: An information acquisition module, configured to identify the patient's outpatient number through the OCR unit and obtain the patient's basic information through outpatient number indexing; Furthermore, the OCR unit is implemented by a convolutional neural network, and the basic patient information obtained includes: The outpatient number of the patient is recognized by the OCR unit, the hospital database is retrieved using the outpatient number as an index, and the basic patient information is extracted and associated with relevant medical record data.

[0021] Specifically, the OCR unit is implemented by a convolutional neural network, where the convolutional neural network can be any network model capable of digital recognition, such as VGG-16, ResNet, etc. In this embodiment, VGG-16 is preferably used.

[0022] Implementing OCR to recognize the patient's outpatient number through a convolutional neural network can effectively improve the accuracy and robustness of character recognition, and reduce misrecognition problems caused by unclear handwritten or printed characters. At the same time, combined with the hospital database indexing mechanism, it can quickly and accurately retrieve the basic patient information and relevant medical record data, ensuring the timeliness and integrity of the data.

[0023] The data acquisition module is used to obtain the patient's vital signs data through the medical device interface, synchronize the medication record and surgical operation node data through the surgical information system interface, extract the voice input of medical staff through the voice command recognition unit, and convert it into structured text data; Through the interface of medical devices (such as multi-parameter monitors, anesthesia machines, electrocardiographs, etc.), it is docked with the hospital information system using the HL7 (Health Level 7) protocol to collect the patient's vital signs data in real time (including but not limited to heart rate, blood pressure, blood oxygen saturation, respiratory rate); through the interface of the surgical information system (OIS, Operation Information System), obtain the medication time, drug name, dose during anesthesia, and key surgical time points (such as skin incision time, suture time, etc.), and synchronize them to the system; collect the voice commands of medical staff, perform speech-to-text processing using an end-to-end deep learning model (such as Wav2Vec2.0 or DeepSpeech), and combine natural language processing (NLP) technology to perform semantic analysis on the text, extract keywords and convert them into standardized structured data (such as operation instructions, medication time, dose, etc.).

[0024] The intelligent processing module is used to extract features from the patient's vital signs data through the time series data analysis unit, perform multi-source data fusion modeling through the medical event extraction unit, construct a medical event causal relationship graph to identify medical events, and then monitor the patient's vital signs data through the anomaly detection unit to evaluate the patient's vital signs status in real time; Furthermore, the time series data analysis unit is implemented by an LSTM network, and the feature extraction of the patient's vital signs data includes: Obtain historical vital sign data from a medical device interface and preprocess the data; Construct a sliding window according to a preset time step to obtain time series data; Input the time series data into an LSTM network, and extract the patient's time series feature data through the forget gate, input gate, and output gate.

[0025] Specifically, obtain the patient's historical vital sign data through the medical device interface, including heart rate, blood oxygen saturation, blood pressure, and respiratory rate, and perform data cleaning on the historical vital sign data, including missing value interpolation filling, outlier detection, and noise filtering, to ensure data quality, and then perform Min-Max normalization to make the data have a unified scale when input into the LSTM network; set the time window size (set to 10s in this embodiment), and slice the time series data in the form of a sliding window. The window interval can be set according to the task requirements (in this embodiment, a 60s sliding window is used), generate time series input data through the sliding window, and each window is used as an independent input sample; finally, input the time series data into the LSTM network, and extract the patient's time series feature data through the forget gate, input gate, and output gate.

[0026] Using an LSTM network to extract features from the patient's vital sign data can make full use of time series information and accurately capture the changing trend of the patient's vital signs. Constructing time series inputs through sliding windows enables the model to have short-term and long-term dependence characteristics and improves the feature expression ability. At the same time, the gating mechanism of the LSTM effectively screens key information, reduces noise interference, and provides high-quality time series feature data for medical event recognition and anomaly detection.

[0027] Furthermore, the medical event extraction unit is implemented by a GNN, and constructing a medical event causal relationship graph includes: Obtain time series feature data, medication record and surgical operation node data, and structured text data, and align the data based on timestamps; Perform medical event modeling through a GNN, define the nodes and edges of the graph, and calculate the node features of each event node through the neighborhood aggregation mechanism; Based on the node features calculated by the GNN, use a causal inference method to extract the causal relationships between medical events and form a medical event causal relationship graph.

[0028] Specifically, based on timestamps, the patient's vital sign data, medication records, surgical operation nodes, and voice instructions are synchronized and aligned to ensure that data from different sources match on the same time axis; a GNN is constructed for medical event modeling, defining nodes and edges, where the nodes represent different medical events, including medication, vital sign fluctuations, surgical steps, medical staff instructions, etc., and the edges connect different nodes, representing the causal relationships between events. The GNN calculates the features of each event node through a neighborhood aggregation mechanism, and the calculation of the neighborhood aggregation mechanism can be expressed as: ; Among them, represents the feature of event , represents the event connected to it, and represent the parameters of the GNN, represents the non-linear activation function; based on the node features calculated by the GNN, a causal reasoning method (in this embodiment, the Structural Causal Model is adopted) is used to extract the causal relationships between medical events, forming a causal relationship map of medical events.

[0029] By using the GNN to uniformly model multi-source medical data, a causal relationship map of medical events is constructed. Through timestamp alignment, it is ensured that the data is accurately associated according to the surgical process, and the neighborhood aggregation mechanism is used to calculate the event node features, extracting the causal relationships between key medical events. Compared with traditional rule matching or statistical analysis methods, this method can automatically learn the potential associations of complex medical events, improve the accuracy of causal relationship reasoning, and thus enhance the intelligent filling ability of the anesthesia record sheet, optimize surgical decision support, and improve the efficiency and accuracy of clinical information management.

[0030] Furthermore, the method for evaluating the patient's vital sign status includes: Obtain historical vital sign data from the medical device interface and preprocess the data; Calculate the mean and standard deviation based on the historical vital sign data as the statistical features of the baseline model; Calculate the sign deviation degree according to the real-time patient's vital sign data, and evaluate the patient's vital sign status according to the deviation degree.

[0031] Furthermore, the calculation formula for the sign deviation degree is: ; Among them, represents the sign deviation degree, represents the th physiological parameter of the patient's vital sign data, represents the mean of the th physiological parameter of the baseline model, Denote the standard deviation of the th physiological parameter of the baseline model, Denote the weight of the th physiological parameter of the baseline model, Denote the total number of physiological parameters of the patient's physical sign data.

[0032] A baseline model is established using historical physical sign data, and the patient's physical sign status is monitored in real time through deviation calculation. Compared with traditional manual observation or single-threshold alarm mechanisms, this method can dynamically adjust the warning criteria, improve the accuracy of anomaly detection, reduce false alarms and missed alarms, and enhance patient safety during surgery.

[0033] An automatic generation module for establishing an event timeline in time series through a timeline construction unit and parsing the event timeline through a structured template engine, and automatically filling the parsing results into the anesthesia record sheet; Furthermore, the establishment of the event timeline includes: Obtaining event nodes and causal relationships from the medical event causal relationship graph, and attaching timestamps to each event node; Sorting all event nodes according to the timestamps to form an initial time series; Establishing a structured timeline containing timestamps, event nodes, and causal relationships, and storing it in a time series database.

[0034] Extract all event nodes (such as medication, vital sign fluctuations, surgical procedures, medical staff instructions, etc.) from the medical event causal relationship graph, the causal relationships between events (such as a certain medication causing a sign change, a certain surgical operation triggering a physiological reaction, etc.), and attach timestamps to the event nodes based on the occurrence time of each event. Use a time synchronization mechanism to align the times of different data sources to ensure that all timestamps are consistent; sort all event nodes in ascending order of timestamps to form an initial time series, ensuring that events are arranged in the actual order of occurrence; establish a structured timeline containing timestamps, event nodes, and causal relationships (in JSON file format), and store it in a time series database.

[0035] By establishing an event timeline based on the medical event causal relationship graph, key events during the surgical process can be accurately captured and sorted according to timestamps, making the order of occurrence and causal relationships of events clearer. This method effectively integrates the patient's physical sign data, medication records, and surgical operation information, making the generation of anesthesia records more systematic and automated, reducing the errors of manual records, and improving the timeliness and traceability of data.

[0036] Furthermore, the parsing process of the event timeline includes: Loading the template corresponding to the surgical type from the template library according to the predefined anesthesia record sheet template; Establish a field mapping relationship based on event nodes and parse the causal relationship; Fill the parsed physical sign data, medication records, and surgical operation nodes into the anesthesia record sheet template item by item in chronological order.

[0037] Specifically, pre-define anesthesia record sheet templates for different surgical types in the system and store them in the template library. According to the specific surgical type, load the matching anesthesia record sheet template from the template library to ensure the standardization of the record format; parse each event node in the event timeline, establish the mapping relationship between the event nodes and the anesthesia record sheet fields to ensure that data from different sources can be correctly filled into the corresponding fields; traverse the event nodes in the event timeline in chronological order, extract relevant data, and fill the parsed patient physical sign data, medication records, and surgical operation nodes into the anesthesia record sheet template item by item in chronological order.

[0038] The parsing process of the event timeline can achieve the automatic filling and standardized recording of the anesthesia record sheet, improving the efficiency and accuracy of data input. Through the pre-defined template library, the system can adapt to different surgical types to ensure the standardization and unity of the record sheet format. Based on the field mapping and causal relationship parsing of event nodes, it can accurately match the key intraoperative events, ensure the integrity and logic of the data, and reduce manual input errors. At the same time, filling the data in chronological order makes the recording of the anesthesia process clearer and traceable, which is helpful for postoperative analysis and medical quality evaluation.

[0039] A secure storage module is used to record the hash value of the anesthesia record sheet through the blockchain evidence storage unit, encrypt the anesthesia record sheet in combination with a multi-level encryption protocol, and store the encrypted anesthesia record sheet in the medical database.

[0040] Furthermore, the encryption process of the anesthesia record sheet includes: Perform a hash calculation on all the data of the anesthesia record sheet through an encryption hash algorithm to generate a unique hash value; Use a multi-level encryption protocol to encrypt the data of the anesthesia record sheet. Among them, the first layer uses a symmetric encryption algorithm to encrypt the data, and the second layer uses an asymmetric encryption algorithm to encrypt the encryption key; Store the generated hash value through the blockchain evidence storage unit, and store the encrypted anesthesia record sheet data in the hospital database.

[0041] Specifically, all data in the anesthesia record sheet is extracted, and the content of the anesthesia record sheet is hashed using a secure hash algorithm (such as SHA-256) to generate a unique hash value to ensure the integrity and immutability of the data; the Advanced Encryption Standard (AES) symmetric encryption algorithm is used to perform the first layer of encryption on the anesthesia record sheet to ensure the security during data storage. A pair of public and private keys is generated through elliptic curve encryption, and the public key is used to perform the second layer of encryption on the symmetric encryption key so that it can only be decrypted by the corresponding private key, enhancing security.

[0042] Generating a unique hash value through an encryption hash algorithm can be used to verify whether the data has been tampered with; adopting a multi-level encryption protocol, combining symmetric encryption and asymmetric encryption, enables the data to have high security during transmission and storage, while ensuring that only authorized parties can decrypt and access it; using blockchain evidence storage technology to record the hash value to ensure the traceability and anti-tampering ability of the data, thereby improving the security management level of the anesthesia record sheet and meeting the requirements of medical data privacy protection. Embodiment 2

[0043] A certain top-three hospital aims to improve the processing efficiency of medical staff for anesthesia record sheets and intends to adopt an intelligent anesthesia record sheet automatic generation system proposed by the present invention. The implementation process of this system is as Figure 2 shown and includes: The outpatient number of the patient is identified through the OCR unit, and the basic information of the patient is obtained through outpatient number indexing. The vital signs data of the patient is obtained through the medical device interface, the medication record and surgical operation node data are synchronized through the surgical information system interface, and the voice input of medical staff is extracted through the voice command recognition unit and converted into structured text data. Feature extraction is performed on the patient's vital signs data through the time series data analysis unit, multi-source data fusion modeling is carried out through the medical event extraction unit to construct a medical event causal relationship map to identify medical events, and then the patient's vital signs data is monitored through the anomaly detection unit to evaluate the patient's vital signs status in real time. An event timeline is established in time series through the timeline construction unit, and the event timeline is parsed through the structured template engine, and the parsing result is automatically filled into the anesthesia record sheet. The hash value of the anesthesia record sheet is recorded through the blockchain evidence storage unit, the anesthesia record sheet is encrypted in combination with the multi-level encryption protocol, and the encrypted anesthesia record sheet is stored in the medical database.

[0044] To verify the index comparison between the technical group of the present invention and the traditional technical group, an anesthesia record process of a simulated laparoscopic cholecystectomy was carried out. The process of the traditional technical group is as follows: Vital signs entry: The anesthesia nurse manually enters the blood pressure value displayed on the monitor every 5 minutes. Medication Record: The anesthesiologist orally states the type and dosage of the administered drugs, and the nurse selects the drugs in the system and fills in the dosage. Time Node Record: The start time of the operation, intubation time, etc. need to be marked by manually clicking the system button. Abnormality Identification: After the anesthesiologist discovers a decrease in the blood pressure curve, manually calculate the decrease amplitude. Decision Support: Consult the paper drug manual to find the usage specifications of vasopressors. Record Supplement: Supplement the event timeline after the operation.

[0045] The comparison data is shown in Table 1. From the comparison data in the table, it can be seen that the technical group of the present invention is significantly superior to the traditional technical group in multiple key indicators. The event record delay is shortened from 3 - 5 minutes in the traditional technical group to 2 seconds, the medication record error rate is reduced from 2% to 0.3%, and the abnormal response time is shortened from 45 - 60 seconds to 3 seconds, greatly improving the processing efficiency of medical staff. In addition, the technical group of the present invention ensures the security and anti-tampering ability of the data through blockchain evidence storage and multiple encryption technologies, while the traditional technical group lacks such protection measures. These advantages make the technical group of the present invention have a significant improvement in the automation, accuracy, and security of anesthesia records.

[0046] Table 1 Comparison Data between the Traditional Technical Group and the Technical Group of the Present Invention

[0047] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent anesthesia record sheet automatic generation system, characterized in that, Including: An information acquisition module, which is used to identify the outpatient number of the patient through the OCR unit and obtain the basic information of the patient through outpatient number indexing; A data acquisition module, which is used to obtain the patient's vital sign data through the medical device interface, synchronize the medication record and surgical operation node data through the surgical information system interface, and extract the voice input of medical staff through the voice command recognition unit and convert it into structured text data; An intelligent processing module, which is used to extract features from the patient's vital sign data through the time series data analysis unit, perform multi-source data fusion modeling through the medical event extraction unit, construct a causal relationship map of medical events to identify medical events, and then monitor the patient's vital sign data through the anomaly detection unit to evaluate the patient's vital sign status in real time; An automatic generation module, which is used to establish an event timeline according to the time series through the timeline construction unit, and parse the event timeline through the structured template engine, and automatically fill the parsing result into the anesthesia record sheet; A secure storage module, which is used to record the hash value of the anesthesia record sheet through the blockchain evidence storage unit, encrypt the anesthesia record sheet in combination with the multi-level encryption protocol, and store the encrypted anesthesia record sheet in the medical database.

2. The intelligent anesthesia record sheet automatic generation system according to claim 1, wherein The OCR unit is implemented by a convolutional neural network. Obtaining the basic information of the patient includes: Identifying the outpatient number of the patient through the OCR unit, retrieving the hospital database with the outpatient number as the index, extracting the basic information of the patient and associating relevant medical record data.

3. The intelligent anesthesia record sheet automatic generation system according to claim 1, wherein The time series data analysis unit is implemented by an LSTM network. Extracting features from the patient's vital sign data includes: Obtaining historical vital sign data from the medical device interface and preprocessing the data; Constructing a sliding window according to the preset time step to obtain time series data; Inputting the time series data into the LSTM network, and extracting the time series feature data of the patient through the forget gate, input gate and output gate.

4. An intelligent anesthesia record sheet automatic generation system according to claim 1, characterized in that, The medical event extraction unit is implemented by a GNN. Constructing a causal relationship map of medical events includes: Obtaining time series feature data, medication records, surgical operation node data and structured text data, and aligning the data based on timestamps; Performing medical event modeling through the GNN, defining the nodes and edges of the graph, and calculating the node features of each event node through the neighborhood aggregation mechanism; Based on the node features calculated by the GNN, using the causal inference method to extract the causal relationship between medical events, forming a causal relationship map of medical events.

5. An intelligent anesthesia record sheet automatic generation system according to claim 1, characterized in that, The evaluation method of the patient's vital sign status includes: Obtaining historical vital sign data from the medical device interface and preprocessing the data; Calculating the mean and standard deviation based on the historical vital sign data as the statistical features of the baseline model; Calculating the sign deviation based on the real-time vital sign data of the patient, and evaluating the patient's vital sign status according to the deviation.

6. The intelligent anesthesia record sheet automatic generation system according to claim 5, characterized in that, The calculation formula of the sign deviation is: ； Among them, represents the sign deviation degree, represents the th physiological parameter of the patient's sign data, represents the mean value of the th physiological parameter of the baseline model, represents the standard deviation of the th physiological parameter of the baseline model, represents the weight of the th physiological parameter of the baseline model, represents the total number of items of the physiological parameters of the patient's sign data.

7. An intelligent anesthesia record sheet automatic generation system according to claim 1, characterized in that, The establishment of the event timeline includes: Obtaining event nodes and causal relationships from the causal relationship map of medical events, and attaching timestamps to each event node; Sorting all event nodes according to timestamps to form an initial time series; Establishing a structured timeline including timestamps, event nodes and causal relationships, and storing it in the time series database.

8. An intelligent anesthesia record sheet automatic generation system according to claim 1, characterized in that, The parsing process of the event timeline includes: Load the template corresponding to the surgical type from the template library according to the pre-defined anesthesia record sheet template; Establish field mapping relationships based on event nodes and parse the causal relationships; Fill the parsed vital sign data, medication records, and surgical operation nodes into the anesthesia record sheet template item by item in chronological order.

9. The intelligent anesthesia record sheet automatic generation system according to claim 1, characterized in that The encryption process of the anesthesia record sheet includes: Perform a hash calculation on all the data of the anesthesia record sheet through an encryption hash algorithm to generate a unique hash value; Use a multi-level encryption protocol to encrypt the data of the anesthesia record sheet. Among them, the first layer uses a symmetric encryption algorithm to encrypt the data, and the second layer uses an asymmetric encryption algorithm to encrypt the encryption key; Store the generated hash value through the blockchain evidence storage unit, and store the encrypted anesthesia record sheet data in the hospital database.

Images