Clinical nursing decision support system and method based on artificial intelligence

Abstract

The invention relates to the technical field of clinical nursing intelligent auxiliary decision making, in particular to a clinical nursing decision supporting system and method based on artificial intelligence, and the system comprises a data layer which obtains and stores patient data; the decision-making rule layer is used for establishing a patient nursing decision-making rule taking a nursing program as a framework based on the patient data in the data layer and data driving; the interoperation layer is bidirectionally connected with the data layer, the decision rule layer and the view interface layer, extracts and processes data in the data layer, generates a response result in combination with the patient data and related patient nursing decision rules, and feeds back the response result to the view interface layer; and the view interface layer is used for executing data visualization of the data layer, the decision rule layer and the interoperation layer. According to the invention, traceability and sustainable evolution of the whole process of nursing decision making are realized, and clinical applicability and man-machine cooperation efficiency are significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of clinical nursing intelligent auxiliary decision-making, and in particular to a clinical nursing decision support system and method based on artificial intelligence. Background Art

[0002] With the rapid development of medical information technology, clinical nursing decision support systems play an increasingly important role in modern medical services. Nursing decision support systems assist nursing staff in making scientific and standardized clinical decisions by integrating basic patient information, test results and vital signs data. At present, medical institutions generally adopt nursing procedures as the basic framework of clinical nursing work, including five stages: assessment, diagnosis, planning, implementation and evaluation. In practical applications, traditional nursing decision support systems are limited by the processing capacity of a single data type and cannot effectively integrate heterogeneous information such as waveform signals and natural language evaluation, resulting in one-sided decision-making basis. On the other hand, the update of static rule bases lags behind the development of clinical practice, which is not only difficult to adapt to the decision-making needs of new nursing problems, but also lacks a dynamic compensation mechanism when rules conflict. In addition, the human-computer interaction interface of existing systems often separates data collection, rule matching and execution feedback, making it difficult for nurses to trace the decision logic in a timely manner, reducing clinical trust. Summary of the invention

[0003] The purpose of the present invention is to provide an artificial intelligence-based clinical nursing decision support system and method, which realizes the full process traceability and continuous evolution of nursing decisions through multimodal data fusion, dynamic decision compensation and closed-loop self-optimization mechanism, and significantly improves clinical applicability and human-computer collaboration efficiency.

[0004] The present invention is achieved through the following technical solutions: A clinical nursing decision support system based on artificial intelligence, comprising: Data layer, which acquires and stores patient data; The decision rule layer is based on the patient data in the data layer and establishes patient care decision rules based on the nursing process framework based on data-driven; An interoperability layer, wherein the interoperability layer is bidirectionally connected to the data layer, the decision rule layer, and the view interface layer, extracts and processes data in the data layer, combines patient data and relevant patient care decision rules, generates response results, and feeds the response results back to the view interface layer; The view interface layer is used to perform data visualization of the data layer, decision rule layer, and interoperability layer.

[0005] Optionally, the specific logic of the data layer is: Connect to hospital information systems, laboratory information systems and bedside equipment to collect patient data, including: patient basic information, test results, and vital signs time series data; The nursing assessment module of the view interface layer collects patient nursing assessment data, including: patient's first nursing assessment data and daily assessment data. The assessment data is stored in JSON format and attached with nurse ID and timestamp; A nursing plan execution record chain is constructed based on patient data and patient nursing assessment data, and stored in the data layer.

[0006] Optionally, the data layer further includes: Perform field verification on the patient's basic information; Convert the units of the test results; The vital sign time series data is averaged according to a time window with a set interval to generate a standardized waveform segment; Complete preprocessing of patient data.

[0007] Optionally, the patient care decision rules based on data-driven establishment with nursing procedures as the framework are specifically: Based on the standardized waveform segments, abnormal events of the patient's vital signs are detected and represented as abnormal event features of the patient; Perform semantic analysis on JSON assessment data, extract nursing assessment keywords, and represent them as patient care semantic features; Taking the abnormal event characteristics of patients' vital signs and the semantic characteristics of patients' nursing as input, and the encoding of patients' nursing decisions as output, a multimodal nursing decision model for patients is established.

[0008] Optionally, the decision rule layer also includes a nursing problem generation and evaluation module, specifically: Matching patient care decision codes with the evidence-based knowledge base in the nursing process to generate a set of candidate patient care decision rules; According to the nursing problem type as the priority, the best candidate patient nursing decision rule is selected to be characterized as the patient nursing decision rule.

[0009] Optionally, the interoperability layer includes: An operation instruction management module receives and analyzes the operation instructions of the view interface layer, and synchronizes the operation results to the data layer through the interoperability layer; A data connection module is connected to the hospital information system, the laboratory information system and the bedside equipment, and generates a response result in combination with the patient care decision rules of the decision rule layer; The operation log storage function saves all operation instructions and data call records issued through the view interface layer in the data layer, forming a nursing operation traceability chain.

[0010] Optionally, the interoperability layer further includes: The real-time status monitoring module continuously obtains the status changes of the nursing plan execution record chain. If the patient care decision rules conflict with the current patient care assessment data, a logic verification alarm is triggered.

[0011] Optionally, the interoperability layer further includes: Dynamic decision compensation module, when the output confidence of the patient multimodal nursing decision model is lower than the set value, the patient nursing decision code is recalculated. If the recalculated result still does not meet the set value, the second best choice in the candidate patient nursing decision rule set is screened out and pushed to the view interface layer.

[0012] Optionally, the view interface layer further includes: The rule matching display module includes a first area and a second area, wherein the first area displays the best candidate patient care decision rule; the second area displays the second best candidate patient care decision rule in parallel; The manual decision feedback interface writes the second-best candidate patient care decision rules into the nursing plan execution record chain as feedback data.

[0013] A clinical nursing decision support method based on artificial intelligence, the method comprising the steps of: Connect to hospital information systems, laboratory information systems and bedside devices to collect patient data, and perform field verification, unit conversion and waveform segment standardization on patient data based on preprocessing; Obtain the patient's first and daily patient nursing assessment data through the nursing assessment module of the view interface layer, and store it in the data layer in JSON format; Extract features from patient data and JSON evaluation data, input abnormal event features of patient vital signs and patient care semantic features into the patient multimodal care decision model, and generate corresponding patient care decision codes; Based on the patient multimodal nursing decision model, the nursing problem generation and evaluation module is called to generate a set of candidate patient nursing decision rules based on the evidence-based knowledge base preset in the nursing procedure; According to the type of nursing problem as the priority, the optimal candidate patient nursing decision rule is selected to be represented as the patient nursing decision rule. The interoperability layer outputs the patient nursing decision rule as a response result and feeds it back to the view interface layer.

[0014] The technical solution of the present invention has at least the following advantages and beneficial effects: The present invention realizes the optimization of the whole process of nursing decision-making through multimodal data fusion and dynamic decision compensation mechanism. The system adopts a layered architecture design, completes the standardized processing of heterogeneous data at the data layer, and ensures the same-dimensional expression of vital sign waveforms and natural language assessment features; the decision rule layer combines data-driven modeling and evidence-based knowledge base to generate interpretable nursing decision coding; the interoperability layer has built-in real-time conflict detection and suboptimal decision compensation algorithms, and automatically triggers the recalculation process when the model confidence is insufficient to ensure the reliability of the decision; the view interface layer provides dual timeline comparison and rule matching dashboards to transform complex decision logic into visual interactive operations. More importantly, through the self-optimization module driven by clinical feedback data, the system can dynamically adjust the rule priority and generate an audit tracking report, forming a complete closed loop of "data collection-intelligent decision-making-effect evaluation-rule iteration", which significantly improves the clinical applicability and continuous evolution capability of nursing decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A logical schematic diagram of a clinical nursing decision support system based on artificial intelligence provided by the present invention; Figure 2 A flowchart of a clinical nursing decision support method based on artificial intelligence provided by the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, what is described is only a part of the present invention, not all of it. Generally, the components of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0017] like Figure 1 As shown, the present invention provides one embodiment: a clinical nursing decision support system based on artificial intelligence, comprising: Data layer, which acquires and stores patient data; The decision rule layer is based on the patient data in the data layer and establishes patient care decision rules based on the nursing process framework based on data-driven; An interoperability layer, wherein the interoperability layer is bidirectionally connected to the data layer, the decision rule layer, and the view interface layer, extracts and processes data in the data layer, combines patient data and relevant patient care decision rules, generates response results, and feeds the response results back to the view interface layer; The view interface layer is used to perform data visualization of the data layer, decision rule layer, and interoperability layer.

[0018] In this embodiment, the data layer is used to store the patient's corresponding nursing assessment, nursing problems, nursing plans, plan execution status and nursing effect evaluation status, including the nursing process generated for individual patients and its time points, including the activation conditions of nursing problems, activation time, the state of nursing measures selected, the time of generating nursing plans, the proposed execution time of nursing plans, the actual execution time of nursing plans, etc. Patient data extracted from other systems are not saved in this system. Interoperability layer: It can accept operation instructions for the view interface, connect to system databases such as HIS and LIS, extract and process data in the data layer, combine patient data and decision rules in other databases, generate response results, and feedback to the view interface layer; all records of view interface layer operations are saved to the data layer through the interoperability layer. Decision rule layer: Based on evidence-based literature, expert experience, data-driven modeling, etc., formulate patient nursing decision rules based on the nursing procedure framework, including nursing problems, activation conditions corresponding to nursing problems, nursing measures associated with nursing problems, and nursing problem outcome evaluation as the basic rules of the decision support system. In addition, the decision rule library can include risk predictions related to machine learning algorithms, further process the data collected in the patient data module, identify, analyze and calculate the patient's nursing sensitive indicators based on deep learning methods, and then output the risk prediction probability of adverse outcomes for patients. View interface layer: used for patient basic information display, nursing assessment, nursing problem selection, nursing plan formulation, nursing measure execution and evaluation, and can also generate records and provide monitoring reminder functions as an interface for nurses to directly contact and operate.

[0019] Specifically, the data layer has the following logic: Connect to hospital information systems, laboratory information systems and bedside equipment to collect patient data, including: patient basic information, test results, and vital signs time series data; The nursing assessment module of the view interface layer collects patient nursing assessment data, including: patient's first nursing assessment data and daily assessment data. The assessment data is stored in JSON format and attached with nurse ID and timestamp; A nursing plan execution record chain is constructed based on patient data and patient nursing assessment data, and stored in the data layer.

[0020] In this embodiment, for the hospital information system, the patient's demographic basic information, such as name, gender, age, hospitalization number, diagnosis information, etc., is mainly obtained; for the laboratory information system, the focus is on collecting test reports and related test indicators, such as blood routine, urine routine and biochemical test results; for the bedside monitoring equipment, blood pressure, heart rate, respiratory rate, blood oxygen saturation and other vital signs time series data are received at a preset frequency. After completing the basic field verification, unit conversion and noise filtering, the system stores the above data in the data layer to provide accurate and traceable data support for the subsequent nursing decision model. In order to collect and store more comprehensive nursing assessment information, this embodiment sets up a dedicated nursing assessment module in the view interface layer for inputting or modifying the patient's first and daily nursing assessment data; the module includes an input interface containing multiple categories of assessment items such as physiological, psychological, social, and functional, and each assessment item is filled in by the nurse manually or by auxiliary questioning. After the evaluation data is generated at the interface layer, it will be saved in a structured JSON format and automatically associated with the nurse ID and the operation timestamp to ensure that the evaluation source and time node can be clearly traced in the subsequent query, audit and decision-making process. At the same time, the system dynamically builds a nursing plan execution record chain based on patient data and nursing evaluation data to reflect the full process information from nursing evaluation to nursing measure specification and execution results. Specifically, whenever a nurse modifies the patient data through the above-mentioned evaluation module or data layer interface, the system will generate a new record node and attach it to the execution record chain of the corresponding patient; the record contains key information such as patient identification, data update content, nurse ID, timestamp and current execution status, thereby forming a complete visual execution context throughout the clinical nursing cycle. This chain storage structure can be interconnected with other clinical data in the data layer, providing an integrated data foundation for intelligent nursing decision-making, execution tracing and statistical analysis.

[0021] Furthermore, the data layer further includes: Perform field verification on the patient's basic information; Convert the units of the test results; The vital sign time series data is averaged according to a time window with a set interval to generate a standardized waveform segment; Complete preprocessing of patient data.

[0022] In this embodiment, after receiving patient data from the hospital information system, laboratory information system and bedside monitoring equipment, the system places it in the data layer for unified preprocessing operations. First, field verification is performed on the patient's basic information: the system verifies the integrity and format of basic fields such as name, gender, age and medical record number according to the internally configured field dictionary, such as detecting whether there are null values, special symbols or strings that do not conform to the expected format. Once an abnormality is found, the system will mark the corresponding record and prompt manual intervention or correction to ensure that the subsequent decision-making process is based on accurate and non-missing patient identity data. The system presets unit mapping tables for common test indicators such as blood routine, blood biochemistry, and urine routine, realizes automatic conversion of multi-source data according to the target unit standard, and coordinates inconsistent or incompatible test items into a common metric. In this process, the system will also perform a numerical rationality check. Once a serious deviation from the laboratory instrument record or beyond the normal physiological range is found, an early warning prompt will be generated to facilitate nursing staff to pay attention to possible extreme value abnormalities in a timely manner. For the vital sign time series data obtained by bedside devices, the system divides the original data into time windows with set intervals, and calculates the mean of the collected multidimensional parameters such as heart rate, respiratory rate, blood pressure, etc. in each window to generate standardized waveform segments. This processing method helps to eliminate the interference of instantaneous fluctuations, while retaining the overall trend and representative characteristics of each monitoring indicator in a specific time period, thereby providing more stable and comparable input data for subsequent multimodal nursing decision models.

[0023] In a further implementation of this embodiment, the patient care decision rules based on data-driven establishment with the nursing procedure as the framework are specifically: Based on the standardized waveform segments, abnormal events of the patient's vital signs are detected and represented as abnormal event features of the patient; Perform semantic analysis on JSON assessment data, extract nursing assessment keywords, and represent them as patient care semantic features; Taking the abnormal event characteristics of patients' vital signs and the semantic characteristics of patients' nursing as input, and the patient nursing decision coding as output, a multimodal nursing decision model for patients is established; Matching patient care decision codes with the evidence-based knowledge base in the nursing process to generate a set of candidate patient care decision rules; According to the nursing problem type as the priority, the best candidate patient nursing decision rule is selected to be characterized as the patient nursing decision rule.

[0024] In this embodiment, after completing the preprocessing of the patient data, the system analyzes the standardized waveform segments to detect abnormal events of the patient's vital signs, and characterizes the detection results as abnormal event features of the patient. For example, when indicators such as blood pressure, heart rate, and blood oxygen saturation exceed or continue to be lower than the reference threshold, the system automatically records and identifies the waveform segment corresponding to the time window as an abnormal segment, so as to provide risk warning signals in subsequent nursing decisions. At the same time, the system performs semantic analysis on the JSON format nursing assessment data from the view interface layer, extracts keywords including psychological, social, and functional aspects, and forms patient nursing semantic features. With the support of natural language processing and keyword index tables, the nursing issues involved in the assessment data are contextually understood to achieve structured expression of multi-dimensional nursing information. On this basis, the abnormal event features of the patient's vital signs and the patient's nursing semantic features are used as inputs, and the output of the model is set as the patient nursing decision code through the patient multimodal nursing decision model to reflect the nursing demand category corresponding to the current patient's comprehensive state. Subsequently, the system will match the decision code with the established evidence-based knowledge base in the nursing program, retrieve and generate a set of candidate patient nursing decision rules. The rules in this set are prioritized according to the type of nursing problem, and the optimal candidate rules are screened out in combination with the level of clinical evidence or nursing guideline requirements. They are finally presented to nursing staff in the form of patient care decision rules, thereby providing more targeted and feasible guidance for the formulation and implementation of nursing measures in actual clinical scenarios.

[0025] During the implementation, historical patient vital sign waveform fragments and nursing assessment data in JSON format are collected and integrated, and the standardized waveform fragments are input into the convolutional neural network module for feature extraction; the nursing semantic text is vectorized and then fed into the long short-term memory network for contextual representation. The waveform feature vector produced by CNN and the nursing semantic text vector produced by LSTM are spliced ​​or weighted fused in the fusion layer to form a multimodal comprehensive representation. Based on the fused features, the patient status is classified or further encoded using a fully connected network, and the patient nursing decision code is output to indicate the possible nursing problem category or priority. Define loss functions such as cross entropy or mean square error, calculate the loss through forward propagation, and then perform back propagation to update the weights of each network, and continue to iterate until the verification index meets the preset threshold. Use the validation set to evaluate the comprehensive performance of the model in the recognition of abnormal vital signs and the determination of nursing problems. If the index is insufficient, perform hyperparameter tuning or network structure fine-tuning, and perform a final evaluation on the test set.

[0026] In this embodiment, the interoperability layer includes: An operation instruction management module receives and analyzes the operation instructions of the view interface layer, and synchronizes the operation results to the data layer through the interoperability layer; A data connection module is connected to the hospital information system, the laboratory information system and the bedside equipment, and generates a response result in combination with the patient care decision rules of the decision rule layer; The operation log storage function saves all operation instructions and data call records issued through the view interface layer in the data layer, forming a nursing operation traceability chain.

[0027] During implementation, when a nurse or other clinical staff issues an operation instruction (such as adding nursing assessment content, modifying nursing measures, viewing patient historical data, etc.) in the view interface layer, the module will receive and parse the instruction, and call the corresponding service interface or submodule for processing according to the instruction type. The execution result of the operation instruction includes information such as success or failure flag, return data, possible abnormal prompts, etc., which are synchronized to the data layer through the interoperability layer with a unified message mechanism to ensure that the data layer can update the record in real time and maintain global data consistency. At the same time, if the instruction is related to the nursing decision-making process, the execution result can also be fed back to the decision rule layer in real time to further trigger or update the patient nursing decision rule status. In addition, the interoperability layer of this embodiment records each operation instruction issued by the view interface layer, as well as the request parameters and return results in the data call process, and binds them to the generated timestamp and operator identification. The log information will be uniformly written to the data layer to form a nursing operation traceability chain.

[0028] Specifically, the interoperability layer also includes: The real-time status monitoring module continuously obtains the status changes of the nursing plan execution record chain. If the patient care decision rules conflict with the current patient care assessment data, a logic verification alarm is triggered.

[0029] Dynamic decision compensation module, when the output confidence of the patient multimodal nursing decision model is lower than the set value, the patient nursing decision code is recalculated. If the recalculated result still does not meet the set value, the second best choice in the candidate patient nursing decision rule set is screened out and pushed to the view interface layer.

[0030] In this embodiment, the view interface layer further includes: The rule matching display module includes a first area and a second area, wherein the first area displays the best candidate patient care decision rule; the second area displays the second best candidate patient care decision rule in parallel; The manual decision feedback interface writes the second-best candidate patient care decision rules into the nursing plan execution record chain as feedback data.

[0031] During implementation, in the view interface, nurses can interact with the software and intuitively complete the above functions; at the same time, the view interface can directly display the statistics, monitoring and reminder functions of some nursing problems or nursing plans, provide an overview of patient data, and prompt patients with nursing risks, unresolved nursing issues, and nursing measures that are about to expire.

[0032] In one application example, after connecting to the hospital information system, laboratory information system and bedside equipment and performing unit conversion, field verification and waveform segment standardization of patient data, the system can use the following four groups of elements as multimodal input sources to trigger and identify patient care problems: Cause (directly related diagnosis): Extract the patient's confirmed disease name or potential diagnosis (such as "myocardial ischemia" or "diabetic neuropathy") from the hospital information system or test results, and use it as the direct cause of the current care problem. Positive indications or abnormal values: Combined with indicators that deviate from the normal range in bedside monitoring equipment and laboratory data (such as decreased blood oxygen saturation and high blood sugar), they are marked as abnormal features in data preprocessing to assist in locating possible care problems (such as hypoxia risk, blood sugar out of control, etc.). Symptoms: The first or daily care assessment data entered by the nursing staff at the view interface layer (stored in JSON format), if it contains subjective symptoms such as pain, chest tightness, nausea, etc. reported by the patient, combined with the above causes and abnormal indicators, can significantly improve the system's accuracy in determining the severity of the disease and the type of care problems. Related or risk factors: Mark risk factors such as comorbidities, advanced age, and long-term bed rest in the assessment records; if the patient has a history of "atherosclerosis" or "high risk of pressure sores", the system can comprehensively evaluate this information to determine whether the nursing issue needs to be prioritized.

[0033] By randomly combining and matching the above four elements to the pre-designed multimodal data interface, the system extracts features from patient data and JSON evaluation data, and maps the identified "cause + positive indication + symptoms + related / risk factors" to the patient care decision code. Subsequently, the nursing problem generation and evaluation module is called based on the decision code, candidate decision rules are generated in the evidence-based knowledge base, and then the priority is set according to the type of nursing problem to select the optimal solution, and the interoperability layer outputs the final nursing decision result to the view interface layer, thereby ensuring that the motivation and trigger logic of the nursing measures are more accurate and efficient.

[0034] like Figure 2 As shown, the present invention provides another embodiment: a clinical nursing decision support method based on artificial intelligence, the steps of the method include: Connect to hospital information systems, laboratory information systems and bedside devices to collect patient data, and perform field verification, unit conversion and waveform segment standardization on patient data based on preprocessing; Obtain the patient's first and daily patient nursing assessment data through the nursing assessment module of the view interface layer, and store it in the data layer in JSON format; Extract features from patient data and JSON evaluation data, input abnormal event features of patient vital signs and patient care semantic features into the patient multimodal care decision model, and generate corresponding patient care decision codes; Based on the patient multimodal nursing decision model, the nursing problem generation and evaluation module is called to generate a set of candidate patient nursing decision rules based on the evidence-based knowledge base preset in the nursing procedure; According to the type of nursing problem as the priority, the optimal candidate patient nursing decision rule is selected to be represented as the patient nursing decision rule. The interoperability layer outputs the patient nursing decision rule as a response result and feeds it back to the view interface layer.

[0035] Specifically, the priority of nursing problems is based on: clinical urgency: such as acute pain, sudden changes in vital signs, etc.; prognostic impact: such as the prevention of pressure sores is particularly critical in long-term bedridden patients; patient complaint intensity and nursing goals: if the patient has a high demand for pain relief, the priority of pain care will be increased; hospital or department resource limitations: when there is a shortage of equipment and manpower required for specific nursing measures, a comprehensive balance is required. This embodiment takes the following types as examples: Level 1 nursing problems: problems that have a significant impact on the patient's life or safety, such as severe pain and risk factors caused by abnormal coagulation function; Level 2 nursing problems: problems that affect the patient's comfort or rehabilitation process, such as physical activity training, nutritional support, etc.; Level 3 nursing problems: problems that require long-term attention but have relatively small effects in the short term, such as long-term nursing interventions for some chronic underlying diseases. Subsequently, based on the abnormal patient vital signs events received by the interoperability layer and the nursing assessment semantic analysis keywords, multiple candidate nursing decision rules related to them are retrieved in the evidence-based knowledge base. When screening and sorting the above candidate rules, the principle of nursing problem type as priority is applied to the scoring model: for level 1 nursing problems, the candidate rules have higher scores; for level 2 or 3 problems, if the system detects concurrent nursing problems, the priority score is lowered to ensure a more reasonable resource allocation and execution order. The above priority score is comprehensively scored with factors such as clinical evidence level, patient complaint intensity, and department resource availability, and the rule with the highest score is selected as the optimal candidate patient nursing decision rule to guide the formulation and scheduling of subsequent nursing measures.

[0036] It can be understood that the artificial intelligence-based clinical nursing decision support method provided in this embodiment and the artificial intelligence-based clinical nursing decision support system provided in the above embodiment are based on the same inventive concept. For more specific working principles of each module in the embodiment of the present invention, please refer to the above embodiment and will not be repeated in the embodiment of the present invention.

[0037] The above is only the preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A clinical nursing decision support system based on artificial intelligence, characterized in that: include: Data layer, which acquires and stores patient data; The decision rule layer is based on the patient data in the data layer and establishes patient care decision rules based on the nursing process framework based on data-driven; An interoperability layer, wherein the interoperability layer is bidirectionally connected to the data layer, the decision rule layer, and the view interface layer, extracts and processes data in the data layer, combines patient data and relevant patient care decision rules, generates response results, and feeds the response results back to the view interface layer; The view interface layer is used to perform data visualization of the data layer, decision rule layer, and interoperability layer.

2. The artificial intelligence-based clinical nursing decision support system according to claim 1, characterized in that: The specific logic of the data layer is as follows: Connect to hospital information systems, laboratory information systems and bedside equipment to collect patient data, including: patient basic information, test results, and vital signs time series data; The nursing assessment module of the view interface layer collects patient nursing assessment data, including: patient's first nursing assessment data and daily assessment data. The assessment data is stored in JSON format and attached with nurse ID and timestamp; A nursing plan execution record chain is constructed based on patient data and patient nursing assessment data, and stored in the data layer.

3. The artificial intelligence-based clinical nursing decision support system according to claim 2, characterized in that: The data layer further includes: Perform field verification on the patient's basic information; Convert the units of the test results; The vital sign time series data is averaged according to a time window with a set interval to generate a standardized waveform segment; Complete preprocessing of patient data.

4. The artificial intelligence-based clinical nursing decision support system according to claim 3, characterized in that: The patient care decision rules based on data-driven establishment with nursing procedures as the framework are specifically: Based on the standardized waveform segments, abnormal events of the patient's vital signs are detected and represented as abnormal event features of the patient; Perform semantic analysis on JSON assessment data, extract nursing assessment keywords, and represent them as patient care semantic features; Taking the abnormal event characteristics of patients' vital signs and the semantic characteristics of patients' nursing as input, and the encoding of patients' nursing decisions as output, a multimodal nursing decision model for patients is established.

5. The artificial intelligence-based clinical nursing decision support system according to claim 4, characterized in that: The decision rule layer also includes a nursing problem generation and evaluation module, specifically: Matching patient care decision codes with the evidence-based knowledge base in the nursing process to generate a set of candidate patient care decision rules; According to the nursing problem type as the priority, the best candidate patient nursing decision rule is selected to be characterized as the patient nursing decision rule.

6. The artificial intelligence-based clinical nursing decision support system according to claim 5, characterized in that: The interoperability layer includes: An operation instruction management module receives and analyzes the operation instructions of the view interface layer, and synchronizes the operation results to the data layer through the interoperability layer; A data connection module is connected to the hospital information system, the laboratory information system and the bedside equipment, and generates a response result in combination with the patient care decision rules of the decision rule layer; The operation log storage function saves all operation instructions and data call records issued through the view interface layer in the data layer, forming a nursing operation traceability chain.

7. The artificial intelligence-based clinical nursing decision support system according to claim 6, characterized in that: The interoperability layer also includes: The real-time status monitoring module continuously obtains the status changes of the nursing plan execution record chain. If the patient care decision rules conflict with the current patient care assessment data, a logic verification alarm is triggered.

8. The artificial intelligence-based clinical nursing decision support system according to claim 7, characterized in that: The interoperability layer also includes: Dynamic decision compensation module, when the output confidence of the patient multimodal nursing decision model is lower than the set value, the patient nursing decision code is recalculated. If the recalculated result still does not meet the set value, the second best choice in the candidate patient nursing decision rule set is screened out and pushed to the view interface layer.

9. The artificial intelligence-based clinical nursing decision support system according to claim 8, characterized in that: The view interface layer also includes: The rule matching display module includes a first area and a second area, wherein the first area displays the best candidate patient care decision rule; the second area displays the second best candidate patient care decision rule in parallel; The manual decision feedback interface writes the second-best candidate patient care decision rules into the nursing plan execution record chain as feedback data.

10. A clinical nursing decision support method based on artificial intelligence, characterized in that: The steps of the method include: Connect to hospital information systems, laboratory information systems and bedside devices to collect patient data, and perform field verification, unit conversion and waveform segment standardization on patient data based on preprocessing; Obtain the patient's first and daily patient nursing assessment data through the nursing assessment module of the view interface layer, and store it in the data layer in JSON format; Extract features from patient data and JSON evaluation data, input abnormal event features of patient vital signs and patient care semantic features into the patient multimodal care decision model, and generate corresponding patient care decision codes; Based on the patient multimodal nursing decision model, the nursing problem generation and evaluation module is called to generate a set of candidate patient nursing decision rules based on the evidence-based knowledge base preset in the nursing procedure; According to the type of nursing problem as the priority, the optimal candidate patient nursing decision rule is selected to be represented as the patient nursing decision rule. The interoperability layer outputs the patient nursing decision rule as a response result and feeds it back to the view interface layer.

Images