Nursing monitoring system and method for patients with severe pancreatitis
Through artificial intelligence and deep learning technology, the vital signs and urine volume time sequence data of patients with severe pancreatitis were analyzed, and the problem of failure to effectively pay attention to dynamic changes in timing and interactive relationships in the existing technology was solved, and personalized nursing monitoring and physiology were realized Efficiency of status abnormality detection.
Patent Information
- Application Number
- CN202411140503.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-20
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-20
AI Technical Summary
The prior art failed to effectively pay attention to the dynamic changes in the timing and implicit interaction relationships between vital sign data in the care of patients with severe pancreatitis, resulting in low accuracy and effectiveness of the analysis results, and no personalized differences were considered. Insufficient generalization ability.
Using data processing algorithms based on artificial intelligence and deep learning, learning between patient vital signs and urinary semantic features through dynamic interaction fusion of vital sign embedding encoding, timing feature enhancement, full-time-sequence correlation feature extraction and multimodal monitoring data. dynamic interactive fusion characteristics for physiological state monitoring and abnormal detection.
It realizes more intelligent and personalized nursing monitoring of patients with severe pancreatitis, improves the accuracy of analysis results and systematic generalization ability, and can promptly detect physiological abnormalities and issue early warnings.
Smart Images

Figure CN118658613B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of intelligent monitoring, and more specifically, to a nursing monitoring system and method for patients with severe pancreatitis. Background Art
[0002] Severe pancreatitis is a complex clinical condition and difficult to treat disease, characterized by acute inflammatory response of pancreatic tissue, which may rapidly progress to systemic inflammatory response syndrome and even multiple organ dysfunction syndrome. The high mortality and complication rate of this disease make close monitoring and timely intervention of patients particularly important.
[0003] Chinese patent CN116741384B discloses a clinical data management method for severe acute pancreatitis based on bedside care, which can collect bedside care data of patients with mild pancreatitis and severe acute pancreatitis, and use HAPS score and Marshall score to process the bedside care data of different types of patients respectively, and then construct a clinical data management model for severe acute pancreatitis based on bedside care evaluation by single factor and multifactor analysis. Further, the area under the ROC curve (AUC) is used to evaluate the accuracy of the model, and the model is classified according to the score. Correlation analysis is performed in combination with HAPS and Marshall scores, and the difference variation coefficient is constructed to optimize the clinical data management model. In this way, the characterization indicators of bedside care can be quickly obtained in the form of constructing a clinical data management model, and the characterization indicators of bedside care can be effectively managed for data information, providing an effective nursing basis for patients with severe pancreatitis, so as to timely reflect the development and changes of pancreatitis symptoms.
[0004] In the above-mentioned clinical data management method for severe acute pancreatitis based on bedside care, the acquired bedside care data is data information collected at the initial visit and 24h time point after the visit, which usually includes the patient's vital signs data, and these data have hidden association patterns and dynamic change characteristics. However, in this scheme, a simple clinical data management model is relied on to perform the development scoring of mild and severe pancreatitis symptoms, as well as the analysis of the differences between the changes in independent variables and dependent variables. This approach does not pay attention to the temporal dynamic change characteristics and implicit interactive correlations between different vital signs data, resulting in low accuracy and effectiveness of the analysis results of the patient's bedside care data. In addition, the clinical data management method usually uses a specific model to analyze bedside care data, which does not take into account the individual differences between different patients, resulting in insufficient generalization ability in different patient groups.
[0005] Therefore, an optimized system and method for monitoring the care of patients with severe pancreatitis is desired. Summary of the invention
[0006] In order to solve the above technical problems, this application is proposed.
[0007] According to one aspect of the present application, a system for monitoring and caring for patients with severe pancreatitis is provided, comprising:
[0008] A vital sign data acquisition module, used to acquire a time series of vital sign data of a monitored severe pancreatitis patient acquired by a sensor group, wherein the vital sign data includes heart rate value, blood pressure value, respiratory rate, body temperature value and blood oxygen saturation;
[0009] A urine volume data acquisition module, used to obtain the time series of urine volume of the monitored severe pancreatitis patient;
[0010] A vital sign embedding mapping association coding module is used to pass each vital sign data in the time series of the monitored severe pancreatitis patient's vital sign data through an embedding encoder based on a fully connected mapping layer to obtain a sequence of embedding coding feature vectors of the severe pancreatitis patient's vital signs;
[0011] A vital sign time series feature enhanced expression module is used to embed the vital signs of the patient with severe pancreatitis into a sequence of encoded feature vectors through a reverse screening and enhancement network based on the vital sign time series features of anchor point centralization to obtain a sequence of semantic feature vectors that enhances the focus on the vital signs of the patient with severe pancreatitis;
[0012] A vital sign full time series correlation feature extraction module is used to perform time series encoding on the sequence of the semantic feature vectors of the vital signs of patients with severe pancreatitis to obtain a time series correlation feature vector of the vital signs of patients with severe pancreatitis;
[0013] A urine volume time series correlation feature extraction module is used to perform time series encoding on the time series of the urine volume of the patient with severe pancreatitis to obtain a urine volume time series correlation feature vector of the patient with severe pancreatitis;
[0014] A vital sign-urine volume time series feature interactive fusion module, which is used to obtain a critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector by combining the critical pancreatitis patient's vital sign time series associated feature vector and the critical pancreatitis patient's urine volume time series associated feature vector through a patient multimodal monitoring data time series feature dynamic interactive fusion module based on a gated attention allocation mechanism;
[0015] The physiological state abnormality detection module is used to determine whether the physiological state of the monitored severe pancreatitis patient is abnormal based on the vital sign-urine volume time series interactive fusion feature vector of the severe pancreatitis patient.
[0016] In the above-mentioned critical pancreatitis patient nursing monitoring system, the vital signs time series feature enhanced expression module includes:
[0017] A position mean calculation unit, used for calculating the position mean vector of the sequence of the embedding encoding feature vector of the vital signs of the patient with severe pancreatitis to obtain a semantic clustering center feature vector of the vital signs of the patient with severe pancreatitis;
[0018] A semantic correlation coefficient calculation unit, used to calculate the semantic correlation coefficient between each critical pancreatitis patient's vital sign embedding coding feature vector in the sequence of critical pancreatitis patient's vital sign embedding coding feature vector and the critical pancreatitis patient's vital sign semantic clustering center feature vector to obtain a sequence of critical pancreatitis patient's vital sign clustering center semantic correlation coefficients;
[0019] A cluster center semantic anti-correlation coefficient extraction unit, used for taking the inverse of each of the critical pancreatitis patients' vital signs cluster center semantic correlation coefficients in the sequence of critical pancreatitis patients' vital signs cluster center semantic correlation coefficients to obtain a sequence of critical pancreatitis patients' vital signs cluster center semantic anti-correlation coefficients;
[0020] A normalization unit, used for normalizing the sequence of semantic anti-correlation coefficients of the vital signs clustering centers of the patients with severe pancreatitis to obtain a sequence of semantic anti-correlation weight coefficients of the vital signs clustering centers of the patients with severe pancreatitis;
[0021] A weighting unit, used for weighting each critical pancreatitis patient vital sign embedded coding feature vector in the sequence of critical pancreatitis patient vital sign embedded coding feature vectors by using each critical pancreatitis patient vital sign cluster center semantic anti-correlation weight coefficient in the sequence of critical pancreatitis patient vital sign cluster center semantic anti-correlation weight coefficient as a weight to obtain a sequence of critical pancreatitis patient vital sign embedded coding feature vectors;
[0022] The enhanced attention unit is used to calculate the positional difference between the sequence of the embedded coding feature vectors of the vital signs of patients with severe pancreatitis and the sequence of the embedded coding feature vectors of the vital signs of patients with severe pancreatitis corresponding to each group and the embedded coding feature vectors of the vital signs of patients with severe pancreatitis to obtain the sequence of the enhanced attention semantic feature vectors of the vital signs of patients with severe pancreatitis.
[0023] In the above-mentioned severe pancreatitis patient nursing monitoring system, the semantic correlation coefficient calculation unit is used to:
[0024] Calculating the logarithmic function value of the feature value at each position in the feature vector of the embedding code of the vital signs of the patient with severe pancreatitis to obtain the semantic logarithmic representation vector of the embedding code of the vital signs of the patient with severe pancreatitis;
[0025] Calculating the positional product between the semantic logarithmic representation vector of the embedding encoding of the vital signs of the patient with severe pancreatitis and the semantic clustering center feature vector of the vital signs of the patient with severe pancreatitis to obtain the vital signs embedding semantic clustering center associated feature vector;
[0026] The semantic correlation coefficient of the vital sign cluster center of the patient with severe pancreatitis is obtained by dividing the associated feature vector of the vital sign embedding semantic cluster center by the number of eigenvalues in the encoding feature vector of the vital sign embedding of the patient with severe pancreatitis.
[0027] In the above-mentioned critical pancreatitis patient care monitoring system, the normalization unit is used to:
[0028] Using each semantic anti-correlation coefficient of the vital signs cluster center of patients with severe pancreatitis in the sequence of semantic anti-correlation coefficients of the vital signs cluster center of patients with severe pancreatitis as an exponent of a natural constant to calculate an exponential function value based on the natural constant according to the position to obtain a sequence of semantic anti-correlation class support coefficients of the vital signs cluster center of patients with severe pancreatitis;
[0029] Calculating the sum of the semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis in the sequence of semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis to obtain the sum of the semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis;
[0030] Calculate the positional division between the semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis and the sum of the semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis to obtain a sequence of semantic anti-correlation weight coefficients of the vital signs clustering centers of patients with severe pancreatitis.
[0031] In the above-mentioned severe pancreatitis patient nursing monitoring system, the vital signs full time series correlation feature extraction module is used to: pass the sequence of the semantic feature vectors of the vital signs of patients with severe pancreatitis that are given enhanced attention through a vital signs time series encoder based on a Bi-LSTM model to obtain the time series correlation feature vectors of the vital signs of patients with severe pancreatitis.
[0032] In the above-mentioned severe pancreatitis patient nursing monitoring system, the urine volume time series association feature extraction module is used to: arrange the time series of urine volume of the monitored severe pancreatitis patient according to the time dimension into a urine volume time series input vector, and then pass the urine volume time series input vector through a urine volume time series encoder based on the Bi-LSTM model to obtain the urine volume time series association feature vector of the severe pancreatitis patient.
[0033] In the above-mentioned critical pancreatitis patient nursing monitoring system, the vital sign-urine volume time series feature interactive fusion module is used to:
[0034] Inputting the critical pancreatitis patient's vital sign time series correlation feature vector and the critical pancreatitis patient's urine volume time series correlation feature vector into a feature combination module for cascade processing to obtain a critical pancreatitis patient's vital sign-urine volume combination feature vector;
[0035] Inputting the joint feature vector of the vital signs and urine volume of the patient with severe pancreatitis into a one-dimensional convolutional layer to extract the associated features to obtain the implicit associated feature vector of the vital signs and urine volume of the patient with severe pancreatitis;
[0036] After calculating the matrix multiplication of the critically ill pancreatitis patient's vital sign-urine volume time series implicit correlation eigenvector and the parameter matrix, the obtained eigenvector is added to the bias vector by position to obtain a linearly transformed critically ill pancreatitis patient's vital sign-urine volume time series implicit correlation eigenvector;
[0037] use The nonlinear activation function activates the linearly transformed critical pancreatitis patient's vital sign-urine volume time series implicit correlation feature vector to obtain the critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector;
[0038] Calculating the position difference between the unit vector and the critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector to obtain the difference critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector;
[0039] use The function performs a normalization operation on the critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector and the difference critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector to obtain a normalized critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector and a normalized difference critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector;
[0040] Calculate the point-by-point multiplication between the normalized critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector and the critical pancreatitis patient's vital sign time series association feature vector to obtain an importance-weighted critical pancreatitis patient's vital sign time series association feature vector;
[0041] Calculate the point-by-point multiplication between the normalized difference critical pancreatitis patient vital sign-urine volume dynamic information nonlinear transformation activation vector and the critical pancreatitis patient urine volume time series associated feature vector to obtain the importance weighted critical pancreatitis patient urine volume time series associated feature vector;
[0042] The positional sum of the importance-weighted critical pancreatitis patient's vital sign time series associated feature vector and the importance-weighted critical pancreatitis patient's urine volume time series associated feature vector is calculated to obtain the critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector.
[0043] In the above-mentioned severe pancreatitis patient care monitoring system, the physiological state abnormality detection module is used to: pass the vital sign-urine volume time series interactive fusion feature vector of the severe pancreatitis patient through a classifier-based abnormality identifier to obtain an identification result, and the identification result is used to indicate whether there is an abnormality in the physiological state of the monitored severe pancreatitis patient.
[0044] According to another aspect of the present application, a method for monitoring the care of a patient with severe pancreatitis is provided, comprising:
[0045] Acquire a time series of vital sign data of a monitored severe pancreatitis patient collected by a sensor group, wherein the vital sign data includes a heart rate value, a blood pressure value, a respiratory rate, a body temperature value, and a blood oxygen saturation;
[0046] Obtaining a time series of urine volume of the monitored severe pancreatitis patient;
[0047] Each vital sign data in the time series of the vital sign data of the monitored severe pancreatitis patient is passed through an embedding encoder based on a fully connected mapping layer to obtain a sequence of embedding encoding feature vectors of the vital signs of the severe pancreatitis patient;
[0048] The sequence of the critical pancreatitis patient's vital signs embedded in the encoding feature vector is subjected to a reverse screening reinforcement network based on the anchor-centered critical pancreatitis patient's vital signs time series feature to obtain a sequence of the critical pancreatitis patient's vital signs semantic feature vector that is enhanced;
[0049] Performing temporal coding on the sequence of the semantic feature vectors of the vital signs of patients with severe pancreatitis that are given intensified attention to obtain a temporal association feature vector of the vital signs of patients with severe pancreatitis;
[0050] Performing time series encoding on the time series of urine volume monitored in patients with severe pancreatitis to obtain a time series correlation feature vector of urine volume in patients with severe pancreatitis;
[0051] The critical pancreatitis patient's vital sign time series correlation feature vector and the critical pancreatitis patient's urine volume time series correlation feature vector are passed through a patient multimodal monitoring data time series feature dynamic interactive fusion module based on a gated attention allocation mechanism to obtain a critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector;
[0052] Based on the vital sign-urine volume time series interactive fusion feature vector of the severe pancreatitis patient, it is determined whether the physiological state of the monitored severe pancreatitis patient is abnormal.
[0053] Compared with the prior art, the present application provides a nursing monitoring system and method for patients with severe pancreatitis. The system and method use data processing and analysis algorithms based on artificial intelligence and deep learning to extract the time-series dynamic change patterns and interactive correlation relationships of the vital signs data and urine volume data of patients with severe pancreatitis, so as to learn and characterize the dynamic interactive fusion characteristics between the patient's vital signs time-series semantics and the urine volume time-series semantics, so as to monitor the physiological state and detect abnormalities of patients with severe pancreatitis, and issue early warnings in time when abnormalities are found. In this way, more intelligent nursing monitoring and management of patients with severe pancreatitis can be achieved. Furthermore, relying on the ability of artificial intelligence and deep learning technology to process large amounts of data and learn the implicit feature patterns therein, the time-series dynamic interactive correlation characteristics and pattern information between the patient's vital signs and urine volume can be learned and extracted, thereby achieving personalized nursing monitoring of patients with severe pancreatitis and improving the generalization ability of the system. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.
[0055] Figure 1 is a block diagram of a nursing monitoring system for patients with severe pancreatitis according to an embodiment of the present application;
[0056] Figure 2 The data flow diagram of the critical pancreatitis patient care monitoring system according to the embodiment of the present application;
[0057] Figure 3 The present invention is a flowchart of a method for monitoring the care of patients with severe pancreatitis according to an embodiment of the present application. DETAILED DESCRIPTION
[0058] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.
[0059] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.
[0060] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.
[0061] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.
[0062] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.
[0063] In the technical solution of the present application, a nursing monitoring system for patients with severe pancreatitis is proposed. Figure 1 4 is a block diagram of a system for monitoring the care of patients with severe pancreatitis according to an embodiment of the present application. Figure 2 FIG. 4 is a data flow diagram of a nursing monitoring system for patients with severe pancreatitis according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to an embodiment of the present application, a nursing monitoring system 300 for severe pancreatitis patients includes: a vital sign data acquisition module 310, which is used to obtain a time series of vital sign data of a monitored severe pancreatitis patient collected by a sensor group, wherein the vital sign data includes a heart rate value, a blood pressure value, a respiratory rate, a body temperature value and a blood oxygen saturation; a urine volume data acquisition module 320, which is used to obtain a time series of urine volume of the monitored severe pancreatitis patient; a vital sign embedding mapping association coding module 330, which is used to pass each vital sign data in the time series of the vital sign data of the monitored severe pancreatitis patient through an embedding encoder based on a fully connected mapping layer to obtain a sequence of embedding encoding feature vectors of the vital signs of the patient with severe pancreatitis; a vital sign time series feature enhancement expression module 340, which is used to pass the sequence of embedding encoding feature vectors of the vital signs of the patient with severe pancreatitis through a reverse screening enhancement network based on the anchor point-centered vital sign time series feature to obtain an enhanced focus on the semantic features of the vital signs of the patient with severe pancreatitis. a sequence of vectors; a vital sign full time series correlation feature extraction module 350, used to perform time series encoding on the sequence of semantic feature vectors of the vital signs of patients with enhanced attention to severe pancreatitis to obtain a time series correlation feature vector of the vital signs of patients with severe pancreatitis; a urine volume time series correlation feature extraction module 360, used to perform time series encoding on the time series of urine volume of the patients with severe pancreatitis to obtain a time series correlation feature vector of urine volume of the patients with severe pancreatitis; a vital sign-urine volume time series feature interactive fusion module 370, used to combine the time series correlation feature vector of the vital signs of the patients with severe pancreatitis and the time series correlation feature vector of urine volume of the patients with severe pancreatitis through a patient multimodal monitoring data time series feature dynamic interactive fusion module based on a gated attention allocation mechanism to obtain a vital sign-urine volume time series interactive fusion feature vector of the patients with severe pancreatitis; a physiological state abnormality detection module 380, used to determine whether there is an abnormality in the physiological state of the monitored severe pancreatitis patient based on the vital sign-urine volume time series interactive fusion feature vector of the patients with severe pancreatitis.
[0064] In particular, the vital sign data acquisition module 310 and the urine volume data acquisition module 320 are used to obtain the time series of vital sign data of the monitored severe pancreatitis patient collected by the sensor group, wherein the vital sign data includes heart rate value, blood pressure value, respiratory rate, body temperature value and blood oxygen saturation; and obtain the time series of urine volume of the monitored severe pancreatitis patient. It should be understood that severe pancreatitis (SAP) is a serious disease characterized by acute inflammation of the pancreas and possibly surrounding tissues. This condition may be accompanied by organ failure, such as kidney failure, and is associated with a higher mortality rate. Urine volume is an important indicator for assessing a patient's renal function and fluid balance. In SAP patients, reduced urine volume (oliguria or anuria) may indicate impaired renal function, which is one of the signs of multiple organ failure and requires urgent intervention and treatment. For example, continuous renal replacement therapy (CRRT) may be used to treat acute renal failure, maintain hemodynamic parameters, and perform blood purification therapy when necessary. Therefore, when treating SAP and monitoring the physiological status of patients with severe pancreatitis, it is necessary not only to monitor the patient's vital signs, but also to monitor the dynamic changes in the patient's urine output over time, while paying attention to the implicit correlation patterns and dynamic interaction effects between the two, because they can reflect the patient's fluid status and organ function.
[0065] In particular, the vital sign embedding mapping association coding module 330 is used to pass each vital sign data in the time series of the vital sign data of the monitored severe pancreatitis patient through an embedding encoder based on a fully connected mapping layer to obtain a sequence of embedding coding feature vectors of the vital signs of the patient with severe pancreatitis. Considering that the vital sign data of the monitored severe pancreatitis patient contains the patient's heart rate value, blood pressure value, respiratory rate, body temperature value and blood oxygen saturation, these vital sign data have mutual hidden association patterns and interactive information. However, since these different types of vital sign data have different dimensions and ranges, it is difficult to provide useful information in the subsequent feature extraction and analysis process. Based on this, in the technical solution of the present application, each vital sign data in the time series of the vital sign data of the monitored severe pancreatitis patient is passed through an embedding encoder based on a fully connected mapping layer to obtain a sequence of embedding coding feature vectors of the vital signs of the patient with severe pancreatitis. Through the processing of the embedding encoder based on the fully connected mapping layer, different types of vital sign data can be embedded and mapped into a common space, and the fully connected mapping layer can be used to capture the implicit association patterns and semantics between different types of vital sign data, so as to convert each vital sign data in the time series of the vital sign data into an embedded coding feature vector of the vital signs of patients with severe pancreatitis, thereby facilitating subsequent processing and detection of abnormal physiological status.
[0066] In particular, the vital signs temporal feature enhanced expression module 340 is used to embed the sequence of the critical pancreatitis patient's vital signs into the coded feature vector through a reverse screening enhanced network based on the anchor-centered vital signs temporal feature to obtain a sequence of enhanced attention to the critical pancreatitis patient's vital signs semantic feature vector. It should be understood that each critical pancreatitis patient's vital signs embedded coding feature vector in the sequence of critical pancreatitis patient's vital signs embedded coding feature vectors respectively contains embedded coding features about multiple vital signs of critical pancreatitis patients at various time points, and these embedded coding feature information of vital signs have temporal correlation relationships and interactive influences, and these association relationships are of great significance for subsequent physiological state abnormality recognition and detection tasks. In addition, the patient's vital signs data embedded coding features at different time points have different importance and contribution to the subsequent physiological state abnormality detection tasks, and there will also be some interference and noise information that is irrelevant to the subsequent. Based on this, in the technical solution of the present application, the sequence of the encoding feature vectors of the vital signs of patients with severe pancreatitis is further embedded in the network through reverse screening based on the anchor-centered vital sign temporal features to obtain a sequence of semantic feature vectors that strengthens the focus on the vital signs of patients with severe pancreatitis.
[0067] In an embodiment of the present application, the sequence of the critical pancreatitis patient's vital signs embedded coding feature vectors is subjected to a reverse screening reinforcement network based on the anchor-centered vital signs temporal features to obtain a sequence of semantic feature vectors of critical pancreatitis patient's vital signs that are strengthened, including: first calculating the position mean vector of the sequence of the critical pancreatitis patient's vital signs embedded coding feature vectors to obtain the critical pancreatitis patient's vital signs semantic cluster center feature vector. Specifically, the anchor-centered critical pancreatitis patient's vital signs temporal features reverse screening reinforcement network can analyze the sequence of the input critical pancreatitis patient's vital signs embedded coding feature vectors through a clustering algorithm to identify the intrinsic structure of each critical pancreatitis patient's vital signs embedded coding feature, and extract a representative cluster center vector. Next, the semantic correlation coefficient between each critical pancreatitis patient's vital sign embedding coding feature vector in the sequence of critical pancreatitis patient's vital sign embedding coding feature vector and the critical pancreatitis patient's vital sign semantic clustering center feature vector is calculated to obtain a sequence of critical pancreatitis patient's vital sign clustering center semantic correlation coefficients. That is, the semantic correlation coefficient between each critical pancreatitis patient's vital sign embedding coding feature vector and the clustering center is calculated to quantify the similarity between them. Then, the reciprocal of each critical pancreatitis patient's vital sign clustering center semantic correlation coefficient in the sequence of critical pancreatitis patient's vital sign clustering center semantic anti-correlation coefficient is taken to obtain a sequence of critical pancreatitis patient's vital sign clustering center semantic anti-correlation coefficient. By taking the reciprocal of these correlation coefficients, the user's critical pancreatitis patient's vital sign semantic anti-correlation coefficient is obtained to emphasize those critical pancreatitis patient's vital sign embedding coding semantic features that differ greatly from the clustering center, which may reveal noise or irrelevant information in the critical pancreatitis patient's vital sign items. Furthermore, the sequence of semantic anti-correlation coefficients of the cluster centers of the vital signs of patients with severe pancreatitis is normalized to obtain a sequence of semantic anti-correlation weight coefficients of the cluster centers of the vital signs of patients with severe pancreatitis. In the technical solution of the present application, by normalizing these anti-correlation coefficients, it is ensured that the weight coefficients are within a reasonable range, thereby providing a balanced weight distribution for the subsequent weighting of semantic features of the embedded coding of the vital signs of patients with severe pancreatitis. Further, using each semantic anti-correlation weight coefficient of the cluster center of the vital signs of patients with severe pancreatitis in the sequence of semantic anti-correlation weight coefficients of the cluster centers of the vital signs of patients with severe pancreatitis as a weight, each critical pancreatitis patient vital sign embedded coding feature vector in the sequence of critical pancreatitis patient vital sign embedded coding feature vectors is weighted to obtain a sequence of suppressed critical pancreatitis patient vital sign embedded coding feature vectors.Among them, the weighting step generates a sequence of suppressed user's critical pancreatitis patient vital signs embedded coding feature vectors by applying the normalized weight coefficient to the original user critical pancreatitis patient vital signs embedded coding feature vectors. This process realizes the reverse screening of feature components to screen out feature components in the feature distribution that are more likely to be interference components or noise distributions. Subsequently, the position difference between each group of corresponding critical pancreatitis patient vital signs embedded coding feature vectors and suppressed critical pancreatitis patient vital signs embedded coding feature vectors in the sequence of critical pancreatitis patient vital signs embedded coding feature vectors and the sequence of suppressed critical pancreatitis patient vital signs embedded coding feature vectors is calculated to obtain the sequence of the enhanced focus on critical pancreatitis patient vital signs semantic feature vectors. Here, the feature components of the semantic feature distribution of the user's critical pancreatitis patients' vital signs that are interference components or noise components are filtered out by calculating the difference between the embedding encoding feature vector of the user's critical pancreatitis patients' vital signs and the suppressed embedding encoding feature vector of the user's critical pancreatitis patients' vital signs, and the sequence of the semantic feature vectors of the critical pancreatitis patients' vital signs that are strengthened is obtained. This fused feature vector not only improves the distinguishability of the semantic features of the embedding encoding of the user's critical pancreatitis patients' vital signs at different time points, but also provides a richer and more distinguishing feature representation for the deep learning model, thereby enhancing the generalization ability of the model and the ability to detect subsequent abnormal physiological states.
[0068] In a specific example, the process of calculating the semantic correlation coefficient between each embedded coding feature vector of the vital signs of patients with severe pancreatitis in the sequence of embedded coding feature vectors of the vital signs of patients with severe pancreatitis and the semantic clustering center feature vector of the vital signs of patients with severe pancreatitis to obtain a sequence of semantic correlation coefficients of the vital signs clustering center of patients with severe pancreatitis includes: calculating the logarithmic function value of the feature value at each position in the embedded coding feature vector of the vital signs of patients with severe pancreatitis to obtain the embedded coding semantic logarithmic representation vector of the vital signs of patients with severe pancreatitis; calculating the positional product between the embedded coding semantic logarithmic representation vector of the vital signs of patients with severe pancreatitis and the semantic clustering center feature vector of the vital signs of patients with severe pancreatitis to obtain the vital signs embedded semantic clustering center associated feature vector; dividing the vital signs embedded semantic clustering center associated feature vector by the number of feature values in the embedded coding feature vector of the vital signs of patients with severe pancreatitis to obtain the semantic correlation coefficient of the vital signs clustering center of patients with severe pancreatitis.
[0069] The sequence of semantic anti-correlation coefficients of the cluster centers of the vital signs of patients with severe pancreatitis is normalized to obtain a sequence of semantic anti-correlation weight coefficients of the cluster centers of the vital signs of patients with severe pancreatitis, including: using each semantic anti-correlation coefficient of the cluster centers of the vital signs of patients with severe pancreatitis in the sequence of semantic anti-correlation coefficients of the cluster centers of the vital signs of patients with severe pancreatitis as an exponent of a natural constant to calculate an exponential function value based on the position with the natural constant as the base to obtain a sequence of semantic anti-correlation class support coefficients of the cluster centers of the vital signs of patients with severe pancreatitis; calculating the semantic anti-correlation weight ... The sum of the semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis in the sequence of semantic anti-correlation class support coefficients is obtained to obtain the total semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis; the positional division between the semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis in the sequence of semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis and the sum of the semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis is calculated to obtain the sequence of semantic anti-correlation weight coefficients of the vital signs clustering centers of patients with severe pancreatitis.
[0070] In summary, in the above embodiment, the sequence of the vital signs of patients with severe pancreatitis embedded in the coded feature vector is processed through a reverse screening and strengthening network based on the anchor-centered vital sign temporal features to obtain a sequence of semantic feature vectors of the vital signs of patients with severe pancreatitis that are strengthened and focused on, including: the sequence of the vital signs of patients with severe pancreatitis embedded in the coded feature vector is processed through the reverse screening and strengthening network based on the anchor-centered vital sign temporal features with the following reverse screening and strengthening formula to obtain the sequence of semantic feature vectors of the vital signs of patients with severe pancreatitis that are strengthened and focused on; wherein the reverse screening and strengthening formula is:
[0071] ;
[0072] ;
[0073] ;
[0074] ;
[0075] ;
[0076] ;
[0077] ;
[0078] in, is a sequence of encoding feature vectors embedded with the vital signs of the patient with severe pancreatitis, represents a sequence of vectors, , , , The first, second, and third sequences of the vital signs of the patient with severe pancreatitis are embedded in the coded feature vector. , The vital signs of patients with severe pancreatitis are embedded in the encoded feature vector, is the number of vectors in the sequence of the vital signs of the patient with severe pancreatitis embedded in the encoding feature vector, is the semantic clustering center feature vector of the vital signs of patients with severe pancreatitis, For the The first part of the feature vector of the vital signs of patients with severe pancreatitis is embedded in the encoding The eigenvalues at the positions, is the first feature vector of the semantic clustering center of the vital signs of patients with severe pancreatitis The eigenvalues at the positions, For the The number of eigenvalues in the feature vector of the vital signs of patients with severe pancreatitis, For the The semantic correlation coefficient between the embedding encoding feature vector of the vital signs of patients with severe pancreatitis and the semantic clustering center feature vector of the vital signs of patients with severe pancreatitis, For the The semantic anti-correlation coefficient corresponding to the embedding encoding feature vector of the vital signs of patients with severe pancreatitis, is the exponential operation, For the The semantic anti-correlation weight coefficient corresponding to the embedding encoding feature vector of the vital signs of patients with severe pancreatitis, For the The suppressed embedding encoding feature vector of the vital signs of patients with severe pancreatitis corresponds to the embedding encoding feature vector of the vital signs of patients with severe pancreatitis. is the first in the sequence of the semantic feature vectors of the vital signs of patients with severe pancreatitis that are given enhanced attention. A semantic feature vector of vital signs of patients with severe pancreatitis that focuses on intensification.
[0079] In particular, the vital signs full temporal correlation feature extraction module 350 is used to perform temporal encoding on the sequence of semantic feature vectors of vital signs of patients with severe pancreatitis that are given enhanced attention to obtain temporal correlation feature vectors of vital signs of patients with severe pancreatitis. In the technical solution of the present application, the sequence of semantic feature vectors of vital signs of patients with severe pancreatitis that are given enhanced attention is encoded in a vital signs temporal encoder based on a Bi-LSTM model to extract the temporal correlation feature information of vital signs of patients with severe pancreatitis after enhanced attention, thereby obtaining the temporal correlation feature vectors of vital signs of patients with severe pancreatitis. It is worth mentioning that Bi-LSTM (Bidirectional Long Short-Term Memory Network) is a deep learning model that is commonly used to process sequence data and has the ability to remember long-term dependencies.
[0080] In particular, the urine volume time series correlation feature extraction module 360 is used to perform time series encoding on the time series of the urine volume of the patient with severe pancreatitis being monitored to obtain a urine volume time series correlation feature vector of the patient with severe pancreatitis. In the technical solution of the present application, after the time series of the urine volume of the monitored patient with severe pancreatitis is arranged as a urine volume time series input vector according to the time dimension, the urine volume time series input vector is encoded by a urine volume time series encoder based on a Bi-LSTM model to extract the patient's urine volume in the time dimension. Time series forward and backward correlation feature information, thereby obtaining a urine volume time series correlation feature vector of the patient with severe pancreatitis.
[0081] In particular, the vital sign-urine volume time series feature interactive fusion module 370 is used to obtain the vital sign-urine volume time series interactive fusion feature vector of the patient with severe pancreatitis through the patient multimodal monitoring data time series feature dynamic interactive fusion module based on the gated attention allocation mechanism. It should be understood that the vital sign time series association feature vector of the patient with severe pancreatitis and the urine volume time series association feature vector of the patient with severe pancreatitis reflect the embedded encoding semantic time series association feature and urine volume time series association feature of the patient with severe pancreatitis, respectively, and the dynamic change characteristics and patterns of the patient's vital signs and urine volume in the time dimension have different contributions and importance to the subsequent patient physiological state monitoring and abnormality detection tasks. Therefore, in order to be able to more comprehensively and accurately monitor and identify abnormalities of the patient's physiological state, in the technical solution of the present application, the time series correlation feature vector of the vital signs of the patient with severe pancreatitis and the time series correlation feature vector of the urine volume of the patient with severe pancreatitis are further processed through a dynamic interactive fusion module of the time series features of the multimodal monitoring data of the patient based on the gated attention allocation mechanism to obtain the vital sign-urine volume time series interactive fusion feature vector of the patient with severe pancreatitis. Through the processing of the dynamic interactive fusion module of the time series features of the multimodal monitoring data of the patient based on the gated attention allocation mechanism, the complex correlation and interaction between the patient's vital sign time series features and the urine volume time series features can be automatically learned and captured, and the importance and contribution of the patient's vital sign time series features and the urine volume time series features in subsequent tasks can be identified, thereby realizing the dynamic interactive fusion of the two, and providing a more comprehensive and accurate basis for subsequent abnormal monitoring of the patient's physiological state.
[0082] In an embodiment of the present application, the time series correlation feature vector of the vital signs of the patient with severe pancreatitis and the time series correlation feature vector of the urine volume of the patient with severe pancreatitis are passed through a dynamic interactive fusion module of the time series features of the multimodal monitoring data of the patient based on the gated attention allocation mechanism to obtain a time series interactive fusion feature vector of the vital signs and urine volume of the patient with severe pancreatitis, including: firstly inputting the time series correlation feature vector of the vital signs of the patient with severe pancreatitis and the time series correlation feature vector of the urine volume of the patient with severe pancreatitis into a feature joint module for cascade processing to obtain a vital sign-urine volume joint feature vector of the patient with severe pancreatitis. That is, the time series semantics of the patient's vital signs and the time series semantics of the urine volume are fused by cascading, and the implicit correlation feature information between the time series features of the patient's vital signs and the time series features of the urine volume is captured in the time dimension by using a one-dimensional convolutional layer. Next, the joint feature vector of the vital signs and urine volume of patients with severe pancreatitis is input into a one-dimensional convolutional layer for correlation feature extraction to obtain the implicit correlation feature vector of the vital signs and urine volume of patients with severe pancreatitis; after calculating the matrix multiplication of the implicit correlation feature vector of the vital signs and urine volume of patients with severe pancreatitis and the parameter matrix, the obtained feature vector is added to the bias vector by position to obtain the linear transformation of the implicit correlation feature vector of the vital signs and urine volume of patients with severe pancreatitis. Then, use The nonlinear activation function activates the linearly transformed critical pancreatitis patient vital sign-urine volume time series implicit correlation feature vector to obtain the critical pancreatitis patient vital sign-urine volume dynamic information nonlinear transformation activation vector; and calculates the position difference between the unit vector and the critical pancreatitis patient vital sign-urine volume dynamic information nonlinear transformation activation vector to obtain the difference critical pancreatitis patient vital sign-urine volume dynamic information nonlinear transformation activation vector. Further, using The function performs normalization operation on the critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector and the difference critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector to obtain the normalized critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector and the normalized difference critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector; then calculates the position point multiplication between the normalized critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector and the critical pancreatitis patient's vital sign time series association feature vector to obtain the importance weighted critical pancreatitis patient's vital sign time series association feature vector; calculates the normalized difference critical pancreatitis patient's vital sign -The nonlinear transformation activation vector of urine volume dynamic information and the urine volume time series correlation feature vector of the patient with severe pancreatitis are multiplied by the position point to obtain the importance-weighted urine volume time series correlation feature vector of the patient with severe pancreatitis, and the position-wise summation between the importance-weighted vital sign time series correlation feature vector of the patient with severe pancreatitis and the importance-weighted urine volume time series correlation feature vector of the patient with severe pancreatitis is calculated to obtain the vital sign-urine volume time series interactive fusion feature vector of the patient with severe pancreatitis; the position-wise summation between the importance-weighted vital sign time series correlation feature vector of the patient with severe pancreatitis and the importance-weighted urine volume time series correlation feature vector of the patient with severe pancreatitis is calculated to obtain the vital sign-urine volume time series interactive fusion feature vector of the patient with severe pancreatitis. That is, after the extracted vital sign-urine volume time series implicit correlation features are linearly transformed and nonlinearly activated modulated for gated attention selection, the extracted vital sign-urine volume time series implicit correlation features are used. The function is normalized to obtain the weight distribution vector of the patient's vital signs and urine volume time series semantics, and the weight distribution vector is used to perform dynamic interactive fusion processing of the vital sign time series associated feature vector of the patient with severe pancreatitis and the urine volume time series associated feature vector of the patient with severe pancreatitis, so as to utilize the fine-grained interactive features and semantics of the patient's vital sign time series features and the urine volume time series features to realize the vital sign-urine volume time series interactive fusion processing of the patient with severe pancreatitis, thereby providing a more comprehensive description of the physiological state based on the interactive complementary information between the patient's vital sign time series feature semantics and the urine volume time series feature semantics, so as to more accurately reflect the patient's actual condition, improve the performance and generalization ability of the model, and provide a more comprehensive and accurate data basis for subsequent detection of abnormal physiological state of patients.
[0083] In summary, in the above embodiment, the time series correlation feature vector of the vital signs of the patient with severe pancreatitis and the time series correlation feature vector of the urine volume of the patient with severe pancreatitis are processed through the dynamic interactive fusion module of the time series features of the multimodal monitoring data of the patient based on the gated attention allocation mechanism to obtain the vital sign-urine volume time series interactive fusion feature vector of the patient with severe pancreatitis, including: the time series correlation feature vector of the vital signs of the patient with severe pancreatitis and the time series correlation feature vector of the urine volume of the patient with severe pancreatitis are processed through the dynamic interactive fusion module of the time series features of the multimodal monitoring data of the patient based on the gated attention allocation mechanism with the following dynamic interactive fusion formula to obtain the vital sign-urine volume time series interactive fusion feature vector of the patient with severe pancreatitis; wherein, the dynamic interactive fusion formula is:
[0084] ;
[0085] ;
[0086] in, and are respectively the time series correlation feature vector of the vital signs of the patients with severe pancreatitis and the time series correlation feature vector of the urine volume of the patients with severe pancreatitis, represents a vector concatenation operation, is a vector concatenation operation, is the parameter matrix, is the bias vector, yes function, is the activation vector of the nonlinear transformation of the vital signs-urine volume dynamic information of patients with severe pancreatitis, yes function, represents a unit vector, It is the point multiplication by position. It is the time series interactive fusion feature vector of vital signs and urine volume of patients with severe pancreatitis.
[0087] In particular, the physiological state abnormality detection module 380 is used to determine whether the physiological state of the monitored severe pancreatitis patient is abnormal based on the vital sign-urine volume time series interactive fusion feature vector of the severe pancreatitis patient. In particular, in a specific example of the present application, the vital sign-urine volume time series interactive fusion feature vector of the severe pancreatitis patient is passed through a classifier-based abnormality identifier to obtain a recognition result, and the recognition result is used to indicate whether the physiological state of the monitored severe pancreatitis patient is abnormal. That is, the interactive fusion feature information between the vital sign embedding coding time series feature and the urine volume time series feature of the severe pancreatitis patient is used for classification processing, so as to identify and detect the abnormal physiological state of the severe pancreatitis patient. In this way, more intelligent nursing monitoring and management of severe pancreatitis patients can be achieved, and the ability of artificial intelligence and deep learning technology to process a large amount of data and learn the implicit feature patterns therein can be used to learn and extract the time series dynamic interactive correlation features and pattern information between the patient's vital signs and urine volume, thereby realizing personalized nursing monitoring for severe pancreatitis patients and improving the generalization ability of the system.
[0088] In a preferred example, the critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector is passed through a classifier-based abnormality identifier to obtain an identification result, including:
[0089] Calculate the characteristic mean of the critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector, and divide the characteristic mean by the difference between the maximum eigenvalue and the minimum eigenvalue of the critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector to obtain the critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature distribution representation value;
[0090] The result is the value of the distribution representation of the vital sign-urine volume time series interaction fusion feature of the patient with severe pancreatitis divided by the distribution representation of the vital sign-urine volume time series interaction fusion feature of the patient with severe pancreatitis to obtain the distribution modulation value of the vital sign-urine volume time series interaction fusion feature of the patient with severe pancreatitis;
[0091] Activating the critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector through a probabilistic function to obtain a probabilistic critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector;
[0092] After performing point-wise subtraction of the probabilistic critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector and the critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature distribution modulation value, taking the absolute value and calculating the negative of the logarithmic value with base 2 to obtain the probabilistic critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature distribution modulation information vector;
[0093] After dividing the distribution representation value of the vital sign-urine volume time series interactive fusion feature of the patient with severe pancreatitis by one minus the difference of each eigenvalue of the probabilistic time series interactive fusion feature vector of the vital sign-urine volume of the patient with severe pancreatitis, summing all eigenvalues of the probabilistic time series interactive fusion feature vector of the vital sign-urine volume of the patient with severe pancreatitis and dividing it by the length of the time series interactive fusion feature vector of the vital sign-urine volume of the patient with severe pancreatitis to obtain the probabilistic time series interactive fusion feature distribution modulation bias value of the vital sign-urine volume of the patient with severe pancreatitis;
[0094] The product of the probabilistic critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature distribution modulation information vector and the probabilistic critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature distribution modulation bias value and the weight as a hyperparameter is added to obtain an optimized critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector;
[0095] The optimized critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector is input into a classifier-based intelligent matcher to obtain a matching result.
[0096] Among them, the optimized vital sign-urine volume time series interactive fusion feature vector of patients with severe pancreatitis is expressed as:
[0097] ;
[0098] And among them:
[0099] ;
[0100] in, is the probabilistic critical pancreatitis patient vital sign-urine volume time series interactive fusion feature vector, The first in the probabilistic critical pancreatitis patient vital sign-urine volume time series interactive fusion feature vector is represented by eigenvalues, is the characteristic distribution value of the vital sign-urine volume time series interactive fusion feature of the patient with severe pancreatitis, is the length of the critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector, is the hyperparameter weight value, It means to make a difference by position. It means adding by position. represents the logarithm operation with base 2, is the optimized critical pancreatitis patient vital sign-urine volume time series interactive fusion feature vector, represents the feature mean of the vital sign-urine volume time series interactive fusion feature vector of the patient with severe pancreatitis, and They respectively represent the maximum eigenvalue and minimum eigenvalue of the vital sign-urine volume time series interactive fusion eigenvector of the patient with severe pancreatitis.
[0101] Considering that the time series association feature vector of the vital signs of patients with severe pancreatitis and the time series association feature vector of the urine volume of patients with severe pancreatitis respectively express the short-range and long-range bidirectional time series association features of the vital signs data and urine volume of the monitored severe pancreatitis patients, when the time series association feature vector of the vital signs of patients with severe pancreatitis and the time series association feature vector of the urine volume of patients with severe pancreatitis are passed through the dynamic interactive fusion module of the time series features of the multimodal monitoring data of patients based on the gated attention allocation mechanism, the obtained time series interactive fusion feature vector of the vital signs and urine volume of patients with severe pancreatitis will also lead to outlier fusion class reasoning mapping deviation due to insufficient coverage of the gated attention dynamic interactive fusion caused by the difference in the distribution of the time series features of the vital signs data and urine volume based on the respective source data representation modalities, thereby affecting the accuracy of the recognition result obtained by the time series interactive fusion feature vector of the vital signs and urine volume of patients with severe pancreatitis through the classifier-based anomaly identifier.
[0102] Based on this, in the above-mentioned preferred example, the eigenvalue-based probability information distribution planning of the vital sign-urine volume time series interaction fusion feature vector of the patient with severe pancreatitis is performed through the Bernoulli probability modulation distribution of the vital sign-urine volume time series interaction fusion feature vector of the patient with severe pancreatitis relative to the eigenvalue distribution, and the probability reverse mapping of the probability characteristics of the vital sign-urine volume time series interaction fusion feature vector of the patient with severe pancreatitis is used as the expanded coverage of the set mapping space of the vital sign-urine volume time series interaction fusion feature vector of the patient with severe pancreatitis, so as to independently understand the interaction path between the intuitive probability information distribution and the abstract probability space mapping of the vital sign-urine volume time series interaction fusion feature vector of the patient with severe pancreatitis, so as to improve the accuracy of the recognition result obtained by inputting the optimized vital sign-urine volume time series interaction fusion feature vector of the patient with severe pancreatitis into the classifier-based anomaly identifier by avoiding the counterfactual reasoning mapping of the outlier feature distribution of the vital sign-urine volume time series interaction fusion feature vector of the patient with severe pancreatitis to the class probability. In this way, the abnormal physiological state of patients with severe pancreatitis can be detected more accurately, thereby realizing personalized and intelligent nursing monitoring of patients with severe pancreatitis and improving the generalization ability of the system.
[0103] As described above, the critical pancreatitis patient care monitoring system 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a critical pancreatitis patient care monitoring algorithm. In one possible implementation, the critical pancreatitis patient care monitoring system 300 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the critical pancreatitis patient care monitoring system 300 can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the critical pancreatitis patient care monitoring system 300 can also be one of the many hardware modules of the wireless terminal.
[0104] Alternatively, in another example, the severe pancreatitis patient care monitoring system 300 and the wireless terminal may also be separate devices, and the severe pancreatitis patient care monitoring system 300 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.
[0105] Furthermore, a nursing monitoring method for patients with severe pancreatitis is also provided.
[0106] Figure 3 FIG. 1 is a flow chart of a method for monitoring the nursing care of a patient with severe pancreatitis according to an embodiment of the present application. Figure 3As shown, according to the embodiment of the present application, the nursing monitoring method for patients with severe pancreatitis includes the following steps: S1, obtaining a time series of vital sign data of the monitored patient with severe pancreatitis collected by a sensor group, wherein the vital sign data include heart rate value, blood pressure value, respiratory rate, body temperature value and blood oxygen saturation; S2, obtaining a time series of urine volume of the monitored patient with severe pancreatitis; S3, passing each vital sign data in the time series of the vital sign data of the monitored patient with severe pancreatitis through an embedding encoder based on a fully connected mapping layer to obtain a sequence of embedding encoding feature vectors of the vital signs of the patient with severe pancreatitis; S4, passing the sequence of embedding encoding feature vectors of the vital signs of the patient with severe pancreatitis through a reverse screening reinforcement network based on the vital sign time series features of anchor point centralization to obtain a reinforcement focus on the vital sign language of the patient with severe pancreatitis. S5, time-series encoding the sequence of semantic feature vectors of the vital signs of patients with enhanced attention to severe pancreatitis to obtain a time-series association feature vector of the vital signs of patients with severe pancreatitis; S6, time-series encoding the time series of urine volume of patients with severe pancreatitis monitoring to obtain a time-series association feature vector of urine volume of patients with severe pancreatitis; S7, the time-series association feature vector of the vital signs of patients with severe pancreatitis and the time-series association feature vector of urine volume of patients with severe pancreatitis are fused through a dynamic interactive fusion module of time-series features of multimodal monitoring data of patients with severe pancreatitis based on a gated attention allocation mechanism to obtain a time-series interactive fusion feature vector of vital signs-urine volume of patients with severe pancreatitis; S8, based on the time-series interactive fusion feature vector of vital signs-urine volume of patients with severe pancreatitis, determine whether there is any abnormality in the physiological state of the monitored severe pancreatitis patient.
[0107] In summary, according to the embodiment of the present application, the nursing monitoring method for patients with severe pancreatitis is explained, which extracts the time-series dynamic change patterns and interactive correlation relationships of the vital signs data and urine volume data of patients with severe pancreatitis by adopting data processing and analysis algorithms based on artificial intelligence and deep learning, so as to learn and characterize the dynamic interactive fusion characteristics between the patient's vital signs time-series semantics and the urine volume time-series semantics, so as to monitor the physiological state and detect abnormalities of patients with severe pancreatitis, and issue early warnings in time when abnormalities are found. In this way, more intelligent nursing monitoring and management of patients with severe pancreatitis can be achieved. Furthermore, relying on the ability of artificial intelligence and deep learning technology to process large amounts of data and learn the implicit feature patterns therein, the time-series dynamic interactive correlation characteristics and pattern information between the patient's vital signs and urine volume can be learned and extracted, thereby realizing personalized nursing monitoring of patients with severe pancreatitis and improving the generalization ability of the system.
[0108] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.
Claims
1. A nursing monitoring system for patients with severe pancreatitis, characterized in that: include: A vital sign data acquisition module, used to acquire a time series of vital sign data of a monitored severe pancreatitis patient acquired by a sensor group, wherein the vital sign data includes heart rate value, blood pressure value, respiratory rate, body temperature value and blood oxygen saturation; A urine volume data acquisition module, used to obtain the time series of urine volume of the monitored severe pancreatitis patient; A vital sign embedding mapping association coding module is used to pass each vital sign data in the time series of the monitored severe pancreatitis patient's vital sign data through an embedding encoder based on a fully connected mapping layer to obtain a sequence of embedding coding feature vectors of the severe pancreatitis patient's vital signs; A vital sign time series feature enhanced expression module is used to embed the vital signs of the patient with severe pancreatitis into a sequence of encoded feature vectors through a reverse screening and enhancement network based on the vital sign time series features of anchor point centralization to obtain a sequence of semantic feature vectors that enhances the focus on the vital signs of the patient with severe pancreatitis; A vital sign full time series correlation feature extraction module is used to perform time series encoding on the sequence of the semantic feature vectors of the vital signs of patients with severe pancreatitis to obtain a time series correlation feature vector of the vital signs of patients with severe pancreatitis; A urine volume time series correlation feature extraction module is used to perform time series encoding on the time series of the urine volume of the patient with severe pancreatitis to obtain a urine volume time series correlation feature vector of the patient with severe pancreatitis; A vital sign-urine volume time series feature interactive fusion module, which is used to obtain a critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector by combining the critical pancreatitis patient's vital sign time series associated feature vector and the critical pancreatitis patient's urine volume time series associated feature vector through a patient multimodal monitoring data time series feature dynamic interactive fusion module based on a gated attention allocation mechanism; A physiological state abnormality detection module, used to determine whether the physiological state of the monitored severe pancreatitis patient is abnormal based on the vital sign-urine volume time series interactive fusion feature vector of the severe pancreatitis patient; The vital signs temporal feature enhanced expression module includes: A position mean calculation unit, used for calculating the position mean vector of the sequence of the embedding encoding feature vector of the vital signs of the patient with severe pancreatitis to obtain a semantic clustering center feature vector of the vital signs of the patient with severe pancreatitis; A semantic correlation coefficient calculation unit, used to calculate the semantic correlation coefficient between each critical pancreatitis patient's vital sign embedding coding feature vector in the sequence of critical pancreatitis patient's vital sign embedding coding feature vector and the critical pancreatitis patient's vital sign semantic clustering center feature vector to obtain a sequence of critical pancreatitis patient's vital sign clustering center semantic correlation coefficients; A cluster center semantic anti-correlation coefficient extraction unit, used for taking the inverse of each of the critical pancreatitis patients' vital signs cluster center semantic correlation coefficients in the sequence of critical pancreatitis patients' vital signs cluster center semantic correlation coefficients to obtain a sequence of critical pancreatitis patients' vital signs cluster center semantic anti-correlation coefficients; A normalization unit, used for normalizing the sequence of semantic anti-correlation coefficients of the vital signs clustering centers of the patients with severe pancreatitis to obtain a sequence of semantic anti-correlation weight coefficients of the vital signs clustering centers of the patients with severe pancreatitis; A weighting unit, used for weighting each critical pancreatitis patient vital sign embedded coding feature vector in the sequence of critical pancreatitis patient vital sign embedded coding feature vectors by using each critical pancreatitis patient vital sign cluster center semantic anti-correlation weight coefficient in the sequence of critical pancreatitis patient vital sign cluster center semantic anti-correlation weight coefficient as a weight to obtain a sequence of critical pancreatitis patient vital sign embedded coding feature vectors; The enhanced attention unit is used to calculate the positional difference between the sequence of the embedded coding feature vectors of the vital signs of patients with severe pancreatitis and the sequence of the embedded coding feature vectors of the vital signs of patients with severe pancreatitis corresponding to each group and the embedded coding feature vectors of the vital signs of patients with severe pancreatitis to obtain the sequence of the enhanced attention semantic feature vectors of the vital signs of patients with severe pancreatitis.
2. The severe pancreatitis patient care monitoring system according to claim 1, characterized in that: The semantic correlation coefficient calculation unit is used to: Calculating the logarithmic function value of the feature value at each position in the feature vector of the embedding code of the vital signs of the patient with severe pancreatitis to obtain the semantic logarithmic representation vector of the embedding code of the vital signs of the patient with severe pancreatitis; Calculating the positional product between the semantic logarithmic representation vector of the embedding encoding of the vital signs of the patient with severe pancreatitis and the semantic clustering center feature vector of the vital signs of the patient with severe pancreatitis to obtain the vital signs embedding semantic clustering center associated feature vector; The semantic correlation coefficient of the vital sign cluster center of the patient with severe pancreatitis is obtained by dividing the associated feature vector of the vital sign embedding semantic cluster center by the number of eigenvalues in the encoding feature vector of the vital sign embedding of the patient with severe pancreatitis.
3. The severe pancreatitis patient care monitoring system according to claim 2, characterized in that: The normalization unit is used for: Using each semantic anti-correlation coefficient of the vital signs cluster center of patients with severe pancreatitis in the sequence of semantic anti-correlation coefficients of the vital signs cluster center of patients with severe pancreatitis as an exponent of a natural constant to calculate an exponential function value based on the natural constant according to the position to obtain a sequence of semantic anti-correlation class support coefficients of the vital signs cluster center of patients with severe pancreatitis; Calculating the sum of the semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis in the sequence of semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis to obtain the sum of the semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis; Calculate the positional division between the semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis and the sum of the semantic anti-correlation class support coefficients of the vital signs clustering centers of patients with severe pancreatitis to obtain a sequence of semantic anti-correlation weight coefficients of the vital signs clustering centers of patients with severe pancreatitis.
4. The severe pancreatitis patient care monitoring system according to claim 3, characterized in that: The vital signs full time series correlation feature extraction module is used to: pass the sequence of the semantic feature vectors of the vital signs of patients with severe pancreatitis who are given intensive attention through a vital signs time series encoder based on a Bi-LSTM model to obtain the vital signs time series correlation feature vectors of the patients with severe pancreatitis.
5. The severe pancreatitis patient care monitoring system according to claim 4, characterized in that: The urine volume time series correlation feature extraction module is used to: arrange the time series of urine volume of the monitored severe pancreatitis patient according to the time dimension into a urine volume time series input vector, and then pass the urine volume time series input vector through a urine volume time series encoder based on a Bi-LSTM model to obtain the urine volume time series correlation feature vector of the severe pancreatitis patient.
6. The severe pancreatitis patient care monitoring system according to claim 5, characterized in that: The vital sign-urine volume time series feature interactive fusion module is used to: Inputting the critical pancreatitis patient's vital sign time series correlation feature vector and the critical pancreatitis patient's urine volume time series correlation feature vector into a feature combination module for cascade processing to obtain a critical pancreatitis patient's vital sign-urine volume combination feature vector; Inputting the joint feature vector of the vital signs and urine volume of the patient with severe pancreatitis into a one-dimensional convolutional layer to extract the associated features to obtain the implicit associated feature vector of the vital signs and urine volume of the patient with severe pancreatitis; After calculating the matrix multiplication of the critically ill pancreatitis patient's vital sign-urine volume time series implicit correlation eigenvector and the parameter matrix, the obtained eigenvector is added to the bias vector by position to obtain a linearly transformed critically ill pancreatitis patient's vital sign-urine volume time series implicit correlation eigenvector; use The nonlinear activation function activates the linearly transformed critical pancreatitis patient's vital sign-urine volume time series implicit correlation feature vector to obtain the critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector; Calculating the position difference between the unit vector and the critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector to obtain the difference critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector; use The function performs a normalization operation on the critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector and the difference critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector to obtain a normalized critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector and a normalized difference critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector; Calculate the point-by-point multiplication between the normalized critical pancreatitis patient's vital sign-urine volume dynamic information nonlinear transformation activation vector and the critical pancreatitis patient's vital sign time series association feature vector to obtain an importance-weighted critical pancreatitis patient's vital sign time series association feature vector; Calculate the point-by-point multiplication between the normalized difference critical pancreatitis patient vital sign-urine volume dynamic information nonlinear transformation activation vector and the critical pancreatitis patient urine volume time series associated feature vector to obtain the importance weighted critical pancreatitis patient urine volume time series associated feature vector; The positional sum of the importance-weighted critical pancreatitis patient's vital sign time series associated feature vector and the importance-weighted critical pancreatitis patient's urine volume time series associated feature vector is calculated to obtain the critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector.
7. The severe pancreatitis patient care monitoring system according to claim 6, characterized in that: The physiological state abnormality detection module is used to: pass the vital sign-urine volume time series interactive fusion feature vector of the severe pancreatitis patient through a classifier-based abnormality identifier to obtain a recognition result, and the recognition result is used to indicate whether the physiological state of the monitored severe pancreatitis patient is abnormal.
8. A nursing monitoring method for patients with severe pancreatitis, using the nursing monitoring system for patients with severe pancreatitis according to claim 1, characterized in that: include: Acquire a time series of vital sign data of a monitored severe pancreatitis patient collected by a sensor group, wherein the vital sign data includes a heart rate value, a blood pressure value, a respiratory rate, a body temperature value, and a blood oxygen saturation; Obtaining a time series of urine volume of the monitored severe pancreatitis patient; Each vital sign data in the time series of the vital sign data of the monitored severe pancreatitis patient is passed through an embedding encoder based on a fully connected mapping layer to obtain a sequence of embedding encoding feature vectors of the vital signs of the severe pancreatitis patient; The sequence of the critical pancreatitis patient's vital signs embedded in the encoding feature vector is subjected to a reverse screening reinforcement network based on the anchor-centered critical pancreatitis patient's vital signs time series feature to obtain a sequence of the critical pancreatitis patient's vital signs semantic feature vector that is enhanced; Performing temporal coding on the sequence of the semantic feature vectors of the vital signs of patients with severe pancreatitis that are given intensified attention to obtain a temporal association feature vector of the vital signs of patients with severe pancreatitis; Performing time series encoding on the time series of urine volume monitored in patients with severe pancreatitis to obtain a time series correlation feature vector of urine volume in patients with severe pancreatitis; The critical pancreatitis patient's vital sign time series correlation feature vector and the critical pancreatitis patient's urine volume time series correlation feature vector are passed through a patient multimodal monitoring data time series feature dynamic interactive fusion module based on a gated attention allocation mechanism to obtain a critical pancreatitis patient's vital sign-urine volume time series interactive fusion feature vector; Based on the vital sign-urine volume time series interactive fusion feature vector of the severe pancreatitis patient, it is determined whether the physiological state of the monitored severe pancreatitis patient is abnormal.
