Intelligent pre-examination and triage system and method for emergency patients

Abstract

The present application discloses an intelligent pre-examination and triage system and method for emergency patients, which relates to the field of intelligent recommendation. It obtains the description of the condition of the emergency patient and collects the time series data of the patient's vital signs at the same time, performs semantic analysis and understanding on the description of the condition of the emergency patient, and performs dynamic analysis and multimodal fusion representation on the time series semantics of the patient's vital signs. Then, the multimodal fusion semantics of the patient's vital signs are used to constrain the expression of the semantics of the description of the condition. In this way, the vital signs can be used as the reference information to automatically identify the symptoms of the patient's condition description, so as to evaluate and grade the patient's condition, so as to automatically match the patient to the corresponding registration department, and determine whether to give priority to the treatment according to the patient's condition. In this way, a more intelligent pre-examination and triage of emergency patients can be achieved, avoiding the accuracy of the pre-examination and triage and the timeliness of treatment affected by the patient's inaccurate description of the condition.

Description

Technical Field

[0001] The present application relates to the field of intelligent recommendation, and more specifically, to an intelligent pre-examination and triage system and method for emergency patients. Background Art

[0002] Emergency medicine is an important part of the medical system. It focuses on dealing with sudden and serious health problems that require rapid response and efficient processing. With the influence of factors such as the aging population and the high incidence of chronic diseases, the number of emergency visits has increased year by year, bringing tremendous pressure to the emergency department. Pre-examination and triage of emergency patients is an important part of the emergency department. Its purpose is to quickly and accurately identify the severity of the patient's condition and divert them to the appropriate department for treatment. The accuracy of pre-examination and triage directly affects the patient's subsequent treatment effect and the rational allocation of medical resources.

[0003] However, traditional emergency pre-examination and triage systems often rely on the experience and subjective judgment of medical staff, which may lead to subjective and inconsistent assessments, which may cause emergency patients in need to be assigned to other departments, resulting in uneven distribution of medical resources and failure of emergency patients to receive timely treatment. In addition, when pre-examination and triage of emergency patients, patients may not be able to accurately describe their condition due to tension, pain or other reasons, and thus cannot accurately identify the severity of the patient's condition, affecting the accuracy of pre-examination and triage and the timeliness of treatment.

[0004] Therefore, an intelligent pre-examination and triage system and method for emergency patients is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed.

[0006] According to one aspect of the present application, an intelligent pre-examination and triage system for emergency patients is provided, comprising:

[0007] Patient condition description collection module, used to obtain the condition description of emergency patients;

[0008] A patient vital sign data acquisition module, used to acquire a time series of vital sign data of the emergency patient acquired by a vital sign monitor, wherein the vital sign data includes heart rate, blood pressure and blood oxygen saturation;

[0009] A multimodal vital sign time series semantic fusion representation module, used for performing time series feature extraction and semantic fusion processing on the time series of the vital sign data to obtain a multimodal time series semantic representation vector of the vital sign;

[0010] A patient condition description word granularity semantic encoding module is used to perform word granularity-based semantic encoding on the emergency patient's condition description to obtain a sequence of condition description word granularity semantic association feature vectors;

[0011] A granular semantic center screening and strengthening module for the condition description words is used to pass the sequence of the granular semantic association feature vectors of the condition description words through a word granular semantic reverse screening and strengthening module based on the description semantic center to obtain a sequence of semantic strengthening expression feature vectors of the emergency patient condition description keywords;

[0012] A module for expressing the semantic constraints of patient condition description guided by vital signs, which is used to use the multimodal time series semantic representation vector of the vital signs as a reference information feature vector, and to perform semantic constraint aggregation processing based on information distillation on the sequence of the semantic enhancement expression feature vectors of the emergency patient condition description keywords to obtain a semantic constraint enhancement expression feature vector of the patient condition description guided by the time series semantics of the vital signs;

[0013] The registration department category recommendation module is used to determine the recommended registration department category and generate a registration number based on the semantic constraint reinforcement expression feature vector of the patient's condition description under the guidance of the vital sign time series semantics.

[0014] In the above-mentioned intelligent pre-examination and triage system for emergency patients, the multimodal vital sign temporal semantic fusion representation module is used to:

[0015] an arranging unit, configured to arrange the time series of the vital sign data according to a sample dimension to obtain a heart rate time series input vector, a blood pressure time series input vector, and a blood oxygen saturation time series input vector;

[0016] A vital sign time series feature extraction unit, used for passing the heart rate time series input vector, the blood pressure time series input vector and the blood oxygen saturation time series input vector through a vital sign time series feature extractor based on a one-dimensional convolutional neural network model to obtain a heart rate time series feature vector, a blood pressure time series feature vector and a blood oxygen saturation time series feature vector;

[0017] The vital sign multimodal time series semantic fusion unit is used to obtain the vital sign multimodal time series semantic representation vector by fusing the heart rate time series feature vector, the blood pressure time series feature vector and the blood oxygen saturation time series feature vector through a vital sign multimodal time series semantic fuser based on a Bayesian probability network.

[0018] In the above-mentioned intelligent pre-examination and triage system for emergency patients, the patient condition description word granularity semantic encoding module is used to: perform word segmentation processing on the emergency patient's condition description and then pass it through a condition description semantic encoder including a word embedding layer to obtain a sequence of the condition description word granularity semantic association feature vectors.

[0019] In the above-mentioned intelligent pre-examination and triage system for emergency patients, the granular semantic center screening and strengthening module for the condition description words includes:

[0020] A mean vector calculation unit, used to calculate the mean vector of the sequence of the granular semantic association feature vectors of the condition description words to obtain a distribution center vector of the granular sequence of the condition description words;

[0021] A condition description word granularity semantic association coefficient calculation unit is used to calculate the correlation coefficient between each condition description word granularity semantic association feature vector in the sequence of the condition description word granularity semantic association feature vector and the condition description word granularity sequence distribution center vector to obtain a sequence of condition description word granularity semantic association coefficients;

[0022] A condition description word granularity semantic anti-correlation coefficient calculation unit, used to calculate the inverse of each condition description word granularity semantic correlation coefficient in the sequence of condition description word granularity semantic correlation coefficients to obtain a sequence of condition description word granularity semantic anti-correlation coefficients;

[0023] A normalization unit, used for normalizing the sequence of the granular semantic anti-correlation coefficients of the condition description words using a Softmax function to obtain a sequence of the granular semantic anti-correlation energy coefficients of the condition description words;

[0024] A condition description reverse inhibition disease unit is used to weight each condition description word granular semantic anti-association energy coefficient in the sequence of condition description word granular semantic anti-association energy coefficients as a weight, and to obtain a sequence of reverse inhibition condition description word granular semantic association feature vectors by weighting each condition description word granular semantic association feature vector in the sequence of condition description word granular semantic association feature vectors;

[0025] The condition description keyword semantic enhancement expression unit is used to calculate the positional difference between each group of corresponding condition description word granularity semantic association feature vectors and the reverse inhibition condition description word granularity semantic association feature vectors between the sequence of the condition description word granularity semantic association feature vectors and the sequence of the reverse inhibition condition description word granularity semantic association feature vectors to obtain the sequence of the emergency patient condition description keyword semantic enhancement expression feature vectors.

[0026] In the above-mentioned intelligent pre-examination and triage system for emergency patients, the granular semantic association coefficient calculation unit of the condition description word is used to:

[0027] Calculate the point-by-point multiplication between the granular semantic association feature vector of the condition description word and the granular sequence distribution center vector of the condition description word to obtain the granular semantic association class support feature vector of the condition description word;

[0028] Calculate the sum of the feature values ​​of each position in the condition description word granular semantic association class support feature vector to obtain the condition description word granular semantic association class support sum value;

[0029] Calculate the square sum of the feature values ​​at each position in the granular semantic association feature vector of the condition description word and the square sum of the feature values ​​at each position in the granular sequence distribution center vector of the condition description word to obtain the granular semantic association square value of the condition description word and the granular sequence distribution center square value of the condition description word;

[0030] Calculate the square root of the sum of the square value of the semantic association of the condition description word granularity and the square value of the center of the condition description word granularity sequence distribution to obtain the condition description word granularity association value;

[0031] The condition description word granular semantic association class support sum value is divided by the condition description word granular association value to obtain the condition description word granular semantic association coefficient.

[0032] In the above-mentioned intelligent pre-examination and triage system for emergency patients, the normalization unit is used to:

[0033] Using each condition description word granularity semantic anti-correlation coefficient of the sequence of condition description word granularity semantic anti-correlation coefficients 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 condition description word granularity semantic anti-correlation exponential coefficients;

[0034] Calculating the sum of the granular semantic anti-correlation index coefficients of each condition description word in the sequence of the granular semantic anti-correlation index coefficients of the condition description word to obtain the total sum of the granular semantic anti-correlation coefficients of the condition description word;

[0035] Calculate the sum of the total sum of the granular semantic anti-association coefficients of the condition description words and a constant of one to obtain the granular semantic anti-association modulation coefficient of the condition description words;

[0036] Each condition description word granularity semantic anti-correlation index coefficient in the sequence of condition description word granularity semantic anti-correlation index coefficients is divided by the condition description word granularity semantic anti-correlation modulation coefficient to obtain the sequence of condition description word granularity semantic anti-correlation energy coefficients.

[0037] In the above-mentioned intelligent pre-examination and triage system for emergency patients, the vital signs-guided patient condition description semantic constraint expression module is used to:

[0038] Calculate each emergency patient condition description keyword semantic enhancement expression feature vector in the sequence of emergency patient condition description keyword semantic enhancement expression feature vectors, multiply it with the corresponding weight matrix, and then add it with the corresponding bias vector by position to obtain a sequence of linearly modulated emergency patient condition description keyword semantic enhancement expression feature vectors;

[0039] Using a sigmoid function to activate the sequence of semantic enhancement expression feature vectors of the emergency patient's condition description keywords, and then multiplying them with the transposed vector of the benchmark information feature vector to obtain a sequence of patient vital signs-condition description semantic relevance;

[0040] Calculating the sum of the semantic relevances of each patient's vital sign-condition description in the sequence of the semantic relevances of the patient's vital sign-condition description to obtain the sum of the semantic relevances of the patient's vital sign-condition description;

[0041] Dividing each patient vital sign-condition description semantic relevance in the patient vital sign-condition description semantic relevance sequence by the sum of the patient vital sign-condition description semantic relevances to obtain a sequence of patient vital sign-condition description semantic association weights;

[0042] Taking each patient's vital sign-condition description semantic association weight in the sequence of the patient's vital sign-condition description semantic association weights as a weighting coefficient, the weighted sum of the sequence of the emergency patient's condition description keyword semantic enhanced expression feature vectors is calculated to obtain the patient's condition description semantic constraint enhanced expression feature vector under the guidance of the vital sign time series semantics.

[0043] In the above-mentioned intelligent pre-examination and triage system for emergency patients, the registration department category recommendation module is used to:

[0044] The semantic constraint enhanced expression feature vector of the patient's condition description guided by the time series semantics of the vital signs is passed through a registration department category recommender based on a classifier to obtain a recommendation result, and the recommendation result is used to represent the recommended registration department category;

[0045] Based on the recommended registration department category, a registration number is generated.

[0046] According to another aspect of the present application, there is provided an intelligent pre-examination and triage method for emergency patients, comprising:

[0047] Obtain descriptions of emergency patients’ medical conditions;

[0048] Acquire a time series of vital sign data of the emergency patient collected by a vital sign monitor, wherein the vital sign data includes heart rate, blood pressure and blood oxygen saturation;

[0049] Performing time series feature extraction and semantic fusion processing on the time series of the vital sign data to obtain a multimodal time series semantic representation vector of the vital sign;

[0050] Performing word-granularity-based semantic encoding on the description of the emergency patient's condition to obtain a sequence of word-granularity semantic association feature vectors of the condition description;

[0051] The sequence of the condition description word granularity semantic association feature vectors is passed through a word granularity semantic reverse screening and strengthening module based on the description semantic center to obtain a sequence of emergency patient condition description keyword semantic strengthening expression feature vectors;

[0052] Taking the multimodal time series semantic representation vector of vital signs as the reference information feature vector, the sequence of the semantic enhancement expression feature vectors of the emergency patient condition description keywords is subjected to semantic constraint aggregation processing based on information distillation to obtain the semantic constraint enhancement expression feature vector of the patient condition description under the guidance of the time series semantics of vital signs;

[0053] Based on the semantically constrained enhanced expression feature vector of the patient's condition description guided by the vital signs temporal semantics, the recommended registration department category is determined and a registration number is generated.

[0054] Compared with the prior art, the present application provides an intelligent pre-examination and triage system and method for emergency patients, which obtains the description of the condition of the emergency patient, collects the time series data of the patient's vital signs, and introduces data processing and semantic understanding algorithms based on artificial intelligence and deep learning at the back end to perform semantic analysis and understanding on the description of the condition of the emergency patient, and dynamically analyzes and multimodal fusion representation of the time series semantics of the patient's vital signs, and then uses the patient's vital signs time series multimodal fusion semantics to constrain the expression of the semantics of the condition description. In this way, the vital signs can be used as the benchmark information to automatically identify the symptoms of the patient's condition description, so as to evaluate and grade the patient's condition, so as to automatically match the patient to the corresponding registration department, and determine whether to give priority to the treatment according to the patient's condition. In this way, a more intelligent pre-examination and triage of emergency patients can be achieved, avoiding the accuracy of the pre-examination and triage and the timeliness of treatment affected by the patient's inaccurate description of the condition. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] 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.

[0056] Figure 1 A block diagram of an intelligent pre-examination and triage system for emergency patients according to an embodiment of the present application;

[0057] Figure 2 A data flow diagram of an intelligent pre-examination and triage system for emergency patients according to an embodiment of the present application;

[0058] Figure 3It is a block diagram of a multimodal vital sign temporal semantic fusion representation module in an intelligent pre-examination and triage system for emergency patients according to an embodiment of the present application;

[0059] Figure 4 The present invention is a flowchart of an intelligent pre-examination and triage method for emergency patients according to an embodiment of the present application. DETAILED DESCRIPTION

[0060] 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.

[0061] 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.

[0062] 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.

[0063] 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.

[0064] 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.

[0065] Traditional emergency pre-examination and triage systems often rely on the experience and subjective judgment of medical staff, which may lead to subjective and inconsistent evaluations, which may cause emergency patients in need to be assigned to other departments, resulting in uneven distribution of medical resources and failure of emergency patients to receive timely treatment. In addition, when pre-examination and triage of emergency patients, patients may not be able to accurately describe their condition due to tension, pain or other reasons, and thus cannot accurately identify the severity of the patient's condition, affecting the accuracy of pre-examination and triage and the timeliness of treatment. Therefore, an intelligent pre-examination and triage system for emergency patients is desired.

[0066] In the technical solution of the present application, an intelligent pre-examination and triage system for emergency patients is proposed. Figure 1 It is a block diagram of an intelligent pre-examination and triage system for emergency patients according to an embodiment of the present application. Figure 2 FIG. 1 is a data flow diagram of an intelligent pre-examination and triage system for emergency patients according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the intelligent pre-examination and triage system 300 for emergency patients includes: a patient condition description acquisition module 310, which is used to obtain the condition description of the emergency patient; a patient vital sign data acquisition module 320, which is used to obtain the time series of the vital sign data of the emergency patient collected by the vital sign monitor, wherein the vital sign data includes heart rate, blood pressure and blood oxygen saturation; a multimodal vital sign time series semantic fusion representation module 330, which is used to perform time series feature extraction and semantic fusion processing on the time series of the vital sign data to obtain a multimodal time series semantic representation vector of the vital sign; a patient condition description word granularity semantic encoding module 340, which is used to perform word granularity-based condition description semantic encoding on the emergency patient's condition description to obtain a sequence of condition description word granularity semantic association feature vectors; condition description word granularity semantic encoding module 340 The center screening and strengthening module 350 is used to pass the sequence of the word-granular semantic association feature vectors of the condition description through the word-granular semantic reverse screening and strengthening module based on the description semantic center to obtain a sequence of the semantic enhancement expression feature vectors of the emergency patient condition description keywords; the vital signs guided patient condition description semantic constraint expression module 360 ​​is used to use the vital signs multimodal temporal semantic representation vector as the benchmark information feature vector, and perform information distillation-based semantic constraint aggregation processing on the sequence of the emergency patient condition description keyword semantic enhancement expression feature vectors to obtain the patient condition description semantic constraint enhancement expression feature vector under the guidance of vital signs temporal semantics; the registration department category recommendation module 370 is used to determine the recommended registration department category based on the patient condition description semantic constraint enhancement expression feature vector under the guidance of vital signs temporal semantics, and generate a registration number.

[0067] In particular, the patient condition description acquisition module 310 and the patient vital sign data acquisition module 320 are used to obtain the condition description of the emergency patient; and to obtain the time series of the vital sign data of the emergency patient collected by the vital sign monitor, wherein the vital sign data includes heart rate, blood pressure and blood oxygen saturation. It is worth mentioning that the vital sign monitor is a medical device used to continuously monitor and record one or more vital signs, such as heart rate, respiratory rate, blood oxygen saturation and body temperature.

[0068] In particular, the multimodal vital sign time series semantic fusion representation module 330 is used to extract time series features and perform semantic fusion processing on the time series of the vital sign data to obtain a multimodal time series semantic representation vector of the vital sign. In particular, in a specific example of the present application, Figure 3 As shown, the multimodal vital sign time series semantic fusion representation module 330 includes: an arrangement unit 331, which is used to arrange the time series of the vital sign data according to the sample dimension to obtain a heart rate time series input vector, a blood pressure time series input vector and a blood oxygen saturation time series input vector; a vital sign time series feature extraction unit 332, which is used to pass the heart rate time series input vector, the blood pressure time series input vector and the blood oxygen saturation time series input vector through a vital sign time series feature extractor based on a one-dimensional convolutional neural network model to obtain a heart rate time series feature vector, a blood pressure time series feature vector and a blood oxygen saturation time series feature vector; a vital sign multimodal time series semantic fusion unit 333, which is used to pass the heart rate time series feature vector, the blood pressure time series feature vector and the blood oxygen saturation time series feature vector through a vital sign multimodal time series semantic fusion unit based on a Bayesian probability network to obtain the vital sign multimodal time series semantic representation vector.

[0069] Specifically, the arrangement unit 331 is used to arrange the time series of the vital signs data according to the sample dimension to obtain a heart rate time series input vector, a blood pressure time series input vector, and a blood oxygen saturation time series input vector. Considering that the patient's vital signs data have a dynamic change law and correlation characteristics in the time dimension, in order to capture the time series correlation relationship and characteristics of each patient's vital signs data in the time dimension, in the technical solution of the present application, the time series of the vital signs data is arranged according to the sample dimension to obtain a heart rate time series input vector, a blood pressure time series input vector, and a blood oxygen saturation time series input vector

[0070] Specifically, the vital sign time series feature extraction unit 332 is used to pass the heart rate time series input vector, the blood pressure time series input vector and the blood oxygen saturation time series input vector through a vital sign time series feature extractor based on a one-dimensional convolutional neural network model to obtain a heart rate time series feature vector, a blood pressure time series feature vector and a blood oxygen saturation time series feature vector. That is, in the technical solution of the present application, the heart rate time series input vector, the blood pressure time series input vector and the blood oxygen saturation time series input vector are subjected to feature mining in a vital sign time series feature extractor based on a one-dimensional convolutional neural network model to respectively extract the time series implicit correlation feature information of the heart rate, blood pressure and blood oxygen saturation of the emergency patient in the time dimension, thereby obtaining a heart rate time series feature vector, a blood pressure time series feature vector and a blood oxygen saturation time series feature vector. It is worth mentioning that a one-dimensional convolutional neural network (1D CNN) is a neural network specifically used to process one-dimensional data, such as time series or signals.

[0071] Specifically, the vital sign multimodal time series semantic fusion unit 333 is used to pass the heart rate time series feature vector, the blood pressure time series feature vector and the blood oxygen saturation time series feature vector through a vital sign multimodal time series semantic fusion device based on a Bayesian probability network to obtain the vital sign multimodal time series semantic representation vector. Since the various vital sign data of emergency patients not only have their own time series dynamic change characteristics and features in the time dimension, but also there are implicit association patterns and complex relationships between these vital sign time series semantics. Based on this, in order to capture the implicit association feature information between the various vital sign time series semantics of emergency patients, so as to more comprehensively reflect the patient's physiological state, so as to help the subsequent description of the patient's condition to constrain the expression and realize intelligent pre-examination and triage, in the technical solution of the present application, the heart rate time series feature vector, the blood pressure time series feature vector and the blood oxygen saturation time series feature vector are further passed through a vital sign multimodal time series semantic fusion device based on a Bayesian probability network to obtain a vital sign multimodal time series semantic representation vector. Through the processing of the vital sign multimodal time series semantic fuser based on the Bayesian probability network, the heart rate time series features, blood pressure time series features and blood oxygen saturation time series features of emergency patients can be well fused, so as to utilize the implicit association patterns and complex dependencies between these vital sign time series semantics to obtain a more robust and interpretable vital sign representation.

[0072] It is worth mentioning that in other specific examples of the present application, the time series of the vital signs data can also be subjected to time series feature extraction and semantic fusion processing in other ways to obtain a multimodal time series semantic representation vector of vital signs, for example: input the time series of the vital signs data; use an autoregressive model (such as ARIMA or LSTM) to capture the dependencies of the time series data and extract time series features; directly connect or splice the extracted features of different vital signs time series data together to form a multimodal time series feature vector; use multiple models (such as LSTM or Transformer) to process different vital signs time series data separately, and then fuse the model outputs together; use an attention mechanism to assign different weights to different vital signs time series data, and fuse them according to their importance; after time series feature extraction and semantic fusion, a multimodal time series semantic representation vector of vital signs is obtained.

[0073] In particular, the patient condition description word granularity semantic encoding module 340 is used to perform word granularity-based semantic encoding of the emergency patient's condition description to obtain a sequence of word granularity semantic association feature vectors of the condition description. In a specific example of the present application, the emergency patient's condition description is processed by word segmentation and then passed through a condition description semantic encoder including a word embedding layer to obtain a sequence of word granularity semantic association feature vectors of the condition description. That is, the emergency patient's condition description is processed by word segmentation and then passed through a condition description semantic encoder including a word embedding layer to perform semantic encoding to extract the word granularity-based semantic encoding features in the emergency patient's condition description, thereby obtaining a sequence of word granularity semantic association feature vectors of the condition description. In a specific example, the description of the emergency patient's condition is segmented to convert the description of the emergency patient's condition into a word sequence composed of multiple words; the embedding layer of the condition description semantic encoder including the word embedding layer is used to map each word in the word sequence into a word embedding vector to obtain a sequence of word embedding vectors; the condition description semantic encoder including the word embedding layer is used to perform global context semantic encoding based on the converter idea on the sequence of word embedding vectors to obtain a sequence of condition description word granularity semantic association feature vectors.

[0074] In particular, the condition description word granularity semantic center screening and strengthening module 350 is used to pass the sequence of the condition description word granularity semantic association feature vectors through a word granularity semantic reverse screening and strengthening module based on the description semantic center to obtain a sequence of emergency patient condition description keyword semantic enhanced expression feature vectors. When performing pre-examination and triage on emergency patients, the patient may not be able to accurately describe the condition when describing his or her condition due to tension, pain or other reasons, thereby affecting the subsequent recognition accuracy of the severity of the patient's condition and the accuracy of pre-examination and triage. Therefore, it is very important to conduct an in-depth analysis of the emergency patient's condition description to accurately understand the patient's condition. Based on this, in the technical solution of the present application, the sequence of the condition description word granularity semantic association feature vectors is further passed through a word granularity semantic reverse screening and strengthening module based on the description semantic center to obtain a sequence of emergency patient condition description keyword semantic enhanced expression feature vectors. Through the processing of the word-granularity semantic reverse screening and reinforcement module based on the description semantic center, the contextual semantic association relationship and importance characteristics based on word-granularity in the patient's condition description can be automatically learned and captured, thereby identifying the keyword-granularity semantics related to the pre-diagnosis and triage tasks, so as to enhance these semantics and enhance the semantic feature expression ability of the emergency patient's condition description.

[0075] In an embodiment of the present application, the sequence of the condition description word granular semantic association feature vectors is subjected to a word granular semantic reverse screening and strengthening module based on the description semantic center to obtain a sequence of emergency patient condition description keyword semantic reinforcement expression feature vectors, including: first, calculating the mean vector of the sequence of the condition description word granular semantic association feature vectors to obtain the condition description word granular sequence distribution center vector; and calculating the correlation coefficient between each condition description word granular semantic association feature vector in the sequence of the condition description word granular semantic association feature vector and the condition description word granular sequence distribution center vector to obtain a sequence of condition description word granular semantic association coefficients, which is the basis for feature selection. Then, calculating the reciprocal of each condition description word granular semantic association coefficient in the sequence of the condition description word granular semantic association coefficient to obtain a sequence of condition description word granular semantic anti-association coefficients, the purpose of which is to reduce the weight of the features with high correlation to the condition description word granular center semantics, in preparation for reverse screening. Next, the Softmax function is used to normalize the sequence of the condition description word granular semantic anti-correlation coefficients to obtain the sequence of the condition description word granular semantic anti-correlation energy coefficients, ensuring that the sum of all coefficients is 1, making the weight distribution smoother and avoiding the appearance of extreme values. Further, each condition description word granular semantic anti-correlation energy coefficient in the sequence of the condition description word granular semantic anti-correlation energy coefficients is used as a weight to weight each condition description word granular semantic association feature vector in the sequence of the condition description word granular semantic association feature vector to obtain a sequence of reverse inhibition condition description word granular semantic association feature vectors; then, the positional difference of each group of corresponding condition description word granular semantic association feature vectors and reverse inhibition condition description word granular semantic association feature vectors between the sequence of the condition description word granular semantic association feature vectors and the sequence of reverse inhibition condition description word granular semantic association feature vectors is calculated to obtain the sequence of the semantic enhancement expression feature vectors of the emergency patient condition description keywords. That is, through weighted and differential processing, the algorithm strengthens the part of the original disease description word granularity semantic features with higher semantic correlation with the central vector of the disease description word granularity semantic sequence distribution, while suppressing the part with lower correlation, thereby achieving feature optimization.

[0076] Among them, the process of calculating the correlation coefficient between each condition description word granularity semantic association feature vector in the sequence of the condition description word granularity semantic association feature vector and the condition description word granularity sequence distribution center vector to obtain a sequence of condition description word granularity semantic association coefficients includes: calculating the position point multiplication between the condition description word granularity semantic association feature vector and the condition description word granularity sequence distribution center vector to obtain the condition description word granularity semantic association class support feature vector; calculating the sum of the feature values ​​of each position in the condition description word granularity semantic association class support feature vector to obtain the condition description word granularity semantic association class support feature vector. Sum value; calculate the sum of the squares of the feature values ​​at each position in the condition description word granularity semantic association feature vector and the sum of the squares of the feature values ​​at each position in the condition description word granularity sequence distribution center vector to obtain the condition description word granularity semantic association square value and the condition description word granularity sequence distribution center square value; calculate the square root of the sum of the condition description word granularity semantic association square value and the condition description word granularity sequence distribution center square value to obtain the condition description word granularity association value; divide the condition description word granularity semantic association class support sum value by the condition description word granularity association value to obtain the condition description word granularity semantic association coefficient.

[0077] More specifically, the process of using the Softmax function to normalize the sequence of the condition description word granularity semantic anti-correlation coefficients to obtain the sequence of the condition description word granularity semantic anti-correlation energy coefficients includes: using each condition description word granularity semantic anti-correlation coefficient of the sequence of the condition description word granularity semantic anti-correlation coefficients as the exponent of the natural constant to calculate the exponential function value with the natural constant as the base by position to obtain the sequence of the condition description word granularity semantic anti-correlation index coefficients; calculating the sum of each condition description word granularity semantic anti-correlation index coefficient in the sequence of the condition description word granularity semantic anti-correlation index coefficients to obtain the total sum of the condition description word granularity semantic anti-correlation coefficients; calculating the sum of the total sum of the condition description word granularity semantic anti-correlation coefficients and the constant one to obtain the condition description word granularity semantic anti-correlation modulation coefficient; dividing each condition description word granularity semantic anti-correlation index coefficient in the sequence of the condition description word granularity semantic anti-correlation index coefficients by the condition description word granularity semantic anti-correlation modulation coefficient to obtain the sequence of the condition description word granularity semantic anti-correlation energy coefficients.

[0078] In summary, in the above embodiment, the sequence of the condition description word granularity semantic association feature vectors is passed through the word granularity semantic reverse screening and strengthening module based on the description semantic center to obtain the sequence of the emergency patient condition description keyword semantic enhancement expression feature vectors, including: the sequence of the condition description word granularity semantic association feature vectors is passed through the word granularity semantic reverse screening and strengthening module based on the description semantic center to be processed with the following word granularity semantic enhancement formula to obtain the sequence of the emergency patient condition description keyword semantic enhancement expression feature vectors; wherein the word granularity semantic enhancement formula is:

[0079] ;

[0080] ;

[0081] ;

[0082] ;

[0083] ;

[0084] ;

[0085] in, is a sequence of granular semantic association feature vectors of the disease description words, represents a sequence of vectors, is the number of feature vectors in the sequence of the granular semantic association feature vectors of the disease description word, is the first in the sequence of the granular semantic association feature vector of the disease description word The granular semantic association feature vector of disease description words, is the first in the sequence of the granular semantic association feature vector of the disease description word The granular semantic association feature vector of disease description words, is the center vector of the granularity sequence distribution of the disease description words, It is the point multiplication by position. is the number of feature values ​​in each of the granular semantic association feature vectors of the disease description words, is the first granular semantic association feature vector of each disease description word The eigenvalues ​​at the positions, is the center vector of the granularity sequence distribution of the disease description word The eigenvalues ​​at the positions, is the first in the sequence of the granular semantic association coefficient of the disease description word The granular semantic correlation coefficient of the disease description words, is the first in the sequence of the granular semantic anti-correlation coefficient of the disease description word The semantic anti-correlation coefficient of the disease description word granularity, is the first in the sequence of the granular semantic anti-correlation energy coefficient of the disease description word The semantic anti-correlation energy coefficient of the disease description word granularity, is a sequence of granular semantic association feature vectors that reversely suppress the disease description word, , , , are the first, second, and third in the sequence of the granular semantic association feature vectors of the reverse inhibition disease description word. and The granular semantic association feature vector of the reverse inhibition disease description word, is a sequence of semantically enhanced expression feature vectors of key words describing the condition of the emergency patient, , , , is the first, second, and third in the sequence of the semantic enhancement expression feature vector of the emergency patient's condition description keyword. and The semantically enhanced expression feature vector of key words in the description of the condition of emergency patients.

[0086] In particular, the vital signs guided patient condition description semantic constraint expression module 360 ​​is used to use the vital signs multimodal time series semantic representation vector as the reference information feature vector, and to perform semantic constraint aggregation processing based on information distillation on the sequence of the emergency patient condition description keyword semantic enhancement expression feature vector to obtain the patient condition description semantic constraint enhancement expression feature vector under the guidance of vital signs time series semantics. It should be understood that since the vital signs multimodal time series semantic representation vector contains the vital signs time series multimodal semantic fusion representation information about the emergency patient, the sequence of the emergency patient condition description keyword semantic enhancement expression feature vector contains the patient condition description word granularity enhancement semantic expression information. In view of the situation that the patient's condition description is not accurate enough due to various factors, in the technical solution of the present application, the vital signs multimodal time series semantic representation vector is further used as the reference information feature vector, and the sequence of the emergency patient condition description keyword semantic enhancement expression feature vector is processed based on information distillation to obtain the patient condition description semantic constraint enhancement expression feature vector under the guidance of vital signs time series semantics. The sequence of the semantically enhanced expression feature vectors of the emergency patient's condition description keywords is processed by semantic constraint aggregation based on information distillation through benchmark information. The multimodal temporal semantics of the emergency patient's vital signs can be used to constrain the expression and aggregation of the emergency patient's condition description keyword semantics. In this way, information distillation is performed on the sequence of the semantically enhanced expression feature vectors of the emergency patient's condition description keywords to pay higher attention to the keyword granularity semantics that are more relevant to subsequent tasks, thereby highlighting the semantic representation information related to the patient's vital signs benchmark information in the semantics of the patient's condition description in the process of word granularity information aggregation, avoiding the accuracy of pre-diagnosis and triage and the timeliness of treatment affected by inaccurate description of the patient's condition during pre-diagnosis and triage.

[0087] In an embodiment of the present application, the multimodal temporal semantic representation vector of vital signs is used as a benchmark information feature vector, and the sequence of the semantically enhanced expression feature vectors of the emergency patient condition description keywords is subjected to semantic constraint aggregation processing based on information distillation to obtain the semantically constrained enhanced expression feature vector of the patient condition description under the guidance of vital sign temporal semantics, including: calculating each of the semantically enhanced expression feature vectors of the emergency patient condition description keywords in the sequence of the semantically enhanced expression feature vectors of the emergency patient condition description keywords, multiplying them by the corresponding weight matrix, and then adding them by position with the corresponding bias vector to obtain a sequence of linearly modulated semantically enhanced expression feature vectors of the emergency patient condition description keywords; using a sigmoid function to activate the sequence of the semantically enhanced expression feature vectors of the emergency patient condition description keywords, and then multiplying them by the transposed vector of the benchmark information feature vector to obtain the patient's vital signs. The invention relates to a sequence of semantic relevance of patient vital signs and condition description; the sum of the natural exponential function values ​​of the semantic relevance of each patient vital sign and condition description in the sequence of semantic relevance of patient vital signs and condition description is calculated to obtain the sum of the semantic relevance of patient vital signs and condition description; the natural exponential function value of each patient vital sign and condition description in the sequence of semantic relevance of patient vital signs and condition description is divided by the sum of the semantic relevance of patient vital signs and condition description to obtain the sequence of semantic association weights of patient vital signs and condition description; each patient vital sign and condition description semantic association weight in the sequence of semantic association weights of patient vital signs and condition description is used as a weighting coefficient to calculate the weighted sum of the sequence of semantic enhanced expression feature vectors of the emergency patient condition description keywords to obtain the semantic constraint enhanced expression feature vector of the patient condition description under the guidance of the vital sign time series semantics.

[0088] In summary, in the above embodiment, the multimodal temporal semantic representation vector of the vital signs is used as the reference information feature vector, and the sequence of the semantically enhanced expression feature vectors of the emergency patient condition description keywords is subjected to semantic constraint aggregation processing based on information distillation to obtain the semantically constrained enhanced expression feature vector of the patient condition description under the guidance of the vital sign temporal semantics, including: using the multimodal temporal semantic representation vector of the vital signs as the reference information feature vector, and using the following semantic constraint aggregation formula to perform semantic constraint aggregation processing based on information distillation on the sequence of the semantically enhanced expression feature vectors of the emergency patient condition description keywords to obtain the semantically constrained enhanced expression feature vector of the patient condition description under the guidance of the vital sign temporal semantics; wherein the semantic constraint aggregation formula is:

[0089] ;

[0090] ;

[0091] ;

[0092] in, is the multimodal temporal semantic representation vector of the vital signs, for each of the emergency patient condition description keyword semantic enhancement expression feature vectors in the sequence of emergency patient condition description keyword semantic enhancement expression feature vectors, is the weight matrix, is the bias vector, is the sigmoid function, is each semantic relevance in the sequence of semantic relevance, is the first in the sequence of semantic relevance Semantic relevance, is the natural exponential function, is the number of semantic relevances in the sequence of semantic relevances, is each semantic association weight in the sequence of semantic association weights, It is a semantic constraint enhanced expression feature vector of the patient's condition description guided by the time series semantics of the vital signs.

[0093] In particular, the registration department category recommendation module 370 is used to determine the recommended registration department category based on the semantic constraint enhanced expression feature vector of the patient's condition description under the guidance of the vital sign time series semantics, and generate a registration number. In the technical solution of the present application, the semantic constraint enhanced expression feature vector of the patient's condition description under the guidance of the vital sign time series semantics is passed through a registration department category recommender based on a classifier to obtain a recommendation result, and the recommendation result is used to represent the recommended registration department category. That is, the classification processing is performed using the semantic constraint expression enhanced features of the condition description under the guidance of the patient's vital sign time series semantics, so as to evaluate and grade the patient's condition so as to automatically match the patient to the corresponding registration department. Then, based on the recommended registration department category, a registration number is generated. In this way, it is possible to decide whether to give priority to the patient's condition when registering. In this way, a more intelligent pre-examination and triage of emergency patients can be achieved, avoiding the patient's inaccurate description of the condition and affecting the accuracy of pre-examination and triage and the timeliness of treatment.

[0094] Preferably, considering that the sequence of the multimodal temporal semantic representation vector of vital signs and the sequence of the semantic enhancement expression feature vector of the key words of the description of the emergency patient's condition respectively express the multimodal Bayesian fusion temporal association characteristics of the vital signs data and the encoded semantic features of the description of the emergency patient's condition based on word-granular semantic reverse screening and enhancement, when the sequence of the key words of the description of the emergency patient's condition is subjected to semantic constraint aggregation processing based on information distillation using the multimodal temporal semantic representation vector of vital signs, the obtained feature vector of the semantic constraint enhancement expression of the patient's condition under the guidance of the vital signs temporal semantics will also lead to outlier reasoning mapping deviation due to insufficient coverage of the distillation constraints of the feature information of different dimensions, thereby affecting the accuracy of the recommendation result obtained by the feature vector of the semantic constraint enhancement expression of the patient's condition under the guidance of the vital signs temporal semantics through the registration department category recommender based on the classifier.

[0095] Therefore, in a preferred example, the feature vector of the semantic constraint-enhanced expression of the patient's condition description under the guidance of the vital sign time sequence semantics is passed through a registration department category recommender based on a classifier to obtain a recommendation result including:

[0096] Calculate the feature mean of the semantic constraint enhanced expression feature vector of the patient's condition description under the guidance of the vital sign time series semantics, and divide the feature mean by the difference between the maximum feature value and the minimum feature value of the semantic constraint enhanced expression feature vector of the patient's condition description under the guidance of the vital sign time series semantics to obtain the distribution representation value of the semantic constraint enhanced expression of the patient's condition description under the guidance of the vital sign time series semantics;

[0097] The distribution representation value of the semantic constraint enhanced expression of the patient's condition description under the guidance of the vital sign time sequence semantics is divided by the distribution representation value of the semantic constraint enhanced expression of the patient's condition description under the guidance of the vital sign time sequence semantics to obtain the distribution modulation value of the semantic constraint enhanced expression of the patient's condition description under the guidance of the vital sign time sequence semantics;

[0098] The semantic constraint enhanced expression feature vector of the patient's condition description under the guidance of the vital signs temporal semantics is activated by a probabilistic function to obtain a probabilistic semantic constraint enhanced expression feature vector of the patient's condition description under the guidance of the vital signs temporal semantics;

[0099] After performing point-wise subtraction of the semantic constraint enhanced expression feature vector of the patient's condition description under the probabilistic vital sign time series semantics and the distribution modulation value of the semantic constraint enhanced expression of the patient's condition description under the probabilistic vital sign time series semantics, taking the absolute value and calculating the negative of the logarithmic value with base 2 to obtain the distribution modulation information vector of the semantic constraint enhanced expression of the patient's condition description under the probabilistic vital sign time series semantics;

[0100] The distribution representation value of the semantic constraint enhanced expression of the patient's condition description under the guidance of the vital signs temporal semantics is divided by one minus the difference of each eigenvalue of the feature vector of the semantic constraint enhanced expression of the patient's condition description under the guidance of the probabilistic vital signs temporal semantics, and then the sum is accumulated and divided by the length of the feature vector of the semantic constraint enhanced expression of the patient's condition description under the guidance of the vital signs temporal semantics to obtain the distribution modulation bias value of the semantic constraint enhanced expression of the patient's condition description under the guidance of the probabilistic vital signs temporal semantics;

[0101] The optimized feature vector of semantic constraint enhanced expression of patient condition description under the guidance of the probabilistic vital sign time series semantics is obtained by multiplying the distribution modulation information vector of semantic constraint enhanced expression of patient condition description under the guidance of the probabilistic vital sign time series semantics by the product of the distribution modulation bias value of semantic constraint enhanced expression of patient condition description under the guidance of the probabilistic vital sign time series semantics and the weight as a hyperparameter;

[0102] The semantically constrained and enhanced expression feature vector of the patient's condition description guided by the optimized vital signs time series semantics is passed through a registration department category recommender based on a classifier to obtain a recommendation result.

[0103] The semantic constraint enhanced expression feature vector of the patient's condition description under the guidance of the optimized vital sign time series semantics is expressed as:

[0104] ;

[0105] ;

[0106] in, is the semantic constraint-enhanced expression feature vector of the patient's condition description under the guidance of the probabilistic time-series semantics of vital signs, represents the semantic constraint-enhanced expression feature vector of the patient's condition description under the guidance of the probabilistic time series semantics of vital signs. eigenvalues, is the distribution representation value of the semantic constraint reinforcement expression of the patient's condition description under the guidance of the temporal semantics of the vital signs, is the length of the semantic constraint enhanced expression feature vector of the patient's condition description under the guidance of the vital signs temporal semantics, is the weight hyperparameter, It means to make a difference by position. It means adding by position. is the semantic constraint enhanced expression feature vector of the patient's condition description under the guidance of the optimized vital sign time series semantics, represents the feature mean of the semantic constraint-enhanced expression feature vector of the patient's condition description under the guidance of the vital signs temporal semantics, and They respectively represent the maximum eigenvalue and the minimum eigenvalue of the semantic constraint enhanced expression feature vector of the patient's condition description under the guidance of the vital signs temporal semantics.

[0107] That is, the probability information distribution planning based on the eigenvalue of the feature vector of the semantic constraint-enhanced expression of the patient's condition description under the guidance of the vital signs temporal semantics is carried out through the Bernoulli probability modulation distribution of the feature vector of the semantic constraint-enhanced expression of the patient's condition description under the guidance of the vital signs temporal semantics relative to the eigenvalue distribution, and the probability reverse mapping of the probability characteristics of the feature vector of the semantic constraint-enhanced expression of the patient's condition description under the guidance of the vital signs temporal semantics is used as the extended coverage of the set mapping space of the feature vector of the semantic constraint-enhanced expression of the patient's condition description under the guidance of the vital signs temporal semantics, so as to independently understand the interaction path between the intuitive probability information distribution and the abstract probability space mapping of the feature vector of the semantic constraint-enhanced expression of the patient's condition description under the guidance of the vital signs temporal semantics, so as to improve the accuracy of the recommendation result obtained by the optimized feature vector of the semantic constraint-enhanced expression of the patient's condition description under the guidance of the vital signs temporal semantics through the classifier-based registration department category recommender by avoiding the outlier feature distribution of the feature vector of the semantic constraint-enhanced expression of the patient's condition description under the guidance of the vital signs temporal semantics to the counterfactual reasoning mapping of the class probability of the feature vector. In this way, more intelligent pre-examination and triage of emergency patients can be achieved, so as to more accurately recommend the corresponding registration department for patients and decide whether to give priority treatment based on the patient's condition.

[0108] As described above, the intelligent pre-examination and triage system 300 for emergency patients according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with an intelligent pre-examination and triage algorithm for emergency patients. In one possible implementation, the intelligent pre-examination and triage system 300 for emergency patients 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 intelligent pre-examination and triage system 300 for emergency patients 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 intelligent pre-examination and triage system 300 for emergency patients can also be one of the many hardware modules of the wireless terminal.

[0109] Alternatively, in another example, the intelligent pre-examination and triage system 300 for emergency patients and the wireless terminal may also be separate devices, and the intelligent pre-examination and triage system 300 for emergency patients 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.

[0110] Furthermore, an intelligent pre-examination and triage method for emergency patients is also provided.

[0111] Figure 4Flow chart of the intelligent pre-examination and triage method for emergency patients according to an embodiment of the present application. Figure 4 As shown, according to the embodiment of the present application, the intelligent pre-examination and triage method for emergency patients includes the following steps: S1, obtaining a description of the emergency patient's condition; S2, obtaining a time series of vital sign data of the emergency patient collected by a vital sign monitor, wherein the vital sign data includes heart rate, blood pressure and blood oxygen saturation; S3, performing temporal feature extraction and semantic fusion processing on the time series of the vital sign data to obtain a multimodal temporal semantic representation vector of the vital signs; S4, performing word-granularity-based semantic encoding on the emergency patient's condition description to obtain a sequence of word-granularity semantic association feature vectors of the condition description; S5, combining the word-granularity semantic encoding of the condition description with the time series of the vital sign data. The sequence of associated feature vectors is filtered and enhanced through a word-granularity semantic reverse screening module based on the description semantic center to obtain a sequence of semantically enhanced expression feature vectors of key words in the description of the emergency patient's condition; S6, using the multimodal temporal semantic representation vector of the vital signs as the benchmark information feature vector, the sequence of semantically enhanced expression feature vectors of key words in the description of the emergency patient's condition is subjected to semantic constraint aggregation processing based on information distillation to obtain a semantically constrained enhanced expression feature vector of the patient's condition description under the guidance of the vital signs temporal semantics; S7, based on the semantically constrained enhanced expression feature vector of the patient's condition description under the guidance of the vital signs temporal semantics, the recommended registration department category is determined, and a registration number is generated.

[0112] In summary, the intelligent pre-examination and triage method for emergency patients according to the embodiment of the present application is explained, which obtains the description of the condition of the emergency patient, collects the time series data of the patient's vital signs, and introduces data processing and semantic understanding algorithms based on artificial intelligence and deep learning at the back end to perform semantic analysis and understanding on the description of the condition of the emergency patient, and dynamically analyzes and multimodal fusion representation of the time series semantics of the patient's vital signs, and then uses the patient's vital signs time series multimodal fusion semantics to constrain the expression of the semantics of the condition description. In this way, vital signs can be used as reference information to automatically identify the symptoms of the patient's condition description, thereby evaluating and grading the patient's condition, so as to automatically match the patient to the corresponding registration department, and determine whether to give priority to the treatment according to the patient's condition. In this way, a more intelligent pre-examination and triage of emergency patients can be achieved, avoiding the patient's inaccurate description of the condition and affecting the accuracy of pre-examination and triage and the timeliness of treatment.

[0113] 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. An intelligent pre-examination and triage system for emergency patients, characterized in that: include: Patient condition description collection module, used to obtain the condition description of emergency patients; A patient vital sign data acquisition module, used to acquire a time series of vital sign data of the emergency patient acquired by a vital sign monitor, wherein the vital sign data includes heart rate, blood pressure and blood oxygen saturation; A multimodal vital sign time series semantic fusion representation module, used for performing time series feature extraction and semantic fusion processing on the time series of the vital sign data to obtain a multimodal time series semantic representation vector of the vital sign; A patient condition description word granularity semantic encoding module is used to perform word granularity-based semantic encoding on the emergency patient's condition description to obtain a sequence of condition description word granularity semantic association feature vectors; A granular semantic center screening and strengthening module for the condition description words is used to pass the sequence of the granular semantic association feature vectors of the condition description words through a word granular semantic reverse screening and strengthening module based on the description semantic center to obtain a sequence of semantic strengthening expression feature vectors of the emergency patient condition description keywords; A module for expressing the semantic constraints of patient condition description guided by vital signs, which is used to use the multimodal time series semantic representation vector of the vital signs as a reference information feature vector, and to perform semantic constraint aggregation processing based on information distillation on the sequence of the semantic enhancement expression feature vectors of the emergency patient condition description keywords to obtain a semantic constraint enhancement expression feature vector of the patient condition description guided by the time series semantics of the vital signs; The registration department category recommendation module is used to determine the recommended registration department category and generate a registration number based on the semantic constraint reinforcement expression feature vector of the patient's condition description under the guidance of the vital sign time series semantics.

2. The intelligent pre-examination and triage system for emergency patients according to claim 1 is characterized in that: The multimodal vital sign temporal semantic fusion representation module is used to: an arranging unit, configured to arrange the time series of the vital sign data according to a sample dimension to obtain a heart rate time series input vector, a blood pressure time series input vector, and a blood oxygen saturation time series input vector; A vital sign time series feature extraction unit, used for passing the heart rate time series input vector, the blood pressure time series input vector and the blood oxygen saturation time series input vector through a vital sign time series feature extractor based on a one-dimensional convolutional neural network model to obtain a heart rate time series feature vector, a blood pressure time series feature vector and a blood oxygen saturation time series feature vector; The vital sign multimodal time series semantic fusion unit is used to obtain the vital sign multimodal time series semantic representation vector by fusing the heart rate time series feature vector, the blood pressure time series feature vector and the blood oxygen saturation time series feature vector through a vital sign multimodal time series semantic fuser based on a Bayesian probability network.

3. The intelligent pre-examination and triage system for emergency patients according to claim 2 is characterized in that: The patient condition description word granularity semantic encoding module is used to: perform word segmentation processing on the emergency patient's condition description and then pass it through a condition description semantic encoder including a word embedding layer to obtain a sequence of the condition description word granularity semantic association feature vectors.

4. The intelligent pre-examination and triage system for emergency patients according to claim 3 is characterized in that: The disease description word granularity semantic center screening and strengthening module includes: A mean vector calculation unit, used to calculate the mean vector of the sequence of the granular semantic association feature vectors of the condition description words to obtain a distribution center vector of the granular sequence of the condition description words; A condition description word granularity semantic association coefficient calculation unit is used to calculate the correlation coefficient between each condition description word granularity semantic association feature vector in the sequence of the condition description word granularity semantic association feature vector and the condition description word granularity sequence distribution center vector to obtain a sequence of condition description word granularity semantic association coefficients; A condition description word granularity semantic anti-correlation coefficient calculation unit, used to calculate the inverse of each condition description word granularity semantic correlation coefficient in the sequence of condition description word granularity semantic correlation coefficients to obtain a sequence of condition description word granularity semantic anti-correlation coefficients; A normalization unit, used for normalizing the sequence of the granular semantic anti-correlation coefficients of the condition description words using a Softmax function to obtain a sequence of the granular semantic anti-correlation energy coefficients of the condition description words; A condition description reverse inhibition disease unit is used to weight each condition description word granular semantic anti-association energy coefficient in the sequence of condition description word granular semantic anti-association energy coefficients as a weight, and to obtain a sequence of reverse inhibition condition description word granular semantic association feature vectors by weighting each condition description word granular semantic association feature vector in the sequence of condition description word granular semantic association feature vectors; The condition description keyword semantic enhancement expression unit is used to calculate the positional difference between each group of corresponding condition description word granularity semantic association feature vectors and the reverse inhibition condition description word granularity semantic association feature vectors between the sequence of the condition description word granularity semantic association feature vectors and the sequence of the reverse inhibition condition description word granularity semantic association feature vectors to obtain the sequence of the emergency patient condition description keyword semantic enhancement expression feature vectors.

5. The intelligent pre-examination and triage system for emergency patients according to claim 4 is characterized in that: The condition description word granularity semantic association coefficient calculation unit is used to: Calculate the point-by-point multiplication between the granular semantic association feature vector of the condition description word and the granular sequence distribution center vector of the condition description word to obtain the granular semantic association class support feature vector of the condition description word; Calculate the sum of the feature values ​​of each position in the condition description word granular semantic association class support feature vector to obtain the condition description word granular semantic association class support sum value; Calculate the square sum of the feature values ​​at each position in the granular semantic association feature vector of the condition description word and the square sum of the feature values ​​at each position in the granular sequence distribution center vector of the condition description word to obtain the granular semantic association square value of the condition description word and the granular sequence distribution center square value of the condition description word; Calculate the square root of the sum of the square value of the semantic association of the condition description word granularity and the square value of the center of the condition description word granularity sequence distribution to obtain the condition description word granularity association value; The condition description word granular semantic association class support sum value is divided by the condition description word granular association value to obtain the condition description word granular semantic association coefficient.

6. The intelligent pre-examination and triage system for emergency patients according to claim 5 is characterized in that: The normalization unit is used for: Using each condition description word granularity semantic anti-correlation coefficient of the sequence of condition description word granularity semantic anti-correlation coefficients 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 condition description word granularity semantic anti-correlation exponential coefficients; Calculating the sum of the granular semantic anti-correlation index coefficients of each condition description word in the sequence of the granular semantic anti-correlation index coefficients of the condition description word to obtain the total sum of the granular semantic anti-correlation coefficients of the condition description word; Calculate the sum of the total sum of the granular semantic anti-association coefficients of the condition description words and a constant of one to obtain the granular semantic anti-association modulation coefficient of the condition description words; Each condition description word granularity semantic anti-correlation index coefficient in the sequence of condition description word granularity semantic anti-correlation index coefficients is divided by the condition description word granularity semantic anti-correlation modulation coefficient to obtain the sequence of condition description word granularity semantic anti-correlation energy coefficients.

7. The intelligent pre-examination and triage system for emergency patients according to claim 6 is characterized in that: The vital signs guide the patient's condition description semantic constraint expression module, which is used to: Calculate each emergency patient condition description keyword semantic enhancement expression feature vector in the sequence of emergency patient condition description keyword semantic enhancement expression feature vectors, multiply it with the corresponding weight matrix, and then add it with the corresponding bias vector by position to obtain a sequence of linearly modulated emergency patient condition description keyword semantic enhancement expression feature vectors; Using a sigmoid function to activate the sequence of semantic enhancement expression feature vectors of the emergency patient's condition description keywords, and then multiplying them with the transposed vector of the benchmark information feature vector to obtain a sequence of patient vital signs-condition description semantic relevance; Calculating the sum of the semantic relevances of each patient's vital sign-condition description in the sequence of the semantic relevances of the patient's vital sign-condition description to obtain the sum of the semantic relevances of the patient's vital sign-condition description; Dividing each patient vital sign-condition description semantic relevance in the patient vital sign-condition description semantic relevance sequence by the sum of the patient vital sign-condition description semantic relevances to obtain a sequence of patient vital sign-condition description semantic association weights; Taking each patient's vital sign-condition description semantic association weight in the sequence of the patient's vital sign-condition description semantic association weights as a weighting coefficient, the weighted sum of the sequence of the emergency patient's condition description keyword semantic enhanced expression feature vectors is calculated to obtain the patient's condition description semantic constraint enhanced expression feature vector under the guidance of the vital sign time series semantics.

8. The intelligent pre-examination and triage system for emergency patients according to claim 7 is characterized in that: The registration department category recommendation module is used to: The semantic constraint enhanced expression feature vector of the patient's condition description guided by the time series semantics of the vital signs is passed through a registration department category recommender based on a classifier to obtain a recommendation result, and the recommendation result is used to represent the recommended registration department category; Based on the recommended registration department category, a registration number is generated.

9. An intelligent pre-examination and triage method for emergency patients, characterized in that: include: Obtain descriptions of emergency patients’ medical conditions; Acquire a time series of vital sign data of the emergency patient collected by a vital sign monitor, wherein the vital sign data includes heart rate, blood pressure and blood oxygen saturation; Performing time series feature extraction and semantic fusion processing on the time series of the vital sign data to obtain a multimodal time series semantic representation vector of the vital sign; Performing word-granularity-based semantic encoding on the description of the emergency patient's condition to obtain a sequence of word-granularity semantic association feature vectors of the condition description; The sequence of the condition description word granularity semantic association feature vectors is passed through a word granularity semantic reverse screening and strengthening module based on the description semantic center to obtain a sequence of emergency patient condition description keyword semantic strengthening expression feature vectors; Taking the multimodal time series semantic representation vector of vital signs as the reference information feature vector, the sequence of the semantic enhancement expression feature vectors of the emergency patient condition description keywords is subjected to semantic constraint aggregation processing based on information distillation to obtain the semantic constraint enhancement expression feature vector of the patient condition description under the guidance of the time series semantics of vital signs; Based on the semantically constrained enhanced expression feature vector of the patient's condition description guided by the vital signs temporal semantics, the recommended registration department category is determined and a registration number is generated.

Images