A method for assessing the risk level of poisoning for emergency department patients

Abstract

The present application relates to the technical field of medical first aid, in particular to a poisoning risk grade evaluation method for patients in emergency department, which comprises the following steps: collecting data of clinical characteristic parameters of patients, constructing a standardized characteristic vector set, constructing a toxicant phenotype space tree structure model based on historical poisoning case samples, inputting the clinical characteristics of the patient to be matched into the model, determining the leaf node it falls into, obtaining the corresponding poisoning sample subset, constructing a leaf vein net unit according to the poisoning sample subset, mapping each leaf vein into a leaf vein point, calculating the spatial distance between each leaf vein point and the coordinate origin, performing hierarchical clustering to generate each microphenotype subcluster, matching similar microphenotype subclusters based on the coordinate position of the leaf vein point of the patient to be matched, and then outputting risk trend label information and determining the poisoning risk type grade, so as to realize more accurate and efficient auxiliary decision support for clinical first aid.

Description

Technical Field

[0001] This invention relates to the field of medical emergency technology, specifically to a method for assessing the risk level of poisoning in emergency department patients. Background Technology

[0002] With the continuous increase in urban population density and the acceleration of industrialization, the widespread use of various chemicals and drugs has led to a year-on-year increase in the number of poisoning patients received by emergency departments. Poisoning emergency events are usually characterized by their suddenness, rapid progression, and hidden risks. If the risk level of the patient cannot be accurately identified in the early stage, the best intervention time can easily be missed, resulting in unnecessary deterioration of the condition, waste of medical resources, or even patient death. Therefore, how to quickly and accurately assess the risk level of poisoning patients is a key issue in clinical emergency care. In existing technologies, clinical decision-making mainly relies on human experience, static indicator judgments, or risk stratification models based on a single algorithm. However, traditional methods generally have the following shortcomings: 1. The lack of structural modeling of the evolutionary paths among historical poisoning cases makes it difficult to fully reflect the impact patterns of different types of toxins on disease development; 2. When faced with new or rare poisons, it is difficult to quickly identify the type of poison and assess the patient's risk level of poisoning, making it difficult to make timely and accurate intervention and treatment decisions. To address the aforementioned shortcomings, a technical solution is provided. Summary of the Invention

[0003] To address the aforementioned shortcomings of existing technologies, this invention provides a method for assessing the poisoning risk level of emergency department patients. This method effectively solves the problems of lacking evolutionary modeling capabilities and being unable to quickly identify the type of toxin and assess the patient's poisoning risk level in existing technologies.

[0004] To achieve the above objectives, the present invention can be implemented through the following technical solutions: This invention provides a method for assessing the poisoning risk level of patients in the emergency department, comprising the following steps: Collect data on patients' clinical characteristic parameters and construct a standardized set of feature vectors; The standardized feature vector set of historically diagnosed poisoning patients is used as training input. Based on the information gain of each clinical feature in historical poisoning samples, the priority of feature dimension division is determined. The standardized feature vectors are then split sequentially to construct a spatial tree structure model of toxic substance phenotype. The standardized feature vector set of the patients to be matched is input into the tree structure model of the toxic phenotype space to determine the leaf node where it falls and obtain the subset of poisoned samples. Based on the subset of poisoned samples, leaf vein network units were constructed, and each leaf vein was mapped to a leaf vein point. The spatial distance between each leaf vein point and the origin of coordinates was calculated. Hierarchical clustering was performed to generate each micro phenotypic sub-cluster, and the corresponding risk trend label information was output. The risk trend label information includes the duration of disease onset, the outcome, and the speed of disease onset. Based on the coordinates of the leaf vein points of the patients to be matched, similar microphenotype subclusters are matched to obtain the risk trend label information of the patients to be matched, and the poisoning risk type level is determined accordingly.

[0005] Furthermore, the specific process of constructing a standardized feature vector set is as follows: Within a set time period, data on patients' clinical characteristic parameters are collected at various monitoring time points over a period of time to construct a dataset of patients' clinical characteristic parameters. The clinical characteristic parameters include basic vital signs parameters, symptom presentation parameters, and biochemical test parameters. The value corresponding to the mode in the dataset of clinical characteristic parameters is used as the representative value of the clinical characteristic parameter. Subsequently, the extracted representative values ​​of clinical characteristic parameters are normalized to construct a standardized feature vector set with a unified format.

[0006] Furthermore, the specific process for obtaining a subset of poisoned samples is as follows: Based on the standardized feature vector set of the patients to be matched, and in accordance with the judgment rules of each split node in the tree structure of the toxic phenotype space, path selection and node traversal are performed layer by layer. When the features of the patients to be matched meet all the partitioning conditions of a certain path, they will eventually fall into a specific leaf node. The feature path represented by this leaf node is a set of poisoning clinical phenotype feature combinations that are highly consistent with the features of the patients to be matched, constituting its preliminary classification result in the known model space. At the same time, the historical poisoning sample subset stored in this leaf node is the poisoning sample subset under the same poisoning clinical phenotype feature combination conditions.

[0007] Furthermore, the specific process of constructing leaf vein network units is as follows: Each poisoned sample is considered as an independent leaf vein, and the temporal evolution relationship between poisoned samples is used to form the connecting fibers between leaf veins, thereby forming a leaf vein network unit.

[0008] Furthermore, the specific process of mapping to leaf vein points is as follows: A two-dimensional coordinate system is constructed with the vein flow velocity v as the abscissa and the vein flow angle d as the ordinate. The two quantities v and d of each vein are mapped to coordinate points in the two-dimensional coordinate system to form a vein point. Among them, the vein flow velocity v is the time window, and the vein flow angle d is the average change trend coefficient. The process of determining the leaf vein flow direction angle is as follows: extract the clinical feature parameters within the time window, and for each clinical feature parameter, use linear regression to fit the slope of the clinical feature change over time as its trend coefficient. Then, calculate the mean of the trend coefficients of all clinical feature parameters to obtain the mean trend coefficient, which is used as the leaf vein flow direction angle within the time window.

[0009] Furthermore, the specific process for generating each microphenotype subcluster is as follows: S4-2: Based on the position coordinates of each leaf vein point and the origin, determine the spatial distance from each leaf vein point to the origin, and divide the data into layers according to the spatial distance to generate each microphenotype sub-cluster; S4-3: The process of dividing the data into layers based on spatial distance is as follows: S4-31: Sort all spatial distances in ascending order, and denote them as spatial distance sequences. Let i represent the spatial distance index, and n represent the total number of spatial distance indexes. Based on this, the kernel density estimation (KDE) method is applied to estimate the density of the spatial distance sequence, resulting in the corresponding density value sequence. And by using interpolation, a continuous density function f(x) is constructed, thus forming a smooth density curve y=f(x) representing the spatial distance distribution characteristics, with the domain being [x1, xn]; S4-32: Divide the interval [x1, xn] into L independent intervals at equal intervals, assign all density points (xi, yi) to their corresponding independent intervals, and accumulate the density values ​​within each independent interval to obtain the histogram vector. , where k represents the index of the kth independent interval, with a value range of k=1, 2…L. In the histogram vector, hk represents the total density value within the kth independent interval; Divide the histogram into the left group interval [1…t] and the right group interval [t+1…L]. Calculate the weights w0(t), w1(t) of the two groups, their respective means u0(t), u1(t), and the overall mean uT, where T represents all independent intervals. Calculate the inter-group variance based on the weights w0(t) and w1(t) of the two groups, their respective means u0(t) and u1(t), and the overall mean uT; Selecting the group variance (t) The largest split point t As the optimal segmentation point, the spatial distance corresponding to the midpoint of its independent interval is denoted as the first-level threshold Z1; Repeat step S4-32 with the spatial distance sequence and the corresponding density value sequence for xi≥Z1 to determine the second layer threshold Z2; The first-level threshold and the second-level threshold constitute the layered interval. The first-level threshold is the minimum value of the layered interval, and the second-level threshold is the maximum value of the layered interval. When the spatial distance is less than the minimum value of the layered interval, it is determined to be the first level. When the spatial distance is within the layered interval, it is determined to be the second level. When the spatial distance is greater than the maximum value of the layered interval, it is determined to be the third level. Each microphenotype subcluster is composed of a first level, a second level, and a third level.

[0010] Furthermore, the specific process of matching similar microphenotype subclusters is as follows: Obtain the distance between the coordinates of the leaf vein points of the patient to be matched and the coordinates of the center points of each microphenotype subcluster. Select the microphenotype subcluster with the smallest distance as the similar microphenotype subcluster.

[0011] Furthermore, the specific process for determining the level of poisoning risk is as follows: Extract the duration of illness from the risk trend label information of the patients to be matched, and combine the poisoning start time and the current time of the patients to be matched to calculate the duration of poisoning. Calculate the ratio of the duration of poisoning to the duration of illness to obtain the illness risk score. Extract the outcome and onset rate from the risk trend label information of the patients to be matched, and perform matching analysis with their corresponding preset risk score tables to obtain the outcome risk score and the onset rate risk score. The onset risk score, outcome risk score, and onset speed risk score of the patients to be matched are multiplied by their respective weighting coefficients and then summed to obtain the comprehensive poisoning risk assessment value. The comprehensive poisoning risk assessment value of the patients to be matched is compared and analyzed with the preset comprehensive poisoning risk assessment threshold to determine the poisoning risk type level of the patients to be matched.

[0012] The technical solution provided by this invention has the following advantages compared with the known prior art: 1. This invention constructs a spatial tree structure model of toxicant phenotypes to perform multi-dimensional segmentation and organization of the clinical characteristics of historical poisoning patients. This not only reveals the effects of various toxins on disease progression but also enhances the model's ability to distinguish between different types of poisoning.

[0013] 2. This invention introduces the concept of "leaf vein network" to map the characteristic change trend in the poisoning evolution process into leaf vein points in a two-dimensional coordinate system. Combined with the hierarchical clustering mechanism of spatial distance, it achieves microscopic characterization and accurate labeling of disease risk trends, which is significantly better than traditional static classification methods. 3. This invention indirectly determines the level of poisoning risk by matching the trend of changes in the clinical characteristics of the patient to be assessed with historical microphenotype subclusters, without relying on known types of poisons. This can be applied to novel poisons or rare poisoning events, enhancing the system's practicality and emergency response capabilities in clinical emergency care. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0017] like Figure 1 As shown, a method for assessing the risk level of poisoning in emergency department patients includes the following steps: S1. Feature Extraction: Collect data on patients' clinical characteristic parameters to construct a standardized feature vector set. The specific construction steps are as follows: Within a defined time period, data on patients' clinical characteristic parameters were collected at various monitoring time points to construct a dataset of these parameters. The clinical characteristic parameters included baseline vital signs, symptom presentation, and biochemical test parameters. The most frequently occurring value (the value with the highest frequency) in the dataset was used as the representative value of each clinical characteristic parameter to enhance its statistical stability. Subsequently, the extracted representative values ​​were normalized to construct a standardized feature vector set with a unified format. , where M represents the number of the standardized feature vector, and M=1,2,3…m, m represents the total number of standardized feature vector numbers, and CM represents the Mth standardized feature vector; Among them, basic vital signs parameters include heart rate, blood pressure, respiratory rate, body temperature, pupillary changes, etc. Symptom parameters include consciousness, convulsions, and epileptic seizures; Biochemical parameters include creatinine, lactate, liver enzymes, and electrolytes; It should be noted that the representative values ​​of basic vital signs parameters and biochemical test parameters are the detected values, while the representative values ​​of symptom parameters are determined based on a preset symptom type status table. This symptom type status table defines a status type (such as graded status and binary status), coding description, and parameter value setting for each symptom parameter. For example, the symptom parameter "consciousness" is set as a graded status, with the coding description "awake / drowsy / comatose" and the corresponding parameter values ​​"0 / 1 / 2" respectively. The symptom parameter "convulsions / epileptic seizures" is set as a binary status, with the coding description "present / absent" and the corresponding parameter values ​​"1 / 0" respectively. S2. Model Construction: The clinical characteristic parameters of historical poisoning cases are used to train a spatial tree-structured model of the toxicant phenotype. The specific construction process is as follows: Collect cases of confirmed poisoning patients within a historical time period and execute step S1 to form a standardized set of feature vectors, which will serve as the training input for the model. Using a standardized feature vector set as the initial input node of the model, the priority of feature dimension division is determined based on the information gain of each clinical feature in historical poisoning samples. The standardized feature vectors are then split layer by layer in this order to construct a tree structure model of toxic phenotype space. Each split node represents a judgment condition of a specific clinical feature, and each path from the root node to the leaf node represents a specific combination of clinical phenotype features of poisoning. Finally, each leaf node stores a subset of poisoning samples that satisfy the feature conditions of that path. The spatial tree structure model of toxic phenotypes constructed based on clinical feature segmentation can effectively identify and mine high-order combination relationships between clinical features of poisoning, thereby realizing hierarchical organization and representation of toxic phenotypes. Example Explanation: Taking "amobarbital poisoning" as an example, suppose we have collected a standardized feature vector set of patients with amobarbital poisoning, where the standardized feature vector set includes lactate, consciousness, and convulsions. The construction process is as follows: The standardized feature vector set of amobarbital poisoning is used as the initial input node: The first level of division is based on lactic acid characteristics: if the representative value of lactic acid is >4.0, then it enters branch A; Under branch A, the second layer is based on the characteristics of consciousness: if the representative value of consciousness = 2, then proceed to branch A1; Under branch A1, the third layer is based on the characteristics of convulsions: if the representative value of the convulsion is 1, then it enters branch A1a; Therefore, the combination of clinical phenotypes of poisoning represented by leaf node A1a is: elevated lactate + coma + convulsions, which can be regarded as the typical phenotype of this type of sedative poisoning.

[0018] S3. Matching Similar Poisoning Cases: Collect the standardized feature vector set of patients to be matched and input it into the constructed toxicant phenotypic spatial tree structure model, and perform the following process: Based on the standardized feature vector set of the patients to be matched, and in accordance with the judgment rules of each split node in the tree structure model of toxic phenotype space, path selection and node traversal are performed layer by layer. When the features of the patients to be matched meet all the partitioning conditions of a certain path, they will eventually fall into a specific leaf node. The feature path represented by this leaf node is a set of poisoning clinical phenotype feature combinations that are highly consistent with the features of the patients to be matched, constituting its preliminary classification result in the known model space. At the same time, the subset of historical poisoning samples stored in this leaf node is the subset of poisoning samples under the same combination of poisoning clinical phenotype feature combinations, which can serve as a pool of candidate similar poisoning cases, providing an accurate comparison basis for subsequent toxic risk level assessment and clinical decision-making.

[0019] S4. Risk Level Trend Deduction: Spatial mapping and micro-phenotypic clustering matching are performed on the dynamic evolution characteristics of the patients to be matched within the leaf nodes. The specific steps are as follows: S4-1: Construct leaf vein network units based on the subset of poisoned samples contained in the leaf nodes, specifically as follows: Each poisoned sample is considered as an independent leaf vein, and the temporal evolution relationship between poisoned samples is used to form the connecting fibers between leaf veins, thereby forming a leaf vein network unit. Construct a two-dimensional coordinate system: The horizontal axis (x-axis) represents the velocity of the leaf veins, denoted as v, and is used as a time window. Vertical axis (y-axis): angle of leaf vein flow, denoted as d, representing the overall directionality of the trend of changes in clinical characteristics; The process of solving the leaf vein flow direction angle is as follows: extract the clinical feature parameters within the time window, and for each clinical feature parameter, use linear regression to fit the slope of the clinical feature change over time as its trend coefficient. Then, calculate the mean of the trend coefficients of all clinical feature parameters to obtain the mean trend coefficient, which is used as the leaf vein flow direction angle within the time window. The two quantities v and d of each leaf vein are mapped to coordinate points in a two-dimensional coordinate system to form a leaf vein point; S4-2: Based on the position coordinates of each leaf vein point and the origin, determine the spatial distance from each leaf vein point to the origin (i.e., the straight-line distance from the leaf vein point to the origin in the two-dimensional coordinate system), and divide the data into layers according to the spatial distance to generate micro-phenotype subclusters. Each micro-phenotype subcluster represents a disease with similar duration of onset, outcome (such as recovery, complications, and death), and disease onset speed. The duration of onset, outcome, and disease onset speed constitute risk trend label information. S4-3: The specific process of dividing the data into layers based on spatial distance is as follows: S4-31: Sort all spatial distances in ascending order, and denote them as spatial distance sequences. Let i represent the spatial distance index, and n represent the total number of spatial distance indexes. Based on this, the kernel density estimation (KDE) method is applied to estimate the density of the spatial distance sequence, resulting in the corresponding density value sequence. And by using interpolation, a continuous density function f(x) is constructed, thus forming a smooth density curve y=f(x) representing the spatial distance distribution characteristics, with the domain being [x1, xn]; The above process transforms discrete spatial distances into continuous probability density distributions, providing a statistical basis for subsequent selection of stratification thresholds. S4-32: Divide the interval [x1, xn] into L independent intervals at equal intervals, assign all density points (xi, yi) to their corresponding independent intervals, and accumulate the density values ​​within each independent interval to obtain the histogram vector. , where k represents the index of the kth independent interval, with a value range of k=1, 2…L. In the histogram vector, hk represents the total density value within the kth independent interval; By accumulating density values ​​in independent intervals, a spatial distribution structure represented by a discrete histogram is constructed, providing an operable numerical basis for the subsequent selection of threshold segmentation points; For each split point t (i.e. after the t-th independent interval), the histogram is divided into the left group interval [1…t] and the right group interval [t+1…L]. The weights w0(t), w1(t) of the two groups and their respective means u0(t), u1(t) and the overall mean uT are calculated, where T represents all independent intervals. The specific solution process for the weights w0(t), w1(t) of the two groups, their respective means u0(t), u1(t), and the overall mean uT is as follows: The histogram vectors are normalized according to the formula: , thus obtaining the probability density value; The weights w0(t) and w1(t) are calculated based on the probability density values, according to the formula: ; Formulas for calculating the group mean u0(t) and u1(t): ; The formula for calculating the overall mean uT is: ; Calculate the between-group variance based on the weights w0(t) and w1(t) of the two groups, their respective means u0(t) and u1(t), and the overall mean uT. (t), the specific formula is: ; Selecting the group variance (t) The largest split point t As the optimal segmentation point, the spatial distance corresponding to the midpoint of its independent interval is denoted as the first-level threshold Z1; It should be noted that determining the optimal split point by maximizing the variance between groups is a statistically effective method for capturing the natural distribution boundary of data. In spatial distance distribution, this method helps to identify the most significant density change points, making the defined threshold have the strongest structural stratification significance, thus laying a solid foundation for the construction of microphenotypic subclusters. Repeat step S4-32 with the spatial distance sequence and the corresponding density value sequence for xi≥Z1 to determine the second layer threshold Z2; The first-level threshold and the second-level threshold constitute the layered interval. The first-level threshold is the minimum value of the layered interval, and the second-level threshold is the maximum value of the layered interval. When the spatial distance is less than the minimum value of the layered interval, it is determined to be the first level. When the spatial distance is within the layered interval, it is determined to be the second level. When the spatial distance is greater than the maximum value of the layered interval, it is determined to be the third level. Each microphenotype subcluster is composed of a first level, a second level, and a third level; The vein flow velocity and vein flow angle of the patient to be matched are mapped onto the corresponding two-dimensional coordinate system inside the leaf node to obtain the vein point of the patient to be matched. Based on the coordinate position of the vein point, similar microphenotypic subclusters are matched, specifically as follows: The distance between the coordinates of the leaf vein points of the patient to be matched and the coordinates of the center points of each microphenotype sub-cluster (i.e., the arithmetic mean of the coordinates of all leaf vein points in the two-dimensional coordinate system in the microphenotype sub-cluster) is obtained. The microphenotype sub-cluster with the smallest distance is selected as the similar microphenotype sub-cluster. Based on this, the risk trend label information of the patient to be matched is obtained. The risk trend label information corresponding to the similar microphenotype sub-cluster is the risk trend label information of the patient to be matched. S5. Risk Level Output: Based on the risk trend label information of the patients to be matched, the poisoning risk type level is determined. The specific analysis process is as follows: Extract the duration of illness from the risk trend label information of the patients to be matched, and combine the poisoning start time and the current time of the patients to be matched to calculate the duration of poisoning. Calculate the ratio of the duration of poisoning to the duration of illness to obtain the illness risk score. Extract the outcome results from the risk trend label information of the patients to be matched, and perform matching analysis between the outcome results and the preset outcome risk score table to obtain the outcome risk score; Extract the onset speed from the risk trend label information of the patients to be matched, and perform matching analysis between the onset speed and the preset onset speed risk score table to obtain the onset speed risk score; The onset risk score, outcome risk score, and onset speed risk score of the patients to be matched are multiplied by their respective weighting coefficients and then summed to obtain the comprehensive poisoning risk assessment value. The comprehensive poisoning risk assessment value of the patients to be matched is compared and analyzed with the preset comprehensive poisoning risk assessment threshold to determine the poisoning risk type level of the patients to be matched. If the comprehensive assessment value of the poisoning risk of the patient to be matched is greater than the preset comprehensive assessment threshold of poisoning risk, it is judged as a level 1 poisoning risk, which means that the risk level is high and immediate intervention is required. If the comprehensive poisoning risk assessment value of the patient to be matched is equal to the preset comprehensive poisoning risk assessment threshold, it is determined to be a level 2 poisoning risk, which means a moderate risk and requires close monitoring. If the comprehensive assessment value of the poisoning risk of the patient to be matched is less than the preset comprehensive assessment threshold of poisoning risk, it is judged as a level three poisoning risk, which means that the risk is low and can be routinely observed. The poisoning risk level of the patients to be matched is composed of Level 1, Level 2, and Level 3 poisoning risks, so as to serve as a basis for clinical decision-making and guide subsequent treatment strategies and intervention pathways.

[0020] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement any of the methods described above; A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any of the methods described above.

[0021] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for assessing the risk level of poisoning in emergency department patients, characterized in that, Includes the following steps: Collect data on patients' clinical characteristic parameters and construct a standardized set of feature vectors; The standardized feature vector set of historically diagnosed poisoning patients is used as training input. Based on the information gain of each clinical feature in historical poisoning samples, the priority of feature dimension division is determined. The standardized feature vectors are then split sequentially to construct a spatial tree structure model of toxic substance phenotype. The standardized feature vector set of the patients to be matched is input into the tree structure model of the toxic phenotype space to determine the leaf node where it falls and obtain the subset of poisoned samples. Based on the subset of poisoned samples, leaf vein network units were constructed, and each leaf vein was mapped to a leaf vein point. The spatial distance between each leaf vein point and the origin of coordinates was calculated. Hierarchical clustering was performed to generate each micro phenotypic sub-cluster, and the corresponding risk trend label information was output. The risk trend label information includes the duration of disease onset, the outcome, and the speed of disease onset. Based on the coordinates of the leaf vein points of the patients to be matched, similar microphenotype subclusters are matched to obtain the risk trend label information of the patients to be matched, and the poisoning risk type level is determined accordingly.

2. The method for assessing the risk level of poisoning in emergency department patients according to claim 1, characterized in that, The specific process of constructing a standardized feature vector set is as follows: Within a set time period, data on patients' clinical characteristic parameters are collected at various monitoring time points over a period of time to construct a dataset of patients' clinical characteristic parameters. The clinical characteristic parameters include basic vital signs parameters, symptom presentation parameters, and biochemical test parameters. The value corresponding to the mode in the dataset of clinical characteristic parameters is used as the representative value of the clinical characteristic parameter. Subsequently, the extracted representative values ​​of clinical characteristic parameters are normalized to construct a standardized feature vector set with a unified format.

3. The method for assessing the risk level of poisoning in emergency department patients according to claim 1, characterized in that, The specific process for obtaining a subset of poisoned samples is as follows: Based on the standardized feature vector set of the patients to be matched, and in accordance with the judgment rules of each split node in the tree structure of the toxic phenotype space, path selection and node traversal are performed layer by layer. When the features of the patients to be matched meet all the partitioning conditions of a certain path, they will eventually fall into a specific leaf node. The feature path represented by this leaf node is a set of poisoning clinical phenotype feature combinations that are highly consistent with the features of the patients to be matched, constituting its preliminary classification result in the known model space. At the same time, the historical poisoning sample subset stored in this leaf node is the poisoning sample subset under the same poisoning clinical phenotype feature combination conditions.

4. The method for assessing the risk level of poisoning in emergency department patients according to claim 1, characterized in that, The specific process of constructing leaf vein network units is as follows: Each poisoned sample is considered as an independent leaf vein, and the temporal evolution relationship between poisoned samples is used to form the connecting fibers between leaf veins, thereby forming a leaf vein network unit.

5. The method for assessing the risk level of poisoning in emergency department patients according to claim 1, characterized in that, The specific process of mapping to leaf vein points is as follows: A two-dimensional coordinate system is constructed with the vein flow velocity v as the abscissa and the vein flow angle d as the ordinate. The two quantities v and d of each vein are mapped to coordinate points in the two-dimensional coordinate system to form a vein point. Among them, the vein flow velocity v is the time window, and the vein flow angle d is the average change trend coefficient. The process of determining the leaf vein flow direction angle is as follows: extract the clinical feature parameters within the time window, and for each clinical feature parameter, use linear regression to fit the slope of the clinical feature change over time as its trend coefficient. Then, calculate the mean of the trend coefficients of all clinical feature parameters to obtain the mean trend coefficient, which is used as the leaf vein flow direction angle within the time window.

6. The method for assessing the risk level of poisoning in emergency department patients according to claim 1, characterized in that, The specific process for generating each microphenotype subcluster is as follows: S4-2: Based on the position coordinates of each leaf vein point and the origin, determine the spatial distance from each leaf vein point to the origin, and divide the data into layers according to the spatial distance to generate each microphenotype sub-cluster; S4-3: The process of dividing the data into layers based on spatial distance is as follows: S4-31: Sort all spatial distances in ascending order, and denote them as spatial distance sequences. Let i represent the spatial distance index and n represent the total number of spatial distance indexes. Based on this, the kernel density estimation method is applied to estimate the density of the spatial distance sequence, resulting in the corresponding density value sequence. And by using interpolation, a continuous density function f(x) is constructed, thus forming a smooth density curve y=f(x) representing the spatial distance distribution characteristics, with the domain being [x1, xn]; S4-32: Divide the interval [x1, xn] into L independent intervals at equal intervals, assign all density points (xi, yi) to their corresponding independent intervals, and accumulate the density values ​​within each independent interval to obtain the histogram vector. , where k represents the index of the kth independent interval, with a value range of k=1, 2…L. In the histogram vector, hk represents the total density value within the kth independent interval; Divide the histogram into the left group interval [1…t] and the right group interval [t+1…L]. Calculate the weights w0(t), w1(t) of the two groups, their respective means u0(t), u1(t), and the overall mean uT, where T represents all independent intervals. Calculate the inter-group variance based on the weights w0(t) and w1(t) of the two groups, their respective means u0(t) and u1(t), and the overall mean uT; Selecting the group variance (t) The largest split point t As the optimal segmentation point, the spatial distance corresponding to the midpoint of its independent interval is denoted as the first-level threshold Z1; Repeat step S4-32 with the spatial distance sequence and the corresponding density value sequence for xi≥Z1 to determine the second layer threshold Z2; The first-level threshold and the second-level threshold constitute the layered interval. The first-level threshold is the minimum value of the layered interval, and the second-level threshold is the maximum value of the layered interval. When the spatial distance is less than the minimum value of the layered interval, it is determined to be the first level. When the spatial distance is within the layered interval, it is determined to be the second level. When the spatial distance is greater than the maximum value of the layered interval, it is determined to be the third level. Each microphenotype subcluster is composed of a first level, a second level, and a third level.

7. The method for assessing the risk level of poisoning in emergency department patients according to claim 1, characterized in that, The specific process of matching similar microphenotype subclusters is as follows: Obtain the distance between the coordinates of the leaf vein points of the patient to be matched and the coordinates of the center points of each microphenotype subcluster. Select the microphenotype subcluster with the smallest distance as the similar microphenotype subcluster.

8. The method for assessing the risk level of poisoning in emergency department patients according to claim 1, characterized in that, The specific process for determining the level of poisoning risk is as follows: Extract the duration of illness from the risk trend label information of the patients to be matched, and combine the poisoning start time and the current time of the patients to be matched to calculate the duration of poisoning. Calculate the ratio of the duration of poisoning to the duration of illness to obtain the illness risk score. Extract the outcome and onset rate from the risk trend label information of the patients to be matched, and perform matching analysis with their corresponding preset risk score tables to obtain the outcome risk score and the onset rate risk score. The onset risk score, outcome risk score, and onset speed risk score of the patients to be matched are multiplied by their respective weighting coefficients and then summed to obtain the comprehensive poisoning risk assessment value. The comprehensive poisoning risk assessment value of the patients to be matched is compared and analyzed with the preset comprehensive poisoning risk assessment threshold to determine the poisoning risk type level of the patients to be matched.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

Images