Intelligent triage method and system based on face recognition

Abstract

The invention provides an intelligent triage method and system based on face recognition, and the method comprises the steps: completing the identity verification through the comparison of face recognition and certificate information, and calling the medical history data of a patient; performing asymmetric fuzzification processing on currently acquired text chief complaint data, vital sign data and medical history data of the patient to generate a department related symptom feature set; constructing a symptom-department association rule base, and based on the symptom-department association rule base, generating department triage suggestions through a dynamic department decision model; and performing confidence coefficient optimization on the department triage suggestion, and outputting a final department triage report. According to the method, multi-symptom co-scene misjudgment and triage deviation of special crowds can be effectively reduced, a high-credibility and high-adaptability pre-examination triage solution is provided for intelligent medical treatment, and compared with traditional manual triage, the triage accuracy and emergency pre-examination efficiency can be greatly improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent triage technology, and in particular to an intelligent triage method and system based on facial recognition. Background Technology

[0002] As the contradiction between the increasing demand for medical care and the uneven distribution of medical resources becomes increasingly prominent, the efficiency and accuracy of triage, as the first link in the medical process, directly affect the timeliness of patient treatment and the operational efficiency of hospitals.

[0003] Traditional manual triage relies on nurses' clinical experience, which is subject to problems such as high subjectivity, low efficiency (average time of 5-10 minutes per case), and susceptibility to fatigue or outdated knowledge. In addition, triage decisions for complex symptoms (such as "chest pain + dizziness") and special populations (such as elderly patients) require extremely high levels of nurse experience, with a misjudgment rate of 15%-20%. Summary of the Invention

[0004] Therefore, the purpose of this invention is to propose an intelligent triage method and system based on facial recognition to solve the problems mentioned above.

[0005] According to the present invention, an intelligent triage method based on face recognition is proposed, the method comprising: Identity verification is completed by comparing facial recognition with document information, and the patient's medical history data is retrieved; The patient's textual complaint data, vital sign data, and medical history data are subjected to asymmetric fuzzification processing to generate a set of department-related symptom features; A symptom-department association rule base is constructed, and based on the symptom-department association rule base, a dynamic department decision model is used to generate department triage suggestions; The confidence level of the departmental triage suggestions is optimized, and the final departmental triage report is output.

[0006] Furthermore, the currently acquired patient's textual complaint data, vital sign data, and medical history data undergo asymmetric fuzzification processing to generate a department-related symptom feature set, including: The medical history data will be converted into departmental risk factors. The vital sign data and departmental risk factors are mapped to departmental symptom membership degrees using an asymmetric membership function; Keywords are extracted from the textual complaint data and mapped to the preset departmental symptom dimension to obtain the current patient's department-related symptom feature set.

[0007] Furthermore, the vital sign data and departmental risk factors are mapped to departmental symptom membership degrees using an asymmetric membership function, as shown in the formula: When x ≤ a, u(x) = 0. When \(a \lt x\leq b\), , When \(x \gt b\), , where \(x\) is the vital sign value or department risk factor of the patient, \(u(x)\) is the department symptom membership value, and \(a\) and \(b\) are the clinical statistical boundary thresholds. , are slope parameters. .

[0008] Furthermore, the construction of the symptom - department association rule base includes: Obtain historical triage data from the hospital information system and count the association strength between symptoms and departments: , where represents the number of times that symptom appears simultaneously and the patient visits department , represents the total number of times that symptom appears in the historical cases; Within the sliding time window \(T\), calculate the information entropy of the symptoms: , \(N\) is the total number of departments, , is the number of times that symptom appears in department within the sliding time window \(T\), is the total number of times that symptom appears within the sliding time window \(T\); Based on the information entropy normalization, calculate the dynamic weight of the symptoms: , where is the symptom weight of symptom in the symptom - department association rule base, \(m\) is the total number of symptoms, is the information entropy of symptom within the window period.

[0009] Furthermore, the generation of department triage suggestions through the dynamic department decision model includes: For each symptom in the set of department - related symptom characteristics of the current patient, obtain its membership degree, where \(n\leq m\); Calculate the triage score of each department : , where is the set of department - related symptom characteristics of the current patient, \(n\) is the total number of symptoms of the current patient, is the fuzzy membership degree of symptom ; Sort according to the triage scores from high to low, and select the top 3 departments with the highest scores as the department triage suggestions, that is, the department score set ; When the difference between the highest score and the second highest score is less than 10%, it is marked as low-confidence triage.

[0010] Furthermore, the confidence level of the departmental triage recommendations is optimized, and a final departmental triage report is output, including: The departmental score set in the departmental triage suggestion is as follows: , ; When the highest score and the second highest score satisfy: If so, it is determined that there is a triage conflict; The patient's age group (A) and biological sex (G) were extracted from the facial recognition results. When triage conflicts exist, the scores of the conflicting departments are adjusted according to predefined correction rules: , ,in, The department's score is The correction factor for the department, The department's score is The correction coefficients for each department range from [0.5 to 1.5]. Generate a final departmental triage report, including recommended departmental sequences and key symptom information.

[0011] This invention also proposes an intelligent triage system based on facial recognition, used to implement the above-mentioned intelligent triage method based on facial recognition, the system comprising: Identity verification module: Used to verify identity by comparing facial recognition with document information and to retrieve the patient's medical history data; Fuzzy processing module: used to perform asymmetric fuzzing processing on the currently acquired patient's textual chief complaint data, vital sign data and medical history data to generate a set of department-related symptom features; The triage suggestion module is used to construct a symptom-department association rule base and generate department triage suggestions based on the symptom-department association rule base through a dynamic department decision model. The triage suggestion optimization module is used to optimize the confidence level of the department triage suggestions and output the final department triage report.

[0012] In summary, the intelligent triage method based on facial recognition of this invention verifies identity by comparing facial recognition with document information to obtain the patient's medical history data. It performs asymmetric fuzzification processing on textual complaints, vital signs, and medical history data, overcoming the limitations of traditional threshold methods in handling ambiguous symptoms (such as "severe headache"), transforming heterogeneous data into a unified set of symptom features. This process both tolerates boundary value noise and aligns with clinical semantic logic. A symptom-department association rule base is constructed, based on historical case statistics and a dynamic sliding window update mechanism, ensuring that triage decisions are both objective and adaptable to current epidemic trends. Finally, confidence optimization corrects and resolves low-confidence triage conflicts, outputting the final departmental triage report. This method effectively reduces misjudgments in scenarios with multiple co-occurring symptoms and triage biases in special populations, providing a highly reliable and adaptable pre-screening and triage solution for smart healthcare. Compared to traditional manual triage, it significantly improves triage accuracy and emergency pre-screening efficiency.

[0013] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by means of embodiments of the invention. Attached Figure Description

[0014] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of an intelligent triage method based on face recognition according to Embodiment 1 of the present invention; Figure 2 This is a system block diagram of the intelligent triage system based on face recognition according to Embodiment 2 of the present invention. Detailed Implementation

[0015] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0016] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0018] Example 1 Please see Figure 1 This invention proposes an intelligent triage method based on facial recognition, which includes steps S101 to S104: S101 verifies identity by comparing facial recognition with document information and retrieves the patient's medical history data.

[0019] Understandably, the facial recognition system collects patients' facial biometrics in real time and matches them with the pre-reserved images in the document information database to verify their identity. After verification, the system automatically links to the hospital information system (HIS) and retrieves medical history data from the structured electronic medical record based on the patient's unique identifier, including but not limited to past medical history, surgical records, drug allergy history, and other core medical records, providing key background information for subsequent triage decisions.

[0020] S102, perform asymmetric fuzzing processing on the currently acquired patient's textual complaint data, vital sign data, and medical history data to generate a set of department-related symptom features.

[0021] Optionally, the process of performing asymmetric fuzzification on the currently acquired patient's textual complaint data, vital sign data, and medical history data to generate a department-related symptom feature set includes: The medical history data will be converted into departmental risk factors. The vital sign data and departmental risk factors are mapped to departmental symptom membership degrees using an asymmetric membership function; Keywords are extracted from the textual complaint data and mapped to the preset departmental symptom dimension to obtain the current patient's department-related symptom feature set.

[0022] Understandably, converting medical history data (such as history of diabetes or hypertension) into departmental risk factors (e.g., history of diabetes → endocrinology risk +0.8) aims to quantify the impact of past illnesses on current triage. By pre-setting mapping rules, qualitative medical history information can be transformed into quantitative departmental association strength, avoiding the problems of ignoring medical history or subjectively assessing its weight in traditional triage. For example, a dizziness complaint in a diabetic patient is more likely to be associated with endocrinology than neurology; quantifying risk factors can clarify this tendency.

[0023] Design an asymmetric membership function for the patient's current vital sign data (such as body temperature, blood pressure) and department risk factors to achieve data fuzzification and alignment with clinical logic. The asymmetric function uses piecewise mapping (rapid rise in the low-value area and gentle rise in the high-value area), which not only tolerates fluctuations in boundary values due to noise (such as the slight difference between 37.1°C and 37.2°C in body temperature), avoiding sudden changes in triage results caused by data noise, but also highlights the contribution of significant abnormalities to department association (such as a blood pressure of 180 mmHg requiring more vigilance in the cardiovascular department than 140 mmHg), conforming to doctors' diagnostic habits of prioritizing attention to abnormalities in key indicators.

[0024] Extract keywords from the patient's text chief complaint (such as chest pain for 3 days) and map them to the preset department symptom dimensions (such as chest pain → cardiovascular symptom dimension). The preset dimensions are constructed based on a clinical knowledge base (such as chest pain being associated with the cardiovascular department, respiratory department, etc.), ensuring the accuracy of the symptom description and department association. At the same time, unstructured text is converted into numerical features that can be calculated by the model, such as the membership degree of chest pain in the cardiovascular dimension being 0.9.

[0025] Further optionally, the vital sign data and department risk factors are mapped to department symptom membership degrees through an asymmetric membership function. The formula is: When x ≤ a, u(x) = 0, When a < x ≤ b, , When x > b, , where x is the patient's vital sign value or department risk factor, u(x) is the department symptom membership degree value, a and b are clinical statistical boundary thresholds, , are slope parameters, .

[0026] It is understandable that through the asymmetric membership function, the vital sign data and department risk factors are mapped to department symptom membership degrees. This asymmetric membership function is divided into three intervals: when the index value x is lower than the clinical normal threshold a, the membership degree u(x) = 0, indicating no department association; when x is between a and the warning threshold b, an S-shaped function with a slope is used to make the membership degree rise rapidly (such as when the body temperature rises from 37.3°C to 38.0°C, the membership degree increases from 0.1 to 0.7), simulating doctors' sensitive attention to indicators close to the abnormal threshold; when x exceeds b, an S-shaped function with a slope is used instead, and the growth rate of the membership degree slows down (such as when the blood pressure rises from 180 mmHg to 190 mmHg, the membership degree only increases from 0.8 to 0.85), avoiding overreacting to indicators that are already clearly abnormal.

[0027] This embodiment addresses the problem of oversensitivity in the low-value region or insufficient response in the high-value region of traditional symmetric membership functions by using segmented processing. For example, a small difference between a body temperature of 37.1℃ and 37.2℃ will not lead to a sudden change in triage results, while a high fever of 39.0℃ will be clearly associated with the infectious disease department or the respiratory department.

[0028] By incorporating risk factors related to medical history (such as a history of diabetes → endocrinology risk +0.8) into the same functional framework, we can ensure that the impact of historical diseases on current triage is neither overemphasized, such as mild risks not dominating decisions, nor ignored, such as high risks significantly increasing the correlation strength.

[0029] S103, construct a symptom-department association rule base, and based on the symptom-department association rule base, generate department triage suggestions through a dynamic department decision-making model.

[0030] Further, optionally, the construction of the symptom-department association rule base includes: Historical triage data is obtained from the hospital information system, and the correlation strength between symptoms and departments is statistically analyzed. ,in, This indicates that symptoms appeared simultaneously in historical cases. And in the department Number of visits Indicates symptoms in historical cases Total number of times; Within the sliding time window T, calculate the information entropy of the symptoms: N is the total number of departments. , Symptoms within the sliding time window T In the department The number of times it appears, Symptoms within the sliding time window T Total number of occurrences; Dynamic weights of symptoms are calculated based on information entropy normalization: ,in, Symptoms in the symptom-department association rule base The symptom weights, where m is the total number of symptoms. For symptoms Information entropy within the window period.

[0031] Understandably, this embodiment uses historical data statistics and real-time information entropy analysis to dynamically adjust the correlation strength, thereby constructing a dynamic symptom-department correlation rule base. Specifically, firstly, based on historical triage data from the hospital information system, the correlation strength between symptoms and departments is calculated. Symptoms In the department The conditional probability of occurrence is used. By quantifying the co-occurrence frequency of historical cases, objective evidence is provided for the association between symptoms and departments (e.g., chest pain has a 90% probability of occurrence in the cardiology department), rather than relying on manual experience. Secondly, a sliding time window (default 30 days) is introduced to calculate the information entropy of symptoms. This measure assesses the uncertainty of symptom distribution across departments. Lower information entropy indicates a stronger concentration of symptoms in a specific department; higher information entropy indicates a more dispersed symptom distribution, such as abdominal pain potentially involving multiple departments including gastroenterology and urology. This allows the association rule base to capture real-time disease trends, such as an increased association between fever and respiratory departments during an epidemic, overcoming the lag of static historical data. Finally, dynamic weights are calculated based on information entropy normalization. Low-entropy symptoms (strong departmental orientation) are given higher weights, while high-entropy symptoms (weak departmental orientation) are given lower weights.

[0032] Further optionally, the generation of departmental triage suggestions through the dynamic departmental decision-making model includes: For each symptom in the current patient's department-related symptom feature set, obtain its membership degree, where n≤m; Calculate each department Triage score: ,in, Let n be the set of department-related symptoms of the current patient, and n be the total number of symptoms of the current patient. For symptoms Fuzzy membership degree; According to triage score Sort the departments from highest to lowest score and select the three highest-scoring departments as the department triage recommendations, i.e., the department score set. ; When the difference between the highest score and the second highest score is less than 10%, it is marked as low-confidence triage.

[0033] Understandably, the core of generating departmental triage suggestions through a dynamic departmental decision-making model lies in achieving accurate triage decisions through multi-dimensional evidence fusion and uncertainty identification. Specifically, based on the fuzzification processing results of step S102, the model obtains the fuzzy membership degree of each symptom in the current patient's symptom feature set (e.g., the membership degree of chest pain is 0.85), ensuring that heterogeneous data such as vital signs and medical history are quantified into computable association strengths. Secondly, a symptom-department association rule base is constructed (…). ) and dynamic weights Calculate the triage score for each department. This formula, through weighted fusion, integrates individual patient characteristics (membership degree) and long-term clinical experience (…). ) and short-term disease trends ( This effectively addresses the problem of traditional techniques relying solely on historical data or ignoring the ambiguity of symptoms. For example, the increased dynamic weighting of "fever" during flu season makes it more significant in respiratory triage, thus adapting to changes in epidemic trends.

[0034] Furthermore, the model selects the top three departments based on triage scores as recommendations, providing both priority references (e.g., cardiology with the highest score) and alternative options (e.g., respiratory medicine, emergency medicine), avoiding rigid decision-making due to a single recommendation. Finally, by setting a confidence threshold (marking low confidence when the difference between the highest and second-highest scores is <10%), potential conflicting cases are identified (e.g., chest pain may be associated with cardiology and gastroenterology), triggering subsequent confidence optimization processes to ensure the reliability of triage results.

[0035] S104, optimize the confidence level of the department triage suggestions and output the final department triage report.

[0036] Further, optionally, the confidence optimization of the departmental triage suggestions, and the output of the final departmental triage report, includes: The departmental score set in the departmental triage suggestion is as follows: , ; When the highest score and the second highest score satisfy: If so, it is determined that there is a triage conflict; The patient's age group (A) and biological sex (G) were extracted from the facial recognition results. When triage conflicts exist, the scores of the conflicting departments are adjusted according to predefined correction rules: , ,in, The department's score is The correction factor for the department, The department's score is The correction coefficients for each department range from [0.5 to 1.5]. Generate a final departmental triage report, including recommended departmental sequences and key symptom information.

[0037] Understandably, confidence optimization methods can be used to address uncertainties in triage decisions and output a final triage report, thereby improving the reliability and clinical acceptability of triage results.

[0038] Specifically, low-confidence triage conflicts are identified by setting a threshold (the difference between the highest and second-highest scores is <10%). For example, when the cardiology score is 90 and the respiratory score is 85 (a difference of 5.56%), the system determines that there is a conflict, avoiding misassignment due to the model's ambiguity in decision-making for scenarios with multiple co-occurring symptoms (such as "chest pain + cough"), and providing triggering conditions for subsequent optimization.

[0039] Based on facial recognition results, patient age group (A) and biological sex (G) are extracted, and the scores of conflicting departments are adjusted using predefined correction rules. Correction coefficients are used. , The value range is [0.5, 1.5], and it is designed based on prior clinical knowledge. For example, chest pain in elderly patients (A≥65 years old) in the cardiology department... It may increase by 1.2 times (enhancing the predisposition to age-related diseases), while abdominal pain in young women (G=female) is more common in gynecological conditions. It could potentially increase by 1.3 times (considering gender-specific diseases). By dynamically adjusting the scores, the triage results can be made more aligned with the clinical patterns of specific populations, reducing misjudgments caused by a one-size-fits-all approach.

[0040] Finally, a final triage report is generated, including a recommended department sequence sorted in descending order of adjusted scores, and the top 3 symptom criteria with the highest affiliation (e.g., chest pain 0.85, cough 0.78, history of diabetes 0.8). This allows doctors to quickly verify the criteria (e.g., a high affiliation for a history of diabetes supports endocrinology), and patients can understand the reasons for triage.

[0041] In summary, the intelligent triage method based on facial recognition of this invention verifies identity by comparing facial recognition with document information to obtain the patient's medical history data. It performs asymmetric fuzzification processing on textual complaints, vital signs, and medical history data, overcoming the limitations of traditional threshold methods in handling ambiguous symptoms (such as "severe headache"), transforming heterogeneous data into a unified set of symptom features. This process both tolerates boundary value noise and aligns with clinical semantic logic. A symptom-department association rule base is constructed, based on historical case statistics and a dynamic sliding window update mechanism, ensuring that triage decisions are both objective and adaptable to current epidemic trends. Finally, confidence optimization corrects and resolves low-confidence triage conflicts, outputting the final departmental triage report. This method effectively reduces misjudgments in scenarios with multiple co-occurring symptoms and triage biases in special populations, providing a highly reliable and adaptable pre-screening and triage solution for smart healthcare. Compared to traditional manual triage, it significantly improves triage accuracy and emergency pre-screening efficiency.

[0042] Example 2 Please see Figure 2 This invention proposes an intelligent triage system based on facial recognition, which includes: Identity verification module: Used to verify identity by comparing facial recognition with document information and to retrieve the patient's medical history data; Fuzzy processing module: used to perform asymmetric fuzzing processing on the currently acquired patient's textual chief complaint data, vital sign data and medical history data to generate a set of department-related symptom features; Triage recommendation module: used to construct a symptom - department association rule base, and based on the symptom - department association rule base, generate department triage recommendations through a dynamic department decision model; Triage recommendation optimization module: used to optimize the confidence level of the department triage recommendations and output the final department triage report.

[0043] Further optionally, the fuzzy processing module is further used for: Convert the medical history data into department risk factors; Map the vital sign data and department risk factors to department symptom membership degrees through an asymmetric membership function; Extract keywords from the text chief complaint data and map them to a preset department symptom dimension to obtain the set of department - related symptom characteristics of the current patient.

[0044] Further optionally, the fuzzy processing module is further used for: The mapping of the vital sign data and department risk factors to department symptom membership degrees through an asymmetric membership function, the formula is: When x ≤ a, u(x) = 0, When a < x ≤ b, , When x > b, , where x is the vital sign value of the patient or the department risk factor, u(x) is the department symptom membership degree value, a and b are clinical statistical boundary thresholds, , are slope parameters, .

[0045] Further optionally, the triage recommendation module is further used for: Obtain historical triage data from the hospital information system and statistically analyze the association strength between symptoms and departments: , where, represents the number of times that symptom appears simultaneously in historical cases and the patient visits department , represents the total number of times that symptom appears in historical cases; Within the sliding time window T, calculate the information entropy of the symptoms: , N is the total number of departments, , is the number of times that symptom appears in department within the sliding time window T, is the total number of times that symptom appears within the sliding time window T; Based on the information entropy normalization, calculate the dynamic weights of the symptoms: ,in, Symptoms in the symptom-department association rule base The symptom weights, where m is the total number of symptoms. For symptoms Information entropy within the window period.

[0046] Further optionally, the triage suggestion module is also used for: For each symptom in the current patient's department-related symptom feature set, obtain its membership degree, where n≤m; Calculate each department Triage score: ,in, Let n be the set of department-related symptoms of the current patient, and n be the total number of symptoms of the current patient. For symptoms Fuzzy membership degree; Based on the triage scores, sorted from highest to lowest, the three departments with the highest scores were selected as the departmental triage recommendations, i.e., the departmental score set. ; When the difference between the highest score and the second highest score is less than 10%, it is marked as low-confidence triage.

[0047] Further optionally, the triage suggestion optimization module is also used for: The departmental score set in the departmental triage suggestion is as follows: , ; When the highest score and the second highest score satisfy: If so, it is determined that there is a triage conflict; The patient's age group (A) and biological sex (G) were extracted from the facial recognition results. When triage conflicts exist, the scores of the conflicting departments are adjusted according to predefined correction rules: , ,in, The department's score is The correction factor for the department, The department's score is The correction coefficients for each department range from [0.5 to 1.5]. Generate a final departmental triage report, including recommended departmental sequences and key symptom information.

[0048] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A smart triage method based on facial recognition, characterized in that, The method includes: Identity verification is completed by comparing facial recognition with document information, and the patient's medical history data is retrieved; The patient's textual complaint data, vital sign data, and medical history data are subjected to asymmetric fuzzification processing to generate a set of department-related symptom features; A symptom-department association rule base is constructed, and based on the symptom-department association rule base, a dynamic department decision model is used to generate department triage suggestions; The confidence level of the departmental triage suggestions is optimized, and the final departmental triage report is output.

2. The intelligent triage method based on face recognition according to claim 1, characterized in that, The process involves asymmetric fuzzification of the currently acquired patient's textual complaint data, vital sign data, and medical history data to generate a set of department-related symptom features, including: The medical history data will be converted into departmental risk factors. The vital sign data and departmental risk factors are mapped to departmental symptom membership degrees using an asymmetric membership function; Keywords are extracted from the textual complaint data and mapped to the preset departmental symptom dimension to obtain the current patient's department-related symptom feature set.

3. The intelligent triage method based on face recognition according to claim 2, characterized in that, The vital sign data and departmental risk factors are mapped to departmental symptom membership degrees using an asymmetric membership function, as follows: When x ≤ a, u(x) = 0. When a < x ≤ b, , When x > b , Where x represents the patient's vital signs or departmental risk factor, u(x) represents the departmental symptom membership value, and a and b are clinical statistical boundary thresholds. , For slope parameter, .

4. The intelligent triage method based on face recognition according to claim 1, characterized in that, The construction of the symptom-department association rule base includes: Historical triage data is obtained from the hospital information system, and the correlation strength between symptoms and departments is statistically analyzed. ,in, This indicates that symptoms appeared simultaneously in historical cases. And in the department Number of visits Indicates symptoms in historical cases Total number of times; Within the sliding time window T, calculate the information entropy of the symptoms: N is the total number of departments. , Symptoms within the sliding time window T In the department The number of times it appears, Symptoms within the sliding time window T Total number of occurrences; Dynamic weights of symptoms are calculated based on information entropy normalization: ,in, Symptoms in the symptom-department association rule base The symptom weights, where m is the total number of symptoms. For symptoms Information entropy within the window period.

5. The intelligent triage method based on face recognition according to claim 3 or 4, characterized in that, The generation of departmental triage suggestions through a dynamic departmental decision-making model includes: For each symptom in the current patient's department-related symptom feature set, obtain its membership degree, where n≤m; Calculate each department Triage score: ,in, Let n be the set of department-related symptoms of the current patient, and n be the total number of symptoms of the current patient. For symptoms Fuzzy membership degree; Based on the triage scores, sorted from highest to lowest, the three departments with the highest scores were selected as the departmental triage recommendations, i.e., the departmental score set. ; When the difference between the highest score and the second highest score is less than 10%, it is marked as low-confidence triage.

6. The intelligent triage method based on face recognition according to claim 1, characterized in that, The process of optimizing the confidence level of the departmental triage recommendations and outputting the final departmental triage report includes: The departmental score set in the departmental triage suggestion is as follows: , ; When the highest score and the second highest score satisfy: If so, it is determined that there is a triage conflict; The patient's age group (A) and biological sex (G) were extracted from the facial recognition results. When triage conflicts exist, the scores of the conflicting departments are adjusted according to predefined correction rules: , ,in, The department's score is The correction factor for the department, The department's score is The correction coefficients for each department range from [0.5 to 1.5]. Generate a final departmental triage report, including recommended departmental sequences and key symptom information.

7. A facial recognition-based intelligent triage system, used to implement the facial recognition-based intelligent triage method according to any one of claims 1 to 6, characterized in that, The system includes: Identity verification module: Used to verify identity by comparing facial recognition with document information and to retrieve the patient's medical history data; Fuzzy processing module: used to perform asymmetric fuzzing processing on the currently acquired patient's textual chief complaint data, vital sign data and medical history data to generate a set of department-related symptom features; The triage suggestion module is used to construct a symptom-department association rule base and generate department triage suggestions based on the symptom-department association rule base through a dynamic department decision model. The triage suggestion optimization module is used to optimize the confidence level of the department triage suggestions and output the final department triage report.

Images