Intelligent hospital guide method based on hospital guide elevator, hospital guide elevator and system

Abstract

The invention provides an intelligent hospital guide method based on a hospital guide elevator, the hospital guide elevator and a system. The method comprises the steps that a patient is guided to enter a preset hospital guide elevator for data acquisition; determining abnormal signs, criticality levels and candidate diseases of the patient, and predicting an association probability; selecting a current optimal doctor-seeing department of the patient, and constructing a doctor-seeing path; screening necessary detection items of the patient and making an appointment in advance; and controlling the hospital guide elevator to arrive at the floor and then guiding the patient to arrive at the optimal doctor-seeing department. According to the scheme, the hospital guide elevator is used as a carrier, the intelligent hospital guide function is integrated in the elevator space, information collection, illness state analysis, department allocation, detection reservation, path guide and other links are organically integrated, an integrated and automatic intelligent hospital guide closed loop is formed, resource loads of all departments are effectively balanced, idle resources are reduced, and the work efficiency is improved. The utilization rate of equipment and the working efficiency of medical staff are improved, and comprehensive improvement of medical service quality and management level is promoted.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent guidance, in particular, to an intelligent guidance method based on a guidance elevator, a guidance elevator and a system. BACKGROUND

[0002] In the current medical service scenario, how to quickly and accurately complete triage and obtain effective treatment after the patient enters the hospital is a key problem to improve the efficiency and quality of medical services. The traditional guidance method mainly relies on manual inquiry and simple guidance signs, which has many drawbacks. On the one hand, manual triage is low in efficiency and is easily affected by subjective factors, making it difficult to quickly and accurately determine the patient's condition and assign to the appropriate department, which can lead to long patient waiting time and misallocation of medical resources. On the other hand, during the guidance process, patients need to go through symptom consultation, physical examination, registration and queuing in different areas, which not only consumes a lot of time and energy, but also may lead to information omission or errors due to the cumbersome process, affecting subsequent diagnosis and treatment.

[0003] With the development of medical informatization and intelligentization, some hospitals have introduced intelligent guidance systems, but existing systems are mostly independent devices or software platforms, and are not closely integrated with the actual medical process of the hospital. For example, some intelligent guidance terminals can only provide simple symptom inquiry and department recommendation, and cannot complete preliminary examination during patient waiting or movement; although some systems have certain data analysis capabilities, they lack comprehensive consideration of multiple factors such as patient condition severity, department real-time patient number, etc., making it difficult to achieve optimal department allocation and medical path planning. In addition, the existing guidance system also cannot effectively combine the patient waiting time in each department to make scientific arrangements in the detection item reservation link, resulting in repeated travel of patients in the hospital and poor medical experience.

[0004] Therefore, there is an urgent need for a guidance method that can integrate multiple resources and achieve full-process intelligentization to optimize patient medical experience and improve medical service efficiency. SUMMARY

[0005] Based on the problems existing in the prior art, the present application provides an intelligent guidance method based on a guidance elevator, a guidance elevator and a system. The specific scheme is as follows: First, the present application provides an intelligent guidance method based on a guidance elevator, which is suitable for a guidance elevator, and the intelligent guidance method comprises: guiding the patient to enter a preset guidance elevator, obtaining the patient's subjective symptom description and the patient's external symptom performance, and making the patient in the guidance elevator perform preliminary detection through a preset preliminary examination medical device to obtain preliminary examination data; analyzing the subjective symptom description, the external symptom manifestation and the preliminary examination data, finding abnormal signs of the patient, delimiting the criticality level of the patient, determining candidate diseases possibly involved by the patient, and predicting the association probability between the candidate diseases and general departments and special departments respectively; selecting an optimal treatment department for the patient according to the criticality level, the association probability corresponding to each department and the real-time number of patients, completing the registration in the optimal treatment department and constructing a treatment path for the patient to reach the optimal treatment department by the guidance elevator; screening necessary detection items of the patient based on the abnormal signs and the detection items associated with the candidate diseases, and making an advance reservation for the necessary detection items according to the treatment waiting time of the patient in the optimal treatment department and the detection waiting time in the department corresponding to the necessary detection items; guiding the patient to reach the optimal treatment department according to the treatment path after the guidance elevator reaches the floor where the optimal treatment department is located.

[0006] In some embodiments, the method further comprises: analyzing the subjective symptom description by a natural language processing algorithm, extracting structured information including symptom keywords, duration and severity, and then generating a symptom vector; analyzing the external symptom manifestation by a visual processing algorithm, extracting quantified features including facial features, body posture features and abnormal area features, and then forming an external feature matrix; retrieving historical treatment data of the patient to construct a time-sequenced medical history atlas including at least past diseases, past detection items and allergy history; determining abnormal indicators by feature analysis on the preliminary examination data, establishing causal association, time sequence dependence between the abnormal indicators and mapping relationship with past diseases, and generating a pathological state vector; generating abnormal signs of the patient based on the symptom vector, the external feature matrix, the time-sequenced medical history atlas and the pathological state vector.

[0007] In some embodiments, the process of obtaining the necessary detection items specifically comprises: finding all diseases associated with the abnormal signs based on a preset association rule base, and calculating the posterior probability of the occurrence of each disease by a Bayesian network, taking diseases with posterior probability exceeding a preset probability threshold as candidate diseases; finding candidate detection items associated with the candidate diseases based on a preset association rule base, and performing contraindication filtering on the candidate detection items to obtain feasible detection items; constructing a multi-objective optimization function for each feasible detection item and arranging to generate a priority list, adjusting the priority list in real time according to the criticality level of the patient, and selecting feasible detection items as the final necessary detection items based on the arrangement order of the priority list.

[0008] In some embodiments, the method further comprises: acquiring patient motion data from entering the guidance elevator from outside to inside of the guidance elevator by the image module, determining gait features of the patient by analyzing the motion data, and predicting the walking speed of the patient in combination with the physiological parameters of the patient; predicting the first walking time required for the patient to reach the optimal treatment department from the guidance elevator according to the walking speed and the treatment path, and predicting the waiting time of the patient in the optimal treatment department according to the real-time number of patients in the optimal treatment department.

[0009] In some embodiments, the method further comprises: constructing a detection path of the patient from the optimal treatment department to the department corresponding to the necessary detection item, and predicting the second walking time required for the patient to reach the department corresponding to the necessary detection item from the optimal treatment department according to the walking speed and the detection path; comprehensively considering the walking time and the waiting time to make a preliminary appointment for the patient in the department corresponding to the necessary detection item, so that the waiting time of the preliminary appointment is not less than the sum of the first walking time, the second walking time and the waiting time; after the diagnosis of the necessary detection item by the optimal treatment department is confirmed, converting the preliminary appointment into an actual appointment, and making the waiting time of the actual appointment not less than the second walking time.

[0010] In some embodiments, the selection of the optimal treatment department specifically comprises: predicting the waiting time of the patient in each special department according to the real-time number of patients in each special department; if the criticality level shows that the patient is currently in a critical state, selecting the general department or the special department with the least waiting time as the optimal treatment department; if the criticality level shows that the patient is currently in a non-critical state, determining whether there is a special department with an association probability exceeding a preset first threshold: if not, selecting the general department as the optimal treatment department; if so, selecting the department with the least waiting time as the optimal treatment department from the special departments with the association probability exceeding the preset first threshold.

[0011] In the second part, the application provides a guidance elevator for implementing the intelligent guidance method of any one of the first part; the guidance elevator comprises: an elevator cabin; a preliminary examination medical device located in the elevator cabin, used for preliminary examination of the patient; a human-computer interaction module located in the elevator cabin, used for symptom input and detection process guidance; an image module, located inside and outside the elevator cabin, configured to acquire images of the patient outside the elevator cabin and inside the elevator cabin; a guiding module, configured to guide the patient; a data processing module, communicatively connected with the elevator cabin, the preliminary examination medical device, the human-computer interaction module, the image module and the guiding module respectively.

[0012] In some specific embodiments, the preliminary examination medical device comprises an integrated vital sign detector, a dry-type microfluidic biochemical analysis device, a portable channel electrocardiograph, a pulse wave detector, an intelligent stethoscope and a skin image acquisition instrument.

[0013] In some specific embodiments, the application further comprises: a moving module, configured to move the patient under the guidance of the guiding module.

[0014] In the third part, the application provides an intelligent guiding system based on a guiding elevator, comprising: a preliminary examination unit, configured to guide the patient to enter a preset guiding elevator, acquire a subjective symptom description of the patient and an external symptom performance of the patient, and make the patient in the guiding elevator perform preliminary examination through a preset preliminary examination medical device to obtain preliminary examination data; an analysis unit, configured to analyze the subjective symptom description, the external symptom performance and the preliminary examination data, find abnormal signs of the patient, delimit a critical level of the patient and determine candidate diseases possibly involved by the patient, and predict an association probability between the candidate diseases and a comprehensive department and a special department respectively; a clinic selection unit, configured to select an optimal treatment department of the patient according to the critical level and an association probability corresponding to each department, a real-time number of patients, complete registration of the optimal treatment department and construct a treatment path of the patient from the guiding elevator to the optimal treatment department; a reservation unit, configured to filter necessary detection items of the patient by comprehensively considering the abnormal signs and detection items associated with the candidate diseases, and make an advance reservation of the necessary detection items according to a treatment waiting time of the patient in the optimal treatment department and a detection waiting time of the patient in a department corresponding to the necessary detection items; a guiding unit, configured to guide the patient to reach the optimal treatment department according to the treatment path after the guiding elevator reaches a floor where the optimal treatment department is located.

[0015] Beneficial effects: The application provides an intelligent guiding method, a guiding elevator and a system based on the guiding elevator, integrates intelligent guiding functions in elevator space taking the guiding elevator as a carrier, realizes organic integration of information collection, illness analysis, department allocation, detection reservation and path guidance, forms an integrated and automated intelligent guiding closed loop, effectively balances resource load of each department, reduces resource idling, improves equipment utilization and work efficiency of medical staff, and promotes comprehensive improvement of medical service quality and management level.

[0016] In order to make the above objectives, characteristics and advantages of the present application more apparent, clear and easy to understand, the following preferred embodiments are specifically described below, and the accompanying drawings are referred to for detailed description. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0018] Figure 1 is a flow chart of the intelligent guiding method of the present application; Figure 2 is a schematic diagram of the principle of the intelligent guiding method of the present application; Figure 3 is a module example diagram of the guiding elevator of the present application; Figure 4 is a structure example diagram of the guiding elevator in the present application; Figure 5 is a module example diagram of the intelligent guiding system in the present application.

[0019] Reference signs: 1-elevator cabin; 2-initial examination medical equipment; 3-human-computer interaction module; 4-image module; 5-guiding module; 6-data processing module; A1-initial examination unit; A2-analysis unit; A3-diagnosis room selection unit; A4-reservation unit; A5-guiding unit. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0021] Embodiment 1 The embodiment proposes an intelligent guiding method based on a guiding elevator, aiming at the problems of low efficiency, poor accuracy and fragmented process of traditional guiding, integrates the intelligent guiding function in the elevator space, realizes multi-dimensional breakthrough through technical innovation, and converts the originally idle elevator time into the pre-link of the diagnosis and treatment process. The process schematic diagram of the intelligent guiding method is shown in the accompanying drawings of the specification Figure 1 , the principle is shown in the accompanying drawings Figure 2 , and the scheme is as follows: An intelligent guiding method based on a guiding elevator is suitable for a guiding elevator, and the method comprises the following steps: 101. guiding a patient to enter a preset guiding elevator, acquiring a subjective symptom description of the patient and an external symptom performance of the patient, and making the patient in the guiding elevator perform preliminary detection through a preset preliminary examination medical device to obtain preliminary examination data; 102. analyzing the subjective symptom description, the external symptom performance and the preliminary examination data, finding abnormal signs of the patient, delimiting a critical level of the patient and determining candidate diseases possibly involved by the patient, and predicting the association probability between the candidate diseases and the comprehensive departments and the special departments respectively; 103. selecting an optimal treatment department of the patient at present according to the critical level and the association probability corresponding to each department, the real-time number of patients, completing the registration of the optimal treatment department and constructing a treatment path of the patient from the guiding elevator to the optimal treatment department; 104. screening necessary detection items of the patient according to the abnormal signs and the detection items associated with the candidate diseases, and making advance reservation of the necessary detection items according to the treatment waiting time of the patient in the optimal treatment department and the detection waiting time in the department corresponding to the necessary detection items; 105. guiding the patient to reach the optimal treatment department according to the treatment path after the guiding elevator reaches the floor where the optimal treatment department is located.

[0022] In the traditional hospital, the elevator only bears the vertical transportation function, and the intelligent guiding function is integrated in the elevator space in the scheme, relying on the moving characteristics of the elevator, seamlessly connecting the whole guiding process of data collection, intelligent analysis, department allocation, registration reservation, path guidance and the like, and converting the originally idle elevator time into the pre-link of the diagnosis and treatment process. Through the composite use of space function, the invalid movement of the patient in the hospital is reduced, the length of the medical path is shortened by an average of 30%, the guiding process efficiency is improved by about 40%, and the efficient use of hospital space resources and the optimization of service process are realized.

[0023] Step 101 utilizes the special carrier of the guidance elevator to efficiently complete the collection of multi-dimensional health information during the patient's journey from the hospital entrance to the treatment department. The intelligent guidance solution of the present application relies on the guidance elevator that has been specially modified and configured, which integrates various hardware facilities such as human-computer interaction modules, image modules, medical devices, etc., and is the core carrier to realize the intelligent guidance function. These elevators are usually deployed near the hospital entrance or in areas with high patient flow, making it convenient for patients to use them at the first time.

[0024] By human-computer interaction, the patient's subjective symptom description is obtained, the external sign performance is captured by the image module, and the preliminary physical detection is performed by the medical device, so as to quickly and comprehensively collect the patient's health data, laying a foundation for subsequent intelligent analysis and accurate guidance. Subjective symptom description is the expression of the patient's own discomfort, covering symptom types (such as pain, fever, cough), duration, severity, and inducing factors. Although these information is subjective, it is crucial for preliminary diagnosis of diseases.

[0025] Relying on the image modules installed in the elevator, such as high-definition cameras, infrared thermal imagers, and depth cameras, the external sign images and video data of the patient are collected from multiple angles. External symptom performance refers to the patient's physical characteristics and state that can be observed visually, including complexion (such as pale, flushed), posture (such as limping, rickets), skin condition (such as rash, jaundice), and limb movement (such as convulsions, tremors). These objective signs can provide intuitive clues for diagnosis and assist in determining the condition.

[0026] In addition, the guidance elevator is equipped with various portable and rapid detection medical devices, such as integrated vital sign detectors (which can simultaneously measure heart rate, blood pressure, blood oxygen saturation, and body temperature), dry-type microfluidic biochemical analysis devices (which can detect blood glucose, blood routine, etc. in a short time), and portable electrocardiogram machines. These devices are carefully selected, easy to operate, and fast to detect, suitable for use in the limited space and time of the elevator. During the operation of the elevator, the patient completes the detection of basic physiological indicators and part of the biochemical indicators under the guidance or simple operation of the device. For example, by non-invasive method to collect fingertip blood for blood glucose detection, or to wear electrode sheet to complete electrocardiogram recording.

[0027] The application integrates the information collection links originally dispersed in different areas (such as the reception desk and examination room) into the elevator process, greatly saving the patient's time. At the same time, subjective symptoms, external manifestations and preliminary examination data are obtained, avoiding the limitations of single data sources, making the collected patient health information more comprehensive and three-dimensional. The fusion of multi-source data helps to more accurately judge the disease during subsequent analysis, reducing the possibility of misdiagnosis and missed diagnosis. Patients do not need to repeatedly run between different places in the hospital for information registration and examination, and can complete the operation in the relatively closed and private elevator space, reducing the anxiety caused by complicated processes and improving the comfort and convenience of medical treatment.

[0028] Step 102 converts the original collected information into diagnostic basis with clinical guidance significance through multi-dimensional data fusion and intelligent algorithm processing. Through multi-modal data cross-validation, the error of single data source is reduced. Through quantitative scoring, high-risk patients (such as myocardial infarction and stroke) are quickly identified, and time is gained for emergency procedures. Based on the probability model, the optimal department is recommended to solve the pain of patients not knowing what department to go to.

[0029] For example, the patient enters the reception elevator, describes sudden chest pain and difficulty breathing for 1 hour, and the external manifestations are pale face and holding chest. The preliminary examination data shows that the heart rate is 110 beats per minute, the blood pressure is 85 / 50 mmHg, and the blood oxygen saturation is 89%. NLP extracts chest pain and difficulty breathing for 1 hour, CV identifies pale face and holding chest action, and preliminary examination data marks low blood pressure and low oxygen. Abnormal signs include chest pain, difficulty breathing, low blood pressure, and low oxygen; MEWS score: heart rate 2 points + systolic blood pressure 3 points + respiratory rate (assuming abnormal) 2 points + blood oxygen 2 points → total score 9 points → critical (need immediate emergency). Regarding the association of candidate diseases and departments: Candidate diseases: acute myocardial infarction (probability 75%), pulmonary embolism (20%), pneumothorax (5%); Department association: acute myocardial infarction → cardiology department (95%) + emergency department (100% due to critical level one), pulmonary embolism → respiratory department (70%) + emergency department (100%).

[0030] Output results: abnormal signs (4), critical level (one), candidate diseases (top three), department association probability (such as emergency department 100%, cardiology department 95%).

[0031] In some embodiments, the acquisition of abnormal signs includes: parsing the subjective symptom description through a natural language processing algorithm, extracting structured information including symptom keywords, duration, and severity, and then generating a symptom vector; analyzing the external symptom manifestations through a visual processing algorithm, extracting quantified features including facial features, body posture features, and abnormal area features, and then forming an external feature matrix; retrieving the patient's historical medical data to construct a time-sequenced medical history atlas containing at least past medical conditions, past detection items, and allergy history; determining abnormal indicators by analyzing the initial detection data, establishing causal relationships, time sequence dependencies, and mapping relationships with past medical conditions between abnormal indicators, and generating a pathological state vector. Based on the symptom vector, the external feature matrix, the time-sequenced medical history atlas, and the pathological state vector, the patient's abnormal signs are generated.

[0032] The naming entity recognition technology in natural language processing is used to identify symptom keywords from the patient's input subjective symptom description text. For example, a dictionary and rule-based method is used, combined with a professional dictionary in the medical field, to match symptom words in the text, such as headache, cough, and abdominal pain; a deep learning model can also be used, such as by training a large amount of labeled medical text data to automatically learn the features of symptom keywords, improving the accuracy and generalization ability of extraction. Regular expressions, semantic role labeling, and other methods are used to extract the duration and severity information of the symptoms. The extracted symptom keywords, duration, and severity information are organized into a vector form in a certain order and format. For example, [headache, 5 days, 3 minutes, cough, 2 days, 2 minutes], so as to facilitate subsequent data analysis and processing.

[0033] The patient's external image is collected by image modules in the guidance elevator, such as high-definition cameras, infrared thermal imagers, and other devices. The collected images are preprocessed, including image enhancement, noise reduction, normalization, and other operations to improve the quality of the images and the recognizability of the features. Computer vision algorithms, such as convolutional neural networks (CNN), are used to extract features from the preprocessed images. For facial features, the color distribution, texture information, and other information of the image are analyzed to identify abnormalities such as pale, redness, and jaundice; for body posture features, human pose estimation algorithms are used to detect the positions and poses of human joints to determine whether there are abnormalities such as limping and hunched back; for abnormal area features, target detection algorithms are used to identify abnormal areas such as rashes, wounds, and lumps on the skin, and extract their shape, size, color, and other features. The extracted facial features, body posture features, and abnormal area features are integrated to form a two-dimensional matrix. For example, each row of the matrix represents a feature type, and each column represents a specific feature value, such as [[pale degree, 0.8], [limp degree, 0.6], [rash area, 10 cm²]].

[0034] The patient's historical medical data, including previous diagnosis records, test results, and allergy history, are retrieved from the hospital's electronic medical record system and database. The retrieved data is cleaned to remove duplicate, incorrect, or incomplete records. The cleaned data is arranged in chronological order based on the time of the visit to construct a time-sequenced medical history graph. In the graph, different conditions, test results, and allergy history are represented as nodes and edges, with nodes representing specific events or information and edges representing their relationships, such as cause-and-effect relationships, temporal relationships, and so on. For example, a cold that led to pneumonia can be represented as a cause-and-effect relationship between the cold node and the pneumonia node.

[0035] Each physiological indicator and test result in the initial examination data is compared with the normal reference range to determine if there is an abnormality. For example, the normal blood pressure range is 90-140 mmHg for systolic pressure and 60-90 mmHg for diastolic pressure. If the test result exceeds this range, it is marked as an abnormal blood pressure indicator. Statistical methods and machine learning algorithms are used to analyze the cause-and-effect relationships and temporal dependencies between abnormal indicators. The identified abnormal indicators, analyzed cause-and-effect relationships and temporal dependencies, and mapping information with previous conditions are organized into a pathological state vector. Each element in the vector represents a specific pathological feature or relationship, such as [abnormal blood sugar, correlation with insulin secretion 0.7, related to previous diabetes history].

[0036] The generated symptom vector, external feature matrix, time-sequenced medical history graph, and pathological state vector are fused. Weighted summation, feature concatenation, and other methods can be used to integrate these multi-dimensional data into a comprehensive data set. Then, based on medical knowledge and pre-set rules, the fused data is analyzed and judged to identify any abnormalities, and the patient's abnormal signs are finally generated. For example, if there is chest pain in the symptom vector and the duration is long, the ECG is abnormal in the pathological state vector, and the patient's face is pale in the external feature matrix, it is concluded that there may be an abnormal sign related to the heart. The patient's health information is converted into structured and quantifiable data, providing an objective basis for subsequent medical decision-making.

[0037] In some embodiments, based on the pre-set association rule base, all conditions associated with the abnormal signs are found, and the posterior probability of each condition is calculated using Bayesian networks. Conditions with posterior probabilities exceeding a pre-set probability threshold are considered candidate conditions. Through the double-layer logic of the pre-set association rule base and Bayesian network probability calculation, high-probability candidate conditions are screened from the patient's abnormal signs. The data sources of the association rule base include medical guidelines, authoritative medical literature, and hospital historical case data.

[0038] The tree index + hash table storage is adopted, and the system (such as cardiovascular system, digestive system) and the sign type (subjective / objective) are classified and classified to improve the retrieval efficiency. The abnormal signs are converted into structured parameters (such as face color pale mapping face feature = pale, blood pressure 160 / 100 mmHg mapping initial detection index = blood pressure, value = 160 / 100) that can be recognized by the rule base. Through vector similarity calculation (such as cosine similarity) and rule condition logic matching (AND / OR / NOT), all rules that meet the conditions are found. If the patient's abnormal signs are chest pain (for 3 days, severity level 4) + pale + blood pressure 160 / 100 mmHg, the rule base matches: coronary heart disease (support degree 35%, confidence 75%).

[0039] Regarding the network structure of the Bayesian network. The parent node is the abnormal sign (symptom vector, external feature, pathological index, etc.). The child node is the candidate disease (such as coronary heart disease, pneumonia, etc.). The weight of the edge is the conditional probability P (sign | disease) between the sign and the disease. Using historical case data, the conditional probability table (CPT) between nodes is calculated by maximum likelihood estimation method.

[0040] For each matched disease D, the posterior probability P (D | S) is calculated, which is equal to the product of two values divided by the third value. The first value is the joint probability P (S | D) of the disease D leading to the occurrence of the abnormal sign set S (calculated by the conditional probability table of the Bayesian network); the second value is the prior probability P (D) of the disease D itself (i.e. the probability of the disease D occurring without considering any signs); and the third value as the divisor is the marginal probability P (S) of the occurrence of the abnormal sign set S, which serves as a normalization constant here (to ensure that the sum of the posterior probabilities of all possible diseases is 1).

[0041] Set the posterior probability threshold P ≥ 0.3, and include the diseases exceeding the threshold in the candidate set, sort the candidate diseases from high to low according to the posterior probability, and generate the candidate disease list. The rule base is used to quickly filter low correlation diseases and narrow down the Bayesian network calculation range. The Bayesian network is used to quantify the probability of the disease and solve the limitations of the rule base being black or white. For incomplete signs (such as patients not providing part of the symptoms), the missing value inference ability of the Bayesian network (such as Markov chain Monte Carlo method) is used; for contradictory signs (such as part of the indicators supporting disease A and part supporting disease B), the posterior probability comparison is used to automatically arbitrate.

[0042] In some embodiments, regarding the correlation probability between the predicted candidate diseases and the comprehensive department and the special department respectively, a disease-department mapping relationship model is constructed to provide a basis for patient triage or department resource scheduling.

[0043] The data sources of the prediction model include historical visit data, medical knowledge graph, and expert annotated data. Historical visit data: the correspondence between symptoms and departments recorded in the hospital information system (HIS), such as pneumonia → respiratory department, diabetes → endocrinology department, abdominal pain → digestive department / surgery department, etc.; medical knowledge graph: such as ICD coding, department responsible for symptom range in clinical guidelines (such as the diagnosis and treatment scope of each department specified in Internal Medicine); expert annotated data: department assignment cases for a small number of complex symptoms (such as cross-department symptoms) annotated by medical experts.

[0044] Map the candidate symptoms to standard terminology uniformly to avoid synonym ambiguity; clarify the scope of comprehensive departments and specialized departments. Organize historical data into symptom-department pairs, and annotate the main and secondary associated departments for each symptom. Extract explicit rules from medical guidelines and department diagnosis and treatment specifications to build a medical rule base. For each candidate symptom, retrieve its associated system or disease type through the rule base to obtain the initial association probability. Take the symptom characteristics as input and output the association probability (0-1) between the symptom and each department. Train the model parameters based on historical data. For cross-department symptoms (such as abdominal pain, which may involve digestive internal medicine, surgery, and gynecology), use a multi-label classification model to allow a symptom to be associated with multiple departments and output the probability of each department. In addition, a department knowledge graph can be constructed to find the shortest path from the symptom to the department, such as pneumonia → respiratory disease → respiratory department. The shorter the path, the higher the association probability. Calculate the semantic similarity between the symptom node and the department node (e.g., based on Word2Vec embedding vector space, calculate the cosine similarity), the higher the similarity, the higher the probability. For each candidate symptom, combine the rule prior probability, statistical model probability, and graph reasoning probability to obtain the final association probability through weighted fusion. The weights can be determined by training historical triage data.

[0045] Step 103 selects the optimal department for the patient based on the criticality level and the associated probability of each department, as well as the real-time number of patients. Registering the patient in the optimal department and constructing the patient's visit path from the guidance elevator to the optimal department is the core decision-making link of the intelligent guidance system, and its core goal is to provide the patient with the optimal department selection and seamless connection of the visit process under multiple constraint conditions.

[0046] Among them, the criticality level is based on the scoring system of step 102, such as MEWS score or ESI classification, which divides patients into red (critical), yellow (moderate), and green (non-urgent) levels as the primary constraint condition for department assignment. The association probability is calculated by Bayesian network to determine the association strength between candidate symptoms and departments (e.g., pneumonia → respiratory department, probability 0.85).

[0047] Step 104 involves necessary detection item screening and advance booking. Based on the preset association rule base, candidate detection items related to abnormal signs and candidate diseases are found, combined with patient individual contraindications (such as allergy history, physical condition) for filtering to obtain feasible detection items. A multi-objective optimization function is constructed, considering factors such as the diagnostic value, timeliness, safety, and cost of the detection items, to calculate the priority score for each feasible detection item and generate a priority list. For example, emergency patients prefer to choose detection items with fast results. According to the predicted waiting time of the patient in the optimal treatment department and the estimated walking time from the treatment department to the detection department (combined with the estimated walking speed and path distance in step 101), the advance booking time of the necessary detection items is determined to ensure seamless integration of the detection time and the treatment process.

[0048] In some embodiments, the process of obtaining necessary detection items specifically includes: based on the preset association rule base, finding candidate detection items associated with candidate diseases, and performing contraindication filtering on the candidate detection items to obtain feasible detection items; constructing a multi-objective optimization function for each feasible detection item and arranging it to generate a priority list, and adjusting the priority list in real time according to the patient's criticality level, and selecting feasible detection items as the final necessary detection items based on the arrangement order of the priority list. Through the logic of rule matching-contraindication filtering-multi-objective optimization-dynamic adjustment, intelligent recommendation of detection items is achieved, which not only avoids misdiagnosis caused by missing key detection, but also reduces the waste of medical resources caused by unnecessary detection, while considering patient safety and diagnostic efficiency. The essence is to convert the clinical decision-making logic into a calculable algorithm process.

[0049] In this application, the mapping relationship between disease and detection item is stored in the preset association rule base, such as pneumonia→chest CT, diabetes→blood glucose detection, etc. These rules are usually based on medical guidelines, clinical pathways or historical case data, and the essence is the knowledge accumulation of what kind of disease needs what kind of detection to confirm the diagnosis. The input is the candidate diseases generated in step 102 (such as pneumonia, diabetes, etc.), and the output is the set of candidate detection items associated with each disease. By searching the association rule base, the corresponding detection items for each candidate disease are extracted. Different diseases may be associated with the same detection item (such as fever, which may correspond to blood routine, and infection, which may also correspond to blood routine), and the detection items of multiple candidate diseases need to be combined to remove duplicates to form a preliminary list of candidate detection items. Candidate detection items may not be implemented due to individual patient conditions (such as allergies, physical contraindications, critical illness, etc.), and need to be excluded by contraindication filtering to ensure detection safety. If the detection item involves an allergen (such as contrast agent, certain detection reagent), and the patient has a corresponding history of allergy, the detection item is excluded. If the detection item may exacerbate the patient's current condition (such as severe heart failure patients cannot tolerate the load of enhanced CT), it is excluded. After contraindication filtering, a list of feasible detection items is obtained.

[0050] wherein the feasible detection items need to be prioritized to balance the multi-objective requirements of diagnostic efficiency, accuracy, patient safety, and medical cost. The multi-objective optimization function of the feasible detection items is used to comprehensively evaluate the pros and cons of the detection items, which is specifically the product of the ratio of the product of the clinical necessity index and the patient-specific index to the resource adaptability penalty factor, and the product between the time sequence dependent product term for processing the detection order.

[0051] Clinical necessity index: measures the necessity of the detection item in clinical decision-making. It is obtained by multiplying the posterior probability of all candidate diseases (combining the probability of the current signs and the existing evidence) by the information entropy reduction of the detection item for the disease (i.e., the uncertainty that the detection can reduce for diagnosis), and then summing up. The larger, the higher the necessity of the detection for distinguishing the candidate diseases.

[0052] Patient-specific index: generated by a logistic regression model, used to evaluate whether the detection item is suitable for a specific patient, with a value of 0-1. It combines two aspects: one is the patient's historical compliance (such as a score reduction for refusing a gastroscopy, e.g., 0.3); the other is the patient's physiological limitations (such as a score reduction for wheelchair users who cannot perform standing light detection, e.g., an impact coefficient of 0.5). The higher the value, the more suitable the detection for the patient.

[0053] Resource adaptability penalty factor: converts the resource adaptability index into a penalty term, the worse the index value, the stronger the penalty. The resource adaptability index is composed of three weighted factors: the first weight x device load (such as ECG machine occupancy rate 0.6), the second weight x estimated time consumption (such as scan time consumption normalized 0.2), and the third weight x (1 - consumable inventory rate) (such as the item increases when the blood oxygen probe is in short supply); the first weight, the second weight, and the third weight can be adjusted according to the priority of the hospital process (such as increasing the second weight to pay more attention to time consumption). The larger the resource adaptability index, the worse the detection item's consumption or adaptability to resources, and its penalty factor will decrease.

[0054] Time sequence dependent product term: used to handle detection sequence constraints. When a detection item is included in the detection sequence and placed in the i-th position, the product term calculates all detection items placed before the i-th position in the sequence, and ensures that detection items with dependent relationships meet the sequence requirements through a logical factor (such as must do before do, if the sequence is reversed, the logical factor will reduce the product term value, or even make the overall result invalid), avoiding logical conflicts.

[0055] The subjective symptom text is parsed by a medical natural language processing model (such as clinical BERT) to extract structured semantic vectors containing symptom types (such as chest pain and fever), temporal attributes (such as lasting for 48 hours), and severity (such as exacerbation at night). The spatiotemporal characteristics of the external manifestation data are modeled, and the infrared thermal imaging temperature sequence, camera-captured body action video frames, and pulse oximetry waveform are converted into multi-dimensional time series tensors, and key features (such as abnormal respiratory rate and skin rash temperature gradient) are extracted through convolutional neural networks. Historical medical data is retrieved from the hospital data platform to construct a patient health knowledge graph containing disease diagnosis (ICD-11 code), medication record (ATC classification), and detection report (DICOM metadata), with node association relationships including symptoms-disease, disease-detection, and detection-contraindication. A graph attention network (GAT) is used to encode the health knowledge graph to generate context embedding representations of symptom entities and detection item entities. A multi-modal association matrix is constructed, and matrix decomposition algorithms are used to extract symptom-detection association features to generate an initial candidate detection item set. Based on the history of medical data, the detection items that are contraindicated are excluded to form a feasible detection item set.

[0056] A ternary evaluation system for detection items is defined. The clinical necessity is calculated by a Bayesian network to determine the diagnostic gain for the current symptom combination, and the resource adaptability is constructed by a constraint function that integrates the equipment load in the elevator, the detection time, and the inventory of consumables. The patient-specific score is generated by combining the detection compliance and physiological limitations in the history of medical data. A Pareto optimal optimization function is constructed, and a non-dominated sorting genetic algorithm (NSGA-II) is used to solve the multi-objective optimization problem to generate a priority sequence containing mandatory and optional detection items. The equipment status in the elevator and the hospital department resources are monitored in real time, and the priority sequence is corrected by a context-aware engine. For detection items with time sensitivity, a time window constraint is introduced to adjust the sequence. The final detection plan is generated, including the detection item list, the execution order, and the equipment matching relationship, and is synchronized to the elevator control system in real time through the WebSocket protocol. The detection plan is visualized and displayed to the patient through the elevator touch screen, supporting natural language-based plan consultation and limited adjustment. After the patient confirms, the system generates a detection order through digital signature technology, triggers the automatic calibration of the modular medical equipment in the elevator, and guides the detection cooperation through voice interaction.

[0057] In some embodiments, further comprising: acquiring motion data of the patient entering the interior of the guidance elevator from the exterior of the guidance elevator, determining gait features of the patient by analyzing the motion data, and estimating a walking speed of the patient in combination with the physiological parameters of the patient; predicting a first walking time required for the patient to reach the optimal treatment department from the guidance elevator according to the walking speed and the treatment path; predicting a waiting time of the patient in the optimal treatment department according to the real-time number of patients in the optimal treatment department; constructing a detection path of the patient from the optimal treatment department to the department corresponding to the necessary detection item, and predicting a second walking time required for the patient to reach the department corresponding to the necessary detection item from the optimal treatment department according to the walking speed and the detection path; making a preliminary appointment for the patient in the department corresponding to the necessary detection item in combination with the walking time and the waiting time, so that the waiting time of the preliminary appointment is not less than the sum of the first walking time, the second walking time and the waiting time; and converting the preliminary appointment to an actual appointment after the diagnosis of the necessary detection item by the optimal treatment department, and making the waiting time of the actual appointment not less than the second walking time. Through the multi-dimensional integration of gait analysis, path planning and time prediction, the intelligentization and dynamicization of detection appointment are realized, and the essence is to combine the patient's movement ability, spatial distance and medical resource scheduling. Through the hospital clock synchronization system and real-time tracking of the patient's moving trajectory, the deviation between the predicted time and the actual time is corrected, and the invalid waiting time of the patient in the detection department is reduced; if the patient's walking speed suddenly decreases (such as taking a break), the system automatically extends the appointment time after detecting by the camera, avoiding the conflict between detection and treatment process.

[0058] Multi-view cameras (such as RGB cameras + depth cameras) are deployed at the entrances and exits of the guidance elevator and the corridors, and devices with a frame rate of ≥30 fps are used to capture the motion trajectory of the patient entering the interior from the exterior of the elevator. The patient target and the environment are separated by background subtraction algorithm (such as Gaussian mixture model); a lightweight neural network is used to process video stream in real time on the edge device (elevator controller), reducing the delay; human pose estimation algorithm (such as OpenPose) is used to extract joint coordinates (ankle, knee, hip joint, etc.), and generate motion trajectory sequence. Then gait feature analysis is performed to extract key features, including: step length: the spatial distance between the positions of the two feet; step frequency: the number of steps per unit time (such as the number of steps per 10 seconds); gait symmetry: the difference between the motion trajectories of the left and right legs (used to determine whether there is a limp); joint activity angle: the flexion and extension amplitude of the knee and hip joint (reflecting the movement ability). Based on the linear relationship between gait features and speed, the speed is adjusted in combination with the pathological state vector (such as heart rate, blood oxygen saturation) of step 102.

[0059] Based on the three-dimensional map of the hospital, the shortest path of the guidance elevator to the optimal treatment department is generated (considering the elevator floor, corridor connectivity); the path is represented as a node sequence (such as elevator 1→corridor A→stair B→clinic C), and each node is labeled with physical distance. The current number of patients waiting in the optimal treatment department and the average treatment time per patient are obtained from the hospital HIS system. Among them, the waiting time is the performance of the three of the fluctuation coefficient, the current number of patients waiting and the average treatment time per patient, and the fluctuation coefficient can be dynamically adjusted according to historical data.

[0060] The time constraint of the preliminary appointment ensures that the patient just reaches the detection after completing the following processes: walking from the elevator to the treatment department (first walking time); waiting in the treatment department and completing the treatment (waiting time); walking from the treatment department to the detection department (second walking time). The patient completes the diagnosis in the treatment department, and the doctor confirms that the necessary detection items need to be performed, triggering the conversion from the preliminary appointment to the actual appointment. If the treatment process ends early (such as the actual waiting time is 25 minutes), the system recalculates the remaining time: remaining available time=preliminary appointment time-used time (1+25=26 minutes); if the remaining time≥second walking time (2 minutes), execute the original appointment; if the remaining time<2 minutes, automatically postpone the appointment time to meet the constraints.

[0061] The construction of the clinic path is the core step of the intelligent guide system to achieve precise navigation. It provides an efficient route from the guide elevator to the optimal clinic department for the patient through multi-source data fusion and intelligent algorithm planning. Through laser scanning or BIM (Building Information Modeling) technology, a three-dimensional spatial model of each floor of the hospital is constructed, and physical nodes such as department location, corridor, elevator, staircase, emergency department area, etc. are labeled to form a spatial database containing coordinates, dimensions, and traffic rules (such as one-way passage, access restrictions). Real-time synchronization of dynamic parameters such as hospital flow density (through camera video analysis or Wi-Fi probe statistics), elevator operating status (stop floor, load condition), and temporary control information of the area where the department is located (such as maintenance, epidemic isolation area). Integrate the patient's criticality level (such as critical state requiring priority use of emergency access), mobility (such as wheelchair patients avoiding stairs), and real-time location (guide elevator current floor) as constraint conditions for path planning. The real-time location of the guide elevator is determined by the floor sensor or GPS (indoor positioning system) as the starting point of the path. At the same time, the system reads the current floor's heat map of people flow to identify congested areas (such as registration area, elevator hall, etc.).

[0062] In some embodiments, the selection of the optimal clinic department specifically includes: predicting the waiting time of the patient in each special department according to the real-time number of patients in each special department; if the criticality level shows that the patient is currently in a critical state, selecting the special department or the comprehensive department with the least waiting time as the optimal clinic department; if the criticality level shows that the patient is currently in a non-critical state, determining whether there is a special department with an association probability exceeding a preset first threshold; if not, selecting the comprehensive department as the optimal clinic department; if there is, selecting the department with the least waiting time among the special departments with an association probability exceeding the preset first threshold as the optimal clinic department.

[0063] The selection of the optimal department is based on three elements: patient criticality level, department waiting time, and the association probability between the disease and the department. Through a hierarchical decision-making process, efficient triage is achieved, which can be divided into two processing logics: critical state and non-critical state. Through the hospital information system (HIS) or the queuing call system, the number of patients in each special department (such as cardiology department, orthopedics department, etc.) is collected in real time, and the waiting time of patients in each department is predicted based on the average consultation time. For example, if there are currently 5 people in a special department, and the average consultation time for each patient is 10 minutes, the predicted waiting time is 50 minutes. Based on patient vital signs (such as heart rate, blood pressure, blood oxygen, etc.), symptom severity (such as chest pain, heavy bleeding), and medical history data, the system or medical staff determines the patient's criticality level, which is usually divided into critical state (such as ICU level, rescue level) and non-critical state (such as general outpatient level).

[0064] When the criticality level of the patient is in a critical state (such as cardiac arrest, severe trauma), the emergency triage process is started. All general departments (such as emergency department) and special departments are traversed to obtain the real-time waiting time of each department. The department with the shortest waiting time is directly selected as the optimal department for treatment, without considering the association probability between the disease and the department, to maximize the reduction of treatment delay.

[0065] When the patient's criticality level is in a non-critical state (such as common cold, chronic disease follow-up), the regular triage logic is started. Step 1: Association probability screening of special departments. Based on the pre-set association rule base or the calculation results of the Bayesian network (such as the association probability between the candidate disease and the department in the previous step), special departments with an association probability exceeding a pre-set first threshold (such as 0.7) are screened out. For example, if the candidate disease is pneumonia, and the association probability of the respiratory department is higher than 0.7, the respiratory department will be included in the candidate special department set. Step 2: Determine whether there is an effective special department. If there is no special department with an association probability exceeding the threshold (such as a rare disease with no specific corresponding specialty), the general department (such as the general outpatient department) is directly selected as the optimal department for treatment to ensure that the patient receives basic diagnosis first. If there is an effective special department, proceed to the next step of screening. Step 3: Optimal selection based on waiting time. Among the special departments with association probability meeting the standard, compare the real-time waiting time of each department, and select the department with the shortest waiting time as the optimal department for treatment. For example, if the respiratory department (association probability 0.8) has a waiting time of 20 minutes and the infectious diseases department (association probability 0.75) has a waiting time of 30 minutes, the respiratory department is selected to ensure professionalism while reducing waiting time.

[0066] In practical applications, a queuing theory model (such as the M / M / c queue model) or a machine learning algorithm (such as LSTM or random forest) can be used to dynamically predict the waiting time of each department based on historical visit data, current queue length, and doctor examination efficiency. Based on a pre-set association rule base (such as the deterministic rule of pneumonia -> respiratory department) or Bayesian network (considering the probability association between symptoms and departments), the association probability of candidate symptoms and specialized departments is calculated. The pre-set first threshold is dynamically adjusted according to the saturation degree of hospital department resources. For example, when the respiratory department is currently full of patients, the association probability threshold is increased to guide patients to preferentially select a sub-optimal department with shorter waiting time, balancing resource allocation.

[0067] Step 105 involves guidance and path guidance. The system controls the guidance elevator to directly reach the floor of the optimal examination department, reducing intermediate stops. After the elevator arrives, the patient is provided with a detailed navigation path from the elevator exit to the target department through the display screen inside the elevator, voice prompts, or a mobile terminal APP, including route map, distance, estimated walking time, and other information. For patients with limited mobility, the intelligent wheelchair or robot inside the hospital can also be linked to provide escort services.

[0068] The present application also proposes a guidance elevator for implementing any of the above intelligent guidance methods; the module schematic diagram of the guidance elevator is shown in FIG. 2, and the structural schematic diagram is shown in FIG. 3, which comprises: Figure 3 Figure 4 The elevator cabin 1; the internal size is expanded by 20%-30% (such as 2.5m x 2.5m) compared with ordinary elevators, reserving medical equipment installation positions and patient activity space; adopting antibacterial stainless steel or PVC interior decoration, equipped with ultraviolet disinfection lamps to meet the infection control requirements of medical environment; additional emergency call buttons, handrails, and seats are suitable for patients with limited mobility.

[0069] The preliminary examination medical equipment 2 is located in the elevator cabin 1 and is used for preliminary examination of the patient; the elevator is converted from a transportation tool to a mobile medical unit to realize pre-positioning of detection.

[0070] The human-computer interaction module 3 is located in the elevator cabin 1 and is used for symptom input and detection process guidance; equipped with a 10-15 inch touch screen (supporting glove operation), integrating a capacitive button and a voice pickup; supporting multi-language switching function (Mandarin, English, dialect, etc.), adapting to different patient groups.

[0071] ​​Image module 4, located inside and outside the elevator cabin 1, is used to obtain images of patients outside and inside the elevator cabin 1. For example, 2 wide-angle cameras are installed on the top of the elevator cabin 1 to capture facial expressions, body movements and other external symptoms of patients; equipped with infrared fill light to support clear shooting in low light environment. A depth camera is installed outside the elevator door to detect the gait characteristics (step length, step frequency, symmetry) of patients entering the elevator; combined with computer vision algorithms, the patient's walking speed and whether they use auxiliary tools (wheelchair, crutch) are identified.

[0072] Guidance module 5 is used to guide patients; a 2D / 3D path map is displayed on the touch screen to mark turn prompts, distance remaining, estimated arrival time, and arrow animations are used to indicate the direction of travel; it can also broadcast voice prompts such as "turn left in 50 meters" and "arrive at the internal medicine diagnosis area". It can also display path dynamics synchronously through the patient's mobile phone APP, support real-time positioning and navigation, and continue to guide even if the patient leaves the guidance elevator (such as the last part of the journey from the elevator to the department).

[0073] Data processing module 6 is respectively connected to elevator cabin 1, preliminary examination medical equipment 2, human-computer interaction module 3, image module 4 and guidance module 5. The module is equipped with an edge computing server that supports real-time processing of multi-modal data and a local storage array, which interfaces with the hospital HIS, LIS, PACS system through the HL7 / FHIR protocol, synchronizes the number of patients in the department, the status of the detection equipment, etc. In addition, natural language processing (NLP), computer vision (CV), and machine learning (ML) models are deployed to achieve functions such as symptom analysis, disease prediction, and path planning. Based on the rule engine and Bayesian network, the optimal department for examination, the priority of detection items, and other decision-making suggestions are generated.

[0074] In some embodiments, the preliminary examination medical device 2 includes an integrated vital sign detector, a dry microfluidic biochemical analysis device, a portable channel electrocardiograph, a pulse wave detector, an intelligent stethoscope, and a skin image acquisition instrument. The integrated vital sign detector synchronously collects basic indicators such as heart rate, blood pressure, blood oxygen saturation, body temperature, and respiratory rate, and supports non-invasive fingertip detection or wrist-worn devices. The dry microfluidic biochemical analysis device rapidly detects indicators such as blood glucose, blood routine (white blood cells, red blood cells, platelets), and electrolytes (potassium, sodium, chlorine) through fingertip blood. The portable channel electrocardiograph synchronously collects 12 leads and supports real-time early warning of arrhythmias such as atrial fibrillation and myocardial infarction. It is equipped with an AI analysis module that automatically marks critical signs such as ST segment elevation and T wave abnormalities. The pulse wave detector analyzes radial artery pulse waves through a photoelectric sensor to evaluate indicators such as arterial stiffness and cardiac output. The intelligent stethoscope supports high-definition collection and spectral analysis of heart sounds and lung sounds through electronic amplification and noise reduction processing. It has an internal abnormal sound library to assist in primary diagnosis. The skin image acquisition instrument supports close-up photography of skin lesions such as rashes, ulcers, and pigmentation using a high-definition camera and LED ring light. It is equipped with an AI image recognition algorithm that automatically labels lesion area size and color characteristics.

[0075] In some embodiments, it further includes a moving module 7 for moving the patient under the guidance of the guiding module 5. Exemplarily, the moving module 7 adopts a wheeled chassis + motor drive, supports autonomous navigation (SLAM algorithm) and remote control, has a maximum load of 150 kg, a speed of 0.5-1.0 m / s, and is equipped with ultrasonic obstacle avoidance sensors.

[0076] The present application proposes an intelligent guiding system based on a guiding elevator, as shown in Figure 5 , which includes: A preliminary examination unit A1 for guiding the patient to enter a preset guiding elevator, obtaining a subjective symptom description of the patient and an external symptom performance of the patient, and making the patient perform preliminary detection in the guiding elevator through a preset preliminary examination medical device to obtain preliminary examination data; An analysis unit A2 for analyzing the subjective symptom description, the external symptom performance, and the preliminary examination data, finding abnormal signs of the patient, classifying a critical level of the patient, determining candidate diseases possibly involved by the patient, and predicting association probabilities between the candidate diseases and comprehensive departments and special departments, respectively; A consultation room selection unit A3 for selecting an optimal consultation department of the patient according to the critical level and the association probabilities corresponding to each department, the real-time number of patients, completing registration in the optimal consultation department, and constructing a consultation path of the patient from the guiding elevator to the optimal consultation department; The reservation unit A4 is configured to screen necessary detection items of the patient based on the detection items associated with the abnormal signs and the candidate diseases, and to make an advance reservation for the necessary detection items according to the waiting time of the patient in the optimal consultation department and the waiting time of the patient in the department corresponding to the necessary detection items; The guiding unit A5 is configured to control the guidance elevator to guide the patient to the optimal consultation department according to the consultation path after the guidance elevator arrives at the floor where the optimal consultation department is located.

[0077] The present application provides an intelligent guidance method, a guidance elevator and a system based on the guidance elevator. The information collection, disease analysis, department allocation, detection reservation and path guidance are organically integrated to form an integrated and automated intelligent guidance closed loop, effectively balancing the resource load of each department, reducing resource idling, improving equipment utilization and medical staff work efficiency, and promoting the overall improvement of medical service quality and management level. Through the deep application of multi-modal data fusion and intelligent algorithms, dynamic optimization of medical resources is realized, which not only ensures that patients can receive timely and accurate medical treatment services, but also effectively alleviates the problems of hospital department resource shortage and long patient waiting time, and promotes the upgrading of medical services to intelligent and humanized. In the guidance elevator, the patient completes the input of subjective symptom description, the collection of external signs and preliminary medical detection. The information collection work originally scattered in different areas and different links is concentrated in the process of taking the elevator, which greatly saves the patient's time and improves the information collection efficiency by more than 50%. The synchronous acquisition of multi-source data verifies the subjective symptoms and objective detection results, avoids misdiagnosis caused by single data, and makes the subsequent disease analysis more comprehensive and accurate. Using natural language processing, computer vision and other technologies to analyze multi-dimensional data can quickly and accurately find abnormal signs, determine the critical level, determine the candidate diseases, and predict the probability of disease and department association. For critical patients, the system can quickly identify and prioritize them to the department with the shortest waiting time to ensure the golden treatment time; for non-critical patients, the system considers the association probability and the real-time number of patients in the department to reduce the waiting time while ensuring professionalism, and the overall triage accuracy is improved to more than 90%, effectively solving the problem of "not knowing what department to go to". According to the patient's condition, the optimal consultation path is constructed, and the real-time flow data is dynamically adjusted to reduce the patient's ineffective movement in the hospital, and the average walking time is shortened by 40%. At the same time, according to the waiting time of the department and the moving time of the patient, the necessary detection items are reserved in advance, so that the patient's consultation and detection process is seamlessly connected, avoiding long waiting and repeated travel, and the consultation efficiency is improved by 30%, greatly improving the patient's medical experience.

[0078] Those skilled in the art will appreciate that the modules of the application described above can be implemented with general computing systems, which can be centralized on a single computing system or distributed across a network of multiple computing systems. Alternatively, they can be implemented with computer system executable program code, which can be stored in a storage system and executed by a computing system, or they can be implemented as individual integrated circuit modules, or multiple modules or steps can be implemented as a single integrated circuit module. Thus, the application is not limited to any particular combination of hardware and software.

[0079] Note that the above only describes the preferred embodiments of the application and the principles of the applied technology. Those skilled in the art will understand that the application is not limited to the specific embodiments described herein, and that various obvious changes, reconfigurations and substitutions can be made without departing from the scope of the application. Therefore, although the application has been described in detail through the above embodiments, the application is not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the application, and the scope of the application is determined by the appended claims.

[0080] The above only discloses several specific implementation scenarios of the application, but the application is not limited thereto, and any changes that can be thought of by those skilled in the art should fall within the protection scope of the application.

Claims

1. A smart patient guidance method based on a patient guidance elevator, characterized in that, Applicable to patient guidance elevators, the intelligent patient guidance method includes: Patients are guided into a pre-set triage elevator to obtain their subjective symptom descriptions and external symptom manifestations. Pre-set initial medical equipment is used to conduct preliminary tests on patients in the triage elevator to obtain initial test data. The subjective symptom descriptions, external symptom manifestations, and initial examination data are analyzed to identify abnormal signs in patients, determine the severity level of patients, identify potential candidate diseases, and predict the probability of association between the candidate diseases and general departments and specialized departments, respectively. Based on the severity level and the correlation probability of each department, as well as the real-time number of patients, the optimal department for the patient is selected, the registration for the optimal department is completed, and the patient's treatment path from the guidance elevator to the optimal department is constructed. Based on the combined abnormal signs and the test items associated with the candidate diseases, the necessary test items for the patient are screened, and the necessary test items are scheduled in advance according to the patient's waiting time in the optimal department and the waiting time in the corresponding department of the necessary test item. After the guided elevator reaches the floor where the optimal department is located, it guides the patient to the optimal department according to the treatment path.

2. The intelligent triage method according to claim 1, characterized in that, Also includes: The subjective symptom description is parsed using natural language processing algorithms to extract structured information, including symptom keywords, duration, and severity, and then a symptom vector is generated. The external symptoms are analyzed using visual processing algorithms, and quantitative features, including facial features, body shape features, and abnormal area features, are extracted to form an external feature matrix. Retrieve patients’ historical medical records to construct a time-series medical history map that includes at least past medical conditions, past test results, and allergy history; The initial detection data is analyzed for features to identify abnormal indicators, and causal relationships, temporal dependencies, and mapping relationships with past symptoms are established among the abnormal indicators to generate a pathological state vector. Abnormal signs of the patient are generated based on the symptom vector, external feature matrix, temporal medical history atlas, and pathological state vector.

3. The intelligent triage method according to claim 2, characterized in that, The process of obtaining the necessary detection items specifically includes: Based on a preset association rule base, all diseases associated with the abnormal signs are searched, and a Bayesian network is used to calculate the posterior probability of each disease. Diseases with posterior probabilities exceeding a preset probability threshold are selected as candidate diseases. Based on a preset association rule base, candidate detection items associated with the candidate disease are searched, and the candidate detection items are filtered for taboos to obtain feasible detection items; A multi-objective optimization function for each feasible test item is constructed and arranged to generate a priority list. The priority list is adjusted in real time according to the patient's severity level. Based on the order of the priority list, feasible tests are selected as the final necessary tests.

4. The intelligent triage method according to claim 1, characterized in that, Also includes: The image module acquires the patient's motion data as they enter the triage elevator from outside. By analyzing the motion data, the patient's gait characteristics are determined, and the patient's walking speed is estimated by combining the patient's physiological parameters. The first walking time is obtained by predicting the time required for the patient to reach the optimal department from the guidance elevator based on the walking speed and the medical route; the waiting time of the patient in the optimal department is predicted based on the real-time number of patients in the optimal department.

5. The intelligent triage method according to claim 4, characterized in that, Also includes: A detection path is constructed for the patient to reach the department corresponding to the necessary detection item from the optimal department, and the second walking time is obtained by predicting the time required for the patient to reach the department corresponding to the necessary detection item from the optimal department based on the walking speed and the detection path. The combined walking time and waiting time are used to make an initial appointment for the patient in the department corresponding to the necessary test items, so that the waiting time for the initial appointment is not less than the sum of the first walking time, the second walking time, and the waiting time. After the optimal department confirms the diagnosis of the necessary tests, the preliminary appointment is converted into an actual appointment, and the waiting time for the actual appointment is not less than the second walking time.

6. The intelligent triage method according to claim 1, characterized in that, The selection of the optimal department for medical treatment specifically includes: Predict patient waiting time in each specialized department based on real-time patient numbers. If the severity level indicates that the patient is currently in a critical condition, then the general or specialized department with the shortest waiting time will be selected as the optimal department for treatment. If the severity level indicates that the patient is currently in a non-critical state, then it is determined whether there is a specialized department with a correlation probability exceeding a preset first threshold: if not, then a general department is selected as the optimal department for treatment; if so, then among the specialized departments with a correlation probability exceeding the preset first threshold, the department with the shortest waiting time is selected as the optimal department for treatment.

7. A patient guidance elevator, characterized in that, For implementing the intelligent patient guidance method according to any one of claims 1-6; the patient guidance elevator includes: Elevator cabin; The initial medical equipment, located inside the elevator car, is used for preliminary testing of patients; The human-computer interaction module, located inside the elevator car, is used for symptom input and guidance of the detection process; An image module, located inside and outside the elevator car, is used to acquire images of the patient both outside and inside the elevator car. The guidance module is used to guide patients. The data processing module is connected to the elevator car, the initial medical equipment, the human-computer interaction module, the image module, and the guidance module.

8. The patient guidance elevator according to claim 7, characterized in that, The initial medical equipment includes an integrated vital signs detector, a dry microfluidic biochemical analyzer, a portable channel electrocardiograph, a pulse wave detector, an intelligent stethoscope, and a skin image acquisition device.

9. The patient guidance elevator according to claim 7, characterized in that, Also includes: A mobility module is used to move the patient under the guidance of the guidance module.

10. A smart patient guidance system based on a patient guidance elevator, characterized in that, include: The initial screening unit is used to guide patients into a pre-set triage elevator, obtain the patient's subjective symptom description and external symptom manifestation, and use pre-set initial screening medical equipment to conduct preliminary tests on the patient in the triage elevator to obtain initial screening data. The analysis unit is used to analyze the subjective symptom descriptions, external symptom manifestations, and initial examination data to find abnormal signs of the patient, determine the patient's critical level, identify the candidate diseases that the patient may be involved in, and predict the probability of association between the candidate diseases and general departments and specialized departments, respectively. The consultation room selection unit is used to select the optimal consultation department for the patient based on the severity level and the correlation probability of each department, as well as the real-time number of patients, to complete the registration of the optimal consultation department and construct the consultation path for the patient to reach the optimal consultation department from the guidance elevator. The appointment unit is used to screen necessary tests for patients by comprehensively considering the abnormal signs and the test items associated with the candidate diseases, and to make advance appointments for necessary tests based on the patient's waiting time in the optimal department and the waiting time in the corresponding department of the necessary test. The guidance unit is used to control the patient guide elevator to reach the floor where the optimal department is located and guide the patient to the optimal department according to the treatment path.

Images