One-way payment in-medicine-hospital full-process method based on non-inductive intelligent handling mode

Abstract

The invention discloses a one-way payment medicine hospital full-process method based on a non-inductive intelligent handling mode, and relates to the technical field of medical informatization, and the method comprises the steps: collecting the facial features of a patient, and generating biological feature information and a hospital medical account; receiving symptom description, extracting feature vectors, and constructing a symptom evolution graph to generate hospital guide information; identity verification is carried out, and hospital guide information is pushed to the doctor terminal; receiving prescription information based on the hospital guide information, and analyzing a drug metabolism pathway network to generate a prescription optimization suggestion; forming dispensing information according to the optimization suggestion, and generating a payment voucher through distributed consensus verification; and verifying the payment voucher and carrying out identity comparison to complete medicine taking. According to the invention, intelligent management of the whole medical treatment process is realized, and the medical treatment efficiency and safety are improved.

Description

Technical Field

[0001] This invention relates to the field of medical information technology, specifically to a method for a single-line payment system for drugs within a hospital based on a seamless intelligent payment model. Background Technology

[0002] Currently, the medical environment in hospitals is complex, and patients need to queue multiple times during the medical process, which is cumbersome and time-consuming, affecting the medical experience. The traditional hospital medical process usually includes multiple steps such as registration, waiting, medical treatment, prescription issuance, payment, and medication collection. Each step requires the patient to be present in person, which not only increases the patient's waiting time but also increases the workload of hospital administrators.

[0003] With the development of information technology, hospitals are gradually adopting intelligent systems to optimize the medical process. Existing technologies already include applications such as intelligent triage and online payment, but these technologies still have the following problems: current identity verification technologies mainly rely on manual verification, which is prone to identity theft; intelligent triage systems provide relatively simple analysis of patient symptoms and cannot accurately grasp the dynamic changes in a patient's condition; the prescription process lacks intelligent drug interaction analysis, which may lead to medication safety issues; and the information security and traceability of the payment process need to be improved.

[0004] Existing systems are often independent, lacking effective information integration and collaboration mechanisms. This makes seamless data integration across the entire medical process difficult, impacting the efficiency and quality of intelligent services. Traditional payment methods pose information security risks, making it difficult to guarantee transaction credibility and traceability. For drugs requiring separate payment, the high unit price and often high reimbursement rates under medical insurance necessitate strict control over application conditions and reimbursement requirements, further complicating the medical process and increasing the demand for a seamless intelligent payment system. Therefore, this invention focuses on such drugs, but the related content can be extended to other types of medical and pharmaceutical processes. Summary of the Invention

[0005] The purpose of this invention is to provide a single-line payment method for the entire in-hospital process of medicines based on a contactless intelligent processing mode, which solves the problems of low efficiency, insecure identity authentication, lack of intelligent assistance in prescription issuance, and insufficient security in the payment process in existing hospital medical treatment processes. It aims to provide an intelligent medical treatment solution that integrates technologies such as biometric recognition, intelligent triage, drug interaction analysis, and blockchain payment, so as to achieve efficient, safe, and convenient processing of the entire patient medical treatment process.

[0006] The single-line payment method for drugs within hospitals based on a seamless intelligent payment model provided in this embodiment of the invention includes the following steps: Collect facial features of the target patient to form biometric information and generate an in-hospital medical account containing medical records; The system receives symptom descriptions from target patients, extracts symptom feature vectors, maps the symptom feature vectors to medical records, constructs a symptom evolution map, determines the diagnosis and treatment time points based on the symptom evolution map and the distribution of doctor resources, and generates triage information. Upon receiving the hospital admission instruction, the system collects facial and biometric information on-site for identity verification. Once the verification is successful, the intelligent triage information is pushed to the doctor's terminal. Based on intelligent triage information, electronic prescription information is received, drug interaction relationships are identified from the electronic prescription information, a drug metabolism pathway network is constructed, overlapping sites of drug metabolism pathways are analyzed, the competition intensity of metabolism pathways is calculated, and prescription optimization suggestions are generated. Modify the electronic prescription information according to the prescription optimization suggestions and generate the medication information. Then, verify the payment amount of the medication information with the in-hospital medical account through distributed consensus and generate a payment voucher containing a trusted timestamp. Verify the payment voucher to complete the payment, collect the target patient's real-time facial features for identity comparison, complete the medication collection after the comparison is successful, and record the medical information.

[0007] Furthermore, by collecting facial features of the target patient to form biometric information, an in-hospital medical account containing medical records is generated, including: Collect static facial features and voice information of the target patient, extract facial structural features to generate static feature vectors, extract voiceprint features to generate dynamic feature vectors; The information redundancy of the static feature vector and the dynamic feature vector is calculated. The information redundancy is sorted in ascending order to obtain the difference. Based on the difference, a partitioning threshold is set to generate multiple complementary sub-intervals. The feature distribution density is calculated according to the degree of feature aggregation within the complementary sub-intervals. The main feature direction is obtained by multi-scale decomposition of the feature distribution density. Based on the static feature vector and the dynamic feature vector, a feature spanning space is constructed. The main feature direction is adaptively projected onto the feature spanning space to obtain the initial basis vector. The initial basis vectors are orthogonalized to generate an orthogonal projection matrix. The static and dynamic eigenvectors are then projected onto the orthogonal projection matrix to generate a fused eigenvector. An environmental compensation matrix is ​​constructed by collecting environmental parameters corresponding to the fused feature vectors. The fused feature vectors and the environmental compensation matrix are then regularized to generate biometric information. Biometric information is encrypted to generate tamper-proof feature records, which are then written into the hospital's medical account to generate an in-hospital medical account containing medical records.

[0008] Furthermore, the system receives symptom descriptions from target patients, extracts symptom feature vectors, maps these vectors to medical records, constructs a symptom evolution map, determines treatment time points based on the symptom evolution map and doctor resource distribution, and generates triage information including: Receive symptom description information from the target patient, extract symptom temporal features, and generate symptom feature vectors; The symptom feature vector is correlated with historical medical records to obtain the symptom correlation degree, and the symptom correlation degree is then subjected to hierarchical clustering to obtain the symptom similarity. Based on symptom similarity, the evolution patterns of symptoms are extracted and constructed into a directed graph structure, where nodes represent symptoms and edges represent the evolutionary correlation strength between adjacent symptoms, thus generating a symptom evolution map. Obtain the distribution of doctor resources, calculate the symptom conversion probability based on the symptom evolution map, and dynamically optimize the allocation of doctor resource distribution based on the symptom conversion probability to obtain doctor resource weights; The symptom treatment priority is calculated based on the strength of the evolution association. The symptom treatment priority is then weighted and combined with the doctor resource weight. The optimal treatment time point is determined according to the temporal distribution of the symptom conversion probability. Based on the optimal treatment time point, triage information is generated.

[0009] Furthermore, upon receiving the hospital admission instruction, the system collects facial and biometric information for identity verification. Once verification is successful, intelligent triage information is pushed to the doctor's terminal, including: Obtain the hospital admission instruction, collect the on-site facial image of the target patient, extract the brightness distribution of the on-site facial image, compensate the on-site facial image based on the brightness distribution, and generate a compensated image; Facial features are extracted from the compensated image to construct a live facial feature vector. Voice information of the target patient is collected to extract voiceprint features to construct a live voiceprint feature vector. The live facial feature vector and the live voiceprint feature vector are combined to construct a feature matching sequence. The feature matching sequence is dynamically planned and matched with the biometric information in the in-hospital medical account to obtain the feature similarity. The distribution data of the feature similarity is calculated to determine the fusion weight. The fusion weight and the feature similarity are weighted and combined to generate the identity verification similarity. Collect on-site environmental parameters to construct an environmental quality score, determine a decision threshold based on the environmental quality score, and compare the identity verification similarity with the decision threshold to generate an identity verification result. Based on the identity verification results, intelligent triage information will be pushed to the doctor's terminal.

[0010] Furthermore, drug interactions are identified from the electronic prescription information, a drug metabolism pathway network is constructed, overlapping sites of drug metabolism pathways are analyzed, the competition intensity of metabolic pathways is calculated, and prescription optimization suggestions are generated, including: Structural information of drugs is extracted from electronic prescription information. The structural information is decomposed into molecular features to obtain a structural feature set. Recursive matching is performed on the structural feature set in a drug database to extract drug target and metabolic site information. Obtain medical records, extract historical medication records, perform time-series mapping between the historical medication records and drug targets to obtain target action intensity, calculate drug metabolism overlap based on the overlap interval between the target action intensity and metabolic site information, and generate drug interaction relationships; A drug metabolism pathway network is constructed based on drug interaction relationships. The network is recursively traversed to obtain node connection data. Overlapping sites of drug metabolism pathways are extracted based on the node connection data and the degree of overlap in drug metabolism. Drug concentration data is extracted from electronic prescription information, and the drug concentration data is accumulated at overlapping sites of drug metabolism pathways to obtain the competition degree of overlapping sites. The competition intensity of metabolic pathways is calculated based on the competition degree of overlapping sites and drug interaction relationships. The competition intensity of the metabolic pathways is compared with the safety thresholds in clinical medication rules to generate drug dosage adjustment schemes as prescription optimization suggestions.

[0011] Furthermore, based on prescription optimization suggestions, the electronic prescription information is modified to generate medication dispensing information. The payment amount of the dispensing information is then verified through distributed consensus with the hospital's medical account to generate a payment voucher containing a trusted timestamp, including: Based on the prescription optimization suggestions, modify the electronic prescription information, extract the drug specifications and dosage from the modified electronic prescription information to generate a drug dispensing matrix, match and optimize the drug dispensing matrix with the pharmacy's regional inventory information to obtain the drug delivery route, generate dispensing information based on the drug delivery route, and calculate the payment amount for the dispensing information. The system retrieves user medical insurance information from the in-hospital medical account, categorizes and matches the reimbursement amount in the user medical insurance information with the payment amount to generate a reimbursement plan, and performs distributed consensus verification on the payment amount based on the reimbursement plan to generate payment data. The transaction information of the payment data is grouped and encrypted according to the block structure to obtain block data. The block data is then verified by nodes to generate verification records. A trusted timestamp is generated based on the verification records. The trusted timestamp and the payment data are hashed together to generate a payment credential.

[0012] Furthermore, the payment voucher is verified to complete the payment, the target patient's real-time facial features are collected for identity verification, and medication is dispensed after successful verification. Medical information is recorded, including: The payment voucher is decrypted in layers to obtain block data and payment data. The verification records in the block data are calculated to obtain the verification weight. The payment data is verified according to the verification weight to generate a payment verification result. The payment is completed based on the payment verification result. Real-time facial images of the target patient are acquired, and the illumination distribution matrix and structural feature points of the real-time facial images are extracted. An environmental compensation coefficient is calculated based on the illumination distribution matrix. The structural feature points are compensated based on the environmental compensation coefficient to generate facial feature data. The facial feature data is combined with the verification weight to generate feature verification information. Biometric information is obtained from the hospital's medical account. The feature verification information is matched with the biometric information to generate an identity verification result. The medication dispensing authority is confirmed based on the identity verification result and the payment verification result. After the medication is dispensed, the payment verification result and the identity verification result are written into the hospital's medical account.

[0013] Furthermore, a drug metabolism pathway network is constructed based on drug interaction relationships. The network is recursively traversed to obtain node connection data. Based on this node connection data and the degree of overlap in drug metabolism, overlapping sites in the drug metabolism pathways are extracted, including: Based on drug interaction relationships, the location and intensity of action points in the drug metabolism process are extracted. The locations of action points are constructed as network nodes. The connection weights between nodes are calculated based on the intensity of action. The connection relationships between the nodes are generated according to the connection weights between the nodes. The connection relationships are combined to form a drug metabolism pathway network. The nodes in the drug metabolism pathway network are recursively traversed, and the predecessor and successor nodes of each node are recorded to generate a node dependency sequence. The node activity is obtained by counting the node frequency in the node dependency sequence. The node activity is combined with the node connection weight to generate node connection data. The overlapping area between nodes is calculated based on the node connection data. The nodes within the overlapping area are matched with the drug metabolism overlap, and the nodes with a matching degree exceeding a preset matching degree threshold are extracted as drug metabolism pathway overlap sites.

[0014] One technical solution provided in this embodiment of the invention is an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0015] One technical solution provided in this embodiment of the invention is a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the steps in any of the aforementioned methods.

[0016] This invention employs biometric identification technology for identity verification, enabling reliable identity confirmation throughout the entire medical process and effectively preventing impersonation. Based on symptom feature vector mapping and atlas construction, it can intelligently track the evolution of the disease, improving the accuracy of diagnosis and treatment. Through drug metabolism pathway network analysis and competition intensity calculation, it can effectively prevent adverse drug interactions and ensure medication safety. Utilizing a distributed consensus verification mechanism and trusted timestamps, it ensures the traceability and immutability of the payment process. The intelligent closed-loop management characteristics of this solution effectively reduce hospital operating costs and improve the efficiency of medical services. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the in-hospital full-process method for single-line payment of drugs based on the contactless intelligent payment mode provided in an embodiment of the present invention; Figure 2 This is a core flowchart of the intelligent medical payment system according to an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. In the following description relating to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements.

[0020] like Figure 1 As shown, Figure 1 This is a flowchart of a single-line payment method for drugs within a hospital based on a contactless intelligent payment mode, provided by an embodiment of the present invention. The method includes the following steps: Collect facial features of the target patient to form biometric information and generate an in-hospital medical account containing medical records; The system receives symptom descriptions from target patients, extracts symptom feature vectors, maps the symptom feature vectors to medical records, constructs a symptom evolution map, determines the diagnosis and treatment time points based on the symptom evolution map and the distribution of doctor resources, and generates triage information. Upon receiving the hospital admission instruction, the system collects facial and biometric information on-site for identity verification. Once the verification is successful, the intelligent triage information is pushed to the doctor's terminal. Based on intelligent triage information, electronic prescription information is received, drug interaction relationships are identified from the electronic prescription information, a drug metabolism pathway network is constructed, overlapping sites of drug metabolism pathways are analyzed, the competition intensity of metabolism pathways is calculated, and prescription optimization suggestions are generated. Modify the electronic prescription information according to the prescription optimization suggestions and generate the medication information. Then, verify the payment amount of the medication information with the in-hospital medical account through distributed consensus and generate a payment voucher containing a trusted timestamp. Verify the payment voucher to complete the payment, collect the target patient's real-time facial features for identity comparison, complete the medication collection after the comparison is successful, and record the medical information.

[0021] Collecting facial features of the target patient to form biometric information, generating an in-hospital medical account containing medical records, including: Collect static facial features and voice information of the target patient, extract facial structural features to generate static feature vectors, extract voiceprint features to generate dynamic feature vectors; The information redundancy of the static feature vector and the dynamic feature vector is calculated. The information redundancy is sorted in ascending order to obtain the difference. Based on the difference, a partitioning threshold is set to generate multiple complementary sub-intervals. The feature distribution density is calculated according to the degree of feature aggregation within the complementary sub-intervals. The main feature direction is obtained by multi-scale decomposition of the feature distribution density. Based on the static feature vector and the dynamic feature vector, a feature spanning space is constructed. The main feature direction is adaptively projected onto the feature spanning space to obtain the initial basis vector. The initial basis vectors are orthogonalized to generate an orthogonal projection matrix. The static and dynamic eigenvectors are then projected onto the orthogonal projection matrix to generate a fused eigenvector. An environmental compensation matrix is ​​constructed by collecting environmental parameters corresponding to the fused feature vectors. The fused feature vectors and the environmental compensation matrix are then regularized to generate biometric information. Biometric information is encrypted to generate tamper-proof feature records, which are then written into the hospital's medical account to generate an in-hospital medical account containing medical records.

[0022] First, static facial features and voice information of the target patient are collected. Static facial features are collected by obtaining a frontal photograph of the patient using a high-definition camera, extracting key facial points, geometric features, and texture features. Voice information is collected by recording the patient reading specific text using a microphone. The collected static facial features are processed to extract structural features, including eye spacing, nose shape, and lip contour, generating a static feature vector. The voice information is then processed to extract voiceprint features, including spectral characteristics, pitch variations, and formant distribution, generating a dynamic feature vector.

[0023] The information redundancy between static and dynamic feature vectors is calculated. Information redundancy is determined by the mutual information between feature vectors; higher mutual information indicates greater redundancy and less difference. The calculated information redundancy is then arranged in ascending order to obtain the difference between features. A partitioning threshold is set based on the difference. For example, features with a difference between 0 and 0.3 are assigned to the first sub-interval, those between 0.3 and 0.6 to the second sub-interval, and those between 0.6 and 1 to the third sub-interval, generating multiple complementary sub-intervals. For each complementary sub-interval, the clustering degree of features within the interval is calculated, i.e., the feature distribution density. The feature distribution density is calculated by dividing the number of feature points within the interval by the interval length. For example, the feature density of the first sub-interval is 42 feature points / 0.3 = 140.

[0024] Multi-scale decomposition of feature distribution density can be performed using methods such as wavelet transform to analyze the feature density distribution at different scales and obtain the main directions of feature change, i.e., the principal feature directions. A feature spanning space is constructed based on static and dynamic feature vectors; this space is generated by a linear combination of the static and dynamic feature vectors. The principal feature directions are adaptively projected onto the feature spanning space to obtain initial basis vectors. Adaptive projection dynamically adjusts the projection parameters according to the distribution characteristics of the feature space, ensuring that the projected initial basis vectors better represent the information of the original features.

[0025] The initial basis vectors are orthogonalized using the Schmidt orthogonalization method, which transforms non-orthogonal initial basis vectors into mutually orthogonal basis vectors, generating an orthogonal projection matrix. The static and dynamic feature vectors are then projected onto this orthogonal projection matrix to achieve feature fusion, generating a fused feature vector. This fused feature vector contains both the patient's static facial features and dynamic speech features, providing a more comprehensive representation of the patient's biometrics.

[0026] Environmental parameters corresponding to the fused feature vectors are collected, including light intensity, background noise level, and parameters of the acquisition device. An environmental compensation matrix is ​​constructed based on the collected environmental parameters. This matrix is ​​used to eliminate or reduce the influence of environmental factors on feature extraction. The fused feature vectors and the environmental compensation matrix are then regularized. This regularization process normalizes the features and suppresses noise interference, improving the robustness of the features and generating biometric information.

[0027] The generated biometric information is encrypted using an asymmetric encryption algorithm, converting it into a tamper-proof feature record. This tamper-proof feature record is then written into the hospital's medical account, generating an account containing the patient's medical records. The tamper-proof feature record in the medical account can be used for patient authentication during subsequent medical visits, eliminating the need for patients to repeatedly provide identification and achieving seamless identification.

[0028] For example, when a patient first visits the hospital, their facial features and voice samples are collected and processed to generate a fused feature vector. If a patient's static feature vector contains 128 feature points and their dynamic feature vector contains 96 feature points, after calculating information redundancy and sorting, and setting difference thresholds of 0.3 and 0.7, three complementary sub-intervals are formed. The feature distribution density of each interval is calculated, and the density is decomposed into principal feature directions using multi-scale decomposition. A feature spanning space is constructed, and the principal feature directions are projected to obtain initial basis vectors, which are then orthogonalized to form a 32-dimensional orthogonal projection matrix. The original feature vector is projected onto this matrix to generate a 64-dimensional fused feature vector. Environmental parameters are collected to construct a compensation matrix, which is then regularized to generate biometric information, encrypted, and written to the patient's in-hospital medical account. When the patient visits again, identity verification can be completed simply through facial or voice recognition, allowing for direct medical procedures such as medication payment.

[0029] This invention improves the accuracy and security of biometric recognition by fusing static facial features and dynamic voice features; it effectively reduces feature redundancy and improves feature representation by employing complementary feature sub-intervals and orthogonal projection technology; it introduces an environmental compensation mechanism to enhance the environmental adaptability and robustness of biometric recognition; it uses encryption and anti-tampering technologies to ensure the security of patient biometric information; and it achieves seamless identification and convenient payment throughout the entire patient medical process, reducing patient waiting time, improving medical service efficiency and patient experience, while providing medical institutions with a safer and more efficient patient management method.

[0030] The system receives symptom descriptions from target patients, extracts symptom feature vectors, maps these vectors to medical records, constructs a symptom evolution map, determines treatment time points based on the symptom evolution map and doctor resource distribution, and generates triage information including: Receive symptom description information from the target patient, extract symptom temporal features, and generate symptom feature vectors; The symptom feature vector is correlated with historical medical records to obtain the symptom correlation degree, and the symptom correlation degree is then subjected to hierarchical clustering to obtain the symptom similarity. Based on symptom similarity, the evolution patterns of symptoms are extracted and constructed into a directed graph structure, where nodes represent symptoms and edges represent the evolutionary correlation strength between adjacent symptoms, thus generating a symptom evolution map. Obtain the distribution of doctor resources, calculate the symptom conversion probability based on the symptom evolution map, and dynamically optimize the allocation of doctor resource distribution based on the symptom conversion probability to obtain doctor resource weights; The symptom treatment priority is calculated based on the strength of the evolution association. The symptom treatment priority is then weighted and combined with the doctor resource weight. The optimal treatment time point is determined according to the temporal distribution of the symptom conversion probability. Based on the optimal treatment time point, triage information is generated.

[0031] Patients can input symptom descriptions via hospital smart terminals or mobile applications, including symptom name, duration, severity, and other information. Natural language processing (NLP) is used to analyze the symptom descriptions, extracting keywords and labeling them with time attributes to form a symptom feature set with time-series markers. These features are then vectorized, with each symptom feature corresponding to a numerical value, and multiple features combined to form a symptom feature vector. For example, for a patient's description of "starting cough three days ago, having a fever of 38.5 degrees Celsius yesterday, and experiencing mild chest tightness this morning," semantic analysis extracts the symptoms "cough," "fever," and "chest tightness," along with their corresponding time points "three days ago," "yesterday," and "this morning," and quantifies each as a feature value to construct a symptom feature vector.

[0032] The acquired symptom feature vectors are correlated with symptom information in the patient's historical medical records. These records contain information such as past diagnoses, symptom descriptions, and treatment plans. The symptom correlation is obtained by calculating the similarity between the current symptom feature vector and historical symptom records. The cosine similarity method is used to calculate the cosine of the angle between the two vectors; a larger value indicates higher similarity. Hierarchical clustering analysis is performed on the calculated symptom correlations, employing a bottom-up aggregation strategy to group symptoms with similarity higher than a user-defined threshold into one category, gradually forming a hierarchical structure of symptoms and obtaining the similarity relationships between them. Correlation analysis is performed on the patient's symptom feature vectors and the symptoms from five previous medical visits in their historical records. The calculated correlation values ​​are 0.82, 0.45, 0.37, 0.91, and 0.29, respectively. After hierarchical clustering, the current symptom is determined to have the highest similarity to the symptom from the fourth medical visit.

[0033] Symptom evolution patterns are extracted based on symptom similarity. By analyzing the time-series changes of highly similar symptom clusters, the development trends and transformation patterns of symptoms are identified. The identified symptom evolution patterns are constructed into a directed graph structure, where each node represents a specific symptom, and edges represent the evolutionary association strength between adjacent symptoms. The evolutionary association strength is calculated based on the frequency and time interval of symptom transformations; the higher the frequency and the shorter the interval, the stronger the association. For this patient, historical record analysis shows that the evolutionary association strength from "cough" to "fever" is 0.75, and the evolutionary association strength from "fever" to "chest tightness" is 0.68. Based on this, a symptom evolution atlas is constructed. The atlas contains the main symptom nodes and their evolutionary paths, clearly presenting the possible directions of symptom development.

[0034] The distribution of physician resources within the hospital is obtained, including the number of doctors in each department, their specialties, and their consultation hours. Based on the previously constructed symptom evolution map, the conversion probability between each symptom node is calculated. The conversion probability is obtained through evolution association strength normalization, representing the likelihood of one symptom developing into another. Physician resource distribution is dynamically optimized based on symptom conversion probabilities. The optimization process considers the development trend and severity of symptoms, tending to allocate medical resources to symptom treatment areas with high conversion probabilities, thus obtaining physician resource weights. In this example, based on the symptom evolution link of "cough-fever-chest tightness," the calculated conversion probabilities are 0.75 and 0.68, respectively. A weight of 0.85 is assigned to respiratory physician resources.

[0035] Based on the strength of evolutionary associations in the symptom evolution map, the treatment priority of each symptom is calculated. Treatment priority considers the severity, urgency, and position of the symptom in the evolutionary chain. The higher the strength of the evolutionary association, the higher the treatment priority of the corresponding symptom. The calculated symptom treatment priorities are weighted and combined with the aforementioned physician resource weights to generate a comprehensive weight value. According to the temporal distribution of symptom conversion probabilities, key time points in symptom development are analyzed to determine the optimal treatment time. The optimal treatment time refers to the time window during symptom development where medical intervention is most effective. Based on the determined optimal treatment time, referral information is generated, including recommended departments, appointment times, and estimated waiting times. For this patient, a comprehensive analysis of the symptom evolution trend and physician resource distribution determines the optimal treatment time to be within 24 hours, generating referral information recommending a visit to the respiratory medicine department and providing optional appointment time slots.

[0036] Once a patient receives triage information and confirms an appointment through the hospital's mobile application, an appointment record is automatically created, and relevant symptom information is pushed to the corresponding doctor's workstation. Upon arrival at the hospital at the appointed time, the patient verifies their identity via facial recognition or other biometrics, bypassing the need to queue for registration and proceeding directly to the designated consultation room. The doctor provides targeted diagnosis and treatment based on the pushed symptom information and pre-analysis results. After the consultation, an electronic prescription is automatically generated, which the patient can view details and confirm payment on the hospital's mobile application. After payment, the pharmacy is notified to prepare the medication, which the patient can then pick up directly without queuing for payment again. The entire medical process achieves seamless identification, intelligent triage, and one-line payment, greatly simplifying traditional medical procedures.

[0037] This invention achieves accurate identification and evolution prediction of patient symptoms through symptom temporal feature extraction and correlation analysis; improves the efficiency of medical resource utilization through symptom conversion probability calculation and dynamic allocation of doctor resources; realizes precise triage and appointment services by combining symptom treatment priority and optimal treatment time point determination; integrates contactless identity recognition and one-line payment functions to simplify the patient's medical process and reduce waiting time; and effectively improves diagnostic efficiency and accuracy through intelligent analysis and prediction capabilities, while providing data support for hospital management, thus achieving improved quality and efficiency of medical services and comprehensive optimization of patient experience.

[0038] Upon receiving the hospital arrival instruction, the system collects facial and biometric information for identity verification. After successful verification, intelligent triage information is pushed to the doctor's terminal, including: Obtain the hospital admission instruction, collect the on-site facial image of the target patient, extract the brightness distribution of the on-site facial image, compensate the on-site facial image based on the brightness distribution, and generate a compensated image; Facial features are extracted from the compensated image to construct a live facial feature vector. Voice information of the target patient is collected to extract voiceprint features to construct a live voiceprint feature vector. The live facial feature vector and the live voiceprint feature vector are combined to construct a feature matching sequence. The feature matching sequence is dynamically planned and matched with the biometric information in the in-hospital medical account to obtain the feature similarity. The distribution data of the feature similarity is calculated to determine the fusion weight. The fusion weight and the feature similarity are weighted and combined to generate the identity verification similarity. Collect on-site environmental parameters to construct an environmental quality score, determine a decision threshold based on the environmental quality score, and compare the identity verification similarity with the decision threshold to generate an identity verification result. Based on the identity verification results, intelligent triage information will be pushed to the doctor's terminal.

[0039] Patients can initiate a medical request through the hospital's mobile application or directly at the hospital's self-service terminal. Upon receiving the medical request, the subsequent process begins. When the patient stands in front of the hospital's self-service terminal, the terminal's camera automatically captures the patient's facial image. During the capture process, the camera continuously captures the patient's face at a rate of 24 frames per second, selecting the image with the highest resolution for processing.

[0040] Brightness distribution extraction is performed on the acquired facial images from the scene. Brightness distribution extraction is achieved by calculating the brightness value of each pixel in the image and statistically analyzing the distribution of different brightness values. Brightness values ​​are generally represented by values ​​from 0 to 255, where 0 represents the darkest and 255 represents the brightest. By analyzing the brightness distribution histogram, the uniformity of image brightness and the presence of overexposed or underexposed areas are identified. Based on the extracted brightness distribution, the facial images from the scene are compensated using adaptive histogram equalization technology. This technology dynamically adjusts compensation parameters according to the brightness characteristics of local areas of the image, improving image contrast and detail. In low-light environments, details in dark areas are enhanced; in high-light environments, overexposure in bright areas is reduced. Through brightness compensation processing, a compensated image is generated to ensure that facial features are clearly presented under different lighting conditions.

[0041] Facial features are extracted from the compensated image, including the location and shape of key points such as facial contours, eyes, nose, and mouth. A deep convolutional neural network (CNN) is used for extraction; this network, trained on large-scale face data, accurately identifies facial feature points and calculates feature descriptors. The extracted facial features are encoded to construct a 128-dimensional on-site facial feature vector. Simultaneously, the patient's voice information is collected via a terminal microphone, requiring the patient to read random text displayed on a screen to prevent voice replay attacks. The collected voice information undergoes preprocessing, including noise reduction, framing, and feature extraction, to extract voiceprint features and construct a 64-dimensional on-site voiceprint feature vector. Voiceprint feature extraction employs Mel-frequency cepstral coefficients (MFCCs), a technique that simulates the human ear's perception of sound and effectively characterizes the speaker's individual acoustic features. The on-site facial feature vector and the on-site voiceprint feature vector are combined according to certain rules to construct a feature matching sequence. The combination method involves concatenating the two feature vectors to generate a 192-dimensional feature matching sequence.

[0042] The constructed feature matching sequence is dynamically matched with the biometric information stored in the hospital's medical account. The dynamic programming matching process treats the feature matching sequence and the stored biometric information as two separate sequences, constructs a matching matrix, finds the optimal matching path, and calculates feature similarity. The dynamic programming algorithm considers the differences in importance of each dimension of the feature vector, assigning higher weights to key feature dimensions to improve matching accuracy. The calculated feature similarity value ranges from 0 to 1, with values ​​closer to 1 indicating a higher matching degree. Distribution data analysis is performed on the calculated feature similarity, including statistical characteristics such as mean, standard deviation, and distribution pattern. Based on the analysis results, the fusion weights of facial features and voiceprint features are determined. The weight values ​​reflect the reliability and importance of each feature in identity recognition. For example, in well-lit but noisy environments, the facial feature weight might be 0.7, and the voiceprint feature weight might be 0.3. The determined fusion weights are then weighted and combined with the feature similarity to calculate the comprehensive identity verification similarity.

[0043] Environmental parameters are collected, including ambient light intensity and background noise level. Ambient light intensity is measured using a photosensitive sensor, in lux; background noise level is measured using a sound level meter, in decibels. An environmental quality score is constructed based on the collected environmental parameters. The environmental quality score considers the impact of multiple environmental factors on recognition accuracy and uses a weighted summation method to calculate the comprehensive score. The score ranges from 0 to 100, with higher scores indicating more favorable environmental conditions for accurate recognition. A decision threshold is dynamically determined based on the environmental quality score. When the environmental quality score is high, the decision threshold can be appropriately increased to reduce the false recognition rate; when the environmental quality score is low, the decision threshold can be appropriately decreased to reduce the rejection rate. The calculated identity verification similarity is compared with the decision threshold. If the similarity is greater than or equal to the decision threshold, the identity verification is considered successful; otherwise, the identity verification is considered unsuccessful. The identity verification result is represented in Boolean form, with 1 for success and 0 for failure.

[0044] Based on the identity verification result, subsequent operations are processed automatically. When identity verification is successful, pre-generated intelligent triage information is pushed to the doctor's terminal. The intelligent triage information includes the patient's basic information, symptom description, summary of historical medical records, and pre-diagnosis analysis results. Doctors can understand the patient's condition in advance through the terminal device and prepare for diagnosis and treatment. The patient's identity verification information is also synchronized to relevant departments of the hospital, providing seamless credentials for subsequent visits, examinations, medication dispensing, etc., realizing "one-time authentication, all-round access" throughout the entire process. If identity verification fails, the patient is guided to re-identify or go to the manual service window for processing.

[0045] For example, a patient stands in front of a hospital self-service terminal, and their facial image is captured. Brightness analysis shows insufficient lighting on the left side of the image, with the brightness distribution concentrated in the range of 80-160. Brightness compensation is applied to the image, expanding the brightness distribution to the range of 40-220, generating a compensated image. A 128-dimensional facial feature vector is extracted from the compensated image. Simultaneously, a voice sample of the patient reading randomly occurring text is collected, and a 64-dimensional voiceprint feature vector is extracted. These are combined to form a 192-dimensional feature matching sequence. Dynamic programming matching is performed with the biometric information stored in the hospital's medical account, resulting in a facial feature similarity of 0.92 and a voiceprint feature similarity of 0.85. Environmental assessment shows that the current illumination is 500 lux, the noise level is 45 dB, and the environmental quality score is 85 points. Based on this, a decision threshold of 0.82 is determined. After weighted combination calculation, the overall identity verification similarity is 0.90, which is greater than the decision threshold, and the identity verification passes. The patient's intelligent triage information is immediately pushed to the corresponding doctor's terminal, and consultation guidance information is also pushed to the patient.

[0046] This invention improves the accuracy of facial feature extraction under different lighting conditions through brightness distribution analysis and compensation technology; enhances the security and reliability of identity verification by combining multimodal fusion recognition of facial and voiceprint features; dynamically adjusts the decision threshold through an environmental quality scoring mechanism, making the identity verification process adaptable to different environmental conditions; and achieves precise delivery of medical information based on the identity verification results, establishing an intelligent connection between patients and medical resources. The entire process eliminates the need for patients to repeatedly present documents or manually enter information, significantly improving medical efficiency and experience, ensuring the security and accuracy of medical information, and realizing intelligent, convenient, and humanized medical services.

[0047] Drug interactions are identified from the electronic prescription information, a drug metabolism pathway network is constructed, overlapping sites of drug metabolism pathways are analyzed, the competition intensity of metabolic pathways is calculated, and prescription optimization suggestions are generated, including: Structural information of drugs is extracted from electronic prescription information. The structural information is decomposed into molecular features to obtain a structural feature set. Recursive matching is performed on the structural feature set in a drug database to extract drug target and metabolic site information. Obtain medical records, extract historical medication records, perform time-series mapping between the historical medication records and drug targets to obtain target action intensity, calculate drug metabolism overlap based on the overlap interval between the target action intensity and metabolic site information, and generate drug interaction relationships; A drug metabolism pathway network is constructed based on drug interaction relationships. The network is recursively traversed to obtain node connection data. Overlapping sites of drug metabolism pathways are extracted based on the node connection data and the degree of overlap in drug metabolism. Drug concentration data is extracted from electronic prescription information, and the drug concentration data is accumulated at overlapping sites of drug metabolism pathways to obtain the competition degree of overlapping sites. The competition intensity of metabolic pathways is calculated based on the competition degree of overlapping sites and drug interaction relationships. The competition intensity of the metabolic pathways is compared with the safety thresholds in clinical medication rules to generate drug dosage adjustment schemes as prescription optimization suggestions.

[0048] Electronic prescription information includes drug name, dosage, and administration method. The corresponding molecular structure data is retrieved from a drug database by searching the drug name. Drug molecular structures are typically stored in standardized formats such as SMILES or InChI. Cheminformatics tools are used to parse these formats, extracting core structural information such as the molecular skeleton, functional groups, and stereoconfiguration. The extracted structural information is then subjected to molecular feature decomposition, and molecular fingerprinting technology is used to convert the drug molecular structure into a feature vector form. A molecular fingerprint is a digital representation describing a molecular structure; commonly used types include MACCS keys, topological fingerprints, and ring system fingerprints. A specific algorithm maps the molecular structure to a fixed-length binary feature vector, forming a structural feature set.

[0049] Recursive matching is performed on drug databases based on the generated structural feature sets. The recursive matching process first searches the database for molecules with high structural similarity, then expands the search space layer by layer to explore molecules with similar structures but diverse functions. The matching algorithm uses the Tanimoto coefficient to calculate the similarity between structural feature sets, with a similarity threshold set to 0.7; molecules exceeding the threshold are considered related molecules. Information on the drug's target and metabolic sites is extracted through recursive matching. Targets are the biomolecules through which drugs exert their therapeutic effects, and metabolic sites are key sites for drug transformation in vivo. This information is derived from pharmacology databases and includes detailed data such as target type, affinity, and metabolic enzyme type.

[0050] Obtain the patient's medical records and extract historical medication records. These records include drug name, administration time, dosage, and treatment duration. Perform a time-series mapping between the historical medication records and the extracted drug targets to analyze the cumulative effect of drugs on the same target at different time points. The time-series mapping establishes a time-decay model based on drug half-life and target recovery time to calculate the intensity of action on each target at different time points. Target intensity represents the degree of influence of a drug on a specific target, ranging from 0 to 1, with higher values ​​indicating stronger effects. Calculate the drug metabolism overlap based on the overlap range between target intensity and metabolic site information. The overlap range is determined by comparing the distribution of metabolic sites of different drugs; the overlap represents the degree of competition among multiple drugs on the same metabolic pathway. Based on the metabolic overlap and drug action mechanisms, generate drug interaction relationships to describe possible synergistic or antagonistic effects between different drugs.

[0051] A drug metabolism pathway network was constructed based on drug interactions. This network starts with drug molecules, uses metabolic intermediates as intermediate nodes, terminates with final metabolites, and uses metabolic enzymes and transport proteins as edges, forming a directed graph structure. The network construction employed a path search algorithm, integrating known drug metabolism pathway information to generate a complete metabolic pathway graph. The constructed drug metabolism pathway network was recursively traversed using a depth-first search algorithm to explore all possible metabolic pathways, recording the connections between nodes. Node connection data included information such as node type, connection strength, and metabolic conversion rate. Based on the node connection data and the previously calculated drug metabolism overlap, overlapping sites in drug metabolism pathways were extracted. Overlapping sites are key nodes shared in the metabolism of multiple drugs and are often sites of potential drug interactions.

[0052] Drug concentration data, including dosage, dosing frequency, and bioavailability, is extracted from electronic prescription information. Pharmacokinetic models are used to predict blood drug concentration curves at different time points in vivo. The predicted drug concentration data are accumulated at the previously identified overlapping sites of metabolic pathways to obtain the competition degree of these overlapping sites. The cumulative calculation considers the relative concentration and binding capacity of each drug at the overlapping sites, and a competitive inhibition model is used to assess the competition of multiple drugs for the same metabolic site. The competition degree of overlapping sites reflects the intensity of competition among multiple drugs for shared metabolic resources; a higher value indicates more intense competition. Based on the competition degree of overlapping sites and drug interactions, the competition intensity of metabolic pathways is calculated. Competition intensity is an overall evaluation index that comprehensively considers the number of overlapping sites, the magnitude of competition, and the type of interaction, used to quantify the degree of metabolic interference that may result from the coexistence of multiple drugs.

[0053] The calculated metabolic pathway competition intensity was compared with safety thresholds in clinical medication guidelines. These safety thresholds, derived from clinical pharmacology research data and medication guidelines, reflect the acceptable range of drug metabolic interference. When the competition intensity exceeds the safety threshold, it indicates a potential risk of drug interactions, necessitating adjustments to the medication regimen. Based on the comparison results, drug dosage adjustment plans were generated as prescription optimization suggestions. These adjustments may include reducing the dosage of certain drugs, adjusting the dosing interval, or replacing drugs with alternatives that do not overlap with the metabolic pathway.

[0054] Prescription optimization suggestions are pushed to the prescribing physician through the hospital information system, and the physician decides whether to adjust the prescription based on the suggestions. If the physician confirms the adjustment, the modified prescription will be automatically updated in the patient's medical record. The patient receives the prescription information through the hospital's mobile application, and the medication cost is automatically calculated and a payment order is generated. After the patient confirms payment, the pharmacy receives the dispensing instruction and automatically prepares the medication. The patient verifies their identity at the pharmacy's dispensing window using their identification information, completing the medication collection process without presenting a paper prescription or payment receipt. The entire process achieves seamless integration from prescription generation, drug interaction analysis, prescription optimization to medication payment and collection, embodying the core concept of a contactless intelligent service model.

[0055] The proposed method for a seamless, intelligent payment system for medication within hospitals demonstrates significant technical advantages. Through drug structure information analysis and recursive matching, it achieves precise identification of drug targets and metabolic sites. By combining historical medication records with temporal mapping, a complete drug interaction network is constructed. Through the construction and recursive traversal of drug metabolism pathway networks, key overlapping sites in drug metabolism are accurately located. Based on drug concentration data and metabolic competition models, the intensity of drug interactions is quantitatively assessed. By comparing the intensity of metabolic pathway competition with clinical safety thresholds, scientifically sound prescription optimization suggestions are generated. This entire technical solution integrates complex drug interaction analysis into a seamless, intelligent payment system, improving prescription safety, reducing the risk of adverse reactions, optimizing the patient experience, and enhancing the efficiency of hospital pharmacy management. It provides strong support for precise drug management and precise medication use in hospitals.

[0056] like Figure 2 As shown, the electronic prescription information is modified according to the prescription optimization suggestions to form medication information. The payment amount of the medication information is verified through distributed consensus with the in-hospital medical account to generate a payment voucher containing a trusted timestamp, including: Based on the prescription optimization suggestions, modify the electronic prescription information, extract the drug specifications and dosage from the modified electronic prescription information to generate a drug dispensing matrix, match and optimize the drug dispensing matrix with the pharmacy's regional inventory information to obtain the drug delivery route, generate dispensing information based on the drug delivery route, and calculate the payment amount for the dispensing information. The system retrieves user medical insurance information from the in-hospital medical account, categorizes and matches the reimbursement amount in the user medical insurance information with the payment amount to generate a reimbursement plan, and performs distributed consensus verification on the payment amount based on the reimbursement plan to generate payment data. The transaction information of the payment data is grouped and encrypted according to the block structure to obtain block data. The block data is then verified by nodes to generate verification records. A trusted timestamp is generated based on the verification records. The trusted timestamp and the payment data are hashed together to generate a payment credential.

[0057] Prescription optimization suggestions are generated based on patient medication safety analysis, including adjustments to drug dosage, usage, or alternative regimens. Doctors modify the original prescription, adjusting dosages or replacing drugs, recording the changes, and generating revised electronic prescription information. Drug specifications and dosage information are extracted from the revised electronic prescription information. Drug specifications include identifiers such as drug name, model number, and manufacturer; dosage includes quantitative indicators such as single dose, frequency of use, and course of treatment. The extracted drug specifications and dosages are organized into a drug dispensing matrix. Rows in the matrix represent different drugs, and columns represent various attribute parameters of the drugs, including drug code, specifications, dosage, quantity, and pharmacy area code, forming a digital description of drug delivery.

[0058] Pharmacy zone inventory information is inventory management data categorized by pharmacy based on factors such as drug category and storage conditions. It includes parameters such as drug inventory quantity, location coordinates, and retrieval efficiency for each zone. The drug dispensing matrix is ​​matched and optimized with the pharmacy zone inventory information, and a multi-objective optimization algorithm is used to calculate the optimal delivery route. Optimization objectives include minimizing delivery distance, minimizing delivery time, and maximizing drug sorting accuracy. The optimization process considers constraints such as physical distance between pharmacy zones, drug retrieval efficiency, and inventory status, comprehensively evaluating the performance indicators of each feasible path, and selecting the path with the highest comprehensive score as the drug delivery route. Based on the determined drug delivery route, dispensing information is generated, including a drug list, retrieval order, and division of labor among dispensing personnel. The payment amount is calculated based on the dispensing information, taking into account factors such as drug unit price, quantity, and hospital pricing strategies to arrive at the final total drug cost.

[0059] The system retrieves user medical insurance information from the hospital's in-hospital medical account. This information includes parameters such as the patient's medical insurance type, reimbursement ratio, reimbursement amount, and co-payment ratio. The reimbursement amount in the user's medical insurance information is then matched with the payment amount. This matching process categorizes drugs into Class A, Class B, and Class C drugs according to the medical insurance policy's reimbursement rules for different categories, each with its own applicable reimbursement ratio. For each class of drug, the reimbursement amount and co-payment amount are calculated, generating a reimbursement plan with detailed reimbursement breakdowns. The reimbursement plan clearly lists the original price, reimbursement amount, co-payment amount, and total amount for each type of drug. The payment amount is then verified through distributed consensus based on the generated reimbursement plan. This verification process uses a multi-node collaborative verification mechanism within the medical insurance settlement network to ensure the accuracy and compliance of the reimbursement calculation. Each verification node independently calculates the reimbursement amount and verifies the consistency of the calculation results using a consensus algorithm. Only when more than a specified percentage of nodes reach consensus is the final payment data generated. The payment data includes information on the medical insurance reimbursement portion and the patient's co-payment portion, as well as key data such as the corresponding payment flow and account information.

[0060] Payment data transaction information is encrypted in blocks according to a block structure, which consists of a block header and a block body. The block header contains metadata such as the hash value, timestamp, and random number of the previous block, while the block body contains the specific transaction information. The transaction information is organized according to a predetermined data structure, including fields such as transaction identifier, transaction amount, payer, payee, and transaction time. Block encryption uses an asymmetric encryption algorithm, employing the patient's public key to encrypt the transaction information, ensuring data security and privacy. The encrypted data forms blocks and is stored in a distributed ledger. The blocks are submitted to a network of verification nodes within the hospital for verification. The verification nodes use a consensus algorithm to verify the validity, integrity, and legality of the blocks. The verification process checks key elements such as the correctness of the transaction information format, the validity of the signature, and the adequacy of the account balance. Upon successful verification, a verification record is generated, containing the verification result, the verification node identifier, and the verification time.

[0061] A trusted timestamp is generated based on the verification record. This trusted timestamp serves as digital proof of the existence of specific data at a specific point in time and is provided by a trusted timestamp service. The timestamp generation process first calculates the hash value of the verification record, then combines the hash value with the current time information, and digitally signs it using the private key of the timestamp authorization authority to form a trusted timestamp. The trusted timestamp comprises three parts: time information, the original data hash value, and the digital signature, ensuring the immutability of the transaction time. The generated trusted timestamp is then hashed and combined with the payment data. This hashing process uses a secure hash algorithm to concatenate the payment data and the timestamp, calculating the hash value to form a unique digital fingerprint. This hash value, along with the original payment data and the timestamp, constitutes a complete payment voucher. This payment voucher serves as electronic proof of the patient's completed medication payment and is used in the subsequent medication collection process, enabling patients to collect medications seamlessly at self-service kiosks or pharmacy windows within the hospital through identity verification.

[0062] Pharmacy staff dispense medications according to optimized delivery routes based on dispensing information, preparing the necessary medications and awaiting patient pickup. Patients receive a notification via the hospital's mobile application that their medication is ready and proceed to the pharmacy pickup area. In the pickup area, patients verify their identity using facial recognition, fingerprints, or their medical insurance card. Once verified, the system automatically retrieves the patient's payment voucher. The payment voucher's integrity and authenticity are ensured through hash value verification. After successful verification, the pharmacy's automated dispensing system or staff delivers the medication to the patient. The entire process eliminates the need for patients to present paper prescriptions or payment vouchers, achieving a seamless medication pickup experience. The system automatically updates patients' medication records and medical insurance reimbursement records, completing a closed-loop management process.

[0063] This invention improves the efficiency and accuracy of pharmacy dispensing by optimizing the drug dispensing matrix and delivery routes; it achieves intelligent processing of medical insurance reimbursement based on the classification and matching of medical insurance information and the generation of reimbursement schemes; it ensures the security and immutability of payment data by employing distributed consensus verification and block structure encryption; and it introduces a trusted timestamp and hash combination mechanism to ensure the uniqueness and verifiability of payment vouchers. The data storage and verification mechanism based on blockchain technology enhances the security and credibility of medical data, providing more transparent and efficient technical support for hospital drug management and medical insurance settlement, and promoting the innovative development of smart healthcare.

[0064] Verify the payment voucher to complete the payment, collect the target patient's real-time facial features for identity verification, and complete the medication collection after successful verification. Record the following medical information: The payment voucher is decrypted in layers to obtain block data and payment data. The verification records in the block data are calculated to obtain the verification weight. The payment data is verified according to the verification weight to generate a payment verification result. The payment is completed based on the payment verification result. Real-time facial images of the target patient are acquired, and the illumination distribution matrix and structural feature points of the real-time facial images are extracted. An environmental compensation coefficient is calculated based on the illumination distribution matrix. The structural feature points are compensated based on the environmental compensation coefficient to generate facial feature data. The facial feature data is combined with the verification weight to generate feature verification information. Biometric information is obtained from the hospital's medical account. The feature verification information is matched with the biometric information to generate an identity verification result. The medication dispensing authority is confirmed based on the identity verification result and the payment verification result. After the medication is dispensed, the payment verification result and the identity verification result are written into the hospital's medical account.

[0065] The payment credential employs a multi-layered encryption structure, comprising an outer encryption layer and an inner encryption layer. The process of decrypting the payment credential layer by layer requires using different decryption keys to sequentially remove each layer of encryption protection. The outer layer decryption uses the hospital's public key to retrieve the block data; the inner layer decryption uses the patient's private key to retrieve the payment data. The block data contains metadata such as block header information, verification records, and timestamps. The verification record records detailed information about the payment data's verification by the verification node network, including the identity of each verification node, the verification result, and the verification time. Verification weights are calculated from the verification records in the block data, and these weights are used to quantify the trustworthiness of each verification node using a weighted algorithm. The weighted algorithm considers factors such as the historical trustworthiness of the verification node, the freshness of the verification time, and verification consistency, assigning a different weight value to each node. The sum of the weight values ​​of all nodes constitutes the overall verification weight.

[0066] The payment data is verified based on the calculated verification weights. The verification process includes three aspects: data integrity verification, signature validity verification, and transaction logic verification. Data integrity verification calculates the hash value of the payment data and compares it with the hash value recorded in the block data to confirm that the data has not been tampered with during transmission. Signature validity verification checks whether the digital signature was generated using a valid private key to confirm the authenticity of the transaction initiator's identity. Transaction logic verification checks whether the payment amount is correct, the account balance is sufficient, and the transaction time is reasonable, among other business logic checks. Based on the combined verification results from these three aspects, a payment verification result is generated, including the verification status (pass / fail) and detailed verification information. Based on the payment verification result, the hospital payment system automatically completes the payment operation, transferring the patient's out-of-pocket portion from the patient's account to the hospital's account, and transferring the medical insurance reimbursement portion from the medical insurance account to the hospital's account, thus completing the settlement of drug costs.

[0067] The process of acquiring real-time facial images of target patients utilizes high-definition camera equipment within the hospital to capture facial images of patients in real time in the medication dispensing area. Image acquisition must consider factors such as lighting conditions, shooting angle, and image sharpness to ensure high-quality facial images. Illumination distribution matrices and structural feature points are extracted from the acquired real-time facial images. The illumination distribution matrix describes the changes in light intensity in different areas of the image. Image processing algorithms are used to divide the facial region into blocks, and the average brightness value of each block is calculated to form the illumination distribution matrix. Structural feature points are the set of coordinates of key facial locations, including the eyes, nose, mouth, and facial contours. The extraction process employs a deep learning model to perform feature detection on the input facial image and accurately locate the coordinates of key facial points.

[0068] An environmental compensation coefficient is calculated based on the extracted illumination distribution matrix. This coefficient is used to eliminate the impact of different illumination conditions on facial feature recognition. The calculation process compares the current illumination distribution matrix with the distribution matrix under standard illumination conditions, analyzes the illumination differences, and generates targeted compensation coefficients. These compensation coefficients include multiple dimensions such as brightness adjustment coefficients, contrast adjustment coefficients, and directional illumination compensation coefficients. Based on the calculated environmental compensation coefficients, structural feature points are compensated. The compensation process adjusts the position and weight of feature points to eliminate the interference of uneven illumination on feature point localization, making the feature points more accurately reflect facial structural information. The compensated feature point set, together with other facial feature parameters (such as texture features, geometric relationships, etc.), constitutes complete facial feature data. The facial feature data is then combined with the previously calculated verification weights. The combination method uses the verification weights as the credibility coefficients of the facial feature data, influencing the decision threshold in the subsequent identity matching process, generating feature verification information containing identity information and credibility assessment.

[0069] Biometric information is retrieved from the patient's in-hospital medical account. This biometric information is the identity verification data entered by the patient during hospital registration, including multimodal biometric data such as facial feature templates, fingerprint features, and iris features. Feature verification information is matched with the biometric information using a multi-level feature comparison strategy. First, a coarse match is performed to quickly filter possible matches; then, a fine match is performed to compare the features of candidate matches in detail; finally, a fusion decision is made, combining the matching results from all feature dimensions to generate the final identity verification result. The identity verification result includes verification status (pass / fail), matching score, and credibility assessment. Medication pickup permission is confirmed based on both the identity verification result and the payment verification result. Patients are only granted medication pickup permission when both payment verification and identity verification are successful.

[0070] After medication dispensing authorization is confirmed, the pharmacy automatically dispenses the medication or pharmacy staff prepares the corresponding medication based on the electronic prescription information and delivers it to the patient through the dispensing window or automated dispensing machine. Upon receiving the medication, the system automatically records the medication receipt status and writes this information, along with the payment verification result and identity verification result, into the patient's in-hospital medical account. The writing process employs a secure data exchange protocol to ensure data security during transmission and storage. After the in-hospital medical account is updated, the relevant information is synchronized to the hospital information system, medical insurance settlement system, and drug management system, completing a closed-loop data process. The updated account information includes the current medical visit record, medication information, payment record, and identity verification record, providing data support for subsequent medical services and drug supervision.

[0071] This invention improves the security and reliability of payment verification by using layered decryption and verification weight calculation of payment vouchers; it solves the accuracy problem of facial recognition under different environmental conditions by employing illumination distribution matrix and environmental compensation technology; it enhances the security and anti-counterfeiting capabilities of identity authentication based on multimodal biometrics; and it effectively prevents the risk of fraudulently obtaining medicines by integrating a dual confirmation mechanism for payment verification and identity verification. Through data encryption, blockchain notarization, and other technologies, it protects patient privacy and medical data security, providing medical institutions with an efficient and secure drug management solution.

[0072] A drug metabolism pathway network is constructed based on drug interaction relationships. The network is recursively traversed to obtain node connection data. Overlapping sites in the drug metabolism pathways are extracted based on the node connection data and the degree of overlap. Based on drug interaction relationships, the location and intensity of action points in the drug metabolism process are extracted. The locations of action points are constructed as network nodes. The connection weights between nodes are calculated based on the intensity of action. The connection relationships between the nodes are generated according to the connection weights between the nodes. The connection relationships are combined to form a drug metabolism pathway network. The nodes in the drug metabolism pathway network are recursively traversed, and the predecessor and successor nodes of each node are recorded to generate a node dependency sequence. The node activity is obtained by counting the node frequency in the node dependency sequence. The node activity is combined with the node connection weight to generate node connection data. The overlapping area between nodes is calculated based on the node connection data. The nodes within the overlapping area are matched with the drug metabolism overlap, and the nodes with a matching degree exceeding a preset matching degree threshold are extracted as drug metabolism pathway overlap sites.

[0073] Based on drug interactions, the sites of action and intensity of action in drug metabolism processes were extracted to construct a drug metabolism pathway network. Drug interaction data were obtained from drug instructions, clinical medication guidelines, and drug metabolism databases, and a structured dataset was formed through data integration and standardization. Sites of action refer to the specific locations in the human body where drug interactions occur, including specific enzyme systems, receptors, transport proteins, and other molecular targets. Text mining techniques were used to extract site of action locations, performing semantic analysis on drug interaction literature to identify and extract the names of molecular targets involved in the interactions and their corresponding tissue and organ information. Intensity of action refers to the degree of interaction between drugs, which can be divided into three levels: strong interaction, moderate interaction, and weak interaction. The quantification of intensity of action is based on clinical observation data and pharmacokinetic parameters, determined by comprehensively assessing factors such as the magnitude of changes in drug blood concentration caused by the interaction, the degree of increased risk of adverse reactions, and the necessity of clinical intervention.

[0074] The extracted sites of action are constructed as network nodes, each representing a site of action in the drug metabolism process. Node attributes include site name, affiliated tissue / organ, and a list of involved drugs. Connection weights between nodes are calculated based on interaction strength, considering multiple parameters such as interaction strength, probability of action, and clinical importance score. Strong interactions are assigned high weights, moderate interactions medium weights, and weak interactions low weights. The initial weights are adjusted according to the level of clinical evidence for the interaction; higher evidence levels result in larger adjustment coefficients. Connection relationships of the drug metabolism pathway are generated based on the calculated connection weights, describing the strength of metabolic associations between different sites of action. For node pairs with connection weights exceeding a set threshold, directed connections are established, with the connection direction representing the direction of metabolite flow or signal transduction, and the connection weight representing the association strength. All connections are combined to form a drug metabolism pathway network, represented by a directed weighted graph where nodes represent sites of action, edges represent metabolic associations, and edge weights represent association strength.

[0075] A recursive traversal of nodes in the drug metabolism pathway network is performed using a depth-first search algorithm, starting from the entry node and exploring the network structure layer by layer along the metabolic pathway. During the traversal, the predecessor and successor nodes of each node are recorded to generate a node dependency sequence. Predecessor nodes are the set of nodes upstream of the current node in the metabolic pathway, and successor nodes are the set of nodes downstream of the current node. The node dependency sequence is stored using a linked data structure, with each sequence element containing a node identifier and dependency type. Node activity is obtained by counting the frequency of nodes in the dependency sequence; node frequency is the total number of times a node appears in all dependency sequences. Node activity calculation considers both node frequency and the node's position weight in the sequence, assigning higher position weights to nodes that appear earlier in the sequence. Node activity is combined with the connection weights between nodes to generate node connection data. The combination method uses node activity as an importance indicator for nodes and connection weights as an importance indicator for edges; both together determine the critical paths in the metabolic pathway.

[0076] The overlapping region between nodes is calculated based on node connection data. This calculation is based on a community detection algorithm in graph theory, identifying tightly connected subsets of nodes in the network. The overlapping region refers to a set of shared action points across different metabolic pathways, representing metabolic processes jointly affected by multiple drugs. The determination of the overlapping region considers multiple factors such as node connection density, connection weight, and node activity. Cluster analysis is used to divide the network into multiple sub-communities, and the intersections between communities represent potential overlapping regions. Nodes within the overlapping region are matched with drug metabolism overlap, which refers to the degree to which a specific node participates in the metabolic processes of multiple drugs. The matching process compares the overlap between the drug list involved in a node and the drug list in the prescription, calculating the overlap coefficient. Nodes with a matching degree exceeding a preset matching degree threshold are extracted as drug metabolism pathway overlap sites. These overlap sites represent key metabolic processes involving multiple drugs and are potential drug interaction risk points.

[0077] In the hospital's entire drug delivery process, the identification of overlapping sites in drug metabolism pathways directly impacts the generation of optimized electronic prescriptions. When high-risk overlapping sites are identified, the system automatically analyzes drug combinations involving those sites and assesses the risk level of interactions. For drug combinations with high risk levels, the system generates specific prescription optimization suggestions, including adjusting dosing intervals, reducing drug dosages, and replacing them with alternative drugs that have the same efficacy but different metabolic pathways. These optimization suggestions are then pushed to doctors through the hospital information system, who modify their electronic prescriptions accordingly. The modified prescription information is then validated again through drug metabolism pathway analysis to ensure that the optimized drug combinations have significantly improved safety. The entire process requires no patient involvement, completing prescription optimization without the patient's awareness and improving medication safety.

[0078] The optimized electronic prescription enters the drug dispensing process. Based on drug metabolism pathway analysis results, the system adds special markers to drug combinations requiring special attention, reminding pharmacy staff to pay attention to them during drug delivery and patient medication guidance. Simultaneously, information on overlapping drug metabolism pathway sites is combined with the patient's personal medication history and physiological parameters to generate personalized medication guidance suggestions. These suggestions are delivered to the patient along with the medication through the hospital's pharmacy system. Upon picking up the medication, the patient verifies their identity via facial recognition, and the system automatically sends medication precautions. The payment process also incorporates drug metabolism pathway analysis results. For prescriptions containing high-risk drug combinations, the system displays risk warning information on the patient's payment confirmation interface to ensure informed consent. After payment, the drug interaction analysis results and medication recommendations are written into the patient's hospital account as an important part of their health record.

[0079] The application of drug metabolism pathway overlap site identification technology in the hospital's seamless intelligent service model enables intelligent management of drug safety throughout the entire process, from prescription issuance to drug payment, delivery, and dispensing. The system continuously collects clinical medication data and adverse reaction reports, dynamically updates the drug metabolism pathway network, and constantly optimizes the accuracy and sensitivity of the overlap site identification algorithm. Simultaneously, based on patient feedback and clinical observation data, the system's actual effectiveness is regularly evaluated, and preset matching thresholds and risk grading standards are adjusted to improve the system's practicality and reliability.

[0080] This invention, by constructing a drug metabolism pathway network and identifying overlapping sites, can accurately predict potential drug interaction risks, providing a scientific basis for prescription optimization. Comprehensive analysis based on node activity and connection weights improves the accuracy and comprehensiveness of risk identification. The application of recursive traversal and community detection algorithms enables the handling of complex multi-drug combination scenarios for interaction analysis. This enhances the accuracy of drug delivery and medication guidance, strengthens patient medication safety, and ensures a seamless, patient-unnoticed experience throughout the entire process, thus improving the patient's healthcare experience.

[0081] One technical solution provided in this embodiment of the invention is an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0082] One technical solution provided in this embodiment of the invention is a computer-readable storage medium storing a computer program, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0083] The specific embodiments described above are preferred embodiments of the present invention and are not intended to limit the specific scope of the present invention. The scope of the present invention includes, but is not limited to, these specific embodiments. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A single-line payment drug in-hospital whole-process method based on a non-sensing intelligent management mode, characterized in that, The method comprises the following steps: Collecting facial features of a target patient to form biometric information, generating an in-hospital medical account containing medical records; Receiving a description of the target patient's symptoms, extracting a symptom feature vector, mapping the symptom feature vector with the medical records, constructing a symptom evolution map, determining a diagnosis and treatment time point according to the symptom evolution map and the distribution of doctor resources, and generating guide diagnosis information; Obtaining an in-hospital medical instruction, collecting on-site facial features and biometric information for identity verification, and pushing intelligent guide diagnosis information to a doctor terminal after verification; Based on the intelligent guide diagnosis information, receiving electronic prescription information, identifying drug interaction relationships from the electronic prescription information, constructing a drug metabolism pathway network, analyzing overlapping sites of the drug metabolism pathway, calculating the competition intensity of the metabolism pathway, and generating prescription optimization suggestions; According to the prescription optimization suggestions, modify the electronic prescription information and form the dispensing information, distribute the payment amount of the dispensing information with the in-hospital medical account for distributed consensus verification, and generate a payment voucher containing a trusted timestamp; Verify the payment voucher to complete the payment, collect the real-time facial features of the target patient for identity comparison, and complete the medicine taking after comparison, and record the diagnosis and treatment information.

2. The method of claim 1, wherein, Collecting facial features of a target patient to form biometric information, generating an in-hospital medical account containing medical records comprises: Collecting the static facial features and voice information of the target patient, extracting the facial structure features to generate a static feature vector, and extracting the voiceprint features to generate a dynamic feature vector; Calculate the information redundancy of the static feature vector and the dynamic feature vector, sort the information redundancy in ascending order to obtain the difference degree, set a division threshold based on the difference degree to generate a plurality of complementary subintervals, and calculate the feature distribution density according to the aggregation degree of the features in the complementary subintervals; Multi-scale decomposition is performed on the feature distribution density to obtain the main feature direction, a feature span space is constructed based on the static feature vector and the dynamic feature vector, and the main feature direction is adaptively projected into the feature span space to obtain an initial basis vector; Orthogonalization processing is performed on the initial basis vector to generate an orthogonal projection matrix, and the static feature vector and the dynamic feature vector are projected into the orthogonal projection matrix to generate a fusion feature vector; Collecting the environmental parameters corresponding to the fusion feature vector to construct an environmental compensation matrix, and regularizing the fusion feature vector and the environmental compensation matrix to generate biometric information; Encrypt the biometric information to generate a tamper-resistant feature record, and write it into the in-hospital medical account to generate an in-hospital medical account containing medical records.

3. The method of claim 1, wherein, Receiving a description of the target patient's symptoms, extracting a symptom feature vector, mapping the symptom feature vector with the medical records, constructing a symptom evolution map, determining a diagnosis and treatment time point according to the symptom evolution map and the distribution of doctor resources, and generating guide diagnosis information comprises: Receiving the symptom description information of the target patient, extracting the symptom time sequence features to generate a symptom feature vector; Correlation analysis is performed on the symptom feature vector and the historical medical records to obtain a symptom correlation degree, and the symptom correlation degree is hierarchically clustered to obtain a symptom similarity; Based on the similarity of symptoms, the evolution rule of symptoms is extracted, the evolution rule of symptoms is constructed as a directed graph structure, the node represents the symptom, the edge represents the evolution correlation strength between adjacent symptoms, and the symptom evolution map is generated; Obtain the distribution of doctor resources, calculate the symptom conversion probability according to the symptom evolution map, and dynamically optimize and allocate the distribution of doctor resources based on the symptom conversion probability to obtain the weight of doctor resources; According to the evolution correlation strength, the priority of symptom treatment is calculated, the symptom treatment priority and the weight of doctor resources are combined, the optimal diagnosis and treatment time point is determined according to the time sequence distribution of the symptom conversion probability, and the guide diagnosis information is generated based on the optimal diagnosis and treatment time point.

4. The method of claim 1, wherein, Get the hospitalization instruction, collect the face feature and biological feature information on the spot for identity verification, and push the intelligent guide diagnosis information to the doctor terminal after the verification is passed, including: Get the hospitalization instruction, collect the face image of the target patient on the spot, extract the brightness distribution of the face image on the spot, compensate the face image on the spot based on the brightness distribution, and generate a compensation image; Extract the face feature of the compensation image to construct a face feature vector on the spot, collect the voice information of the target patient to extract the voiceprint feature to construct a voiceprint feature vector on the spot, and combine the face feature vector and the voiceprint feature vector on the spot to construct a feature matching sequence; The feature matching sequence and the biological feature information in the hospital medical account are dynamically programmed and matched to obtain the feature similarity, the distribution data of the feature similarity is calculated to determine the fusion weight, and the fusion weight and the feature similarity are combined to generate an identity verification similarity; Collect the environmental parameters on the spot to construct an environmental quality score, determine a decision threshold according to the environmental quality score, compare the identity verification similarity with the decision threshold to generate an identity verification result; Based on the identity verification result, the intelligent guide diagnosis information is pushed to the doctor terminal.

5. The method of claim 1, wherein, Identify the drug interaction relationship from the electronic prescription information, construct the drug metabolism pathway network, analyze the drug metabolism pathway overlap site, calculate the competition strength of the metabolism pathway, and generate the prescription optimization suggestion, including: Extract the structural information of the drug in the electronic prescription information, decompose the structural information into a structural feature set, recursively match the structural feature set in the drug database, extract the drug action target and metabolism site information; Get the medical record information, extract the history of drug use, time sequence map the history of drug use with the drug action target to obtain the target action strength, calculate the drug metabolism coincidence degree according to the overlap interval of the target action strength and the metabolism site information, and generate the drug interaction relationship; Based on the drug interaction relationship, the drug metabolism pathway network is constructed, the node connection data is obtained by recursively traversing the drug metabolism pathway network, and the drug metabolism pathway overlap site is extracted according to the node connection data and the drug metabolism coincidence degree; Extract the drug concentration data in the electronic prescription information, accumulate and calculate the drug concentration data at the drug metabolism pathway overlap site to obtain the overlap site competition degree, and calculate the competition strength of the metabolism pathway according to the overlap site competition degree and the drug interaction relationship; The metabolic pathway competition intensity is compared with a safety threshold in a clinical medication rule to generate a drug dosage adjustment scheme as a prescription optimization suggestion.

6. The method of claim 1, wherein, The electronic prescription information is modified according to the prescription optimization suggestion, and dispensing information is formed, the payment amount of the dispensing information is verified by distributed consensus with a hospital account, and a payment voucher including a trusted timestamp is generated including: The electronic prescription information is modified according to the prescription optimization suggestion, and the drug specifications and dosages in the modified electronic prescription information are extracted to generate a drug allocation matrix, the drug allocation matrix is matched and optimized with pharmacy partition inventory information to obtain a drug distribution path, and dispensing information is generated according to the drug distribution path, and the payment amount of the dispensing information is calculated; User medical insurance information is obtained from the hospital account, the reimbursement amount in the user medical insurance information is classified and matched with the payment amount to generate a reimbursement scheme, and the payment amount is verified by distributed consensus according to the reimbursement scheme to generate payment data; The transaction information of the payment data is grouped and encrypted according to the block structure to obtain block data, the block data is verified by nodes to generate verification records, the trusted timestamp is generated according to the verification records, and the payment voucher is generated by hashing the trusted timestamp and the payment data.

7. The method of claim 1, wherein, The payment voucher is decrypted to obtain block data and payment data, the verification records in the block data are calculated to obtain verification weights, the payment data is verified according to the verification weights to generate a payment verification result, and payment is completed based on the payment verification result; The payment voucher is decrypted to obtain block data and payment data, the verification records in the block data are calculated to obtain verification weights, the payment data is verified according to the verification weights to generate a payment verification result, and payment is completed based on the payment verification result; Real-time facial images of the target patient are collected, the illumination distribution matrix and structural feature points of the real-time facial images are extracted, the environmental compensation coefficient is calculated according to the illumination distribution matrix, the structural feature points are compensated based on the environmental compensation coefficient to generate facial feature data, and the facial feature data and the verification weights are combined to generate feature verification information; Biometric information is obtained from the hospital account, the feature verification information is matched with the biometric information to generate an identity verification result, the dispensing authority is confirmed according to the identity verification result and the payment verification result, and the payment verification result and the identity verification result are written into the hospital account after dispensing.

8. The method of claim 5, wherein, A drug metabolism pathway network is constructed based on drug interaction relationships, node connection data is obtained by recursively traversing the drug metabolism pathway network, and drug metabolism pathway overlapping sites are extracted according to the node connection data and drug metabolism overlap degree including: The action point position and action strength in the drug metabolism process are extracted according to the drug interaction relationship, the action point position is constructed as a network node, the connection weight between nodes is calculated based on the action strength, the connection relationship of the drug metabolism pathway is generated according to the connection weight between nodes, and the connection relationship is combined to form a drug metabolism pathway network; recursively traversing nodes in the drug metabolism pathway network, recording pre-order nodes and post-order nodes of each node to generate a node dependency sequence, counting node frequencies in the node dependency sequence to obtain node activity, and combining the node activity and inter-node connection weights to generate node connection data; calculating an overlapping area range between nodes according to the node connection data, matching nodes in the overlapping area range with drug metabolism coincidence degrees, and extracting nodes with a matching degree exceeding a preset matching degree threshold as drug metabolism pathway overlapping sites.

9. An electronic device, comprising: comprise: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps in the method according to any one of claims 1 to 8 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer program instructions, and the computer program instructions are executed by the processor to implement the steps in the method according to any one of claims 1 to 8.

Images