A method and device for assessing rabies exposure risk

Abstract

This invention discloses a method and apparatus for assessing rabies exposure risk. The method includes: generating a patient information set based on the patient's answers to questions about rabies exposure; analyzing the patient information set using a pre-trained question-and-answer analysis model, and acquiring the patient's wound visual content if the analysis result exceeds a pre-set threshold; assessing the rabies exposure risk based on the wound visual content and the patient information set, thereby enabling a more scientific and accurate rabies exposure risk assessment, providing corresponding treatment suggestions, and generating patient medical records for physician reference, thereby improving physician work efficiency and providing patients with safer, more accurate, and faster medical services.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medicine, in particular to a rabies exposure risk assessment method and device. BACKGROUND

[0002] Rabies is an animal-borne infectious disease caused by infection with a virus of the rabies virus genus, with central nervous system symptoms as the main symptom, and the mortality rate is almost 100%. There are about 59,000 rabies deaths worldwide each year, which is a global public health problem. China is still one of the countries with rabies epidemic. It is estimated that about 40 million people are bitten or scratched by dogs and cats each year in China, and 134 cases of rabies were reported in 16 provinces nationwide in 2022. The World Health Organization (WHO) believes that post-exposure prophylaxis (PEP) is the most effective strategy for preventing rabies, and timely and standardized PEP can almost 100% prevent the disease. PEP mainly includes three aspects: first, timely and standardized wound disposal; second, rabies vaccination; third, correct application of rabies passive immunization agents. Our country attaches great importance to standardizing PEP, and in 2023, the National Disease Control Bureau issued the "Rabies Exposure Prevention and Disposal Standard (2023 Edition)" in conjunction with the National Health and Health Bureau, guiding the national rabies prevention work.

[0003] However, in the process of implementing this important prevention strategy, there are some challenges, and some doctors have misunderstandings about rabies exposure risk assessment, which leads to the problem of non-standard PEP disposal. In addition, rabies exposure has obvious time distribution characteristics, especially in summer and autumn and in the afternoon to evening period of the day, which often leads to queuing at large medical centers during these periods, and patients have to wait for a long time, resulting in a poor medical experience.

[0004] Therefore, there is an urgent need for a method to assist doctors in assessing the risk of rabies exposure, improve the accuracy and efficiency of rabies prevention work, and focus on enhancing the capabilities of medical professionals and optimizing the medical process. SUMMARY

[0005] The present application aims to provide a rabies exposure risk assessment method and device, which can solve the problems of inaccurate exposure risk assessment, non-standard PEP disposal, low efficiency of medical treatment during peak hours, and poor patient experience in the prior art.

[0006] According to one aspect of the present invention, a method for assessing rabies exposure risk is provided, comprising: generating a patient information set based on the patient's answers to questions about rabies exposure; analyzing the patient information set using a pre-trained question-and-answer analysis model, and obtaining the patient's wound visual content if the analysis result is greater than a preset threshold; and assessing the rabies exposure risk based on the wound visual content and the patient information set.

[0007] Preferably, the patient information set is analyzed using a pre-trained question-answering analysis model according to the following formula: ( );in, For the results of this analysis, For this pre-trained question-answering analysis model, For this patient information set, each This represents the patient's answer to questions about rabies exposure.

[0008] Preferably, before analyzing the patient information set using the pre-trained question-answering analysis model, the method further includes: acquiring medical knowledge input text information and general large language model pre-training model information; extracting features from the pre-processed medical knowledge input text information, constructing a feature knowledge base, and integrating it into the large language model; generating new questions that the patient may ask, testing the medical diagnostic effect of the question-answering analysis model, and iteratively optimizing the question-answering analysis model based on the medical diagnostic effect.

[0009] Preferably, after acquiring the visual content of the patient's wound, the method further includes: processing the visual content of the wound according to a pre-trained visual analysis model to obtain a judgment result on the wound type, wherein the pre-trained visual analysis model includes a loss function. : ;in, It is the number of categories. It is a unique hot encoding of the real label. This is the model's predicted probability for the category; when the wound type is a puncture wound or a laceration, the model determines the wound size based on a pre-trained visual analysis model, which also includes a loss function. : ;in, This is the actual size of the wound. It is to predict the size of the wound. This refers to the number of samples; if the judgment result of the wound size does not meet the requirements, a suggestion to retake the photo will be given.

[0010] Preferably, when the wound size determination meets the requirements, the wound location determination is obtained based on a pre-trained visual analysis model; wherein, the pre-trained visual analysis model also includes a loss function. : ;in, Indicates the true label, The label indicates the prediction; if the judgment result of the wound location does not meet the requirements, a suggestion to retake the photo is given; if the judgment result of the wound location meets the requirements, the steps to assess the risk of rabies exposure are performed.

[0011] Preferably, after acquiring the visual content of the patient's wound, the method further includes: processing the visual content of the wound according to a pre-trained visual analysis model to obtain a judgment result on the wound type, wherein the pre-trained visual analysis model includes a loss function. ; ;in, It is the number of categories. It is a unique hot encoding of the real label. This represents the model's predicted probability for the category. When the wound type is a scratch or bite mark, the model uses a pre-trained visual analysis model to determine the bleeding status of the wound. This pre-trained visual analysis model also includes a loss function. : ;in, Indicates the true label, The label indicates the prediction; if the assessment of the bleeding condition of the wound does not meet the requirements, a suggestion to reshoot is given.

[0012] Preferably, when the assessment of the bleeding condition of the wound meets the requirements, the assessment of the wound location is obtained based on a pre-trained visual analysis model; wherein, the pre-trained visual analysis model also includes a loss function. : ;in, Indicates the true label, The label indicates the prediction; if the judgment result of the wound location does not meet the requirements, a suggestion to retake the photo is given; if the judgment result of the wound location meets the requirements, the steps to assess the risk of rabies exposure are performed.

[0013] Preferably, a judgment result on the presence or absence of biological residues is obtained based on a pre-trained visual analysis model; if biological residues are present, the type and amount of biological residues are determined. and perform steps to assess the risk of rabies exposure; the bioresidue. Calculate using the following formula: ;in The magnitude of the observed signal remaining on the surface by this organism. For error intensity, It is a proportionality constant. These are nonlinear coefficients. , , It is derived through training a neural network.

[0014] Preferably, the risk of rabies exposure is assessed according to the following formula: ;in, , , These are weighting coefficients. , , These represent functions for animal characteristics and injury status, wound severity, and patient immune status, respectively. Calculate using the following formula: ;in , , , , It is the weight of each factor; Calculate using the following formula: E= (T, L, D, B, R); where E is the injury score, T is the wound type, L is the wound location, D is the wound size, B is the bleeding level, and R is the amount of biological residue. Calculate using the following formula:

[0015] According to another aspect of the present invention, a rabies exposure risk assessment device is provided, comprising: an information generation module for generating a patient information set based on the patient's answers to questions about rabies exposure; a visual acquisition module for analyzing the patient information set using a pre-trained question-and-answer analysis model, and acquiring the patient's wound visual content if the analysis result is greater than a preset threshold; and a risk assessment module for assessing the rabies exposure risk based on the wound visual content and the patient information set.

[0016] This invention provides a method and apparatus for assessing rabies exposure risk. The method includes: generating a patient information set based on the patient's answers to questions about rabies exposure; analyzing the patient information set using a pre-trained question-and-answer analysis model, and acquiring the patient's wound visual content if the analysis result exceeds a pre-set threshold; assessing the rabies exposure risk based on the wound visual content and the patient information set, thereby enabling a more scientific and accurate rabies exposure risk assessment, providing corresponding treatment suggestions, and generating patient medical records for physician reference, thereby improving physician efficiency and providing patients with safer, more accurate, and faster medical services. Attached Figure Description

[0017] The accompanying drawings, which are provided to further illustrate the invention and constitute a part of this invention, are illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention.

[0018] Figure 1 This is a flowchart of a rabies exposure risk assessment method according to an embodiment of the present invention;

[0019] Figure 2 A flowchart illustrating the specific operation of the rabies exposure risk assessment method according to an embodiment of the present invention; and

[0020] Figure 3 This is a schematic diagram of a rabies exposure risk device according to an embodiment of the present invention. Detailed Implementation

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] Reference will now be made in detail to various embodiments of the invention, examples of which are shown in the accompanying drawings and described below. For ease of interpretation and precise definition in the appended claims, the terms “upper,” “lower,” “inner,” and “outer” are used to describe features with reference to their location in the exemplary embodiments shown in the figures.

[0023] This invention provides a method for assessing rabies exposure risk, such as... Figure 1 As shown, the process includes: generating a patient information set based on the patient's answers to questions about rabies exposure; analyzing the patient information set using a pre-trained question-and-answer analysis model, and obtaining the patient's wound visual content if the analysis result exceeds a pre-set threshold; and assessing the rabies exposure risk based on the wound visual content and the patient information set.

[0024] In related technologies, the treatment of PEP (Penalty Pulmonary Epilepsy) faces challenges due to uneven distribution of medical resources and varying levels of physician expertise, leading to non-standard treatment, especially during peak hours, resulting in low efficiency and a poor patient experience. This invention addresses these issues by automating the collection of patient information sets, analyzing them using pre-trained models, and acquiring visual wound information when necessary. This enables rapid and accurate risk assessment, improving physician efficiency, optimizing the patient experience, and providing a new, efficient, and accurate strategy for rabies prevention.

[0025] According to an embodiment of the present invention, the patient information set is analyzed using a pre-trained question-answering analysis model according to the following formula: ( );in, For the results of this analysis, For this pre-trained question-answering analysis model, For this patient information set, each This represents the patient's answer to questions about rabies exposure.

[0026] According to an embodiment of the present invention, before analyzing the patient information set using a pre-trained question-answering analysis model, the method further includes: acquiring medical knowledge input text information and general large language model pre-training model information; extracting features from the pre-processed medical knowledge input text information, constructing a feature knowledge base, and integrating it into the large language model; generating new questions that the patient may raise, testing the medical diagnostic effect of the question-answering analysis model, and iteratively optimizing the question-answering analysis model based on the medical diagnostic effect.

[0027] This invention improves the accuracy and efficiency of assessment by automating the collection of patient information sets and analyzing them using a pre-trained question-answering analysis model. The method also includes constructing a feature knowledge base and integrating it into a large language model to generate new questions that patients might ask, testing the medical diagnostic effectiveness of the question-answering analysis model, and iteratively optimizing it.

[0028] According to an embodiment of the present invention, after acquiring the visual content of the patient's wound, the method further includes: processing the visual content of the wound according to a pre-trained visual analysis model to obtain a judgment result on the wound type, wherein the pre-trained visual analysis model includes a loss function. : ;in, It is the number of categories. It is a unique hot encoding of the real label. This is the model's predicted probability for the category; when the wound type is a puncture wound or a laceration, the model determines the wound size based on a pre-trained visual analysis model, which also includes a loss function. : ;in, This is the actual size of the wound. It is to predict the size of the wound. This refers to the number of samples; if the judgment result of the wound size does not meet the requirements, a suggestion to retake the photo will be given.

[0029] In related technologies, wound types are generally not classified in PEP (Penetration, Excision, and Prophylaxis) treatment, resulting in inconsistent and untargeted treatment of different wound types. This invention classifies wound types into puncture wounds, lacerations, and scratches / bites. Specifically, it uses a pre-trained visual analysis model to process the acquired visual content of the wound to determine the wound type.

[0030] Furthermore, based on the assessment results, for puncture wounds or lacerations, the wound size is determined, and if it does not meet the requirements, a re-evaluation is recommended, thereby improving the accuracy of wound assessment and increasing the efficiency of the physician's work. Determining the size of puncture or laceration wounds is crucial, as these wounds are usually more serious, involving deep tissue damage, and carry a higher risk of rabies exposure. Puncture wounds are caused by sharp objects penetrating the skin and tissues; they generally have a small entry point but a deep depth, as seen in cat bite wounds. Lacerations are caused by external force tearing human tissue; the wound edges are irregular, tear-like, and the wound is deeper and wider, easily leading to bleeding and infection, as seen in dog bite wounds.

[0031] According to an embodiment of the present invention, when the determination result of the wound size meets the requirements, the determination result of the wound location is obtained based on a pre-trained visual analysis model; wherein, the pre-trained visual analysis model further includes a loss function. : ;in, Indicates the true label, The label indicates the prediction; if the judgment result of the wound location does not meet the requirements, a suggestion to retake the photo is given; if the judgment result of the wound location meets the requirements, the steps to assess the risk of rabies exposure are performed.

[0032] In this embodiment of the invention, when the wound size assessment meets the requirements, the model further determines the wound location to ensure accurate wound identification and positioning. If the wound location assessment does not meet the requirements, a re-enactment is prompted to obtain a more accurate wound image. This step ensures the accuracy of wound assessment and provides a reliable basis for subsequent rabies exposure risk assessment.

[0033] According to an embodiment of the present invention, after acquiring the visual content of the patient's wound, the method further includes: processing the visual content of the wound according to a pre-trained visual analysis model to obtain a judgment result on the wound type, wherein the pre-trained visual analysis model includes a loss function. ; ;in, It is the number of categories. It is a unique hot encoding of the real label. This represents the model's predicted probability for the category. When the wound type is a scratch or bite mark, the model uses a pre-trained visual analysis model to determine the bleeding status of the wound. This pre-trained visual analysis model also includes a loss function. : ;in, Indicates the true label, The label indicates the prediction; if the assessment of the bleeding condition of the wound does not meet the requirements, a suggestion to reshoot is given.

[0034] In related technologies, wound types are generally not classified in PEP (Penetration, Excision, and Prophylaxis) treatment, resulting in inconsistent and untargeted treatment of different wound types. This invention classifies wound types into puncture wounds, lacerations, and scratches / bites. Specifically, it uses a pre-trained visual analysis model to process the acquired visual content of the wound to determine the wound type.

[0035] Furthermore, based on the assessment results, for scratches and bite marks, the extent of bleeding is determined, and if the results are unsatisfactory, a re-enhancing of the wound is recommended. This improves the accuracy of wound assessment and increases the efficiency of the physician's work. Assessing the extent of bleeding in scratch or bite wounds is crucial, as it reflects the depth and severity of the wound and has significant clinical implications for rabies exposure risk stratification.

[0036] According to an embodiment of the present invention, when the judgment result of the bleeding condition of the wound meets the requirements, the judgment result of the wound location is obtained based on the pre-trained visual analysis model; wherein, the pre-trained visual analysis model further includes a loss function. : ;in, Indicates the true label, The label indicates the prediction; if the judgment result of the wound location does not meet the requirements, a suggestion to retake the photo is given; if the judgment result of the wound location meets the requirements, the steps to assess the risk of rabies exposure are performed.

[0037] In this embodiment of the invention, when the judgment result of the wound bleeding meets the requirements, the model will further determine the wound location, thereby ensuring the accuracy of wound assessment. For medical professionals, this can provide more accurate wound size and location information, improve medical efficiency, and ensure that patients receive timely and appropriate medical treatment.

[0038] According to an embodiment of the present invention, a judgment result on the presence or absence of biological residues is obtained based on a pre-trained visual analysis model; when biological residues are present, the type and amount of biological residues are determined. and perform steps to assess the risk of rabies exposure; the bioresidue. Calculate using the following formula: ;in The magnitude of the observed signal remaining on the surface by this organism. For error intensity, It is a proportionality constant. These are nonlinear coefficients. , , It is derived through training a neural network.

[0039] In related technologies, the treatment of wounds after rabies exposure lacks precise assessment of biological residues, especially when differentiating between puncture wounds, lacerations, scratches, and bite marks, leading to non-standard wound treatment. The embodiments of this invention can determine the presence or absence of biological residues. When biological residues are present, the type and amount R of the residues are determined, and steps to assess the risk of rabies exposure are performed.

[0040] According to an embodiment of the present invention, the risk of rabies exposure is assessed according to the following formula: ;in, , , These are weighting coefficients. , , These represent functions for animal characteristics and injury status, wound severity, and patient immune status, respectively. Calculate using the following formula: ;in , , , , It is the weight of each factor; Calculate using the following formula: E= (T, L, D, B, R); where E is the injury score, T is the wound type, L is the wound location, D is the wound size, B is the bleeding level, and R is the amount of biological residue. Calculate using the following formula:

[0041] According to another embodiment of the present invention, a rabies exposure risk assessment device is provided, such as... Figure 3 As shown, it includes: an information generation module, used to generate a patient information set based on the patient's answers to questions about rabies exposure; a visual acquisition module, used to analyze the patient information set using a pre-trained question-and-answer analysis model, and to acquire the patient's wound visual content if the analysis result is greater than a preset threshold; and a risk assessment module, used to assess the rabies exposure risk based on the wound visual content and the patient information set.

[0042] In related technologies, the treatment of PEP (Penalty Pulmonary Epilepsy) faces challenges due to uneven distribution of medical resources and varying levels of physician expertise, leading to non-standard treatment, especially during peak hours, resulting in low efficiency and a poor patient experience. This invention addresses this by automating the collection of patient information sets, analyzing them using pre-trained models, and acquiring visual wound information when necessary. This enables rapid and accurate risk assessment, improving physician efficiency, optimizing the patient experience, and providing a new, efficient, and accurate strategy for rabies prevention.

[0043] Figure 2 This diagram illustrates a detailed operational flowchart of the rabies exposure risk assessment method according to an embodiment of the present invention. Figure 2 As can be seen, the present invention mainly includes three parts: intelligent question answering system, intelligent recognition system, and intelligent decision-making system.

[0044] I. Intelligent Question Answering System

[0045] The intelligent question-answering system is an auxiliary medical question-answering system. This system uses artificial intelligence to build a knowledge base containing information related to rabies, diagnostic criteria, clinical manifestations and corresponding treatment suggestions, rabies prevention methods, etc. The knowledge base is input into the question-answering system as training data to participate in the training of the question-answering system. After the training is completed, the question-answering system will automatically generate questions for the intelligent question-answering system. After the patient provides feedback, the question-answering system will generate a preliminary judgment on the patient.

[0046] (1) The data acquisition method is as follows:

[0047] Medical dictionaries, manually reviewed medical information, medical records annotated by physicians, medical imaging materials, etc.;

[0048] (2) The training method is as follows:

[0049] Step 1: Obtain medical knowledge input text information and acquire information from the pre-trained general language model. Generally, the pre-trained general language model already contains the entire general knowledge base. Therefore, in this process, the input medical-specific information needs to be processed. The processing methods are: a. Obtain the input medical professional knowledge information; b. Preprocess the medical professional knowledge information, including question-and-answer settings, output generation, and diversification; c. Extract feature words and keywords and generate a relevant knowledge base.

[0050] Step 2: Extract features from the preprocessed patient input information, construct a feature knowledge base, and integrate it into the large language model. By fine-tuning the large language model using the preprocessed information, feature values ​​are constructed by combining the patient input data with the data within the large language model itself, generating a new large language model weight library.

[0051] Step 3: Test the training effect. Generate new questions that new patients might ask to test the medical diagnostic effectiveness of the large language model. Based on the test results, iteratively optimize the large language model.

[0052] (3) The method of use is as follows:

[0053] When a patient begins using the intelligent question-and-answer system, the system first generates a series of preliminary questions based on the medical knowledge questions entered by the patient. These questions aim to gather key information such as the reason for the patient's visit, the types of animals the patient was exposed to, the manner of exposure, the patient's health status, previous rabies vaccination history, and allergy history. As the patient responds, the system dynamically adjusts its questioning style using natural language processing technology to gain a deeper understanding of the patient's specific situation.

[0054] Assume the patient information set collected by the system is Each of them This represents the patient's answers to questions about rabies exposure. The system analyzes this information using a pre-trained model M to generate an exposure assessment score R, calculated as follows: ( )

[0055] This exposure assessment score R will be used to determine whether the patient is at risk of rabies exposure. The system sets a threshold T to distinguish between cases with no rabies exposure risk and cases requiring further investigation.

[0056] If R ≤ T, the system concludes that there is "no risk of rabies exposure." In this case, the system will advise the patient that no further PEP treatment is needed and submit the relevant information to the physician for review to ensure that no potential problems have been overlooked, thus completing the entire process.

[0057] When R ≤ T, the intelligent question-and-answer system will also pay special attention to whether the patient belongs to a high-risk group for rabies. This includes laboratory staff engaged in rabies research, individuals with prolonged contact with wild or stray animals due to their profession (such as veterinarians and animal rescuers), those residing in or traveling to areas with a high incidence of rabies, and staff who have contact with rabies patients. If the patient belongs to these high-risk groups, the system will proactively recommend pre-exposure prophylaxis (PrEP) to ensure the patient receives more adequate protection in potentially risky environments. Furthermore, the system can provide more detailed rabies-related scientific information based on the patient's needs, including the transmission routes of rabies, clinical manifestations, severity after onset, and the importance and methods of prevention. Simultaneously, the system will encourage patients to consult a professional physician when unsure of their situation to ensure the most appropriate medical advice is given and to avoid ignoring any potential health risks. Through these measures, the intelligent question-and-answer system not only provides scientific advice under specific conditions but also improves public awareness of rabies, comprehensively protecting patient safety.

[0058] If R>T, the system considers the patient's potential exposure cannot be ruled out and further diagnosis is required. In this case, the system will invoke the intelligent recognition system for more in-depth analysis and evaluation.

[0059] The intelligent recognition system includes more sophisticated professional image analysis to obtain more comprehensive risk assessment information, thereby providing more accurate medical advice.

[0060] In this way, the intelligent question-and-answer system can not only quickly determine whether a patient is at risk of rabies exposure, but also flexibly call other intelligent systems for in-depth analysis based on the complexity of the situation, ensuring that patients receive timely and accurate medical advice.

[0061] Once an exposure risk is identified, if a skin wound exists, the intelligent recognition system will be used to further identify the wound and determine its severity before proceeding to the intelligent decision-making system. If no skin wound exists, the patient will be directly referred to the intelligent decision-making system based on the patient's information from the intelligent question-and-answer system.

[0062] II. Intelligent Recognition System

[0063] The intelligent recognition system is a medical auxiliary recognition system based on images. The system collects the patient's image information through image sensors or information transmission interfaces. After collection, the system processes the image according to the neural network. After processing, it obtains detailed information about the patient's wound, including: (1) wound type (puncture wound, laceration, scratch or bite mark); (2) wound location (head and face, trunk, limbs); (3) wound size (length, width, depth); (4) wound bleeding status; and (5) biological residue.

[0064] 1. Wound type (T): A categorical variable, using numerical values ​​to represent different types (e.g., puncture wound = 1, laceration = 2, scratch / bite = 3).

[0065] 2. Wound location (L): Specifically, whether it includes the head and face (1: include, 0: exclude); whether it includes the trunk (1: include, 0: exclude); whether it includes the limbs (1: include, 0: exclude).

[0066] 3. Wound size (D): includes length, width, and depth, expressed in L. d W d H d (Unit: cm)

[0067] 4. Wound bleeding status (B): Whether there is obvious bleeding, a binary variable (1: bleeding, 0: no bleeding).

[0068] 5. Biological Residue (R): Is there any biological residue, such as teeth, hair, saliva, etc.? If biological residue is present, determine the type and amount of biological residue. :

[0069]

[0070] in The magnitude of the observed signal of the biological residue on the surface. For error intensity, It is a proportionality constant. These are nonlinear coefficients. , , It is derived through training a neural network.

[0071] Define a severity score E to quantify the severity of a wound. This scoring formula will consider all features and assign a weight to each feature:

[0072] E= The (T, L, D, B, R) injury score is determined by a neural network. Simultaneously, if the intelligent recognition system detects localized deformities, suspected fractures, or suspected foreign body residue in the wound, it will request further imaging examinations. The results of these imaging examinations are then input into the intelligent decision-making system.

[0073] The training method for neural networks is as follows:

[0074] Data collection: Due to the high privacy and sensitivity of this type of medical image data, it is different from traditional collection methods. It is necessary to provide a large amount of desensitized and professionally labeled medical image data. The following methods can be used for collection: (1) Cooperate with hospitals or medical institutions to obtain de-identified medical image data. De-identification means removing all information that can identify the patient; (2) Physician annotation: On the obtained images, the physician manually annotates the image information to be identified in different dimensions; (3) In addition, since it is necessary to obtain the size information of the image, it is also necessary to collect the camera parameters when the image was taken, such as aperture and focal length.

[0075] Designing the loss function: In intelligent recognition systems, to output information in multiple dimensions, different loss functions need to be designed for different dimensions, as described below:

[0076] For each wound type, the loss function is designed.

[0077]

[0078] Where C is the number of categories. It is a unique hot encoding of the real label. It is the model's predicted probability for the category.

[0079] For the wound location, the loss function is designed as follows:

[0080]

[0081] Where y represents the true label (1: include, 0: exclude). Labels indicating predictions.

[0082] For wound size information, the loss function is designed as mean squared error:

[0083]

[0084] in, This is the actual size of the wound. It predicts the wound size, where n is the number of samples.

[0085] (4) For bleeding wounds, the loss function is similar to that of the wound location.

[0086] Training the Neural Network: The main network structure of this type of recognition system uses a convolutional neural network as the backbone. Since it needs to detect both the location and size of the wound, target detection models (such as YOLO, SSD, Faster R-CNN, etc.) can be selected. These models can simultaneously predict the location and category of the target. The overall network is built by combining pre-processing and post-processing. Based on the diversity of the target data to be acquired, the neural network can be trained single-word or multiple times using a single-structure neural network, or it can be trained separately using neural networks with different structures. Different loss functions can be set for different training tasks, including mean squared error loss, cross-entropy loss, etc. Therefore, overall, the neural network structure is a single-input multiple-output (SMILE) or single-input multiple-output (SMILE) network. Simultaneously, pre-trained models with medical image information (such as VGG, ResNet, etc.) can be used, and fine-tuned on top of them to save training time and improve model performance.

[0087] Meanwhile, the input to the neural network can also be video. By adding timeline data, the video data can be analyzed to obtain specific target data.

[0088] How to use the system:

[0089] If the intelligent question-and-answer system needs to further assess the risk of rabies exposure and there is an unhealed skin wound, it will invoke the intelligent recognition system for processing. The processing flow of the intelligent recognition system is as follows: First, it acquires images. The image acquisition rules of the intelligent recognition system are: ① Take and upload at least one clear photo taken from approximately 15cm away from the wound, showing the entire wound; ② Apply appropriate force in a direction perpendicular to the long axis of the wound to open the wound to its maximum extent tolerable for the patient, then take and upload at least one clear photo as before. If a ruler is available, it can be placed before taking the photo (the system screen can display an illustration of the wound opening and ruler placement method).

[0090] After capturing images or videos, the acquired images are processed through multiple neural networks to obtain different attribute outputs from the neural networks, and these attributes are transmitted to the intelligent decision-making system for further decision-making.

[0091] III. Intelligent Decision-Making System

[0092] Intelligent decision-making systems are decision-making systems that take medical information as input. By acquiring information such as the characteristics of the animal that caused the injury, the circumstances of the injury (such as whether the animal actively attacked or whether one dog injured multiple people), the patient's immune status, and the severity of the wound, the intelligent decision-making system will provide two types of information after judgment: first, the patient's rabies exposure classification, and second, the patient's treatment suggestions, and generate medical records for doctors' reference.

[0093] How intelligent decision-making systems work:

[0094] The input to the intelligent decision-making system consists of three parts: (1) characteristics of the animal that caused the injury, (2) severity of the wound, and (3) the patient's immune status. The output of the intelligent decision-making system consists of two parts: (1) rabies exposure classification and (2) clinical treatment recommendations.

[0095] The risk assessment score S is a weighted sum of three main factors, each evaluated through a specific function. , , To indicate:

[0096]

[0097] Here, α, β, and γ are weighting coefficients, used to represent the proportion of each factor in the overall assessment. These weights can be determined through clinical data or expert opinions.

[0098] 1. Characteristics of the animal causing injury and the function of causing injury

[0099] The primary assessment should consider the likelihood of the injured animal being infected with rabies virus, taking into account the following factors:

[0100] The type of animal that caused the injury (e.g., wild / stray animal = 1, owned animal = 0).

[0101] Aggressive behavior (initiative aggression = 1, provoked aggression = 0)

[0102] Number of people injured by animals recently (value is 1 if ≥ 3 people, otherwise 0).

[0103] Laboratory testing of the animal that caused the injury (confirmed rabies = 1, no testing or rabies exclusion after testing = 0)

[0104] Animal behavior that causes injury (obvious abnormal behavior = 1, none = 0)

[0105] definition The formula is:

[0106]

[0107] Where c1, c2, c3, c4, and c5 are the weights of each factor.

[0108] 2. Wound severity function This item is the output E of the intelligent recognition system.

[0109] 3. Patient's immune status function

[0110] Assess the patient's immune status, especially their rabies immunity, including:

[0111] Rabies vaccination status (incomplete vaccination = 1, complete vaccination = 0)

[0112] Does the patient have severe immunodeficiency? (Yes = 1, No = 0)

[0113] definition The formula is:

[0114] =e1 × Rabies vaccination status + e2 × Immune function status

[0115] The risk assessment score S is based on a point system, with a total of 10 points. A higher score indicates a higher risk of rabies exposure. The threshold for defining the exposure level is derived from neural network training.

[0116] By using the exposure grading threshold provided by the neural network and the calculated risk assessment score, a final rabies exposure grading can be determined. Combining this rabies exposure grading with specific information obtained from an intelligent question-and-answer system, the intelligent decision-making system will provide specific treatment recommendations.

[0117] It should be noted that this system does not have prescription authority and cannot replace a doctor's diagnosis. However, as an auxiliary tool for doctors' diagnosis and treatment, it can provide treatment suggestions for doctors to refer to and selectively adopt.

[0118] If, after a physician's assessment, the system finds an error in its diagnostic conclusion, the relevant diagnostic data will be automatically added to the database as a bad case for subsequent training.

[0119] Example 1

[0120] The patient came to the hospital after being scratched on the right calf by a dog. The system works as follows:

[0121] I. Intelligent Question Answering System

[0122] The intelligent question-and-answer system automatically generates questions, and during the interaction between the patient and the system, it absorbs the following information:

[0123] The patient presented with a chief complaint of a dog scratch on his right calf 15 hours prior; the dog was his own pet; the attack was not unprovoked; the dog had only injured one other person recently; the dog had not been tested for rabies; the patient had washed the wound with water and disinfected it with alcohol after the injury; the patient had been previously healthy with no history of chronic diseases; one year prior, he was bitten on his right thumb by a dog, and received wound treatment, rabies immunoglobulin, and one dose of rabies vaccine at Hepingli Hospital, but did not receive subsequent vaccinations as prescribed; he had no history of allergies.

[0124] Based on the above information, the exposure assessment score R is calculated. Since R > T, the assessment is that the possibility of exposure cannot be ruled out, and further judgment is required.

[0125] II. Intelligent Recognition System

[0126] After taking a picture of the wound and inputting it into the intelligent recognition system, the following conclusions were drawn:

[0127] Wound type (T): Scratch and bite marks = 3

[0128] Wound location (L): Does it include the head and face: Excluded = 0, Does it include the trunk: Excluded = 0, Does it include the limbs: Included = 1

[0129] Wound dimensions (D): Length 7cm, Width 0.1cm, Depth 0.1cm

[0130] Oral bleeding status (B): No bleeding = 0

[0131] Biological residue status (R): No residue = 0

[0132] III. Intelligent Decision-Making System

[0133] The intelligent decision-making system calculated a risk assessment score S, classifying the patient as having a Category II rabies exposure. Simultaneously, based on the patient's information, the intelligent decision-making system provided the following treatment recommendations:

[0134] Diagnosis: Dog scratch

[0135] Management: 1. The patient has been exposed to rabies again (Level II). Since the previous rabies vaccination course was not completed, this case should be treated the same as the first Level II rabies exposure. 2. The wound was rinsed with professional equipment and disinfected with iodine. The patient was instructed to keep the wound clean and dry. The wound dressing was changed at the trauma emergency department 2 days later. 3. Rabies vaccine was administered using the 5-dose regimen or the 2-1-1 regimen. The patient was instructed to receive the subsequent doses of the vaccine according to the schedule. 4. Follow up if any discomfort occurs.

[0136] Example 2

[0137] The patient came to the hospital for treatment after being bitten by a cat. The system works as follows:

[0138] I. Intelligent Question Answering System

[0139] The intelligent question-and-answer system automatically generates questions, and during the interaction between the patient and the system, it absorbs the following information:

[0140] The patient presented with a chief complaint of being bitten on the right thumb by a cat one hour prior; the cat was a stray; the attack was not unprovoked; the cat had only injured one person recently; the cat had not been tested for rabies virus; the patient did not treat the injury on their own; the patient had a history of diabetes with poor blood sugar control; the patient had received a full course of rabies vaccination two years prior due to a dog bite; and the patient had no history of allergies.

[0141] Based on the above information, the exposure assessment score R is calculated. Since R > T, the assessment is that the possibility of exposure cannot be ruled out, and further judgment is required.

[0142] III. Intelligent Recognition System

[0143] After taking a picture of the wound and inputting it into the intelligent recognition system, the following conclusions were drawn:

[0144] Wound type (T): Puncture wound = 1

[0145] Wound location (L): Does it include the head and face: Excluded = 0, Does it include the trunk: Excluded = 0, Does it include the limbs: Included = 1

[0146] Wound dimensions (D): Length 0.2cm, Width 0.2cm, Depth 0.5cm

[0147] Bleeding status of the wound (B): Bleeding = 1

[0148] Biological residue status (R): No residue = 0

[0149] III. Intelligent Decision-Making System

[0150] The intelligent decision-making system calculated a risk assessment score S, classifying the patient as a Category III rabies exposure. Based on the patient's information, the system also provided the following treatment recommendations:

[0151] Diagnosis: Cat bite

[0152] Treatment: 1. Re-exposure to rabies (Level III); 2. Rinse the wound with professional irrigation equipment, disinfect with iodine, and instruct the patient to keep the wound clean and dry. Change the dressing at the trauma emergency department 2 days later; 3. Administer two booster doses of rabies vaccine using the 0-3 method, and instruct the patient to receive the subsequent doses according to the schedule; 4. Administer oral antibiotics to prevent wound infection; 5. Seek medical attention promptly if any discomfort occurs.

[0153] This invention integrates artificial intelligence technology to develop a phased automated processing flow for rabies prevention and treatment clinics, significantly improving physician efficiency and optimizing the patient experience. This invention enables more efficient and humanized medical services, and can provide more scientific medical advice based on wound type (such as puncture wounds or lacerations, scratches or bite marks), location, bleeding, and biological residue.

[0154] This invention's patented solution deeply considers the system's practicality and feasibility, going beyond a purely algorithmic level. Through the design of a deep neural network algorithm and its close integration with the hardware environment and medical practice scenarios, the comprehensiveness and practicality of the solution are ensured. Furthermore, this invention is specifically optimized for emergency treatment of rabies, enabling it to rapidly provide accurate medical advice and buy valuable treatment time for patients.

[0155] The above embodiments are merely examples to clearly illustrate the present invention and are not intended to limit the implementation of the invention. Those skilled in the art can make other variations or modifications based on the above description, and these variations, modifications, substitutions, and alterations arising from the principles and spirit of the present invention still fall within the protection scope of the present invention.

Claims

1. A method of rabies exposure risk assessment, characterized in that, The method comprises the following steps: generating a patient information set according to the answers of the patient to the rabies exposure questions; analyzing the patient information set using a pre-trained question and answer analysis model, and obtaining the visual content of the wound of the patient when the analysis result is greater than a pre-set threshold; According to the pre-trained visual analysis model, the wound visual content is processed to obtain a judgment result of the wound type, wherein the pre-trained visual analysis model comprises a loss function : ； wherein, is the number of classes, is a one-hot encoding of the true label, is the predicted probability of the class by the visual analysis model; When the wound type is a puncture wound or a laceration, a judgment result of the wound size is obtained according to a pre-trained visual analysis model, wherein the pre-trained visual analysis model further comprises a loss function : ； wherein, is the true wound size, is the predicted wound size, is the number of samples; when the judgment result of the wound size does not meet the requirements, a re-shooting suggestion is given; when the judgment result of the wound size meets the requirements, a judgment result of the wound position is obtained according to a pre-trained visual analysis model; The pre-trained visual analysis model further comprises a loss function : ; wherein, represents the true label, represents the predicted label; when the judgment result of the wound position does not meet the requirements, a re-shooting suggestion is given; when the judgment result of the wound position meets the requirements, the step of evaluating the rabies exposure risk is performed; when the wound type is scratch or bite, a judgment result of the wound bleeding condition is obtained according to a pre-trained visual analysis model; when the judgment result of the wound bleeding condition does not meet the requirements, a re-shooting suggestion is given; when the judgment result of the wound bleeding condition meets the requirements, a judgment result of the wound position is obtained according to a pre-trained visual analysis model; when the judgment result of the wound position does not meet the requirements, a re-shooting suggestion is given; when the judgment result of the wound position meets the requirements, the step of evaluating the rabies exposure risk is performed; According to the pre-trained visual analysis model, a judgment result of whether there is biological residue is obtained; when there is biological residue, the type and amount of biological residue are determined and performing the step of assessing the risk of exposure to rabies; the amount of biological residue is calculated according to the following formula: ； wherein is the observed signal size of the biological residue on the surface, is the error intensity, is the proportionality constant, is the non-linear coefficient, , , is derived by neural network training; the rabies exposure risk is evaluated according to the following formula: ； wherein, , , are weight coefficients, , , represent respectively the injury animal characteristics and injury situation function, the wound severity function, the patient immunity function; And wherein, This is calculated according to the following formula: ； wherein, for the animal species, wild / stray animal=1, owned animal=0, for the attack behavior, active attack=1, attack after provocation=0, for the number of injured people, if the number of injured people is greater than or equal to 3, the value is 1, otherwise the value is 0, for the rabies condition, detection of rabies=1, no detection or detection of no rabies=0, for the behavior abnormality, obvious behavior abnormality=1, no obvious behavior abnormality=0, , , , , are the weights of the respective factors; and wherein, E = 0.5 * T + 0.5 * L + 0.5 * D + 0.5 * B + 0.5 * R (T, L, D, B, R), where E is the injury score, T is the wound type, L is the wound location, D is the wound size, B is the wound bleeding, and R is the biological residue; And wherein, This is calculated according to the following formula: ； wherein, for the rabies vaccination, no full vaccination=1, full vaccination=0, for the immune function, severe immune dysfunction=1, no severe immune dysfunction=0, , e are weights for the respective factors.

2. The method of claim 1, wherein, the patient information set is analyzed using a pre-trained question and answer analysis model according to the following formula: ( )； wherein, is the analysis result, is the pre-trained question and answer analysis model, is the patient information set, and n is the number of answers in the patient information set.

3. The method of claim 2, wherein, before analyzing the patient information set using the pre-trained question and answer analysis model, the method further comprises the following steps: obtaining medical knowledge input text information and general large language model pre-training model information; extracting features from the pre-processed medical knowledge input text information, constructing a feature knowledge base, and integrating the large language model; generating new patient questions, testing the medical diagnosis effect of the question and answer analysis model, and iteratively optimizing the question and answer analysis model according to the medical diagnosis effect.

4. A rabies exposure risk assessment device, characterized in that, The method comprises the following steps: an information generation module is configured to generate a patient information set according to the answers of the patient to the rabies exposure questions; a visual acquisition module is configured to analyze the patient information set using a pre-trained question and answer analysis model, and obtain the visual content of the wound of the patient when the analysis result is greater than a pre-set threshold; The first judging module is configured to process the visual content of the wound according to a pre-trained visual analysis model to obtain a judgment result of the wound type, wherein the pre-trained visual analysis model comprises a loss function : ； wherein, is the number of classes, is a one-hot encoding of the true label, is the predicted probability of the class by the visual analysis model; When the wound type is a puncture wound or a laceration wound, a judgment result of the wound size is obtained according to a pre-trained visual analysis model, wherein the pre-trained visual analysis model further comprises a loss function : ； wherein, is the true wound size, is the predicted wound size, is the number of samples; when the judgment result of the wound size does not meet the requirements, a re-shooting suggestion is given; when the judgment result of the wound size meets the requirements, a judgment result of the wound position is obtained according to a pre-trained visual analysis model; The pre-trained visual analysis model further comprises a loss function : ; wherein, represents the true label, represents the predicted label; When the judgment result of the wound position does not meet the requirement, a suggestion of rephotographing is given; when the judgment result of the wound position meets the requirement, the step of evaluating the risk of rabies exposure is executed; When the wound type is scratch or bite, according to the pre-trained visual analysis model, a judgment result of the bleeding condition of the wound is obtained; when the judgment result of the bleeding condition of the wound does not meet the requirement, a suggestion of rephotographing is given; when the judgment result of the bleeding condition of the wound meets the requirement, according to the pre-trained visual analysis model, a judgment result of the wound position is obtained; when the judgment result of the wound position does not meet the requirement, a suggestion of rephotographing is given; when the judgment result of the wound position meets the requirement, the step of evaluating the risk of rabies exposure is executed; The second judging module is configured to obtain a judging result of whether there is biological residue according to a pre-trained visual analysis model; when there is biological residue, determine a type of the biological residue and a biological residue amount and perform the step of evaluating the risk of rabies exposure; the biological residue amount is calculated according to the following formula: ； wherein is the observed signal size of the biological residue on the surface, is the error intensity, is the proportionality constant, is the non-linear coefficient, , , derived by neural network training; The risk evaluation module is used for evaluating the risk of rabies exposure according to the following formula: ； wherein, , , are weight coefficients, , , represent respectively the injury animal characteristics and injury situation function, the wound severity function, the patient immunity function; And wherein, This is calculated according to the following formula: ； Wherein, For the animal species, wild / wandering animal = 1, and owned animal = 0, For the attack behavior, active attack = 1, and provoked attack = 0, For the number of injured people, if the number of injured people is greater than or equal to 3, the value is 1, otherwise, the value is 0, For the rabies condition, detection confirmation of rabies = 1, and no detection or detection exclusion of rabies = 0, , , , , are the weights of the respective factors; and wherein, E = 0.5 * T + 0.5 * L + 0.5 * D + 0.5 * B + 0.5 * R (T, L, D, B, R), where E is the injury score, T is the wound type, L is the wound location, D is the wound size, B is the wound bleeding, and R is the biological residue; And wherein, This is calculated according to the following formula: ； For the behavior abnormality, obvious behavior abnormality = 1, and no obvious behavior abnormality = 0, Wherein, For the rabies vaccination condition, no full vaccination = 1, and full vaccination = 0, For the immune function condition, serious immune function deficiency = 1, and no serious immune function deficiency = 0, , e is the weight of each factor.

Images