A method and system for early prediction of post-herpetic neuralgia

Abstract

The present application relates to the technical field of post-neuralgia risk prediction evaluation, and particularly relates to a method and system for early prediction of post-herpetic neuralgia. The present application first analyzes and obtains herpes characteristic coefficients of each historical patient, then compares the herpes position and neuralgia position of each historical patient at each follow-up visit, and the herpes pain grade and neuralgia grade, obtains neuralgia characteristic parameters at each follow-up visit, and screens reference post-neuralgia patients based on the time sequence changes thereof, and finally constructs a prediction model based on the time sequence changes of the neuralgia characteristic parameters of the reference post-neuralgia patients to perform subsequent early prediction. The present application combines the feature that the pain position of post-neuralgia is concentrated in the original herpes lesion position, analyzes the herpes outbreak performance and neuralgia changes of historical patients, and filters out part of the historical patients who are not post-herpes neuralgia, thereby improving the prediction model construction effect and the early prediction effect of post-herpes neuralgia of patients.

Description

Technical Field

[0001] This invention relates to the field of postherpetic neuralgia risk prediction and assessment technology, specifically to an early prediction method and system for postherpetic neuralgia. Background Technology

[0002] Herpes zoster is an acute viral skin disease caused by the varicella-zoster virus. Postherpetic neuralgia is one of the typical complications of herpes zoster. It refers to the lesions caused by the residual virus in the neurons of the affected area after the herpes zoster rash has healed, resulting in persistent and severe pain in the nerve area. Postherpetic neuralgia usually occurs within a few weeks to a few months after the rash has healed, and some patients may even suffer from it for a long time. Therefore, early prediction and timely intervention or treatment are crucial.

[0003] Existing methods construct predictive models for postherpetic neuralgia based on the acute symptoms and postherpetic neuralgia outcomes of a large number of historical herpes zoster patients, aiming to predict postherpetic neuralgia in current patients at an early stage. However, because postherpetic neuralgia primarily manifests as neuropathic pain without obvious skin lesions during its course, it is highly similar in symptoms to primary neuropathy such as sciatica. This leads to the easy misdiagnosis of primary neuropathy as postherpetic neuralgia, resulting in poor early prediction effectiveness for postherpetic neuralgia. Summary of the Invention

[0004] To address the technical problem of poor early prediction efficacy for postherpetic neuralgia, the present invention aims to provide a method and system for early prediction of postherpetic neuralgia, the specific technical solution of which is as follows: A method for early prediction of postherpetic neuralgia, the method comprising: Obtain basic information of all historical patients, herpes information during the herpes outbreak, and follow-up information at each follow-up visit after the herpes healed; the herpes information includes at least the herpes pain level, herpes symptom information, and herpes location, and the follow-up information includes at least the neuralgia level and neuralgia location; Based on the herpes symptom information and the basic information, the herpes characteristic coefficient of each historical patient is obtained; for each historical patient, the neuralgia characteristic parameters at each follow-up visit are obtained based on the difference between the location of the herpes and the location of the neuralgia at each follow-up visit, and the difference between the herpes pain level and the neuralgia level. Based on the temporal changes of the neuralgia characteristic parameters of each historical patient during follow-up visits, combined with the herpes zoster characteristic coefficient, reference postherpetic neuralgia patients are selected from all historical patients. Based on the temporal changes of the neuralgia characteristic parameters of all reference postherpetic neuralgia patients, a prediction model is constructed to make early predictions of postherpetic neuralgia for the current patient.

[0005] Furthermore, the information on herpes symptoms includes at least the density and area of ​​the blisters, and the basic information includes at least age.

[0006] Furthermore, the method for obtaining the herpes characteristic coefficient includes: For each level of herpes pain, based on the density and concentration characteristics of the herpes area of ​​all historical patients, the characteristic parameters of herpes symptoms for the corresponding level of herpes pain are obtained. Herpes risk weights are determined based on the age of each historical patient. The herpes symptom characteristic parameters under the herpes pain level of the historical patient are weighted using the herpes risk weights, and the weighted result is used as the herpes characteristic coefficient of the corresponding historical patient.

[0007] Furthermore, the method for obtaining the herpes symptom characteristic parameters includes: For each level of herpes pain, the density of herpes in all historical patients is averaged to obtain a density representative parameter, and the area of ​​herpes in all historical patients is averaged to obtain an area representative parameter; the weighted sum of the density representative parameter and the area representative parameter is used as a characteristic parameter of herpes symptoms.

[0008] Furthermore, the method for obtaining the neuralgia characteristic parameters includes: The location descriptions of the herpes location and the neuralgia location are extracted from the electronic medical records, and the word vectors of the corresponding description words are obtained based on the BERT model. For each historical patient, the cosine similarity between the word vectors of the neuralgia location at each follow-up visit and the herpes location during the herpes outbreak is used as the location association parameter. For each historical patient, the negative correlation mapping result between the difference between the neuropathic pain level at each follow-up visit and the herpes pain level during the herpes outbreak is used as the pain association parameter; By fusing the location-related parameters and the pain-related parameters, neuropathic pain characteristic parameters of the corresponding historical patients are obtained.

[0009] Furthermore, the method for obtaining the reference postneuropathy patient includes: Based on the temporal changes of the neuralgia characteristic parameters of each historical patient and the herpes characteristic coefficient, the confidence coefficient of postneuropathy is obtained, and historical patients whose postneuropathy confidence coefficient is greater than a preset threshold are identified as postneuropathy patients. Based on the differences in the confidence coefficients of postneuropathy among different postneuropathy patients, and the differences in the temporal changes of the corresponding neuropathic characteristic parameters during follow-up visits, reference postneuropathy patients were selected from all postneuropathy patients.

[0010] Furthermore, the method for obtaining the confidence coefficient of postneuropathy includes: Based on the time-series change curve of the neuralgia characteristic parameters of each historical patient at each follow-up visit, the postneural pain reference coefficient is determined based on the slope of the time-series change curve; the postneural pain reference coefficient is weighted using the herpes characteristic coefficient, and the normalized value of the weighted result is used as the postneural pain confidence coefficient.

[0011] Furthermore, based on the differences in the confidence coefficients of postneuropathy among different postneuropathy patients, and the temporal changes in the corresponding neuropathic characteristic parameters during follow-up visits, reference postneuropathy patients are selected from all postneuropathy patients, including: Among different patients with postneuropathy, the difference between the confidence coefficients corresponding to the postneuropathy and the DTW distance between the time-series change curves corresponding to the neuropathic feature parameters are fused to obtain a metric distance; all patients with postneuropathy are clustered based on the metric distance, and the clusters are sorted in descending order based on the cluster size; all patients with postneuropathy in the first preset number of clusters are used as reference patients with postneuropathy.

[0012] Furthermore, early prediction methods for postherpetic neuralgia include: Based on the time series of the herpes feature coefficients and the neuralgia feature parameters of each reference postneuropathy patient, a sample set is constructed, and a prediction model for postneuropathy is trained. The herpes feature coefficients of the current patient are input into the trained prediction model, and the time series of neuralgia feature parameters output by the model is used as the early prediction result of the current patient's postneuropathy.

[0013] An early prediction system for postherpetic neuralgia, the system comprising 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 of the early prediction method for postherpetic neuralgia.

[0014] The present invention has the following beneficial effects: This invention first acquires basic information, herpes information during the herpes outbreak, and follow-up information from each subsequent visit after the herpes heals, allowing for the analysis of herpes onset and postherpetic neuralgia in a large number of historical herpes zoster patients. This provides a historical reference for predicting postherpetic neuralgia in current patients. Then, based on herpes symptom information and basic information, it obtains a herpes characteristic coefficient reflecting the risk of postherpetic neuralgia in each historical patient. Furthermore, considering the characteristic that postherpetic neuralgia pain is concentrated at the original herpes lesion site, it analyzes the correlation between the herpes location and the neuralgia location at each follow-up visit for each historical patient. The differences between herpes zoster pain levels and neuralgia levels were analyzed to obtain neuralgia characteristic parameters at each follow-up visit. These neuralgia characteristic parameters reflect the association between neuralgia and herpes zoster to some extent, indirectly indicating the possibility of postherpetic neuralgia. Furthermore, based on the temporal changes in neuralgia characteristic parameters of each historical patient during follow-up visits, combined with herpes zoster characteristic coefficients, reference postherpetic neuralgia patients were selected from all historical patients. Based on the temporal changes in neuralgia characteristic parameters of all reference postherpetic neuralgia patients, a predictive model was constructed to predict postherpetic neuralgia in current patients at an early stage. This invention improves the predictive model construction effect and the early prediction effect of postherpetic neuralgia in patients by analyzing the herpes zoster manifestations and neuralgia changes in historical patients, and by combining the characteristic that the pain location of postherpetic neuralgia is concentrated at the original herpes zoster lesion site, filtering out some historical patients who do not have postherpetic neuralgia. Attached Figure Description

[0015] Figure 1 A flowchart illustrating an early prediction method for postherpetic neuralgia provided in one embodiment of the present invention; Figure 2 This is a flowchart illustrating a method for obtaining patients with post-neuralgia as provided in one embodiment of the present invention. Detailed Implementation

[0016] The following description, in conjunction with the accompanying drawings, details the specific scheme of the early prediction method and system for postherpetic neuralgia provided by this invention.

[0017] Please see Figure 1 The diagram illustrates a flowchart of an early prediction method for postherpetic neuralgia according to an embodiment of the present invention, specifically including: Step S1: Obtain basic information of all historical patients, herpes information during the outbreak, and follow-up information at each follow-up visit after the herpes heals; the herpes information includes at least the herpes pain level, herpes symptom information, and herpes location, and the follow-up information includes at least the neuralgia level and neuralgia location.

[0018] In one embodiment of the present invention, the electronic medical records of all patients with a history of shingles are first retrieved from the existing hospital information management system. The basic information of the patients, the information of herpes during the outbreak, and the follow-up visit information after the herpes heals are obtained. The herpes information includes at least the level of herpes pain, the information of herpes symptoms, and the location of the herpes. The follow-up visit information includes at least the level of neuralgia and the location of neuralgia. This allows for the analysis of the herpes onset information and postherpetic neuralgia of a large number of patients with a history of shingles, providing a historical reference for predicting the postherpetic neuralgia of current patients.

[0019] It should be noted that, in order to ensure the reliability of historical data, implementers also need to select historical patients who have regular follow-up visits and complete medical records for subsequent analysis.

[0020] In a preferred embodiment of the present invention, considering that the larger and denser the herpes lesions, the more severe the symptoms and the greater the likelihood of postherpetic neuralgia; and considering that older people often have some neurodegenerative changes, decreased immune system function, or chronic diseases, which may increase the risk of postherpetic neuralgia; based on this, the herpes symptom information includes at least the density and area of ​​the herpes lesions, and the basic information includes at least age; implementers may also add or adjust these themselves.

[0021] Specifically, in the electronic medical records of each patient with a history of shingles, the patient's age is extracted from the patient's basic information field; the herpes information text during the shingles outbreak is extracted from the symptom information field, which includes the patient's complaints of herpes pain and the herpes symptom information recorded by the doctor after consultation; and the follow-up visit record text after the herpes healed is extracted from the follow-up visit record field, which includes the time of each follow-up visit and the follow-up visit record. This is already an existing technical method, and the specific process will not be described in detail. Then, natural language processing tools are used to extract word segments describing the pain sensation of herpes, the distribution of herpes, the density of herpes, and the area of ​​herpes from the herpes information text, so as to determine the level of herpes pain, the density of herpes, the area of ​​herpes, and the location of herpes. This is a well-known technology and will not be elaborated further. Among them, the pain sensation segmentation of herpes may include any one of the following: mild stinging, mild stinging with burning sensation, moderate stinging, moderate stinging with burning sensation, severe stinging, and severe stinging with burning sensation. The pain sensation segmentation of herpes is divided into six levels of herpes pain in the above order. The more severe the pain sensation, the higher the level. The description of the distribution location of herpes may include left waist, right waist, left buttock, right buttock, left face, right face, left chest and right chest, etc. The density of herpes lesions can be described by terms such as scattered, sparse, clustered, dense, and very tight. These terms are then assigned parameter labels in the order listed above, such as 1-5, to represent the density of the herpes lesions. The denser the herpes lesions, the larger the parameter label. The area of ​​herpes lesions can be described by terms such as percentage of the area, for example, herpes lesions covering 10% of the body surface area. It should be noted that electronic medical records are recorded by doctors, so the text of the medical records of patients with different histories has a certain degree of standardization; however, there may still be slight differences. Implementers can preprocess the text descriptions in the medical record information. For example, if the medical record uses other range descriptions such as the size of a palm to represent the area of ​​herpes, these can be uniformly converted into the percentage of the area of ​​herpes to the body surface area, and the dimensions of the herpes area can be removed for subsequent analysis and calculation.

[0022] In other embodiments of the present invention, the implementer may also extract herpes images from electronic medical records, and then automatically mark the area and density of herpes based on a pre-trained recognition model. Furthermore, the implementer may also set a preset dictionary to perform emotional recognition on the subjective pain sensation, and then classify the pain sensation. The training and application of the recognition model are well-known technical means and will not be described in detail here.

[0023] Meanwhile, using the same methods described above, the patient's description of neuralgia sensation and neuralgia distribution location were extracted from the follow-up visit records for each visit. The neuralgia sensation description was then graded, and six neuralgia levels were determined, which will not be elaborated further. Among them, neuralgia may be postherpetic neuralgia, sciatica, or other types of neuralgia. The neuralgia distribution location description also includes left waist, right waist, left buttock, right buttock, left face, right face, back, left chest, and right chest.

[0024] Step S2: Based on the herpes symptom information and basic information, obtain the herpes characteristic coefficient for each historical patient; for each historical patient, based on the difference between the location of the herpes and the location of the neuralgia at each follow-up visit, and the difference between the herpes pain level and the neuralgia level, obtain the neuralgia characteristic parameters at each follow-up visit.

[0025] Considering that the risk and changes in postherpetic neuralgia may vary among patients with different degrees of herpes history, and that the herpes symptom information and basic information of patients with a history of herpes can help assess the severity of herpes and the risk of postherpetic neuralgia to some extent, this embodiment of the invention first obtains the herpes characteristic coefficient for each patient based on the herpes symptom information and basic information. The herpes characteristic coefficient reflects the severity of herpes and the risk of postherpetic neuralgia in patients with a history of herpes.

[0026] Preferably, in one embodiment of the present invention, considering that the pain level of herpes also reflects the severity of herpes to some extent, the concentrated characteristics of the area and density of herpes in historical patients under each pain level can characterize the general situation of herpes symptoms. The more severe the herpes symptoms and the older the historical patient, the greater the possibility of subsequent neuralgia. Therefore, the method for obtaining the herpes characteristic coefficient includes: For each level of herpes pain, based on the density and area of ​​herpes in all historical patients, characteristic parameters of herpes symptoms at the corresponding level of herpes pain were obtained. Herpes risk weights are determined based on the age of each historical patient. Herpes symptom characteristic parameters under the herpes pain level of the historical patient are weighted using the herpes risk weights, and the weighted result is used as the herpes characteristic coefficient of the corresponding historical patient.

[0027] As an example, the general characteristics of herpes symptoms are first comprehensively assessed at each level of herpes pain to determine the characteristic parameters of herpes symptoms. These characteristic parameters reflect the severity of symptoms from the perspective of herpes manifestations and indirectly provide parameters for the risk of postherpetic neuralgia. In a preferred embodiment of the present invention, the method for obtaining herpes symptom characteristic parameters includes: For each level of herpes pain, the density of herpes in all historical patients is averaged to obtain a density representative parameter, and the area of ​​herpes in all historical patients is averaged to obtain an area representative parameter; the weighted sum of the density representative parameter and the area representative parameter is used as a characteristic parameter of herpes symptoms. The dense representative parameter and the area representative parameter are weighted with 0.6 and 0.4 respectively and then summed to obtain the herpes symptom characteristic parameters; the implementer can also adjust the weights themselves; then the age of the historical patients is mapped to the sigmoid function, and the mapping result is used as the herpes risk weight; then the herpes risk weight is multiplied by the herpes symptom characteristic parameters of the herpes pain level of the historical patients, and the product is used as the herpes characteristic coefficient of the corresponding historical patients.

[0028] Considering that postherpetic neuralgia is a sequela after the healing period of herpes, when the patient's immunity is low, the remaining virus is reactivated. At this time, although the primary lesion of herpes will not produce new herpes symptoms, the patient will continue to feel obvious burning pain. That is, the location of postherpetic neuralgia is mainly concentrated at the location of the primary lesion of herpes, and the degree of pain of postherpetic neuralgia is closely related to the severity of herpes. The more severe the herpes and the lower the patient's immunity, the higher the risk and degree of pain of postherpetic neuralgia. Furthermore, considering that for each patient, the neuralgia they complain of during their follow-up visit after herpes recovery may not be postherpetic neuralgia, but rather neuralgia caused by other conditions such as lumbar disc herniation, which may interfere with the diagnosis and reduce the predictive value of neuralgia provided by follow-up visits; while the location of the herpes and the pain at the time of follow-up visit can provide some reference information for assessing whether it is postherpetic neuralgia. Based on this, the embodiments of the present invention will obtain the neuralgia characteristic parameters of each patient at each follow-up visit based on the differences between the location of herpes and the location of neuralgia, as well as the differences between the level of herpes pain and the level of neuralgia; the neuralgia characteristic parameters preliminarily reflect the possibility that the neuralgia is postherpetic neuralgia.

[0029] Preferably, in one embodiment of the present invention, considering that the herpes location and neuralgia location extracted from the electronic medical record are text descriptions and segmented words, direct comparison is difficult to quantify the location differences, while the BERT model can be used to perform text similarity calculation, thereby helping to assess the location association between herpes and neuralgia; furthermore, considering that pain level is determined based on pain perception, the herpes pain level can reflect the risk of nerve damage or post-neuralgia to a certain extent, and comparing pain levels can help assess the association between herpes and neuralgia to a certain extent; therefore, the method for obtaining neuralgia feature parameters includes: The location descriptions of herpes location and neuralgia location are extracted from electronic medical records, and word vectors of the corresponding description words are obtained based on the BERT model. For each historical patient, the cosine similarity between the word vectors of the neuralgia location at each follow-up visit and the herpes location during the herpes outbreak is used as the location association parameter. For each patient with a history of illness, the negative correlation between the difference between the neuropathic pain level at each follow-up visit and the herpes pain level during the outbreak of herpes was used as the pain association parameter. By fusing location-related parameters and pain-related parameters, neuropathic pain characteristic parameters of the corresponding historical patients are obtained.

[0030] As an example, for each historical patient, the word vectors of the descriptive words corresponding to the location of herpes and the location of neuralgia at each follow-up visit are obtained based on the BERT model. Then, the cosine similarity between the two word vectors is used as the location association parameter. The larger the cosine similarity, the higher the consistency between the location of neuralgia at each follow-up visit and the location of herpes during the outbreak of herpes. Then, the difference is measured by the absolute value of the difference. The absolute value of the difference between the neuralgia level at each follow-up visit and the herpes pain level during the outbreak is mapped to the function exp(-x) with the natural constant e as the base to obtain the pain association parameter. The larger the difference in pain level, the less likely the neuralgia is caused by the postherpetic neuralgia virus. The smaller the association, the smaller the pain association parameter is obtained by adjusting the logical relationship through the negative exponential function. Finally, the location-related parameters and pain-related parameters were assigned weights of 0.7 and 0.3 respectively and then summed using a weighted average. The implementer can also adjust the weights themselves. The weighted sum was then linearly normalized to obtain the neuropathic pain characteristic parameters of the corresponding historical patients. The larger the neuropathic pain characteristic parameters, the stronger the association between neuropathic pain and herpes, and the greater the likelihood of postneuropathy.

[0031] It should be noted that linear normalization can specifically employ max-min normalization, which is performed across all historical patient dimensions. This is a well-known technique and will not be elaborated upon further.

[0032] Step S3: Based on the temporal changes of neuropathic pain characteristic parameters of each historical patient during the follow-up visit, combined with the herpes characteristic coefficient, reference postherpetic neuralgia patients are selected from all historical patients. Based on the temporal changes of neuropathic pain characteristic parameters of all reference postherpetic neuralgia patients, a prediction model is constructed to make early predictions of postherpetic neuralgia for the current patient.

[0033] After obtaining the neuropathic pain characteristic parameters of each historical patient, historical patients with larger neuropathic pain characteristic parameters can be identified as herpetic neuralgia patients to provide a reference for the changes in postneurotic pain during follow-up visits, thus preparing for subsequent prediction. Considering that the characteristic parameters of neuralgia reflect the association between neuralgia and herpes to a certain extent, and indirectly reflect the possibility that it is postneuralgia, there is still the possibility that neuralgia caused by lumbar disc herniation may overlap with the location of herpes lesions, that is, neuralgia may be mistaken for herpetic neuralgia. Furthermore, considering that the pain of postherpetic neuralgia is mainly distributed at the site of the herpes lesions, while the pain intensity of other neuralgia, such as sciatica, usually changes gradually with the degree of nerve compression and stimulation, and may spread, for example, starting from the waist or lower back and radiating down the buttocks, changes in the intensity and location of pain will lead to changes in the characteristic parameters of neuralgia. Among them, although the pain intensity of postherpetic neuralgia also changes with treatment, the location of pain remains unchanged, so the changes in the characteristic parameters of neuralgia are relatively smaller than those of sciatica. Furthermore, considering that the herpes characteristic coefficient reflects the risk of postherpetic neuralgia to a certain extent, the smaller the temporal variation of the neuralgia characteristic parameters of a historical patient and the larger their herpes characteristic coefficient, the greater the likelihood that the neuralgia they have is postherpetic neuralgia. Based on this, this embodiment of the invention selects reference postherpetic neuralgia patients from all historical patients based on the temporal variation of the neuralgia characteristic parameters of each historical patient during the follow-up visit, combined with the herpes characteristic coefficient, so as to provide a more accurate historical reference for subsequent prediction.

[0034] Preferably, in one embodiment of the present invention, the method for obtaining a reference postneuropathy patient includes: Please see Figure 2 The diagram illustrates a flowchart of a method for obtaining a patient with post-neuralgia according to an embodiment of the present invention, specifically including: Step S301: Based on the temporal changes of neuropathic pain characteristic parameters and herpes characteristic coefficients of each historical patient, obtain the postneural pain confidence coefficient, and identify historical patients whose postneural pain confidence coefficient is greater than a preset threshold as postneural pain patients.

[0035] Since the herpes characteristic coefficient reflects the risk of postherpetic neuralgia to a certain extent, the larger the herpes characteristic coefficient of a historical patient and the smaller the temporal variation of its neuralgia characteristic parameters, the greater the likelihood that the neuralgia is postherpetic neuralgia. Therefore, the confidence coefficient of postherpetic neuralgia for the corresponding historical patient can be obtained. The confidence coefficient of postherpetic neuralgia further quantifies the likelihood of it as postherpetic neuralgia from the perspective of the dynamic change of neuralgia, so as to screen patients with postherpetic neuralgia.

[0036] In a preferred embodiment of the present invention, the method for obtaining the confidence coefficient of postneuropathy includes: Based on the time-series change curves of neuropathic pain characteristic parameters fitted for each historical patient at each follow-up visit, the post-neuralgia reference coefficient is determined based on the slope of the time-series change curves; the post-neuralgia reference coefficient is weighted using the herpes characteristic coefficients, and the normalized value of the weighted result is used as the post-neuralgia confidence coefficient.

[0037] As an example, we first fit the time-series change curve of the neuropathic pain characteristic parameters of each historical patient at each follow-up visit based on the least squares method. The greater the downward trend of the time-series change curve, the greater the trend of the neuropathic pain characteristic parameters decreasing. The more likely the pain intensity of the neuropathic pain is to spread, and the more likely it is to be other types of neuropathic pain such as sciatica. Then, using the first and last two points of the time-series change curve, the slope of the time-series change curve is calculated based on the two-point formula. The slope is then mapped to the exponential function exp(x) with the natural constant e as the base and the range is adjusted to obtain the postneural pain reference coefficient. Then, the herpes characteristic coefficient is multiplied by the postneural pain reference coefficient, and the product is linearly normalized, such as by maximum and minimum normalization, to obtain the postneural pain confidence coefficient of the corresponding historical patient. Finally, the preset threshold is set to 0.4, which can also be adjusted by the implementer. The higher the confidence coefficient of postneuropathy, the greater the possibility of postneuropathy. Therefore, all historical patients with a confidence coefficient of postneuropathy greater than the preset threshold can be regarded as postneuropathy patients.

[0038] Step S302: Based on the differences in confidence coefficients of postneuropathy among different postneuropathy patients and the temporal changes in corresponding neuropathic characteristic parameters during follow-up visits, reference postneuropathy patients are selected from all postneuropathy patients.

[0039] Considering that the manifestations of postneuropathy may vary among patients, the postneuropathy manifestations of some patients selected in step S301 may deviate from the general characteristics, which may affect the construction effect of the subsequent prediction model. Clustering can group patients with similar postneuropathy changes into a cluster, and the size of the cluster can be used to characterize the general characteristics of postneuropathy changes. Based on this, in a preferred embodiment of the present invention, among different postneuropathy patients, the difference between the corresponding postneuropathy confidence coefficients and the DTW distance between the time-series change curves of the corresponding neuropathic characteristic parameters are fused to obtain a metric distance; all postneuropathy patients are clustered based on the metric distance, and the clusters are sorted in descending order based on the cluster size; all postneuropathy patients in the first preset number of clusters are used as reference postneuropathy patients.

[0040] As an example, the difference between different patients with postneuropathy is first measured by the absolute value of the difference. The absolute value of the difference between the confidence coefficients of the corresponding postneuropathy is used as the first distance. The larger the first distance, the greater the difference in the probability that the two are patients with postneuropathy. This can provide a certain similarity reference for clustering to some extent. Then, the DTW distance between the time-series change curves of the corresponding neuralgia characteristic parameters of different postneuropathy patients was used as the second distance. The larger the second distance, the greater the difference in the changes of postneuropathy between the two. The first and second distances are then weighted with 0.4 and 0.6 respectively for a weighted sum. The implementer can also adjust the weights and use the weighted sum as the metric distance between two patients with postneuropathy. Based on the metric distance and the preset K value determined by the elbow method, the K-means algorithm is used to cluster all patients with postneuropathy. The larger the cluster, the more patients with postneuropathy it contains, which indicates that the postneuropathy changes in patients within the cluster have more general characteristics. The clusters are then sorted in descending order of size, and the preset number is set to 0.5 times the preset K value. That is, all patients with postneuropathy in all clusters in the first half of the sorted data are used as reference patients with postneuropathy. The implementer can also adjust the preset number.

[0041] It should be noted that the calculation of DTW distance, the preset K value determined by the elbow method, and the K-means algorithm are all well-known techniques and will not be elaborated further.

[0042] At this point, all reference postherpetic neuralgia patients have been identified. Furthermore, based on the temporal changes in the neuralgia characteristic parameters of all reference postherpetic neuralgia patients, a predictive model can be constructed to make early predictions of postherpetic neuralgia in the current patient.

[0043] Preferably, in one embodiment of the present invention, the method for early prediction of postherpetic neuralgia includes: A sample set is constructed based on the time series of herpes characteristic coefficients and neuralgia characteristic parameters of each reference postneuropathy patient, and a prediction model for postneuropathy is trained. The herpes characteristic coefficients of the current patient are input into the trained prediction model, and the time series of neuralgia characteristic parameters output by the model is used as the early prediction result of postneuropathy for the current patient.

[0044] It should be noted that the prediction model is internally configured with a feature hidden state mapping layer and carries a decoder with a recurrent temporal unit (such as LSTM or GRU architecture). During training and deployment, the herpes feature coefficients are fed into the decoder as the initial dimensionality reduction state for conditional sequence generation. The decoder decodes and outputs a set of neuropathic pain feature parameters according to the temporal sequence rules through its internal step-size recurrent deduction mechanism. The training and application of the prediction model are well-known technologies and will not be elaborated further. The early prediction results of postneuropathy in the current patient output by the model can provide a certain reference to assist relevant medical personnel in early intervention.

[0045] Based on the same inventive concept, the present invention also proposes an early prediction system for postherpetic neuralgia. The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the early prediction method for postherpetic neuralgia described in steps S1-S3 above.

[0046] In summary, this invention first obtains basic information about all historical patients, herpes information during the herpes outbreak, and follow-up information at each subsequent visit after the herpes healed. Based on the herpes symptom presentation and basic information, it obtains the herpes characteristic coefficient for each historical patient. Then, it analyzes the differences between the location of the herpes and the location of the neuralgia at each follow-up visit, as well as the differences between the herpes pain level and the neuralgia level, obtaining neuralgia characteristic parameters at each follow-up visit. Based on the temporal changes of these neuralgia characteristic parameters, reference postherpetic neuralgia patients are selected. Finally, based on the temporal changes of the neuralgia characteristic parameters of all reference postherpetic neuralgia patients, a predictive model is constructed to predict postherpetic neuralgia in current patients at an early stage. This invention improves the predictive model construction effect and the early prediction effect of postherpetic neuralgia by analyzing the herpes outbreak presentation and neuralgia changes in historical patients, combined with the characteristic that the pain location of postherpetic neuralgia is concentrated at the original herpes lesion location, and filtering out some historical patients who do not have postherpetic neuralgia.

Claims

1. A method for early prediction of postherpetic neuralgia, characterized in that, The method includes: Obtain basic information of all historical patients, herpes information during the herpes outbreak, and follow-up information at each follow-up visit after the herpes healed; the herpes information includes at least the herpes pain level, herpes symptom information, and herpes location, and the follow-up information includes at least the neuralgia level and neuralgia location; Based on the herpes symptom information and the basic information, the herpes characteristic coefficient of each historical patient is obtained; for each historical patient, the neuralgia characteristic parameters at each follow-up visit are obtained based on the difference between the location of the herpes and the location of the neuralgia at each follow-up visit, and the difference between the herpes pain level and the neuralgia level. Based on the temporal changes of the neuralgia characteristic parameters of each historical patient during follow-up visits, combined with the herpes zoster characteristic coefficient, reference postherpetic neuralgia patients are selected from all historical patients. Based on the temporal changes of the neuralgia characteristic parameters of all reference postherpetic neuralgia patients, a prediction model is constructed to make early predictions of postherpetic neuralgia for the current patient.

2. The method for early prediction of postherpetic neuralgia according to claim 1, characterized in that, The information on herpes symptoms includes at least the density and area of ​​the blisters, and the basic information includes at least age.

3. The method for early prediction of postherpetic neuralgia according to claim 2, characterized in that, The method for obtaining the herpes characteristic coefficient includes: For each level of herpes pain, based on the density and concentration characteristics of the herpes area of ​​all historical patients, the characteristic parameters of herpes symptoms for the corresponding level of herpes pain are obtained. Herpes risk weights are determined based on the age of each historical patient. The herpes symptom characteristic parameters under the herpes pain level of the historical patient are weighted using the herpes risk weights, and the weighted result is used as the herpes characteristic coefficient of the corresponding historical patient.

4. The method for early prediction of postherpetic neuralgia according to claim 3, characterized in that, The methods for obtaining the herpes symptom characteristic parameters include: For each level of herpes pain, the density of herpes in all historical patients is averaged to obtain a density representative parameter, and the area of ​​herpes in all historical patients is averaged to obtain an area representative parameter; the weighted sum of the density representative parameter and the area representative parameter is used as a characteristic parameter of herpes symptoms.

5. The method for early prediction of postherpetic neuralgia according to claim 1, characterized in that, The method for obtaining the neuralgia characteristic parameters includes: The location descriptions of the herpes location and the neuralgia location are extracted from the electronic medical records, and the word vectors of the corresponding description words are obtained based on the BERT model. For each historical patient, the cosine similarity between the word vectors of the neuralgia location at each follow-up visit and the herpes location during the herpes outbreak is used as the location association parameter. For each historical patient, the negative correlation mapping result between the difference between the neuropathic pain level at each follow-up visit and the herpes pain level during the herpes outbreak is used as the pain association parameter; By fusing the location-related parameters and the pain-related parameters, neuropathic pain characteristic parameters of the corresponding historical patients are obtained.

6. The method for early prediction of postherpetic neuralgia according to claim 1, characterized in that, The method for obtaining the reference postneuropathy patients includes: Based on the temporal changes of the neuralgia characteristic parameters of each historical patient and the herpes characteristic coefficient, the confidence coefficient of postneuropathy is obtained, and historical patients whose postneuropathy confidence coefficient is greater than a preset threshold are identified as postneuropathy patients. Based on the differences in the confidence coefficients of postneuropathy among different postneuropathy patients, and the differences in the temporal changes of the corresponding neuropathic characteristic parameters during follow-up visits, reference postneuropathy patients were selected from all postneuropathy patients.

7. The method for early prediction of postherpetic neuralgia according to claim 6, characterized in that, The method for obtaining the confidence coefficient of postneuropathy includes: Based on the time-series change curve of the neuralgia characteristic parameters of each historical patient at each follow-up visit, the postneural pain reference coefficient is determined based on the slope of the time-series change curve; the postneural pain reference coefficient is weighted using the herpes characteristic coefficient, and the normalized value of the weighted result is used as the postneural pain confidence coefficient.

8. The method for early prediction of postherpetic neuralgia according to claim 7, characterized in that, Based on the differences in confidence coefficients for postneuropathy among different patients and the temporal changes in the corresponding neuropathic pain characteristic parameters during follow-up visits, reference postneuropathy patients are selected from all postneuropathy patients, including: Among different patients with postneuropathy, the difference between the confidence coefficients corresponding to the postneuropathy and the DTW distance between the time-series change curves corresponding to the neuropathic feature parameters are fused to obtain a metric distance; all patients with postneuropathy are clustered based on the metric distance, and the clusters are sorted in descending order based on the cluster size; all patients with postneuropathy in the first preset number of clusters are used as reference patients with postneuropathy.

9. The method for early prediction of postherpetic neuralgia according to claim 1, characterized in that, Early prediction methods for postherpetic neuralgia include: Based on the time series of the herpes feature coefficients and the neuralgia feature parameters of each reference postneuropathy patient, a sample set is constructed, and a prediction model for postneuropathy is trained. The herpes feature coefficients of the current patient are input into the trained prediction model, and the time series of neuralgia feature parameters output by the model is used as the early prediction result of the current patient's postneuropathy.

10. An early prediction system for postherpetic neuralgia, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of an early prediction method for postherpetic neuralgia as described in any one of claims 1 to 9.

Images