Endometriosis patient lactation review follow-up data processing system

Abstract

The application discloses a kind of endometriosis patient lactation review follow-up data processing system, it is related to medical data processing technical field, the system passes through acquisition patient hospital information, follow-up record, self-report data and equipment monitoring data, and using BERT medical NLP model is carried out vectorization processing to text illness record, standardization storage is realized in combination with NER, and patient data set is arranged according to time sequence;Abnormal data is scored using local anomaly factor and isolation forest algorithm, and provides abnormal data processing strategy;Based on time series analysis and EWMA method, calculate trend change coefficient BQS, monitor the trend of illness change and early warning;Through Euclidean distance, cosine similarity and clustering algorithm analysis patient illness trajectory, calculate similarity score Sxs, judge illness similarity, and provide corresponding strategy to high similarity patient.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical data processing, in particular to a data processing system for follow-up review of endometriosis patients during lactation. BACKGROUND

[0002] Endometriosis is a common gynecological disease, and its main pathological feature is that the endometrium, including glands and stroma, grows abnormally outside the uterine cavity. This lesion can cause dysmenorrhea, menstrual abnormalities, infertility and other symptoms. The occurrence and development of endometriosis are closely related to lactation and menstrual patterns. Due to the changes in lactation and menstrual patterns during lactation, the disease may be alleviated or developed, but there is currently a lack of lactation management guidelines for this disease at home and abroad, resulting in a lack of effective management mode for the disease during the special physiological stage of female lactation. Close follow-up during lactation is a crucial link in the treatment process. When processing the review follow-up data of patients, the data can be classified according to the differences in menstrual patterns, lactation and treatment plans of patients during lactation, as well as symptom manifestations and other aspects.

[0003] However, endometriosis is a disease that is prone to recurrence and has dynamic changes. With regular review and follow-up of patients, the disease may exhibit new symptom characteristics at different review stages, so the patient's condition has certain volatility during lactation, resulting in dynamic changes in review follow-up data. This change makes it difficult for traditional data classification methods to accurately reflect the patient's condition and reduces the accuracy of data classification results. If data anomalies cannot be effectively detected, disease trend analysis and monitoring of patient condition trajectories and similarities to historical cases are lacking, and many other problems may occur. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a data processing system for follow-up review of endometriosis patients during lactation to solve the problems mentioned in the background art.

[0005] To achieve the above purpose, the present application realizes the following technical scheme: a data processing system for follow-up review of endometriosis patients during lactation, comprising a data acquisition module, a data structured classification module, a data anomaly detection module, a disease trend analysis module and a disease similarity monitoring module.

[0006] The data acquisition module is used to acquire hospital information data, follow-up information data, self-reporting information data and equipment monitoring data of patients.

[0007] The data structured classification module is used for vectorizing the text class condition record, traditional Chinese medicine data and lactation related health data by a BERT medical NLP model, extracting semantic features, and converting condition description, drug name and symptom into standard terminology by using part-of-speech tagging and named entity recognition (NER), adopting ICD-10 coding standard for disease name, and using ATC classification system for drug name; the processed text data is stored in a structured database format, and a patient data set is established; all time stamps are converted into a standard format, and the condition data is arranged in order of patient review time, the lactation related data is involved, the time markers related to lactation period are added, including colostrum period, lactation period and back-milk period, and the follow-up record and physiological data are matched;

[0008] The data anomaly detection module is used for scoring abnormal data points in real time by using a local anomaly factor and an isolation forest algorithm, detecting abnormal data in real time, calculating a comprehensive data score Stotal in combination with patient data set data, and comparing and analyzing the data with a first threshold Q1 to determine whether the data conforms to a distribution, and giving a strategy for abnormal data;

[0009] The condition trend analysis module is used for real-time monitoring of dynamic changes of patient conditions by using time series analysis, sliding window technology and exponential weighted moving average (EWMA) method, calculating a condition trend change coefficient BQS, and comparing and analyzing the coefficient with a second threshold Q2 to determine whether the patient condition change trend is normal, and giving a strategy for abnormal trend;

[0010] The condition similarity monitoring module is used for comparing the condition change trends of different patients by using a plurality of similarity measurement methods including Euclidean distance and cosine similarity, grouping the patients, and real-time monitoring of condition and similar historical cases of the grouped patient population, calculating a similarity score coefficient Sxs, and comparing and analyzing the coefficient with a third threshold Q3 to determine whether the patient condition trajectory is similar to that of a certain type of historical case, and giving a strategy for the patient with a similar condition trajectory.

[0011] Preferably, the data collection module includes a hospital information collection unit, a follow-up collection unit, a patient self-reporting unit, a device collection unit and a traditional Chinese medicine data collection unit.

[0012] The hospital information collection unit is used for extracting patient condition records, examination reports and medication data from electronic medical records (EMR) in a hospital.

[0013] The follow-up collection unit is used for collecting recovery, symptom change, patient feedback data, ultrasound examination data, prolactin test data, lactation volume, menstrual pattern, lactation related condition description, medication and breastfeeding impact data recorded by doctors during patient follow-up.

[0014] The patient self-reporting unit is used for the patient to submit daily symptoms, subjective disease changes, medication information, menstrual patterns and breastfeeding conditions through a mobile terminal application and an online questionnaire.

[0015] The device acquisition unit is used for acquiring physiological indicators of the patient using a physiological monitoring device, including heart rate, blood pressure, body temperature and respiratory rate.

[0016] The traditional Chinese medicine data acquisition unit is used for collecting traditional Chinese medicine four-examination information, including inspection data, interrogation data and palpation data, and collecting traditional Chinese medicine prescriptions, acupuncture and massage therapy data.

[0017] Preferably, the data structuring classification module includes a data processing unit and a time series processing unit.

[0018] The data processing unit is used for vectorizing the text type disease records and follow-up contents of the patient by using a BERT pre-trained medical NLP model, extracting semantic features, and storing them in a structured manner; combining part-of-speech tagging and named entity recognition NER to convert disease descriptions, drug names and symptom contents into standard terminologies to reduce text ambiguity; using ICD-10 international disease classification coding to standardize mapping of symptom names and disease names to make all disease descriptions consistent, using ATC anatomical therapy chemical classification system coding to unify different sources of drug names; storing the processed text data in a structured database format and establishing a patient data set.

[0019] The time series processing unit is used for unifying the time format of the data, converting all time stamps into a standardized format, and arranging the patient's disease data in sequence according to the patient's review time points, so that the time sequence of the disease data is correct and continuous.

[0020] Preferably, the data anomaly detection module includes a first calculation unit and a first analysis unit.

[0021] The first calculation unit is used for scoring abnormal data points and identifying possible unreasonable data in real time by using a local anomaly factor and an isolation forest algorithm, and calculating a comprehensive score Stotal of the data after dimensionless processing in combination with the patient data set, according to the following formula:

[0022]

[0023] In the formula, n represents the total number of data indicators, Av i represents the i-th data indicator, f(Av i ) represents an anomaly score function corresponding to the data indicator, a1 represents a weight coefficient of the indicator score, which is obtained by model training, m represents the total number of features used to calculate the data features, and D jrepresents the jth data feature score, a2 represents the weight coefficient of the feature score, which is obtained by model training, w1 represents the weight coefficient for adjusting the influence of the data index score, and w2 represents the weight coefficient for adjusting the influence of the data feature score.

[0024] Preferably, the first analysis unit is used to preset a first threshold Q1 in advance, and the data anomaly coefficient SJY is compared and analyzed with the first threshold Q1 to obtain a first evaluation result, which includes:

[0025] When the data comprehensive score Stotal is less than the first threshold Q1, it indicates that the data conforms to the distribution and is normal, and continuous monitoring is performed.

[0026] When the data comprehensive score Stotal is greater than the first threshold Q1, it indicates that the data does not conform to the distribution, there is a data error, there is a risk of disease progression and detection error, a first warning instruction is triggered, and a first strategy is generated: starting manual review of medical history and clinical examination.

[0027] Preferably, the disease trend analysis module includes a second calculation unit and a second analysis unit.

[0028] The second calculation unit is used to adopt time series analysis, sliding window technology and exponential weighted moving average (EWMA) method to monitor the dynamic change of the patient's disease in real time, and combine the patient data set data to calculate and obtain a disease trend change coefficient BQS after dimensionless processing, and the formula is as follows:

[0029]

[0030] In the formula, M represents the total number of disease characteristics, t represents a review time point, T represents the length of a time window period, d is a small change amount on a differential time, Bv k (t) represents the value of the kth disease index at time t, γ represents a time decay factor, and a k represents the weight coefficient of the kth disease index.

[0031] Preferably, the second analysis unit is used to preset a second threshold Q2 in advance, and the disease trend change coefficient BQS is compared and analyzed with the second threshold Q2 to obtain a second evaluation result, which includes:

[0032] When the disease trend change coefficient BQS is less than the second threshold Q2, it indicates that the disease change trend is normal and there is no risk of disease deterioration, and continuous monitoring is performed.

[0033] When the disease trend change coefficient BQS is greater than or equal to the second threshold Q2, it indicates that the disease change trend is abnormal and there is a risk of disease deterioration, a second warning instruction is triggered, and a second strategy is generated: giving an adjustment treatment scheme suggestion.

[0034] Preferably, the illness similarity monitoring module comprises a grouping unit, a third calculation unit and a third analysis unit.

[0035] The grouping unit is used for comparing the illness change trends between different patients using a plurality of similarity measurement methods including Euclidean distance and cosine similarity, grouping the patients by the similarity between illness characteristics using nonlinear clustering algorithms K-means and DBSCAN, and establishing a patient group of similar illness trajectories.

[0036] Preferably, the third calculation unit is used for monitoring the illness of the patient group and similar historical cases in real time, combining the patient data set data, performing non-dimensional processing, and calculating and obtaining a similarity score coefficient Sxs, with the formula as follows:

[0037]

[0038] In the formula, M represents the total number of illness characteristics, Bv k (t) represents the value of the kth illness index at time t, represents the illness index value of the historical case at the same time point t, and k represents the weight coefficient of the kth illness index, represents the maximum value of the kth illness index, represents the minimum value of the kth illness index.

[0039] Preferably, the third analysis unit is used for presetting a third threshold Q3 in advance, comparing and analyzing the similarity score coefficient Sxs with the third threshold Q3, and obtaining a third evaluation result, including:

[0040] When the similarity score coefficient Sxs is less than the third threshold Q3, it indicates that the illness trajectory of the patient is not similar to the trajectory of a certain type of historical case, and continuous monitoring is required.

[0041] When the similarity score coefficient Sxs is greater than or equal to the third threshold Q3, it indicates that the illness trajectory of the patient is similar to the trajectory of a certain type of historical case, a third early warning instruction is triggered, and a third strategy is generated: prompting the patient to pay special attention to the illness change, and recommending corresponding treatment measures or intervention schemes.

[0042] The present application provides an endometriosis patient lactation review follow-up data processing system. It has the following beneficial effects:

[0043] (1) The endometriosis patient lactation review follow-up data processing system can effectively reduce ambiguity in text data and ensure accurate classification of review follow-up data by using a BERT medical NLP model to structure patient condition records and using ICD-10 coding to standardize disease descriptions and ATC classification system to unify drug names. This improvement makes the classification results more reliable and avoids the accuracy problems of traditional classification methods when dealing with dynamically changing patient conditions.

[0044] (2) The endometriosis patient lactation review follow-up data processing system can detect data anomalies in real time by using local anomaly factor and isolation forest algorithms and combining comprehensive scoring methods. By comparing with the set threshold, the system can trigger an early warning when data anomalies occur, effectively avoiding the failure to discover anomalies in condition data in a timely manner, thereby improving the accuracy and response speed of condition monitoring.

[0045] (3) The endometriosis patient lactation review follow-up data processing system can track the dynamic changes of patient conditions in real time and calculate the condition trend change coefficient BQS by using time series analysis, sliding window technology and exponential weighted moving average (EWMA) method. When the condition change trend is abnormal, the system can automatically trigger an early warning and give suggestions for adjusting the treatment plan, thereby ensuring the individualization and timeliness of the patient treatment process.

[0046] (4) The endometriosis patient lactation review follow-up data processing system can compare the condition change trends of different patients and group patients with similar condition trajectories by using similarity measurement methods such as Euclidean distance and cosine similarity. By monitoring the condition similarity in real time, the system can identify patients similar to historical cases and provide special attention and corresponding treatment intervention plans, thereby optimizing the treatment effect of patients. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The endometriosis patient lactation review follow-up data processing system block diagram flowchart. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0049] Embodiment 1

[0050] Referring to Figure 1 The application provides a follow-up data processing system for lactation review of endometriosis patients, comprising a data acquisition module, a data structured classification module, a data anomaly detection module, a disease trend analysis module and a disease similarity monitoring module.

[0051] The data acquisition module is used to acquire hospital information data, follow-up information data, self-reporting information data and equipment monitoring data of the patient.

[0052] The data structured classification module is used to perform vectorization processing on text type disease records, traditional Chinese medicine data and lactation related health data through a BERT medical NLP model, extract semantic features, and convert disease descriptions, drug names and symptoms into standard terms using part-of-speech tagging and named entity recognition NER, standardize disease names using ICD-10 coding standards, and unify drug names using the ATC classification system; the processed text data is stored in a structured database format, and a patient data set is established; all time stamps are converted into a standard format, the disease data is arranged in chronological order according to the patient review time, the lactation period data is involved, the time markers related to the lactation period are added, including the initial lactation period, the lactation period and the back lactation period, and the follow-up records and physiological data are matched;

[0053] The data anomaly detection module is used to score abnormal data points in real time by using a local anomaly factor and an isolation forest algorithm, detect abnormal data in real time, calculate a comprehensive data score Stotal in combination with the patient data set data, and compare and analyze the data with a first threshold Q1 to determine whether the data conforms to the distribution, and give a strategy for the abnormal data.

[0054] The disease trend analysis module is used to monitor the dynamic changes of the patient's disease in real time by using time series analysis, sliding window technology and exponential weighted moving average EWMA method, calculate a disease trend change coefficient BQS, compare and analyze the coefficient with a second threshold Q2 to determine whether the patient's disease change trend is normal, and give a strategy for the abnormal trend.

[0055] The disease similarity monitoring module is used to compare the disease change trends between different patients by using a plurality of similarity measurement methods including Euclidean distance and cosine similarity, group the patients, and monitor the disease of the grouped patient population and similar historical cases in real time, calculate a similarity score coefficient Sxs, compare and analyze the coefficient with a third threshold Q3 to determine whether the patient's disease trajectory is similar to that of a certain type of historical case, and give a strategy for the patient with a similar disease trajectory.

[0056] In this embodiment, through the comprehensive use of the data acquisition module, the data structured classification module, the data anomaly detection module, the disease trend analysis module and the disease similarity monitoring module, the system can accurately collect and process the patient's disease data, detect abnormal conditions in real time and analyze the trend of disease changes. At the same time, through the similarity analysis of the disease trajectory of different patients, patients similar to historical cases can be identified, and then the individualized treatment plan is optimized. The intelligent data processing and prediction analysis of the system can effectively improve the accuracy and timeliness of disease monitoring, and ensure that the patient receives the best individualized adjustment and intervention during treatment.

[0057] Embodiment 2,

[0058] This embodiment is an explanation and description in embodiment 1. Specifically, the data acquisition module includes a hospital information acquisition unit, a follow-up acquisition unit, a patient self-reporting unit, a device acquisition unit and a traditional Chinese medicine data acquisition unit.

[0059] The hospital information acquisition unit is used to extract the patient's disease records, examination reports, medication data from the hospital electronic medical record (EMR);

[0060] The follow-up acquisition unit is used to collect the recovery, symptom changes, patient feedback data, ultrasound examination data, prolactin test data, milk secretion, menstrual pattern, lactation-related disease description, medication and breastfeeding impact data recorded by the doctor during the patient's follow-up;

[0061] The patient self-reporting unit is used for patients to submit daily symptoms, subjective disease changes, medication information, menstrual patterns and breastfeeding through mobile applications and online questionnaires;

[0062] The device acquisition unit is used to use physiological monitoring devices to collect physiological indicators of patients, including heart rate, blood pressure, body temperature and respiratory rate;

[0063] The traditional Chinese medicine data acquisition unit is used to collect traditional Chinese medicine four diagnostic information, including diagnosis data, interrogation data and incision data, and collect traditional Chinese medicine prescriptions, acupuncture and massage therapy data.

[0064] In this embodiment, by integrating the hospital information acquisition unit, the follow-up acquisition unit, the patient self-reporting unit, the device acquisition unit and the traditional Chinese medicine data acquisition unit, the system can comprehensively and in real time collect the patient's disease data, follow-up records, symptom changes and physiological indicators. The feedback information provided by the patient through the mobile application and the online questionnaire is combined with the monitoring data of the medical equipment to ensure accurate monitoring and dynamic tracking of the patient's health status. This all-round data acquisition method not only improves the coverage and real-time nature of the data, but also better supports disease analysis and individualized treatment decisions, promoting the refinement and scientific nature of patient health management.

[0065] Embodiment 3,

[0066] This embodiment is an explanation and illustration in embodiment 2, specifically, the data structured classification module includes a data processing unit and a time series processing unit;

[0067] The data processing unit is used to adopt the BERT pre-training medical NLP model to perform vectorization processing on the text disease record and follow-up content of the patient, extract semantic features, and perform structured storage; combined with part-of-speech tagging and named entity recognition NER, the disease description, drug name and symptom content are converted into standard terminology to reduce text ambiguity; for symptoms and disease names, ICD-10 international disease classification coding is used for standardized mapping to make all disease descriptions consistent, and for drug names and treatment plans, ATC anatomical treatment chemical classification system coding is used to unify different sources of drug names; the processed text data is stored in a structured database format, and a patient data set is established;

[0068] The time series processing unit is used to unify the time format of the data, convert all time stamps to a standardized format, and arrange the patient's disease data in sequence according to the patient's review time points, so that the time sequence of the disease data is correct and continuous.

[0069] In this embodiment, the BERT pre-training medical NLP model is used to perform vectorization processing on the patient's disease record and follow-up content, combined with part-of-speech tagging and named entity recognition technology NER, the system can effectively extract the semantic features of disease description, drug name and symptom content, automatically convert the text data into standard terminology, and use ICD-10 and ATC coding system to realize the standardization of disease and treatment information. Combined with the time series processing unit for timestamp standardization and sorting of review data, the accuracy and continuity of the disease data are ensured. This process not only improves the readability and consistency of the disease data, but also provides high-quality structured data support for subsequent data analysis and decision-making.

[0070] Embodiment 4,

[0071] This embodiment is an explanation and illustration in embodiment 3, specifically, the data anomaly detection module includes a first calculation unit and a first analysis unit;

[0072] The first calculation unit is used to adopt the local anomaly factor and the isolated forest algorithm to score the abnormal data points, identify possible unreasonable data, and detect abnormal data in real time, combined with the patient data set data, after dimensionless processing, the data comprehensive score Stotal is calculated and obtained, the formula is as follows:

[0073]

[0074] wherein n represents the total number of data indicators, Av i represents the i-th data indicator, f(Av i ) represents the abnormal score function corresponding to the data indicator, a1 represents the weight coefficient of the indicator score, which is obtained by model training, m represents the total number of characteristics used to calculate the data characteristics, D j represents the j-th data characteristic score, a2 represents the weight coefficient of the characteristic score, which is obtained by model training, w1 represents the weight coefficient for adjusting the influence of the data indicator score, and w2 represents the weight coefficient for adjusting the influence of the data characteristic score.

[0075] In this embodiment, the data anomaly detection module can perform real-time abnormal scoring on the patient's illness data through the local anomaly factor and the isolation forest algorithm, and identify possible unreasonable data. In combination with the patient data set, the comprehensive score Stotal is calculated after dimensionless processing, the data is accurately analyzed whether it conforms to the normal distribution, and compared with the preset threshold value, the potential abnormal data is found in time. This function effectively improves the sensitivity and accuracy of data monitoring, can trigger an early warning when the data is abnormal, supports medical staff to quickly take countermeasures, and reduces misdiagnosis or treatment delay caused by data problems.

[0076] Embodiment 5,

[0077] This embodiment is an explanation and description in embodiment 4. Specifically, the first analysis unit is used to preset a first threshold Q1 in advance, and compare and analyze the data anomaly coefficient SJY with the first threshold Q1 to obtain a first evaluation result, including:

[0078] When the data comprehensive score Stotal is less than or equal to the first threshold Q1, it indicates that the data conforms to the distribution and has no anomaly, and continues to be monitored;

[0079] When the data comprehensive score Stotal is greater than the first threshold Q1, it indicates that the data does not conform to the distribution, there is a data error, and there is a risk of disease progression and detection error, a first early warning instruction is triggered, and a first strategy is generated: starting artificial review of medical history and clinical examination.

[0080] In this embodiment, the data comprehensive score Stotal is compared and analyzed in real time by the set first threshold Q1, which can accurately judge whether the patient data conforms to the normal distribution. When the score exceeds the threshold value, the system can timely identify the potential data error, disease progression or detection error risk, and trigger the early warning instruction. This mechanism effectively reduces the risk of misdiagnosis caused by not timely finding data problems, and at the same time starts artificial review of medical history and clinical examination, ensures accurate adjustment of the treatment plan, and improves the safety and reliability of medical decision-making.

[0081] Embodiment 6,

[0082] The embodiment is an explanation of embodiment 3. Specifically, the disease trend analysis module includes a second calculation unit and a second analysis unit.

[0083] The second calculation unit is configured to use time series analysis, sliding window technology and exponential weighted moving average (EWMA) method to monitor the dynamic changes of the patient's disease in real time. After combining the patient data set data and performing dimensionless processing, the disease trend change coefficient BQS is calculated and obtained. The formula is as follows:

[0084]

[0085] In the formula, M represents the total number of disease characteristics, t represents the review time point, T represents the length of the time window period, d is the small change amount on the differential time, Bv k (t) represents the value of the kth disease index at time t, γ represents the time decay factor, α k represents the weight coefficient of the kth disease index.

[0086] In the embodiment, the disease trend analysis module can monitor the dynamic changes of the patient's disease in real time by using time series analysis, sliding window technology and exponential weighted moving average (EWMA) method. By calculating the disease trend change coefficient BQS, the system can timely identify the small fluctuations of the disease changes, so as to predict the development trend of the patient's disease. This function helps doctors more accurately master the real-time changes of the disease, and provides personalized treatment plan for patients, ensures early intervention, and improves treatment effect and patient prognosis.

[0087] Embodiment 7,

[0088] The embodiment is an explanation of embodiment 6. Specifically, the second analysis unit is configured to preset a second threshold Q2 in advance, and compare and analyze the disease trend change coefficient BQS with the second threshold Q2 to obtain a second evaluation result, including:

[0089] When the disease trend change coefficient BQS is less than the second threshold Q2, it indicates that the disease change trend is normal, and there is no risk of disease deterioration. Continuous monitoring is performed.

[0090] When the disease trend change coefficient BQS is greater than or equal to the second threshold Q2, it indicates that the disease change trend is abnormal, and there is a risk of disease deterioration. A second warning instruction is triggered, and a second strategy is generated: adjustment of the treatment plan is suggested.

[0091] In this embodiment, through the combination of the disease trend analysis module and the second analysis unit, the system can automatically monitor the disease trend change coefficient BQS and compare it with the preset second threshold Q2. When the disease change trend is abnormal, the system will trigger an early warning and automatically generate a suggestion to adjust the treatment plan. This function helps doctors identify and take intervention measures in the early stage of disease deterioration, thereby maximizing the risk reduction of patients and ensuring the timely optimization of the treatment plan, as shown in the following table:

[0092]

[0093] Table 1.

[0094] Embodiment 8,

[0095] This embodiment is an explanation and illustration in embodiment 7. Specifically, the disease similarity monitoring module includes a grouping unit, a third calculation unit, and a third analysis unit.

[0096] The grouping unit is used to compare the disease change trends between different patients using various similarity measurement methods including Euclidean distance and cosine similarity. Through the similarity between disease characteristics, nonlinear clustering algorithms K-means and DBSCAN are used to group patients and establish patient groups with similar disease trajectories.

[0097] In this embodiment, through the grouping unit of the disease similarity monitoring module, the system can compare the disease change trends of patients using various similarity measurement methods such as Euclidean distance and cosine similarity, and group patients with similar disease trajectories through nonlinear clustering algorithms K-means and DBSCAN. This function enables the system to accurately identify patient groups with similar disease development, providing a basis for the development of individualized treatment plans.

[0098] Embodiment 9,

[0099] This embodiment is an explanation and illustration in embodiment 8. Specifically, the third calculation unit is used to monitor the disease of the patient group and the similar historical cases in real time. After combining the patient data set data and performing dimensionless processing, the similarity score coefficient Sxs is calculated and obtained, with the formula as follows:

[0100]

[0101] In the formula, M represents the total number of disease characteristics, Bv k (t) represents the value of the kth disease index at time t, represents the disease index value of the historical case at the same time point t, α k represents the weight coefficient of the kth disease index, represents the maximum value of the kth disease index, represents the minimum value of the kth illness indicator.

[0102] In this embodiment, the system can calculate and obtain the similarity score coefficient Sxs by monitoring the illness changes of the patient population in real time and comparing with similar historical cases through the third calculation unit. This score coefficient can reflect the similarity between the current illness of the patient and the historical cases, thereby helping doctors to identify whether the patient's illness conforms to certain typical disease course patterns and providing real-time data support for personalized treatment intervention. This monitoring and scoring mechanism helps to identify potential illness risks in advance and optimize treatment decisions.

[0103] Embodiment 10,

[0104] This embodiment is an explanation and illustration in embodiment 9. Specifically, the third analysis unit is used to preset a third threshold Q3 in advance, and the similarity score coefficient Sxs is compared and analyzed with the third threshold Q3 to obtain a third evaluation result, including:

[0105] When the similarity score coefficient Sxs is less than the third threshold Q3, it indicates that the illness trajectory of the patient is not similar to the trajectory of a certain type of historical cases, and continuous monitoring is required.

[0106] When the similarity score coefficient Sxs is greater than or equal to the third threshold Q3, it indicates that the illness trajectory of the patient is similar to the trajectory of a certain type of historical cases, a third warning instruction is triggered, and a third strategy is generated: prompting the patient to pay special attention to the illness changes, and recommending corresponding treatment measures or intervention schemes.

[0107] In this embodiment, by setting the third threshold Q3 through the third analysis unit and comparing the similarity score coefficient Sxs, the system can determine whether the illness trajectory of the patient is similar to the trajectory of a certain type of historical cases. When the illness trajectory is similar to the historical cases, the system will trigger an early warning and recommend corresponding treatment or intervention schemes, thereby ensuring that the illness changes of the patient are timely paid attention to and handled. This mechanism can significantly improve the accuracy of disease management and reduce the risk of illness deterioration, as shown in the following table:

[0108]

[0109] Table 2.

[0110] The size of the threshold is set for easy comparison. The size of the threshold depends on the amount of sample data and the base number set by the person skilled in the art for each group of sample data; as long as it does not affect the proportional relationship between the parameters and the quantized values.

[0111] The above formulas are obtained by collecting a large amount of data for software simulation and selecting a formula close to the true value, and the coefficients in the formula are set by the person skilled in the art according to the actual situation. The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can make equivalent replacement or change according to the technical scheme and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A data processing system for follow-up examinations of patients with endometriosis during lactation, characterized in that, The system comprises a data collection module, a data structured classification module, a data anomaly detection module, a disease trend analysis module, and a disease similarity monitoring module. The data collection module is configured to collect hospital information data, follow-up information data, self-reported information data, and equipment monitoring data of a patient. The data structured classification module is configured to perform vectorization processing on text-based disease records, traditional Chinese medicine data, and lactation-related health data by using a BERT medical NLP model, extract semantic features, and convert disease descriptions, drug names, and symptoms into standard terminologies by using part-of-speech tagging and named entity recognition (NER). Disease names are standardized by using ICD-10 coding, and drug names are unified by using the ATC classification system. The processed text data is stored in a structured database format, and a patient data set is established. All timestamps are converted into a standard format, and disease data is arranged in chronological order according to patient review times. Lactation-related data is involved, and time markers related to the lactation period, including the initial lactation period, the lactation period, and the back lactation period, are added. Follow-up records and physiological data are matched. The data anomaly detection module is configured to score abnormal data points in real time by using a local anomaly factor and an isolation forest algorithm, detect abnormal data in real time, calculate a comprehensive data score Stotal in combination with patient data set data, and compare and analyze the data score with a first threshold Q1 to determine whether the data conforms to a distribution and provide a strategy for abnormal data. The disease trend analysis module is configured to monitor the dynamic changes of a patient's disease in real time by using time series analysis, sliding window technology, and the exponential weighted moving average (EWMA) method, calculate a disease trend change coefficient BQS, compare and analyze the coefficient with a second threshold Q2 to determine whether the patient's disease trend is normal, and provide a strategy for abnormal trends. The disease similarity monitoring module is configured to compare the disease trend changes of different patients by using various similarity measurement methods including Euclidean distance and cosine similarity, group the patients, monitor the disease of the grouped patient population and similar historical cases in real time, calculate a similarity score coefficient Sxs, compare and analyze the coefficient with a third threshold Q3 to determine whether the patient's disease trajectory is similar to that of a certain type of historical case, and provide a strategy for patients with similar disease trajectories. The disease trend analysis module comprises a second calculation unit and a second analysis unit. The second calculation unit is configured to monitor the dynamic changes of a patient's disease in real time by using time series analysis, sliding window technology, and the exponential weighted moving average (EWMA) method, calculate a disease trend change coefficient BQS in combination with patient data set data after dimensionless processing, and the formula is as follows: ； In the formula, M represents the total number of disease characteristics, t represents the review time point, T represents the length of the time window period, d is a small change amount on the differential time, represents the value of the kth disease index at time t, represents a time decay factor, represents the weight coefficient of the kth disease index; The second analysis unit is configured to preset a second threshold Q2 in advance, compare and analyze the disease trend change coefficient BQS with the second threshold Q2, and obtain a second evaluation result including: When the disease trend change coefficient BQS is less than the second threshold Q2, it indicates that the disease trend is normal and there is no risk of disease deterioration, and continuous monitoring is required. When the disease trend change coefficient BQS is greater than or equal to the second threshold Q2, it indicates that the disease trend is abnormal, there is a risk of disease deterioration, a second early warning instruction is triggered, and a second strategy is generated: an adjusted treatment plan is recommended.

2. The endometriosis patient lactation review follow-up data processing system according to claim 1, characterized in that, The data collection module includes a hospital information collection unit, a follow-up collection unit, a patient self-reporting unit, a device collection unit, and a traditional Chinese medicine data collection unit. The hospital information collection unit is configured to extract the patient's disease records, examination reports, and medication data from the electronic medical record (EMR) of the hospital. The follow-up collection unit is configured to collect recovery data, symptom changes, patient feedback data, ultrasound examination data, prolactin test data, milk secretion, menstrual patterns, descriptions of lactation-related symptoms, medication, and breastfeeding impact data recorded by doctors during patient follow-up. The patient self-reporting unit is configured for patients to submit daily symptoms, perceived disease changes, medication information, menstrual patterns, and breastfeeding information through a mobile application and an online questionnaire. The device collection unit is configured to use physiological monitoring devices to collect physiological indicators of patients, including heart rate, blood pressure, body temperature, and respiratory rate. The traditional Chinese medicine data collection unit is configured to collect traditional Chinese medicine four diagnostic information, including visual diagnosis data, interrogation data, and palpation data, and collect traditional Chinese medicine prescriptions, acupuncture, and massage therapy data.

3. The endometriosis patient lactation review follow-up data processing system of claim 2, wherein, The data structured classification module includes a data processing unit and a time series processing unit. The data processing unit is configured to use a BERT pre-trained medical NLP model to vectorize the patient's text-based disease records and follow-up content, extract semantic features, and store them in a structured format. By combining part-of-speech tagging and named entity recognition (NER), disease descriptions, drug names, and symptom content are converted into standard terminology to reduce text ambiguity. For symptoms and disease names, ICD-10 international disease classification coding is used for standardized mapping to ensure consistency in disease descriptions. For drug names and treatment plans, ATC anatomical, therapeutic, and chemical classification system coding is used to unify different sources of drug names. The processed text data is stored in a structured database format, and a patient data set is established. · The time series processing unit is configured to unify the time format of the data, convert all timestamps to a standardized format, and arrange the patient's disease data in chronological order based on the patient's review time points to ensure that the disease data is in the correct and continuous order.

4. The endometriosis patient lactation review follow-up data processing system according to claim 3, characterized in that, The data anomaly detection module includes a first calculation unit and a first analysis unit. The first calculation unit is configured to use the local outlier factor and the isolation forest algorithm to score abnormal data points, identify possible unreasonable data, and detect abnormal data in real time. After dimensionless processing, the data comprehensive score Stotal is calculated and obtained based on the patient data set, as follows: ； In the formula, n represents the total number of data indicators, represents the i-th data indicator, represents the abnormal score function corresponding to the data indicator, represents the weight coefficient of the indicator score, which is obtained by model training, and m represents the total number of characteristics used to calculate the data characteristics, represents the j-th data characteristic score, a2 represents the weight coefficient of the characteristic score, which is obtained by model training, w1 represents the weight coefficient for adjusting the influence of the data indicator score, and w2 represents the weight coefficient for adjusting the influence of the data characteristic score, , , and .

5. The endometriosis patient lactation review follow-up data processing system of claim 4, wherein, The first analysis unit is configured to preset a first threshold Q1 in advance and compare the data comprehensive score Stotal with the first threshold Q1 to obtain a first evaluation result, including: When the data comprehensive score Stotal is less than or equal to the first threshold Q1, it indicates that the data is consistent with the distribution and there is no anomaly, and continuous monitoring is required. When the data comprehensive score Stotal is greater than a first threshold Q1, it indicates that the data is not consistent with the distribution, there is a data error, there is a risk of disease progression and detection error, a first early warning instruction is triggered, and a first strategy is generated: starting manual review of medical history and clinical examination.

6. The endometriosis patient lactation review follow-up data processing system of claim 1, wherein, The disease similarity monitoring module comprises a grouping unit, a third calculation unit and a third analysis unit. The grouping unit is configured to compare the disease change trends between different patients using a plurality of similarity measurement methods including Euclidean distance and cosine similarity, group the patients by similarity between disease characteristics using a nonlinear clustering algorithm K-means and DBSCAN, and establish a patient group of similar disease trajectories.

7. The endometriosis patient lactation review follow-up data processing system of claim 6, wherein, The third calculation unit is configured to monitor the disease of the patient group and the similar historical cases in real time, combine the patient data set data, perform non-dimensional processing, and calculate a similarity score coefficient Sxs according to the following formula: ； where M represents the total number of disease characteristics, represents the value of the kth disease indicator at time t, represents the value of the kth disease indicator at time t, represents the weight coefficient of the kth disease indicator, represents the maximum value of the kth disease indicator, represents the minimum value of the kth disease indicator.

8. The endometriosis patient lactation review follow-up data processing system of claim 7, wherein, The third analysis unit is configured to preset a third threshold Q3 in advance, compare the similarity score coefficient Sxs with the third threshold Q3, and obtain a third evaluation result including: When the similarity score coefficient Sxs is less than the third threshold Q3, it indicates that the disease trajectory of the patient is not similar to the trajectory of a certain type of historical case, and continuous monitoring is required. When the similarity score coefficient Sxs is greater than or equal to the third threshold Q3, it indicates that the disease trajectory of the patient is similar to the trajectory of a certain type of historical case, a third early warning instruction is triggered, and a third strategy is generated: prompting the patient to pay special attention to the change in the disease, and recommending corresponding treatment measures or intervention schemes.

Images