A personalized prevention treatment plan recommendation system for urinary stone recurrence

Abstract

The application relates to the technical field of medical data processing, and discloses a personalized prevention and treatment scheme recommendation system for urinary stone recurrence, which comprises the following steps: collecting ideal monitoring data of various medication indexes in a prevention and treatment scheme interacted by a doctor terminal and a patient terminal, and collecting actual monitoring data of various actual medication indexes of the patient; obtaining a plurality of monitoring time periods of the patient; obtaining a treatment reference factor of a historical monitoring time period; obtaining a personalized feedback adjustment coefficient of each medication index; obtaining prediction data; obtaining a final prediction matrix of the patient based on the ideal monitoring data and the prediction data adjusted multiple times; and obtaining a recommendation score of other patients for the current patient through collaborative filtering based on the final prediction matrix of the current patient and the monitoring matrix of the other patients, and recommending a treatment scheme in combination with the treatment scheme of the other patients. The application aims to solve the problem of prevention effect deviation caused by patient compliance with medical orders and personalized scheme treatment.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, specifically to a personalized prevention and treatment recommendation system for recurrent urinary tract stones. Background Technology

[0002] The interaction between doctors and patients forms the core closed-loop mechanism of a personalized prevention plan recommendation system. The bidirectional data flow, collaborative decision-making, and iterative feedback between the two directly determine the accuracy, feasibility, adherence, and dynamic optimization capabilities of the recommended plan. This interaction is not simply "information transmission," but rather a deep integration of "clinical expertise + actual patient scenarios" that corrects the limitations of AI models, transforming "algorithm-generated" plans into effective interventions that are "clinically usable and patient-executable."

[0003] For high-risk groups of urinary tract stone recurrence (such as patients with a history of stone disease, metabolic abnormalities, or family history), due to the two-way data flow between doctors and patients, there is a possibility that the preventive treatment plan corresponding to the doctor's prescription may differ from the actual preventive treatment measures taken by the patient. That is, there are differences in patient compliance with the doctor's prescription and individualized treatment of the condition. Therefore, it is necessary to compare the changes in the patient's data on multiple medication-related indicators with the ideal data changes under the doctor's prescription in order to analyze the patient's individualized preventive treatment measures, and then adaptively iterate the preventive treatment plan to assist clinicians in quickly making personalized treatment decisions and reducing the stone recurrence rate. Summary of the Invention

[0004] This invention provides a personalized prevention and treatment recommendation system for recurrent urinary tract stones, addressing the existing problem of discrepancies in preventative effects caused by patient adherence to medical advice and personalized treatment plans. The specific technical solution adopted is as follows: This invention also proposes a personalized prevention and treatment recommendation system for recurrent urinary tract stones, the system comprising: The patient data acquisition module is used to collect ideal monitoring data of various medication indicators in the prevention and treatment plan between the doctor and the patient, as well as actual monitoring data of various medication indicators of the patient, and to record historical monitoring data of various medication indicators and treatment plans of several patients. The patient data analysis module is used to analyze the periodic changes of monitoring data of various medication indicators of patients, and obtain several monitoring periods and historical monitoring periods of patients; based on the historical monitoring data, it analyzes the correlation of indicator data changes in different monitoring periods under the same medication indicators, and obtains the treatment reference coefficient of historical monitoring periods; combined with the similarity of the overall changes of monitoring data in different treatment plans, it obtains the treatment reference factor of historical monitoring periods. The correlation between changes in monitoring data of different medication indicators was analyzed to obtain the personalized feedback adjustment coefficient for each medication indicator; reference data was constructed based on historical monitoring data and treatment reference factors, and the difference between actual monitoring data and reference data was iteratively analyzed. Predictive data was obtained by combining the personalized feedback adjustment coefficient; based on ideal monitoring data and multiple adjusted predictive data, the overall similarity under the same monitoring period was analyzed by matrix analysis to obtain the final prediction matrix for patients. The treatment plan recommendation module is used to obtain recommendation scores from other patients for the current patient based on the current patient's final prediction matrix and the monitoring matrices of other patients, and then make recommendations based on the treatment plans of other patients.

[0005] Optionally, the specific method for obtaining the patient's several monitoring periods and historical monitoring periods is as follows: The actual monitoring data of any medication indicator in the current patient's treatment plan constitutes the monitoring sequence of that medication indicator for the current patient. The monitoring sequence is decomposed into a trend term, a periodic term, and a residual term by STL time series decomposition. The periodic term is subjected to Fourier transform to obtain the spectrum of the periodic term of the monitoring sequence. The amplitude of each frequency in the spectrum is obtained. The reciprocal of the frequency corresponding to the maximum amplitude is taken as the monitoring period of that medication indicator in the current patient's treatment plan. Obtain the monitoring cycle of each medication indicator in the current patient's treatment plan, and obtain the minimum value among all monitoring cycles as the current patient's monitoring period; divide all the time points corresponding to the monitoring sequence of each medication indicator in the current patient's treatment plan according to the monitoring period to obtain several monitoring time periods for the current patient; Historical monitoring data of each patient's medication indicators were used to divide the monitoring period for each patient's treatment plan according to the corresponding treatment plan, thus obtaining several historical monitoring periods for each patient.

[0006] Optionally, the method for analyzing the correlation of changes in indicator data under the same medication indicators in different monitoring periods based on historical monitoring data to obtain the treatment reference coefficient for the historical monitoring period includes the following specific methods: By standardizing all monitoring data of the same medication indicator for patients, several indicator data of each medication indicator for patients are obtained. The current monitoring period is defined as any one of several monitoring periods obtained from the current patient's monitoring data at the corresponding time. For any medication indicator of the current patient in the current monitoring period, several indicator data are constructed according to time sequence to form an indicator sequence for the current patient's medication indicator in the current monitoring period. Several indicator data of any medication indicator of any historical patient in any historical monitoring period are obtained and constructed according to time sequence to form a historical indicator sequence for the patient's medication indicator in that historical monitoring period. The indicator sequence and the historical indicator sequence are matched by DTW to obtain the DTW distance. The inversely proportional normalized result of the obtained DTW distance is used as a reference factor for the change of the patient's medication indicator in that historical monitoring period to the current patient's current monitoring period. Based on the trend changes of current patients’ actual monitoring data and historical monitoring data under the same medication indicators, and the time distribution of the corresponding monitoring periods, the period reference weights of the same medication indicators for different patients in history for each monitoring period are obtained. The product of the period reference weight of the patient's medication indicator and the historical monitoring period with respect to the current patient's current monitoring period and the change reference factor is used as the treatment reference coefficient of the patient's medication indicator and the historical monitoring period with respect to the current patient's current monitoring period.

[0007] Optionally, the specific methods for obtaining several index data of various medication indicators for patients include: All actual and historical monitoring data of any medication indicator for all patients were standardized, and the results were used as the indicator data for each monitoring period and historical monitoring period of that medication indicator.

[0008] Optionally, the period reference weights for the same medication indicators in different historical patients at different monitoring periods are obtained using the following method: Based on the current monitoring period of any medication indicator of the current patient, the trend sequence of the current patient's medication indicator during the current monitoring period and the trend sequence of the adjacent previous monitoring period are extracted from the trend item of the monitoring sequence of the medication indicator. The difference sequence obtained by subtracting the trend sequence of the adjacent previous monitoring period from the trend sequence of the current monitoring period is used as the trend change sequence of the current monitoring period. Obtain the trend sequence of any patient's medication indicator for any historical monitoring period, and the corresponding trend change sequence; perform DTW matching between the trend change sequence of the current monitoring period and the trend change sequence of the patient's medication indicator for the historical monitoring period to obtain the DTW distance, and use the inversely proportional normalized result of the obtained DTW distance as the period reference weight of the patient's medication indicator for the current monitoring period to the current patient's current monitoring period.

[0009] Optionally, the treatment reference factors for the historical monitoring period are obtained using the following method: For the current patient's treatment plan, several monitoring periods corresponding to any treatment plan for any patient are used as several historical monitoring periods corresponding to the reference treatment plan for the reference patient. The monitoring periods in the current patient's treatment plan are arranged in chronological order to form the monitoring period sequence of the current patient's treatment plan. The historical monitoring period sequence of the reference treatment plan for the reference patient is obtained. The monitoring period sequence and the historical monitoring period sequence are matched by DTW based on the change reference factor of the medication index between the monitoring period and the historical monitoring period, and the DTW distance is obtained. The inverse proportional normalization result of the obtained DTW distance is used as the treatment similarity factor between the current patient's treatment plan and the reference treatment plan for the reference patient. The treatment similarity factor is assigned to several historical monitoring periods in the reference treatment plan of the reference patient, as the treatment similarity factor between each historical monitoring period in the reference treatment plan of the reference patient and each monitoring period in the current patient's treatment plan; The treatment similarity factor is multiplied by the treatment reference coefficient, and the product is used as the treatment reference factor for the current patient's current monitoring period for each historical monitoring period of the medication index in the reference treatment plan of the reference patient.

[0010] Optionally, the specific method for obtaining the personalized feedback adjustment coefficients for each medication indicator includes: For any medication indicator of a current patient, several indicator data during the current monitoring period are arranged in chronological order to form an indicator sequence for that medication indicator during the current monitoring period. The cosine similarity of the indicator sequences of any two medication indicators of the current patient during the current monitoring period is obtained as the correlation between the changes of those two medication indicators during the current monitoring period. The mean of the correlation between the changes of those two medication indicators of all patients during all monitoring periods is calculated as the correlation between the changes of those two medication indicators. Any medication indicator is used as the current medication indicator, and the mean of the correlation between the changes of the current medication indicator and other medication indicators is used as the personalized feedback adjustment coefficient of the current medication indicator.

[0011] Optionally, the specific methods for constructing reference data based on historical monitoring data and treatment reference factors, iteratively analyzing the differences between actual monitoring data and reference data, and obtaining predicted data by combining personalized feedback adjustment coefficients include: For any actual monitoring data of current medication indicators, the corresponding indicator data of the actual monitoring data in the same position of each patient's current medication indicators in each historical monitoring period are weighted and averaged, with the corresponding treatment reference factor in the historical monitoring period as the weight. The result is used as the reference data of the actual monitoring data. The absolute value of the difference between the actual monitoring data and its reference data is obtained as the initial treatment difference of the actual monitoring data. The initial treatment differences of each actual monitoring data in the current monitoring period are obtained. If there are actual monitoring data with an initial treatment difference greater than a preset difference threshold, the actual monitoring data is multiplied by the personalized feedback adjustment coefficient of the current medication index to obtain the adjusted actual monitoring data, which is used as the prediction data for the corresponding adjustment. The treatment difference is recalculated until there is no prediction data greater than the difference threshold. The final prediction data is used as the final prediction data of the current patient's current medication index in the current monitoring period.

[0012] Optionally, the final prediction matrix for the patient can be obtained using the following method: For several predicted data and unadjusted actual monitoring data for the current monitoring period under the same adjustment, a matrix is ​​constructed with time as the column and medication index as the row, which serves as the effective monitoring matrix for the patient in the corresponding adjustment; based on the ideal monitoring data of multiple medication indicators corresponding to the current patient's treatment plan, an ideal monitoring matrix is ​​constructed; based on the ideal monitoring matrix and the effective monitoring matrix of the corresponding adjustment, the feedback optimization coefficient of each adjustment is obtained; Starting from the first adjustment, after the threshold judgment, the feedback optimization coefficient is subtracted from 1 and then multiplied with the personalized feedback adjustment coefficient to obtain the new adjustment coefficient. The second adjustment is then performed, and so on, until the final prediction data after the final adjustment is obtained, which constitutes the final prediction matrix.

[0013] Optionally, the feedback optimization coefficients for each adjustment are obtained using the following method: For the ideal monitoring matrix and the corresponding effective monitoring matrix, a sequence is formed by connecting the first and last rows of the matrix to calculate the similarity. The obtained similarity is used as the feedback optimization coefficient for the corresponding adjustment.

[0014] The beneficial effects of this invention are as follows: This invention comprehensively analyzes actual monitoring data of multiple medication indicators under a patient's treatment plan, combined with historical monitoring data of multiple medication indicators under several patient treatment plans. It obtains a final prediction matrix by considering the ideal monitoring data changes of multiple medication indicators under the comprehensive treatment plan, as well as the relevant performance of multiple medication indicators in similar physiological states in historical monitoring data. This matrix is ​​then used for subsequent collaborative filtering to achieve personalized preventive treatment recommendations based on past treatment plans. The invention considers the periodic changes in the monitoring data of each medication indicator, and the trend and overall data change relationship between the monitoring period and historical monitoring periods. This yields a treatment reference coefficient between the monitoring period and historical monitoring periods. Further, it performs data change similarity analysis based on the overall treatment reference coefficient changes in each monitoring period under the treatment plan, thereby obtaining treatment reference factors for the actual monitoring data from historical monitoring data of the same medication indicators. Subsequently, reference data is constructed to reflect physiological states similar to the current treatment plan. This data is then iteratively adjusted in conjunction with ideal monitoring data, while also considering the correlation changes between different medication indicators, ultimately yielding the final prediction data and constructing the final prediction matrix. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a structural block diagram of a personalized prevention and treatment recommendation system for recurrent urinary tract stones, provided in one embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 The diagram illustrates a structural block diagram of a personalized prevention and treatment recommendation system for recurrent urinary tract stones, provided by an embodiment of the present invention. The system includes: The patient data acquisition module 101 collects ideal monitoring data of various medication indicators in the prevention and treatment plan that is exchanged between the doctor's end and the patient's end, as well as actual monitoring data of various medication indicators of the patient, and records historical monitoring data of various medication indicators of several patients and their treatment plans.

[0019] The purpose of this embodiment is to address the discrepancy between the patient's personalized treatment approach and adherence to medical advice after a preventive treatment plan for recurrent urinary tract stones has been developed through doctor-patient interaction. This allows for iterative adjustments to the preventive treatment plan to improve its matching effect, ensure recommendation accuracy, and reduce the stone recurrence rate. After obtaining a preventive treatment plan for recurrent urinary tract stones through doctor-patient interaction, patient adherence to medical advice and the individualization of the condition can lead to differences between the actual treatment effect and the expected ideal treatment effect. Therefore, the treatment plan needs to be adaptively and iteratively adjusted to achieve personalized recommendations for preventive treatment.

[0020] Specifically, the collected data includes clinical data, patient-reported data, and environmental data. Clinical data includes past history of kidney stones (component analysis results are crucial), imaging reports, laboratory tests (blood / urine metabolic indicators: calcium, phosphorus, uric acid, oxalate, citrate, etc.), comorbidities (diabetes, hypertension, gout, etc.), and medication history, which are entered manually by doctors through the hospital's HIS / LIS / PACS system interface. Patient-reported data includes dietary structure (daily water intake, calcium / oxalate / purine intake), lifestyle habits (exercise frequency, sedentary time, urinary retention habits), and symptom feedback (lower back and abdominal pain, hematuria, etc.), which are filled out by patients through an app / mini-program and pushed through intelligent questionnaires. Environmental data includes the hardness of the water in the place of residence and family history of kidney stones, which are entered through system forms and genetic testing data. The actual monitoring data of various medication indicators are all time-series data, and the collection frequency of all medication indicators is consistent. The specific collection frequency is set by the patient according to the treatment plan, which will not be elaborated in this embodiment.

[0021] Furthermore, after interaction between the doctor and patient, a preliminary prevention and treatment plan is generated. This plan is then analyzed using big data from the hospital database to obtain ideal monitoring data (time series) for various medication indicators after the implementation of this prevention and treatment plan. These medication indicators correspond to various types of collected data. Simultaneously, patients are continuously tracked, and the actual monitoring data of various medication indicators are recorded as monitoring data for various medication indicators in the patient's current treatment plan. Historical monitoring data on the prevention and treatment of urinary stones for several patients, including their corresponding prevention and treatment plans, are extracted from the hospital database.

[0022] Patient data analysis module 102: (1) Analyze the periodic changes of monitoring data of patients’ medication indicators to obtain several monitoring periods and historical monitoring periods; based on historical monitoring data, analyze the correlation of indicator data changes in different monitoring periods under the same medication indicators to obtain the treatment reference coefficient of historical monitoring periods; combine the similarity of overall changes in monitoring data in different treatment plans to obtain the treatment reference factor of historical monitoring periods.

[0023] It should be noted that, due to the periodic changes in monitoring data under treatment regimens, it is first necessary to analyze these periodic changes to extract several monitoring periods. Simultaneously, since the same medication indicator may exhibit significant differences in data across different patients and multiple treatment regimens, standardization is used to eliminate these differences and obtain indicator data. This provides a basis for subsequent comparisons of monitoring data for different indicators, eliminating the influence of dimensions. Furthermore, based on the indicator data, an analysis of the similarity in trend changes between monitoring periods and historical monitoring periods, as well as an analysis of the similarity in overall data changes, is conducted to construct treatment reference factors. These factors are used to obtain reference data from a large number of patients for monitoring various medication indicators under their treatment regimens.

[0024] Preferably, in one embodiment of the present invention, the periodic changes in the monitoring data of various medication indicators of the patient are analyzed to obtain several monitoring periods and historical monitoring periods for the patient. The specific method includes: It should be noted that the monitoring of relevant medication indicators for patients with urinary tract stones will show periodic changes as the treatment plan progresses. At the same time, there are differences in the periodic changes among various medication indicators. By decomposing the time series to extract the periodic items, the monitoring period is initially determined. Furthermore, by combining the monitoring periodic changes of multiple medication indicators, the minimum value is taken as the monitoring period to divide the monitoring time period. At the same time, the historical monitoring data is extracted as a whole according to the corresponding treatment plan, and the historical monitoring time period is further divided.

[0025] Specifically, for any medication indicator in the current patient's treatment plan, the actual monitoring data constitutes the monitoring sequence for that medication indicator. This monitoring sequence is decomposed using STL time series decomposition to obtain a trend term, a periodic term, and a residual term. A Fourier transform is performed on the periodic term to obtain its spectrum, and the 0Hz DC component is filtered out. The amplitude of each frequency in the spectrum is obtained, and the reciprocal of the frequency corresponding to the maximum amplitude is taken as the monitoring period for that medication indicator in the current patient's treatment plan. The monitoring periods for each medication indicator in the current patient's treatment plan are obtained, and the minimum value among all monitoring periods is taken as the current patient's monitoring period. All times corresponding to the monitoring sequence of each medication indicator in the current patient's treatment plan are divided according to the monitoring period to obtain several monitoring time periods for the current patient. STL time series decomposition and Fourier transform are well-known techniques and will not be elaborated upon in this embodiment. Historical monitoring data for each medication indicator of each patient are divided into monitoring periods for each patient's treatment plan based on the corresponding treatment plan, thereby obtaining several historical monitoring time periods for each patient.

[0026] It should be further explained that in the process of analyzing the similarity between the monitoring period and the historical monitoring period, an overall change analysis is performed based on the indicator data to eliminate the huge differences in magnitude caused by the fluctuation of data amplitude. At the same time, the local trend changes corresponding to the monitoring period are considered, that is, the influence of periodic terms and residual terms are avoided. The correlation of trend changes between the monitoring period and the historical monitoring period is further analyzed from the perspective of trend changes, and this is used as the period reference weight. The treatment reference coefficient is obtained by comprehensively analyzing the overall changes of the indicators.

[0027] Preferably, in one embodiment of the present invention, the method for analyzing the correlation of changes in indicator data during different monitoring periods under the same medication indicator based on historical monitoring data to obtain the treatment reference coefficient for the historical monitoring period includes: Standardize all actual and historical monitoring data for any medication indicator for all patients, and use the results as the indicator data for each monitoring period and historical monitoring period of that medication indicator. Take any one of several monitoring periods corresponding to the current patient's monitoring data as the current monitoring period for that patient. For several indicator data for the current patient's medication indicator in the current monitoring period, construct an indicator sequence for the current patient's medication indicator in the current monitoring period according to time sequence. Obtain several indicator data for any historical monitoring period of the medication indicator for any historical patient, and construct a historical indicator sequence for that patient's medication indicator in that historical monitoring period according to time sequence. Perform DTW matching between the indicator sequence and the historical indicator sequence to obtain the DTW distance. Use the inversely proportional normalized result of the obtained DTW distance as a reference factor for the change of the patient's medication indicator in that historical monitoring period from the current patient's current monitoring period.

[0028] Furthermore, based on the current monitoring period of the current patient's medication indicator, the trend sequence of the current patient's medication indicator during the current monitoring period and the trend sequence of the adjacent previous monitoring period are extracted from the trend term of the monitoring sequence of the medication indicator. The difference sequence obtained by subtracting the trend sequence of the adjacent previous monitoring period from the trend sequence of the current monitoring period is used as the trend change sequence of the current monitoring period. It should be noted that if the current monitoring period is the first monitoring period in the time corresponding to the current patient's monitoring data, the trend sequence of the adjacent previous monitoring period is supplemented by linear interpolation. The trend sequence of any patient's medication indicator for any historical monitoring period and the corresponding trend change sequence are obtained. The trend change sequence of the current monitoring period is matched with the trend change sequence of the patient's medication indicator for that historical monitoring period and the DTW distance is obtained. The inversely proportional normalized result of the obtained DTW distance is used as the period reference weight of the patient's medication indicator for that monitoring period to the current patient's current monitoring period.

[0029] Furthermore, the product of the period reference weight of the patient's medication indicator and the historical monitoring period with respect to the current patient's current monitoring period and the change reference factor is used as the treatment reference coefficient of the patient's medication indicator and the historical monitoring period with respect to the current patient's current monitoring period.

[0030] It should be noted that this embodiment adopts... The model is used to represent the inverse proportional relationship and for normalization processing. As input to the model, It represents an exponential function with the natural constant as the base. Implementers can set inverse proportional functions and normalization functions according to the actual situation. The smaller the DTW distance between the sequences composed of indicator data, the more similar the overall changes, and the greater the reference value of relevant data in historical monitoring periods. At the same time, the smaller the DTW distance between trend terms, the more similar the change trends after eliminating the influence of periodic changes and residuals, the greater the period reference weight, and thus the greater the treatment reference coefficient.

[0031] It should be further noted that after analyzing the treatment reference coefficients of the monitoring period and the historical monitoring period, the similarity of changes in the overall treatment plan across multiple monitoring periods is further considered, and a treatment similarity factor is obtained based on the treatment reference coefficients.

[0032] Preferably, in one embodiment of the present invention, the method for obtaining treatment reference factors for historical monitoring periods by combining the overall similarity of changes in monitoring data across different treatment regimens includes: For the current patient's treatment plan, several monitoring periods corresponding to any treatment plan for any patient are used as several historical monitoring periods corresponding to the reference treatment plan for the reference patient. The monitoring periods in the current patient's treatment plan are arranged in chronological order to form the monitoring period sequence of the current patient's treatment plan. The historical monitoring period sequence of the reference treatment plan for the reference patient is obtained. The monitoring period sequence and the historical monitoring period sequence are matched by DTW based on the change reference factor of the medication index in the monitoring period and the historical monitoring period, and the DTW distance is obtained. The inverse proportional normalization result of the obtained DTW distance is used as the treatment similarity factor between the current patient's treatment plan and the reference treatment plan for the reference patient.

[0033] It should be noted that, since the duration of treatment regimens may vary, it is necessary to conduct an overall reference analysis of the changes in similarity factors corresponding to the monitoring periods in the two treatment regimens through DTW matching.

[0034] Furthermore, the treatment similarity factor is assigned to several historical monitoring periods in the reference treatment plan of the reference patient, as the treatment similarity factor of each historical monitoring period in the reference treatment plan of the reference patient for each monitoring period (current monitoring period) in the current patient's treatment plan; then the treatment similarity factor is multiplied by the treatment reference coefficient, and the product is used as the treatment reference factor of the historical monitoring period of the medication index in the reference treatment plan of the reference patient for the current monitoring period of the current patient.

[0035] (2) Analyze the correlation between the changes in monitoring data of different medication indicators to obtain the personalized feedback adjustment coefficient of each medication indicator; construct reference data based on historical monitoring data and treatment reference factors, iteratively analyze the difference between actual monitoring data and reference data, and obtain prediction data by combining personalized feedback adjustment coefficient; based on ideal monitoring data and multiple adjusted prediction data, analyze the overall similarity under the same monitoring period through matrix analysis to obtain the final prediction matrix of the patient.

[0036] It should be noted that when adjusting the monitoring data of the patient's current treatment plan to obtain predictive data that is close to the ideal monitoring data and the historical monitoring data of similar physiological states, in addition to referring to the historical monitoring data and combining treatment reference factors, it is also necessary to consider the correlation between the data changes of different medication indicators. This should be used as the adjustment weight to avoid the loss of correlation between the monitoring data of multiple medication indicators after adjustment.

[0037] Preferably, in one embodiment of the present invention, the method for analyzing the correlation between changes in monitoring data of different medication indicators to obtain personalized feedback adjustment coefficients for each medication indicator includes: For any medication indicator of a current patient, several indicator data points during the current monitoring period are used to construct an indicator sequence for that medication indicator during the current monitoring period, arranged chronologically. The cosine similarity of the indicator sequences of any two medication indicators of the current patient during the current monitoring period is obtained and used as the correlation between the changes in those two medication indicators during the current monitoring period. The mean of the correlation between the changes in those two medication indicators of all patients across all monitoring periods is calculated and used as the correlation coefficient between the changes in those two medication indicators. Any medication indicator is used as the current medication indicator, and the mean of the correlation coefficients between the current medication indicator and other medication indicators is used as the personalized feedback adjustment coefficient for the current medication indicator.

[0038] It should be further explained that, based on historical monitoring data and treatment reference factors, the reference data is constructed by weighting the data at the same location in each historical monitoring period, and the initial treatment difference is obtained from this data. If the treatment difference is large, the adjustment is more necessary. During the adjustment process, the personalized feedback adjustment coefficient is considered to reduce the difference while avoiding serious damage to the correlation.

[0039] Preferably, in one embodiment of the present invention, reference data is constructed based on historical monitoring data and treatment reference factors, the difference between actual monitoring data and reference data is iteratively analyzed, and predicted data is obtained by combining personalized feedback adjustment coefficients. The specific method includes: For any actual monitoring data of the current medication index, the corresponding index data at the same position in each historical monitoring period of each patient's current medication index in the historical monitoring data (since the length of the historical monitoring period may be inconsistent, if it does not exist, it is supplemented by linear interpolation) is weighted and averaged with the corresponding treatment reference factor of the historical monitoring period as the weight. The result is used as the reference data of the actual monitoring data. The absolute value of the difference between the actual monitoring data and the reference data is obtained as the initial treatment difference of the actual monitoring data. The initial treatment difference of each actual monitoring data in the current monitoring period is obtained, and a difference threshold is preset. In this embodiment, the ratio of the initial treatment difference to the monitoring data is 10%. If there is actual monitoring data with an initial treatment difference greater than the difference threshold, the actual monitoring data is multiplied by the personalized feedback adjustment coefficient of the current medication index to obtain the adjusted actual monitoring data, which is used as the prediction data for the corresponding adjustment. The treatment difference is recalculated until there is no prediction data greater than the difference threshold. The final prediction data is used as the final prediction data of the current patient's current medication index in the current monitoring period.

[0040] It should be further explained that the process of adjusting any monitoring data multiple times until the final predicted data is obtained is actually a process of multiple iterations to eliminate the differences between individual differences and adherence to medical advice. Ideal monitoring data, on the other hand, is the data performance of multiple medication indicators under the initial treatment plan, which varies with time. By performing similarity analysis between the effective monitoring matrix and the ideal monitoring matrix, a feedback optimization coefficient is introduced on the basis of the predicted data adjustment. Starting from the first adjustment, after threshold judgment, some monitoring data need to be adjusted to obtain the predicted data, and this is done through personalized feedback adjustment coefficients. At the same time, the corresponding adjustment has obtained the feedback optimization coefficient. The larger the feedback optimization coefficient, the smaller the adjustment range required, and finally the final prediction matrix is ​​obtained.

[0041] Preferably, in one embodiment of the present invention, based on ideal monitoring data and multiple adjusted prediction data, the final prediction matrix for the patient is obtained by matrix analysis of the overall similarity within the same monitoring period. The specific method includes: For several predicted data and unadjusted actual monitoring data for the current monitoring period under the same adjustment, a matrix is ​​constructed with time as the column and medication index as the row, which serves as the effective monitoring matrix for the corresponding adjustment for the patient. Based on the ideal monitoring data of multiple medication indicators corresponding to the current patient's treatment plan, an ideal monitoring matrix is ​​constructed. The ideal monitoring matrix and the effective monitoring matrix for the corresponding adjustment are connected row by row to form a sequence for similarity calculation, and the obtained similarity is used as the feedback optimization coefficient for the corresponding adjustment.

[0042] Furthermore, starting from the first adjustment, after threshold judgment, the feedback optimization coefficient is subtracted from 1 and then multiplied by the personalized feedback adjustment coefficient to obtain the new adjustment coefficient. This is then used for the second adjustment, and so on, until the final prediction data after the final adjustment is obtained, forming the final prediction matrix. It is worth noting that a preset adjustment threshold is used. In this embodiment, the adjustment threshold is described as 0.05. If the obtained adjustment coefficient is less than the adjustment threshold, the adjustment threshold is used as the adjustment coefficient to avoid the adjustment process being affected by an excessively small adjustment coefficient.

[0043] The treatment plan recommendation module 103, based on the current patient's final prediction matrix and the monitoring matrices of other patients, obtains the recommendation scores of other patients for the current patient through collaborative filtering, and makes recommendations in combination with the treatment plans of other patients.

[0044] It should be noted that during the personalized, iterative adjustment process, the similarity between the adjusted predicted data and the ideal monitoring data is considered, and this similarity is used in the feedback iterative adjustment process to make the monitoring data gradually approach the ideal monitoring data. Subsequently, the final prediction matrix and the ideal monitoring matrix are combined with the monitoring matrix performance of other patients to achieve personalized recommendations for the current patient's prevention and treatment plan through a collaborative filtering algorithm.

[0045] Specifically, several historical patients are treated as other patients. For any one of these other patients, historical monitoring data of various medication indicators at multiple time points corresponding to their treatment plan are used to construct a matrix with time points as columns and medication indicators as rows, thus obtaining the monitoring matrix for that other patient. The monitoring matrices for all other patients are obtained using the same method. A similarity analysis is performed between the monitoring matrices of the other patients and the final prediction matrix of the current patient. A collaborative filtering algorithm is used to obtain the recommendation scores for the monitoring matrices of the other patients and the final prediction matrix of the current patient based on similarity. The matrix similarity calculation and collaborative filtering for obtaining the recommendation scores are existing technologies corresponding to the algorithm and will not be elaborated in this embodiment. The treatment plans corresponding to each other patient are displayed to the doctor in descending order of recommendation scores. This assists the doctor in adjusting and optimizing the personalized prevention and treatment plan for the current patient based on the patient's adherence to medical advice and the individualization of the condition, according to the displayed treatment plan. The specific adjustment process is not the focus of this embodiment and will not be elaborated in this embodiment.

[0046] This concludes the embodiment.

[0047] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A personalized prevention and treatment recommendation system for recurrent urinary tract stones, characterized in that, The system includes: The patient data acquisition module is used to collect ideal monitoring data of various medication indicators in the prevention and treatment plan between the doctor and the patient, as well as actual monitoring data of various medication indicators of the patient, and to record historical monitoring data of various medication indicators and treatment plans of several patients. The patient data analysis module is used to analyze the periodic changes of monitoring data of various medication indicators of patients, and obtain several monitoring periods and historical monitoring periods of patients; based on the historical monitoring data, it analyzes the correlation of indicator data changes in different monitoring periods under the same medication indicators, and obtains the treatment reference coefficient of historical monitoring periods; combined with the similarity of the overall changes of monitoring data in different treatment plans, it obtains the treatment reference factor of historical monitoring periods. The correlation between changes in monitoring data of different medication indicators was analyzed to obtain the personalized feedback adjustment coefficient for each medication indicator; reference data was constructed based on historical monitoring data and treatment reference factors, and the difference between actual monitoring data and reference data was iteratively analyzed. Predictive data was obtained by combining the personalized feedback adjustment coefficient; based on ideal monitoring data and multiple adjusted predictive data, the overall similarity under the same monitoring period was analyzed by matrix analysis to obtain the final prediction matrix for patients. The treatment plan recommendation module is used to obtain recommendation scores from other patients for the current patient based on the current patient's final prediction matrix and the monitoring matrices of other patients, and then make recommendations based on the treatment plans of other patients.

2. The personalized prevention and treatment recommendation system for recurrent urinary tract stones according to claim 1, characterized in that, The specific method for obtaining the patient's monitoring periods and historical monitoring periods is as follows: The actual monitoring data of any medication indicator in the current patient's treatment plan constitutes the monitoring sequence of that medication indicator for the current patient. The monitoring sequence is decomposed into a trend term, a periodic term, and a residual term by STL time series decomposition. The periodic term is subjected to Fourier transform to obtain the spectrum of the periodic term of the monitoring sequence. The amplitude of each frequency in the spectrum is obtained. The reciprocal of the frequency corresponding to the maximum amplitude is taken as the monitoring period of that medication indicator in the current patient's treatment plan. Obtain the monitoring cycle of each medication indicator in the current patient's treatment plan, and obtain the minimum value among all monitoring cycles as the current patient's monitoring period; divide all the time points corresponding to the monitoring sequence of each medication indicator in the current patient's treatment plan according to the monitoring period to obtain several monitoring time periods for the current patient; Historical monitoring data of each patient's medication indicators were used to divide the monitoring period for each patient's treatment plan according to the corresponding treatment plan, thus obtaining several historical monitoring periods for each patient.

3. The personalized prevention and treatment recommendation system for recurrent urinary tract stones according to claim 1, characterized in that, The method for analyzing the correlation of changes in indicator data of the same medication indicators across different monitoring periods based on historical monitoring data to obtain the treatment reference coefficient for the historical monitoring period includes the following specific methods: By standardizing all monitoring data of the same medication indicator for patients, several indicator data of each medication indicator for patients are obtained. The current monitoring period is defined as any one of several monitoring periods obtained from the current patient's monitoring data at the corresponding time. For any medication indicator of the current patient in the current monitoring period, several indicator data are constructed according to time sequence to form an indicator sequence for the current patient's medication indicator in the current monitoring period. Several indicator data of any medication indicator of any historical patient in any historical monitoring period are obtained and constructed according to time sequence to form a historical indicator sequence for the patient's medication indicator in that historical monitoring period. The indicator sequence and the historical indicator sequence are matched by DTW to obtain the DTW distance. The inversely proportional normalized result of the obtained DTW distance is used as a reference factor for the change of the patient's medication indicator in that historical monitoring period to the current patient's current monitoring period. Based on the trend changes of current patients’ actual monitoring data and historical monitoring data under the same medication indicators, and the time distribution of the corresponding monitoring periods, the period reference weights of the same medication indicators for different patients in history for each monitoring period are obtained. The product of the period reference weight of the patient's medication indicator and the historical monitoring period with respect to the current patient's current monitoring period and the change reference factor is used as the treatment reference coefficient of the patient's medication indicator and the historical monitoring period with respect to the current patient's current monitoring period.

4. The personalized prevention and treatment recommendation system for recurrent urinary tract stones according to claim 3, characterized in that, The specific methods for obtaining several indicator data of various medication indicators for patients include: All actual and historical monitoring data of any medication indicator for all patients were standardized, and the results were used as the indicator data for each monitoring period and historical monitoring period of that medication indicator.

5. The personalized prevention and treatment recommendation system for recurrent urinary tract stones according to claim 3, characterized in that, The specific method for obtaining the period reference weights of the same medication indicators for different patients in history for each monitoring period is as follows: Based on the current monitoring period of any medication indicator of the current patient, the trend sequence of the current patient's medication indicator during the current monitoring period and the trend sequence of the adjacent previous monitoring period are extracted from the trend item of the monitoring sequence of the medication indicator. The difference sequence obtained by subtracting the trend sequence of the adjacent previous monitoring period from the trend sequence of the current monitoring period is used as the trend change sequence of the current monitoring period. Obtain the trend sequence of any patient's medication indicator for any historical monitoring period, as well as the corresponding trend change sequence; The trend change sequence of the current monitoring period is matched with the trend change sequence of the patient's medication index in the historical monitoring period using DTW to obtain the DTW distance. The inversely proportional normalized result of the obtained DTW distance is used as the period reference weight of the patient's medication index in the current monitoring period to the current patient's current monitoring period.

6. The personalized prevention and treatment recommendation system for recurrent urinary tract stones according to claim 3, characterized in that, The specific method for obtaining the treatment reference factors for the historical monitoring period is as follows: For the current patient's treatment plan, several monitoring periods corresponding to any treatment plan for any patient are used as several historical monitoring periods corresponding to the reference treatment plan for the reference patient. The monitoring periods in the current patient's treatment plan are arranged in chronological order to form the monitoring period sequence of the current patient's treatment plan. The historical monitoring period sequence of the reference treatment plan for the reference patient is obtained. The monitoring period sequence and the historical monitoring period sequence are matched by DTW based on the change reference factor of the medication index between the monitoring period and the historical monitoring period, and the DTW distance is obtained. The inverse proportional normalization result of the obtained DTW distance is used as the treatment similarity factor between the current patient's treatment plan and the reference treatment plan for the reference patient. The treatment similarity factor is assigned to several historical monitoring periods in the reference treatment plan of the reference patient, as the treatment similarity factor between each historical monitoring period in the reference treatment plan of the reference patient and each monitoring period in the current patient's treatment plan; The treatment similarity factor is multiplied by the treatment reference coefficient, and the product is used as the treatment reference factor for the current patient's current monitoring period for each historical monitoring period of the medication index in the reference treatment plan of the reference patient.

7. The personalized prevention and treatment recommendation system for recurrent urinary tract stones according to claim 4, characterized in that, The specific methods for obtaining the personalized feedback adjustment coefficients for each medication indicator are as follows: For any medication indicator of a current patient, several indicator data during the current monitoring period are arranged in chronological order to form an indicator sequence for that medication indicator during the current monitoring period. The cosine similarity of the indicator sequences of any two medication indicators of the current patient during the current monitoring period is obtained as the correlation between the changes of those two medication indicators during the current monitoring period. The mean of the correlation between the changes of those two medication indicators of all patients during all monitoring periods is calculated as the correlation between the changes of those two medication indicators. Any medication indicator is used as the current medication indicator, and the mean of the correlation between the changes of the current medication indicator and other medication indicators is used as the personalized feedback adjustment coefficient of the current medication indicator.

8. The personalized prevention and treatment recommendation system for recurrent urinary tract stones according to claim 7, characterized in that, The specific methods for constructing reference data based on historical monitoring data and treatment reference factors, iteratively analyzing the differences between actual monitoring data and reference data, and obtaining predicted data by combining personalized feedback adjustment coefficients include: For any actual monitoring data of current medication indicators, the corresponding indicator data of the actual monitoring data in the same position of each patient's current medication indicators in each historical monitoring period are weighted and averaged, with the corresponding treatment reference factor in the historical monitoring period as the weight. The result is used as the reference data of the actual monitoring data. The absolute value of the difference between the actual monitoring data and its reference data is obtained as the initial treatment difference of the actual monitoring data. The initial treatment differences of each actual monitoring data in the current monitoring period are obtained. If there are actual monitoring data with an initial treatment difference greater than a preset difference threshold, the actual monitoring data is multiplied by the personalized feedback adjustment coefficient of the current medication index to obtain the adjusted actual monitoring data, which is used as the prediction data for the corresponding adjustment. The treatment difference is recalculated until there is no prediction data greater than the difference threshold. The final prediction data is used as the final prediction data of the current patient's current medication index in the current monitoring period.

9. The personalized prevention and treatment recommendation system for recurrent urinary tract stones according to claim 8, characterized in that, The final prediction matrix for the patient is obtained using the following method: For several predicted data and unadjusted actual monitoring data for the current monitoring period under the same adjustment, a matrix is ​​constructed with time as the column and medication index as the row, which serves as the effective monitoring matrix for the patient in the corresponding adjustment; based on the ideal monitoring data of multiple medication indexes corresponding to the current patient's treatment plan, an ideal monitoring matrix is ​​constructed. Based on the ideal monitoring matrix and the effective monitoring matrix of the corresponding adjustment, the feedback optimization coefficient of each adjustment is obtained; Starting from the first adjustment, after the threshold judgment, the feedback optimization coefficient is subtracted from 1 and then multiplied with the personalized feedback adjustment coefficient to obtain the new adjustment coefficient. The second adjustment is then performed, and so on, until the final prediction data after the final adjustment is obtained, which constitutes the final prediction matrix.

10. A personalized prevention and treatment recommendation system for recurrent urinary tract stones according to claim 9, characterized in that, The specific method for obtaining the feedback optimization coefficients for each adjustment is as follows: For the ideal monitoring matrix and the corresponding effective monitoring matrix, a sequence is formed by connecting the first and last rows of the matrix to calculate the similarity. The obtained similarity is used as the feedback optimization coefficient for the corresponding adjustment.

Images