Physical examination multi-dimensional data analysis method and system based on deep learning

Abstract

The application discloses a physical examination multi-dimensional data analysis method and system based on deep learning, relates to the technical field of physical examination analysis, and has the technical scheme as follows: collecting multi-source physical examination original data, wherein the multi-source physical examination original data are associated with a physical examination implementation subject and a subject individual identifier; screening attention physical examination data of the multi-source physical examination original data, extracting actual detection characteristic values of the attention physical examination data, obtaining deviation data according to the actual detection characteristic values and dynamic reference standards corresponding to health monitoring dimension types, collecting risk information parameters of the deviation data corresponding to a risk subject individual, obtaining a health evaluation index of the risk subject individual according to a risk correlation coefficient of the risk information parameters and the deviation data; determining a service optimization level of the physical examination implementation subject according to the comprehensive evaluation index, and formulating a health management scheme of the risk subject individual according to the health evaluation index, so that the corresponding service optimization level can be matched conveniently.

Description

Technical Field

[0001] This invention relates to the field of physical examination analysis technology, and more specifically, to a method and system for multi-dimensional data analysis of physical examinations based on deep learning. Background Technology

[0002] Traditional deep learning-based multi-dimensional data analysis of health checkups often only draws conclusions about whether a single indicator of an individual is abnormal. It lacks the ability to integrate multi-source health checkup data and struggles to correlate it with the service characteristics of the health checkup providers. For example, differences in the accuracy of testing equipment and data collection standards among different health checkup institutions result in a lack of quantitative basis for assessing the quality of health checkup services. Furthermore, traditional health risk assessment methods rely on fixed reference thresholds, failing to consider characteristic differences and thus prone to misjudgment. Traditional methods also fail to establish a data link between the health checkup provider and the individual, making it difficult for health checkup institutions to identify areas for service improvement from group data. Consequently, individuals cannot receive health risk warnings through health checkup data. Summary of the Invention

[0003] In view of the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for multi-dimensional data analysis of physical examinations based on deep learning.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A deep learning-based method for multi-dimensional data analysis of physical examinations, comprising the following steps: Collect raw data from multiple sources of physical examinations, and the raw data from multiple sources of physical examinations are associated with the physical examination implementer and the examinee's identifier; The system filters the original data from multiple sources of physical examinations, extracts the actual detection feature values ​​of the physical examination data, obtains the deviation data based on the actual detection feature values ​​and the dynamic reference standards of the corresponding health monitoring dimensions, and obtains the risk correlation coefficient based on the actual detection feature values ​​and the deviation data. The first evaluation coefficient of the physical examination implementer is obtained based on the deviation parameters and risk correlation coefficients of the deviation data; the examinees corresponding to the deviation data are marked as high-risk examinees, and the second evaluation coefficient of the physical examination implementer is obtained based on the risk characteristic parameters of the high-risk examinees; the comprehensive evaluation index of the physical examination implementer is obtained based on the first evaluation coefficient and the second evaluation coefficient. Collect risk information parameters corresponding to deviation data of risky individuals, and obtain the health evaluation index of risky individuals based on the risk information parameters and the risk correlation coefficient of deviation data. The service optimization level of the physical examination implementer is determined based on the comprehensive evaluation index, and a health management plan is formulated for the individuals at risk of examination based on the health evaluation index.

[0005] Preferably, the dynamic reference standard includes dynamic threshold ranges of physiological parameters corresponding to the health monitoring dimension types.

[0006] Preferably, the screening of relevant physical examination data from multiple sources includes the following steps: Determine the direction of health impact corresponding to each physical examination indicator in the original data of multi-source physical examinations, and establish the correspondence between physical examination indicators and the direction of health impact; Identify the health risk-related indicators corresponding to the types of risky health problems; Based on the correspondence between physical examination indicators and the direction of health impact, target physical examination indicators corresponding to health risk-related indicators are matched. Extract the relevant health check data corresponding to the target health check indicators from the original data of multi-source health checkups.

[0007] Preferably, the actual detection feature values ​​of the health checkup data are extracted, and deviation data are obtained based on the actual detection feature values ​​and the dynamic reference standards of the corresponding health monitoring dimension type. This specifically includes the following steps: The actual detection feature values ​​of the physical examination data of interest are extracted according to the health monitoring dimension type; wherein, the actual detection feature values ​​include physiological test data, and the actual detection feature values ​​correspond to the dynamic threshold range of physiological parameters of the health monitoring dimension type; Actual detected feature values ​​that exceed the dynamic threshold range of physiological parameters are judged as deviation data.

[0008] Preferably, the risk correlation coefficient is obtained based on the actual detected feature values ​​and deviation data, specifically including the following steps: Determine the health monitoring dimension corresponding to the deviation data, and clarify the direction of deviation of the actual detected feature value under that health monitoring dimension; Retrieve the deviation level classification criteria corresponding to this health monitoring dimension, and classify the deviation level of the deviation data according to the deviation level classification criteria; Match the degree of correlation between the deviation level and the health risk. The risk correlation coefficient for this health monitoring dimension is determined by combining the deviation direction with the degree of correlation with health risk.

[0009] Preferably, the first evaluation coefficient of the physical examination implementer is obtained based on the deviation parameters and risk correlation coefficients of the deviation data, specifically including the following steps: The dimension risk weights of the deviation data are set according to the health monitoring dimension type corresponding to the deviation data; the dimension risk weights correspond to the priority of the impact of the health monitoring dimension type on human health. The comprehensive risk coefficient of the deviation data is obtained based on the risk correlation coefficient and dimensional risk weight of the deviation data; among which, the deviation parameters include the number of deviation data in the physical examination implementation entities, the distribution range of the deviation data, and the severity level of the deviation data; The first risk assessment coefficient for the physical examination implementation entity is obtained based on the number, distribution range, severity level, and comprehensive risk coefficient of the deviation data among the implementation entities. The total amount of original data from multiple sources of physical examinations is obtained from the physical examination implementation entities. The second risk assessment coefficient is obtained based on the total amount of data and the number of deviation data. The first evaluation coefficient of the physical examination implementer is obtained based on the first risk evaluation coefficient and the second risk evaluation coefficient.

[0010] Preferably, the second evaluation coefficient of the physical examination implementer is obtained based on the risk characteristic parameters of the high-risk individuals among the implementers of the physical examination, specifically including the following steps: The risk characteristic parameters include the current number of individuals at risk of examination in the physical examination implementation entity, the number of individuals in the same period in history, the overlap rate of risk symptoms, and the age and gender distribution characteristics of the at-risk population. The variation in the number of individuals under risk testing was determined by comparing the historical number of individuals in the same period with the current number. A comprehensive index of at-risk populations is obtained by combining the overlap rate of risk symptoms and the characteristics of age and gender distribution. The second evaluation coefficient for the physical examination implementer is obtained based on the magnitude of change and the comprehensive index of the risk population.

[0011] Preferably, the comprehensive evaluation index of the physical examination implementer is obtained based on the first evaluation coefficient and the second evaluation coefficient, specifically including the following steps: Set a first evaluation weight and a second evaluation weight; the first evaluation weight and the second evaluation weight are set according to the needs of physical examination service quality assessment; The comprehensive evaluation index corresponding to the physical examination implementation entity is obtained based on the first evaluation weight, the first evaluation coefficient, the second evaluation weight, and the second evaluation coefficient.

[0012] Preferably, the health evaluation index of the risk-examined individual is obtained based on the risk information parameters and the risk correlation coefficient of the deviation data, specifically including the following steps: The risk information parameters include the number of items in the deviation data corresponding to the risk-tested individual, the duration of the deviation, the correlation with past medical history, and the influence coefficient of lifestyle habits; The individual cumulative risk coefficient of the risk subject is obtained based on the duration of deviation, correlation with past medical history, and influence coefficient of lifestyle habits; Based on the number of items that deviate from the data, the cumulative risk coefficient of the risk-examined individual is obtained by calculating the comprehensive risk coefficient corresponding to the deviation data in the risk-examined individual. The health assessment index of the risk-examined individual is obtained based on the individual risk cumulative coefficient and the risk cumulative coefficient.

[0013] A deep learning-based multi-dimensional data analysis system for physical examinations includes: Data Acquisition Module: Collects raw data from multiple sources of physical examinations. The raw data from multiple sources of physical examinations is associated with the physical examination implementer and the examinee's identifier. The first processing module filters the health examination data of interest from the multi-source original health examination data, extracts the actual detection feature values ​​of the health examination data of interest, obtains the deviation data based on the actual detection feature values ​​and the dynamic reference standards of the corresponding health monitoring dimension type, and obtains the risk correlation coefficient based on the actual detection feature values ​​and the deviation data. The second processing module: Obtains the first evaluation coefficient of the physical examination implementer based on the deviation parameters and risk correlation coefficients of the deviation data; identifies the examinees corresponding to the deviation data as high-risk examinees; obtains the second evaluation coefficient of the physical examination implementer based on the risk characteristic parameters of the high-risk examinees; and obtains the comprehensive evaluation index of the physical examination implementer based on the first and second evaluation coefficients. The third processing module collects risk information parameters corresponding to the deviation data of the risk-tested individuals, and obtains the health evaluation index of the risk-tested individuals based on the risk information parameters and the risk correlation coefficient of the deviation data. The module develops a health management plan for individuals at risk based on the health evaluation index, determining the service optimization level of the physical examination implementation entity and the health management plan for individuals at risk based on the health evaluation index.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This invention collects multi-source data by associating it with the identifiers of the entities implementing health checkups, and quantifies the service quality of these institutions through a comprehensive evaluation index. For example, if a health checkup center has a low comprehensive evaluation index, it can be clearly identified that its service shortcomings are concentrated in the testing items where the data deviates significantly, avoiding the ambiguity of previous service evaluations and making service optimization more targeted. The combination of the first and second evaluation coefficients covers risk characteristics at the data level and incorporates the distribution patterns of at-risk examinees, making the comprehensive evaluation more complete. This helps health checkup providers understand their overall service level and thus match the corresponding service optimization level, thereby improving the effectiveness of health management. Attached Figure Description

[0015] Figure 1 A schematic diagram illustrating the steps of a deep learning-based multi-dimensional data analysis method for physical examinations, as provided in this embodiment of the invention; Figure 2 This is a schematic diagram of a deep learning-based multi-dimensional data analysis system for physical examinations, provided for embodiments of the present invention. Detailed Implementation

[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0018] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0019] Reference Figures 1-2 As shown.

[0020] The embodiments further illustrate the deep learning-based multi-dimensional data analysis method and system for physical examinations proposed in this invention.

[0021] A deep learning-based method for multi-dimensional data analysis of physical examinations, comprising the following steps: Collect raw data from multiple sources of physical examinations. The raw data from multiple sources of physical examinations are associated with the physical examination implementer and the examinee's identifier. First, multi-source physical examination raw data is collected. This multi-source physical examination raw data includes various physiological test results and basic information of the examinee. At the same time, each piece of data is associated with the corresponding physical examination implementation entity and examinee's identifier. For example, if a community physical examination center is the physical examination implementation entity, the blood routine, blood pressure, and blood sugar data collected by it for a 35-year-old examinee will be marked with the unique identifier of the physical examination center and the examinee's personal identifier.

[0022] This association method clearly identifies the source institution of each piece of data and also corresponds to a specific examinee. For example, when it is necessary to analyze the overall testing data quality of a certain health checkup center, all data of that institution can be filtered out by the identifier of the health checkup implementation entity; when it is necessary to track the health changes of a particular examinee, relevant records of their previous health checkups can be retrieved by the examinee's individual identifier, ensuring the relevance and traceability of subsequent analysis.

[0023] The system filters the original data from multiple sources of physical examinations, extracts the actual detection feature values ​​of the physical examination data, obtains the deviation data based on the actual detection feature values ​​and the dynamic reference standards of the corresponding health monitoring dimensions, and obtains the risk correlation coefficient based on the actual detection feature values ​​and the deviation data. The first evaluation coefficient of the physical examination implementer is obtained based on the deviation parameters and risk correlation coefficients of the deviation data; the examinees corresponding to the deviation data are marked as high-risk examinees, and the second evaluation coefficient of the physical examination implementer is obtained based on the risk characteristic parameters of the high-risk examinees; the comprehensive evaluation index of the physical examination implementer is obtained based on the first evaluation coefficient and the second evaluation coefficient. Collect risk information parameters corresponding to deviation data of risky individuals, and obtain the health evaluation index of risky individuals based on the risk information parameters and the risk correlation coefficient of deviation data. The service optimization level of the physical examination implementer is determined based on the comprehensive evaluation index, and a health management plan is formulated for the individuals at risk of examination based on the health evaluation index.

[0024] The comprehensive evaluation index is a quantitative result that integrates the deviation risk characteristics of the physical examination provider from the data and the characteristics of the at-risk examinee. Different index ranges correspond to different service optimization levels. For example, if the comprehensive evaluation index is 0.8 or above, the service optimization level is Level 1, indicating that the service quality of the physical examination provider is high, requiring only routine process calibration. If the index is between 0.5 and 0.8, the service optimization level is Level 2, requiring optimization of the accuracy of testing equipment or addition of verification steps to the testing process for items that deviate from the data focus. If the index is below 0.5, the service optimization level is Level 3, requiring a comprehensive review of core aspects such as the physical examination item settings and data collection standards. Taking a certain physical examination provider as an example, its comprehensive evaluation index is 0.48, corresponding to a Level 3 service optimization level. At this point, it is necessary to adjust the coverage of physical examination items and upgrade some testing equipment to improve data accuracy.

[0025] The Health Assessment Index (HAI) is a quantitative result integrating individual risk characteristics and deviations from data risk characteristics. A higher HAI value indicates a higher health risk, and differentiated plans are developed based on different index ranges. If the HAI is above 0.7, it indicates a high risk, and the plan includes recommendations for frequent follow-up examinations, such as requiring the individual to have monthly blood glucose and blood lipid tests, and incorporating the influence coefficient of lifestyle habits, adding a dietary adjustment recommendation to reduce daily refined sugar intake by 50 grams. If the index is between 0.4 and 0.7, it indicates a moderate risk, and the plan focuses on lifestyle interventions, such as recommending at least three 30-minute aerobic exercise sessions per week and a targeted physical examination every quarter. If the index is below 0.4, it indicates a low risk, and the plan should include reminders for daily healthy lifestyle habits. For example, for a risky individual with a HAI of 0.57, corresponding to moderate risk, the plan recommends brisk walking three times a week, and quarterly blood pressure and blood glucose checks, supplemented by guidance on a low-salt diet.

[0026] The dynamic reference standard includes the dynamic threshold range of physiological parameters corresponding to the health monitoring dimension types.

[0027] Health monitoring dimensions are a classification and integration of physical examination items. For example, blood glucose-related tests are categorized under the blood glucose monitoring dimension, and blood pressure-related tests are categorized under the blood pressure monitoring dimension. Different dimensions correspond to different physiological function modules of the human body. The dynamic threshold range of physiological parameters is a normal physiological parameter range that is dynamically adjusted for each health monitoring dimension in combination with multiple factors, including the age, gender, and regional population characteristics of the individual being examined.

[0028] In blood pressure monitoring, the dynamic threshold range for physiological parameters is set at 90-120 mmHg systolic and 60-80 mmHg diastolic for young adults. For older adults, the dynamic threshold range is appropriately widened, for example, 90-140 mmHg systolic and 60-90 mmHg diastolic, because changes in vascular elasticity with age alter the normal blood pressure range. In blood glucose monitoring, the dynamic threshold range for fasting blood glucose is lower for pregnant women than for the general adult population, to accommodate the physiological characteristics of this special group.

[0029] In actual analysis, the type of health monitoring dimension corresponding to the individual being tested is first determined, and then the dynamic threshold range of physiological parameters that are suitable for the individual characteristics under that dimension is retrieved. The actual detected characteristic value is compared with the dynamic threshold range of physiological parameters. If the actual detected value exceeds the dynamic threshold range of physiological parameters, it is judged as deviation data.

[0030] The process of filtering relevant physical examination data from multiple sources includes the following steps: Determine the direction of health impact corresponding to each physical examination indicator in the original data of multi-source physical examinations, and establish the correspondence between physical examination indicators and the direction of health impact; Identify the health risk-related indicators corresponding to the types of risky health problems; Based on the correspondence between physical examination indicators and the direction of health impact, target physical examination indicators corresponding to health risk-related indicators are matched. Extract the relevant health check data corresponding to the target health check indicators from the original data of multi-source health checkups.

[0031] First, determine the direction of health impact of each health checkup indicator and establish a corresponding relationship. Multi-source health checkup data includes numerous indicators, such as blood glucose, blood pressure, and blood lipids, each with a different direction of health impact. For example, blood glucose levels that are too high or too low will negatively affect health; similarly, high-density lipoprotein cholesterol (HDL-C) levels that are too low increase the risk of cardiovascular disease, while levels within a reasonable range are beneficial to health. By establishing this correspondence between each health checkup indicator and its direction of health impact, the positive and negative health ranges for each indicator can be clearly defined.

[0032] Identify the health risk-related indicators corresponding to different types of risky health problems. These types of health problems include diabetes and cardiovascular disease. For each type, identify the indicators directly related to it. For example, for the risky health problem type of diabetes, the corresponding health risk-related indicators include fasting blood glucose, two-hour postprandial blood glucose, and glycated hemoglobin; for cardiovascular disease, the corresponding health risk-related indicators include blood pressure, total cholesterol, and triglycerides.

[0033] Matching target health checkup indicators with health risk-related indicators. Based on the correspondence between health checkup indicators and the direction of their health impact, identified health risk-related indicators are matched. For example, for fasting blood glucose, a health risk-related indicator for diabetes, the corresponding fasting blood glucose level is found from the health checkup indicators. Simultaneously, it is confirmed that both excessively high and excessively low levels are detrimental to health, thus determining this indicator as the target health checkup indicator. Similarly, blood pressure, a health risk-related indicator for cardiovascular disease, is also determined as a target health checkup indicator through this matching process.

[0034] After determining the target health check indicators, the corresponding test data for all targets are screened from the multi-source raw health check data. This data is the focus of subsequent analysis. For example, if the target health check indicators include fasting blood glucose and blood pressure, the fasting blood glucose and blood pressure test results for all examinees are extracted from the raw data.

[0035] Extracting the actual detection feature values ​​from the health checkup data of interest, and obtaining the deviation data based on the actual detection feature values ​​and the dynamic reference standard of the corresponding health monitoring dimension type, specifically includes the following steps: The actual detection feature values ​​of the physical examination data of interest are extracted according to the health monitoring dimension type; among them, the actual detection feature values ​​include physiological test data, and the actual detection feature values ​​correspond to the dynamic threshold range of physiological parameters of the health monitoring dimension type; Actual detected feature values ​​that exceed the dynamic threshold range of physiological parameters are judged as deviation data.

[0036] Extracting actual detection feature values ​​from the health checkup data of interest. Health monitoring dimension types categorize health checkup items, such as blood glucose monitoring and blood pressure monitoring. Based on these dimension types, the corresponding actual detection feature values ​​are extracted from the health checkup data of interest. These actual detection feature values ​​are essentially physiological test data and are matched one-to-one with the dynamic threshold range of the physiological parameters for the corresponding health monitoring dimension type. For example, for the blood glucose monitoring dimension, the actual detection feature values ​​extracted from the health checkup data of interest are the individual's fasting blood glucose value and the blood glucose value two hours after a meal; for the blood pressure monitoring dimension, the extracted actual detection feature values ​​are the measured values ​​of systolic and diastolic blood pressure. These values ​​all correspond to the dynamically adjusted threshold range of the physiological parameters for their respective dimensions.

[0037] The actual measured value is compared with the dynamic threshold range of the physiological parameters for the corresponding health monitoring dimension. If the actual measured value exceeds the dynamic threshold range, it is considered off-target data. For example, if a young subject's blood glucose monitoring dimension dynamic threshold range is 3.9 to 6.1 mmol / L for fasting blood glucose, but their actual measured value is 7.2 mmol / L, then 7.2 mmol / L is considered off-target data. Similarly, if an elderly subject's blood pressure monitoring dimension dynamic threshold range is 90 to 140 mmHg for systolic blood pressure, but their actual measured systolic blood pressure is 150 mmHg, this value is considered off-target data.

[0038] The risk correlation coefficient is obtained based on the actual detected feature values ​​and deviation data, specifically including the following steps: Determine the health monitoring dimension corresponding to the deviation data, and clarify the direction of deviation of the actual detected feature value under that health monitoring dimension; Retrieve the deviation level classification criteria corresponding to this health monitoring dimension, and classify the deviation level of the deviation data according to the deviation level classification criteria; Match the degree of correlation between the deviation level and the health risk. The risk correlation coefficient for this health monitoring dimension is determined by combining the deviation direction with the degree of correlation with health risk.

[0039] First, determine the health monitoring dimension and direction of the deviation data. Each deviation data corresponds to a health monitoring dimension. For example, if a deviation data point is an abnormal fasting blood glucose value, it corresponds to the blood glucose monitoring dimension. First, clarify the blood glucose monitoring dimension, and then determine the deviation direction of the actual measured characteristic value. Taking the blood glucose monitoring dimension as an example, if the actual measured value is higher than the upper limit of the dynamic threshold of physiological parameters, the deviation direction is positive (too high); if it is lower than the lower limit of the threshold, the deviation direction is negative (to too low). In the blood pressure monitoring dimension, if the systolic blood pressure is higher than the upper limit of the threshold, the deviation direction is positive (to too high).

[0040] The deviation level classification criteria for the corresponding health monitoring dimension are retrieved. These criteria are based on the dimension's impact on health. For example, the classification criteria for blood glucose monitoring are: a difference of less than 10% between the actual measured value and the threshold range is considered mild deviation; a difference of 10% to 30% is considered moderate deviation; and a difference exceeding 30% is considered severe deviation. Assuming a subject's upper limit for fasting blood glucose is 6.1 mmol / L, and the actual measured value is 7.3 mmol / L, the difference is approximately 19.7%, then this deviation is classified as moderate.

[0041] The system matches the degree of health risk associated with each deviation level. Different deviation levels correspond to different degrees of health risk association; for example, a mild deviation corresponds to a low degree of health risk association, a moderate deviation to a medium degree, and a severe deviation to a high degree. Taking the moderately deviated blood glucose data as an example, its corresponding health risk association is medium.

[0042] The risk correlation coefficient is calculated by combining the weight of the deviation direction with the quantitative value of the degree of correlation with health risk. The formula is Ri = Dw × Hr, where Ri is the risk correlation coefficient for that health monitoring dimension, Dw is the weight of the deviation direction, and Hr is the quantitative value of the degree of correlation with health risk. Taking the blood glucose monitoring dimension as an example, the weight of a positive deviation that is too high is set to 1.0, and the quantitative value of the degree of correlation with health risk corresponding to a moderate deviation is 0.5. Then, the risk correlation coefficient Ri corresponding to this deviation data is 1.0 × 0.5 = 0.5. If the deviation level is severe, the corresponding Hr is 1.0, then Ri = 1.0 × 1.0 = 1.0. This intuitively shows that the more severe the deviation, the higher the risk correlation coefficient.

[0043] The first evaluation coefficient for the physical examination implementer is obtained based on the deviation parameters and risk correlation coefficients of the deviation data, specifically including the following steps: The dimension risk weights of the deviation data are set according to the health monitoring dimension type corresponding to the deviation data; the dimension risk weights correspond to the priority of the impact of the health monitoring dimension type on human health. The comprehensive risk coefficient of the deviation data is obtained based on the risk correlation coefficient and dimensional risk weight of the deviation data; among which, the deviation parameters include the number of deviation data in the physical examination implementation entities, the distribution range of the deviation data, and the severity level of the deviation data; The first risk assessment coefficient for the physical examination implementation entity is obtained based on the number, distribution range, severity level, and comprehensive risk coefficient of the deviation data among the implementation entities. The total amount of original data from multiple sources of physical examinations is obtained from the physical examination implementation entities. The second risk assessment coefficient is obtained based on the total amount of data and the number of deviation data. The first evaluation coefficient of the physical examination implementer is obtained based on the first risk evaluation coefficient and the second risk evaluation coefficient.

[0044] First, set the dimensional risk weights for the deviation data. Assign dimensional risk weights to the health monitoring dimensions corresponding to the deviation data. The dimensional risk weight directly corresponds to the priority of the dimension's impact on human health; the higher the priority, the larger the weight. For example, the cardiovascular-related blood pressure monitoring dimension has a high priority in its impact on health, so its dimensional risk weight is set to 0.8; while the trace element monitoring dimension has a lower priority, so its weight is set to 0.3.

[0045] The comprehensive risk coefficient of the deviation data is calculated by combining the risk correlation coefficient and the dimensional risk weight. The formula is Cr = Ri × Wi, where Cr is the comprehensive risk coefficient, Ri is the risk correlation coefficient corresponding to the deviation data, and Wi is the dimensional risk weight. Taking a deviation data point in the blood pressure monitoring dimension as an example, if its risk correlation coefficient Ri is 0.6 and the dimensional risk weight Wi is 0.8, then the comprehensive risk coefficient Cr = 0.6 × 0.8 = 0.48. The deviation parameters include the quantity, distribution range, and severity level of the deviation data in the physical examination implementation entity.

[0046] The first risk assessment coefficient is calculated by integrating the quantity, distribution range, severity level, and comprehensive risk coefficient of the deviation data within the physical examination provider. For the quantity of deviation data, a corresponding weight is assigned based on its proportion of the total number of tests conducted by the physical examination provider; for example, the more deviation data there are, the higher the corresponding weight coefficient. If a physical examination provider has 100 test data items, and 20 of them are deviation data, the weight coefficient for that quantity is set to 0.3; if the number of deviation data items reaches 40, the weight coefficient increases to 0.5.

[0047] The distribution range of the deviation data reflects the number of health monitoring dimensions covered by the deviation data. If the deviation data is concentrated in only one dimension (such as only blood glucose monitoring), the weight coefficient corresponding to the distribution range is low, and the distribution range weight coefficient is set to 0.2; if the deviation data covers three or more dimensions (such as blood glucose, blood pressure, and blood lipid monitoring), the distribution range weight coefficient is increased to 0.4.

[0048] The percentage of deviation data at different levels is statistically analyzed. For example, in the deviation data of a certain subject, mild deviation accounts for 40%, moderate deviation accounts for 30%, and severe deviation accounts for 30%. Quantitative values ​​are assigned to different levels, such as 0.2 for mild deviation, 0.5 for moderate deviation, and 1.0 for severe deviation. The weighted average of the severity is used as the quantitative parameter corresponding to the severity level. Assuming that the severity quantification value of this subject is 0.2×0.4+0.5×0.3+1.0×0.3=0.43.

[0049] The comprehensive risk coefficient is obtained by calculating the average of the comprehensive risk coefficients of all deviation data of the subject. For example, if the comprehensive risk coefficients of 5 deviation data of a certain subject are 0.4, 0.5, 0.6, 0.3 and 0.7 respectively, the average value is (0.4+0.5+0.6+0.3+0.7) / 5=0.5.

[0050] Each feature is assigned a corresponding weight, with the total weight being 1. For example, the weight for quantity is 0.2, the weight for distribution range is 0.2, the weight for severity level is 0.3, and the weight for comprehensive risk coefficient is 0.3. The first risk assessment coefficient Fr = quantity weight × quantity parameter + distribution range weight × distribution range parameter + severity weight × severity quantification value + comprehensive risk coefficient weight × comprehensive risk coefficient mean = 0.2 × 0.3 + 0.2 × 0.4 + 0.3 × 0.43 + 0.3 × 0.5 = 0.06 + 0.08 + 0.129 + 0.15 = 0.419, thus comprehensively reflecting the overall risk level of the physical examination implementation entity deviating from the data.

[0051] The second risk assessment coefficient is derived from the ratio of the total data volume of the physical examination implementer to the number of deviation data points. The calculation formula is Sr = Np / Nt, where Sr is the second risk assessment coefficient, Np is the number of deviation data points, and Nt is the total data volume of the original multi-source physical examination data. For example, if a physical examination implementer has a total data volume Nt of 1000 records and a deviation data volume Np of 50 records, then the second risk assessment coefficient Sr = 50 / 1000 = 0.05. The second risk assessment coefficient reflects the proportion of deviation data in the total data.

[0052] The first evaluation coefficient is obtained by combining the first and second risk evaluation coefficients. The calculation formula is E1 = a × Fr + b × Sr, where E1 is the first evaluation coefficient, Fr is the first risk evaluation coefficient, Sr is the second risk evaluation coefficient, and a and b are the weights of the two (a + b = 1). Assuming a is set to 0.7, b to 0.3, Fr to 0.65, and Sr to 0.05, then E1 = 0.7 × 0.65 + 0.3 × 0.05 = 0.455 + 0.015 = 0.47. The first evaluation coefficient comprehensively reflects the degree and proportion of risk of the physical examination implementation entity deviating from the data.

[0053] The second evaluation coefficient for the physical examination implementer is obtained based on the risk characteristic parameters of the high-risk individuals among the implementers of the physical examination. This process includes the following steps: Risk characteristic parameters include the current number of individuals at risk of examination among the entities conducting the physical examination, the number of individuals in the same historical period, the overlap rate of risk symptoms, and the age and gender distribution characteristics of the at-risk population. The variation in the number of individuals under risk testing was determined by comparing the historical number of individuals in the same period with the current number. A comprehensive index of at-risk populations is obtained by combining the overlap rate of risk symptoms and the characteristics of age and gender distribution. The second evaluation coefficient for the physical examination implementer is obtained based on the magnitude of change and the comprehensive index of the risk population.

[0054] The current number of individuals under risk assessment is the number of individuals with health deviation data detected by the physical examination entity at present; the number of individuals under risk assessment in the same historical period is the number of individuals under risk assessment detected by the entity in the same period in the past; the risk symptom overlap rate refers to the proportion of multiple individuals under risk assessment who simultaneously exhibit the same health symptoms corresponding to the deviation data; the age and gender distribution characteristics of the risk population is the degree of concentration of individuals under risk assessment in different age groups and genders.

[0055] The change in the number of individuals at risk of infection is calculated by comparing the historical number of individuals in the same period with the current number. The formula is V = (CH) / H, where V is the change, C is the current number of individuals, and H is the historical number of individuals in the same period. For example, if a medical examination provider currently has 50 individuals at risk (C) and 40 individuals in the same period (H) historically, then the change is V = (50-40) / 40 = 0.25, representing a 25% increase in the current number of individuals at risk compared to the historical period. If the current number is 30 individuals, then the change is V = (30-40) / 40 = -0.25, representing a 25% decrease in the number of individuals.

[0056] A comprehensive risk index is derived by combining the risk symptom overlap rate and age / gender distribution characteristics. For the risk symptom overlap rate, if 60 out of 100 at-risk individuals simultaneously exhibit symptoms of abnormal blood sugar levels, the overlap rate is 0.6. For the age / gender distribution characteristics, if the at-risk individuals are concentrated in men over 50 years old, this group has a higher priority for health risk, and the quantitative value for the age / gender distribution characteristic is set to 0.8; if the distribution is more dispersed, the quantitative value for the age / gender distribution characteristic is set to 0.4. Weights are assigned to these two characteristics, for example, an overlap rate weight of 0.6 and a distribution characteristic weight of 0.4. The comprehensive risk index is obtained by weighted summation, using the formula I = 0.6 × G + 0.4 × D, where I is the comprehensive index, G is the risk symptom overlap rate, and D is the quantitative value of the age / gender distribution characteristic. Substituting the data, we get I = 0.6 × 0.6 + 0.4 × 0.8 = 0.36 + 0.32 = 0.68.

[0057] A second evaluation coefficient is obtained by combining the magnitude of change and the comprehensive risk index of the population. Corresponding weights are assigned to both, for example, the magnitude of change is weighted at 0.4, the comprehensive index at 0.6, and the total weight is 1. The formula is E2 = 0.4 × V + 0.6 × I. If the magnitude of change V is 0.25 and the comprehensive risk index I is 0.68, then the second evaluation coefficient E2 = 0.4 × 0.25 + 0.6 × 0.68 = 0.1 + 0.408 = 0.508. This second evaluation coefficient comprehensively reflects the changing trend and risk level of the population being examined by the health checkup provider.

[0058] The comprehensive evaluation index of the physical examination implementer is obtained based on the first evaluation coefficient and the second evaluation coefficient, which specifically includes the following steps: Set a first evaluation weight and a second evaluation weight; the first evaluation weight and the second evaluation weight are set according to the needs of physical examination service quality assessment; The comprehensive evaluation index corresponding to the physical examination implementation entity is obtained based on the first evaluation weight, the first evaluation coefficient, the second evaluation weight, and the second evaluation coefficient.

[0059] First, set the primary evaluation weight and the secondary evaluation weight. The values ​​of these two weights are determined based on the specific needs of the health checkup service quality assessment. For example, when the assessment focuses more on the risk characteristics of the health checkup data itself, the primary evaluation weight will be set higher; if more attention is paid to changes in the risky population, the secondary evaluation weight will be larger. Furthermore, the sum of the primary and secondary evaluation weights must be 1. For instance, if a scenario prioritizes the risk characteristics of the data, the primary evaluation weight will be set to 0.6 and the secondary evaluation weight to 0.4; if more attention is paid to the risk characteristics of the population, the primary evaluation weight will be set to 0.4 and the secondary evaluation weight to 0.6.

[0060] The comprehensive evaluation index of the physical examination implementer is obtained by weighted summation of the first evaluation weight, the first evaluation coefficient, the second evaluation weight, and the second evaluation coefficient. The calculation formula is EI=a1×E1+a2×E2, where EI is the comprehensive evaluation index, a1 is the first evaluation weight, E1 is the first evaluation coefficient, a2 is the second evaluation weight, and E2 is the second evaluation coefficient.

[0061] Assuming that the first evaluation coefficient E1 of a certain physical examination implementation entity is 0.47 and the second evaluation coefficient E2 is 0.508, and the current assessment needs focus more on data risk characteristics, the first evaluation weight a1 is set to 0.6 and the second evaluation weight a2 is set to 0.4. Then the comprehensive evaluation index EI = 0.6 × 0.47 + 0.4 × 0.508 = 0.282 + 0.2032 = 0.4852.

[0062] If a1 is set to 0.4 and a2 is set to 0.6, then the comprehensive evaluation index EI = 0.4 × 0.47 + 0.6 × 0.508 = 0.188 + 0.3048 = 0.4928.

[0063] The health assessment index of the risk-examined individual is obtained based on the risk information parameters and the risk correlation coefficient of the deviation data, specifically including the following steps: Risk information parameters include the number of deviation data items for each risk-examined individual, the duration of the deviation, the correlation with past medical history, and the influence coefficient of lifestyle habits; The individual cumulative risk coefficient of the risk subject is obtained based on the duration of deviation, correlation with past medical history, and influence coefficient of lifestyle habits; Based on the number of items that deviate from the data, the cumulative risk coefficient of the risk-examined individual is obtained by calculating the comprehensive risk coefficient corresponding to the deviation data in the risk-examined individual. The health assessment index of the risk-examined individual is obtained based on the individual risk cumulative coefficient and the risk cumulative coefficient.

[0064] The number of items deviating from the data is the number of physical examination items with abnormalities for the examinee; the duration of deviation is the duration for which the same physical examination item has abnormalities for the examinee; the correlation of past medical history is the degree of correlation between the deviation data and the examinee's past diseases; and the influence coefficient of lifestyle habits is the degree of influence of the examinee's lifestyle habits on the deviation data.

[0065] The individual risk cumulative coefficient is the result of integrating the duration of deviation, correlation with past medical history, and influence coefficient of lifestyle habits. A quantitative value and weight are assigned to each dimension, with a total weight of 1, adjusted according to the degree of health risk impact. For example, a deviation duration of 3 months is quantified as 0.5 with a weight of 0.3; a high correlation with past medical history is quantified as 0.9 with a weight of 0.4; and a high influence coefficient of lifestyle habits is quantified as 0.8 with a weight of 0.3. The calculation formula is Ci = 0.3 × Td + 0.4 × Hd + 0.3 × Lh, where Ci is the individual risk cumulative coefficient, Td is the quantitative value of the deviation duration, Hd is the quantitative value of the correlation with past medical history, and Lh is the quantitative value of the influence coefficient of lifestyle habits. Substituting the data, we get Ci = 0.3 × 0.5 + 0.4 × 0.9 + 0.3 × 0.8 = 0.15 + 0.36 + 0.24 = 0.75.

[0066] The risk accumulation coefficient needs to be calculated by combining the number of items deviating from the data with the corresponding comprehensive risk coefficient. Assuming an individual has 2 items deviating from the data, with corresponding comprehensive risk coefficients of 0.6 and 0.7, the average comprehensive risk coefficient is calculated as (0.6 + 0.7) / 2 = 0.65. Weights are then assigned to the number of items; for example, the weight corresponding to item 2 is 0.6. The formula is Ar = Np × (ΣC) / n, where Ar is the risk accumulation coefficient, Np is the weight corresponding to the number of items, ΣC is the sum of the comprehensive risk coefficients of all deviating items, and n is the number of deviating items. Substituting the data, we get Ar = 0.6 × 0.65 = 0.39.

[0067] The health assessment index is obtained by weighting and integrating the individual risk cumulative coefficient and the risk cumulative coefficient. Weights are set for both, for example, the individual risk cumulative coefficient has a weight of 0.5, and the risk cumulative coefficient has a weight of 0.5. The calculation formula is Hi = 0.5 × Ci + 0.5 × Ar, where Hi is the health assessment index. Substituting the data, we get Hi = 0.5 × 0.75 + 0.5 × 0.39 = 0.375 + 0.195 = 0.57. The health assessment index directly reflects the degree of health risk of the individual being examined; a higher value indicates a higher health risk.

[0068] A deep learning-based multi-dimensional data analysis system for physical examinations includes: Data Acquisition Module: Collects raw data from multiple sources of physical examinations. The raw data from multiple sources of physical examinations is associated with the physical examination implementer and the examinee's identifier. The first processing module filters the health examination data of interest from the multi-source original health examination data, extracts the actual detection feature values ​​of the health examination data of interest, obtains the deviation data based on the actual detection feature values ​​and the dynamic reference standards of the corresponding health monitoring dimension type, and obtains the risk correlation coefficient based on the actual detection feature values ​​and the deviation data. The second processing module: Obtains the first evaluation coefficient of the physical examination implementer based on the deviation parameters and risk correlation coefficients of the deviation data; identifies the examinees corresponding to the deviation data as high-risk examinees; obtains the second evaluation coefficient of the physical examination implementer based on the risk characteristic parameters of the high-risk examinees; and obtains the comprehensive evaluation index of the physical examination implementer based on the first and second evaluation coefficients. The third processing module collects risk information parameters corresponding to the deviation data of the risk-tested individuals, and obtains the health evaluation index of the risk-tested individuals based on the risk information parameters and the risk correlation coefficient of the deviation data. The module develops a health management plan for individuals at risk based on the health evaluation index, determining the service optimization level of the physical examination implementation entity and the health management plan for individuals at risk based on the health evaluation index.

[0069] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0070] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as RGM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A deep learning-based multi-dimensional data analysis method for physical examinations, characterized in that, The method includes the following steps: Collect raw data from multiple sources of physical examinations, and the raw data from multiple sources of physical examinations are associated with the physical examination implementer and the examinee's identifier; The system filters the original data from multiple sources of physical examinations, extracts the actual detection feature values ​​of the physical examination data, obtains the deviation data based on the actual detection feature values ​​and the dynamic reference standards of the corresponding health monitoring dimensions, and obtains the risk correlation coefficient based on the actual detection feature values ​​and the deviation data. The first evaluation coefficient of the physical examination implementer is obtained based on the deviation parameters and risk correlation coefficients of the deviation data; the examinees corresponding to the deviation data are marked as high-risk examinees, and the second evaluation coefficient of the physical examination implementer is obtained based on the risk characteristic parameters of the high-risk examinees; the comprehensive evaluation index of the physical examination implementer is obtained based on the first evaluation coefficient and the second evaluation coefficient. Collect risk information parameters corresponding to deviation data of risky individuals, and obtain the health evaluation index of risky individuals based on the risk information parameters and the risk correlation coefficient of deviation data. The service optimization level of the physical examination implementer is determined based on the comprehensive evaluation index, and a health management plan is formulated for the individuals at risk of examination based on the health evaluation index.

2. The method for multi-dimensional data analysis of physical examinations based on deep learning according to claim 1, characterized in that, The dynamic reference standard includes the dynamic threshold range of physiological parameters corresponding to the health monitoring dimension types.

3. The method for multi-dimensional data analysis of physical examinations based on deep learning according to claim 1, characterized in that, The process of filtering relevant physical examination data from multiple sources includes the following steps: Determine the direction of health impact corresponding to each physical examination indicator in the original data of multi-source physical examinations, and establish the correspondence between physical examination indicators and the direction of health impact; Identify the health risk-related indicators corresponding to the types of risky health problems; Based on the correspondence between physical examination indicators and the direction of health impact, target physical examination indicators corresponding to health risk-related indicators are matched. Extract the relevant health check data corresponding to the target health check indicators from the original data of multi-source health checkups.

4. The method for multi-dimensional data analysis of physical examinations based on deep learning according to claim 2, characterized in that, Extracting the actual detection feature values ​​from the health checkup data of interest, and obtaining the deviation data based on the actual detection feature values ​​and the dynamic reference standard of the corresponding health monitoring dimension type, specifically includes the following steps: The actual detection feature values ​​of the physical examination data of interest are extracted according to the health monitoring dimension type; wherein, the actual detection feature values ​​include physiological test data, and the actual detection feature values ​​correspond to the dynamic threshold range of physiological parameters of the health monitoring dimension type; Actual detected feature values ​​that exceed the dynamic threshold range of physiological parameters are judged as deviation data.

5. The method for multi-dimensional data analysis of physical examinations based on deep learning according to claim 4, characterized in that, The risk correlation coefficient is obtained based on the actual detected feature values ​​and deviation data, specifically including the following steps: Determine the health monitoring dimension corresponding to the deviation data, and clarify the direction of deviation of the actual detected feature value under that health monitoring dimension; Retrieve the deviation level classification criteria corresponding to this health monitoring dimension, and classify the deviation level of the deviation data according to the deviation level classification criteria; Match the degree of correlation between the deviation level and the health risk. The risk correlation coefficient for this health monitoring dimension is determined by combining the deviation direction with the degree of correlation with health risk.

6. The method for multi-dimensional data analysis of physical examinations based on deep learning according to claim 5, characterized in that, The first evaluation coefficient for the physical examination implementer is obtained based on the deviation parameters and risk correlation coefficients of the deviation data, specifically including the following steps: The dimension risk weights of the deviation data are set according to the health monitoring dimension type corresponding to the deviation data; the dimension risk weights correspond to the priority of the impact of the health monitoring dimension type on human health. The comprehensive risk coefficient of the deviation data is obtained based on the risk correlation coefficient and dimensional risk weight of the deviation data; among which, the deviation parameters include the number of deviation data in the physical examination implementation entities, the distribution range of the deviation data, and the severity level of the deviation data; The first risk assessment coefficient for the physical examination implementation entity is obtained based on the number, distribution range, severity level, and comprehensive risk coefficient of the deviation data among the implementation entities. The total amount of original data from multiple sources of physical examinations is obtained from the physical examination implementation entities. The second risk assessment coefficient is obtained based on the total amount of data and the number of deviation data. The first evaluation coefficient of the physical examination implementer is obtained based on the first risk evaluation coefficient and the second risk evaluation coefficient.

7. The method for multi-dimensional data analysis of physical examinations based on deep learning according to claim 6, characterized in that, The second evaluation coefficient for the physical examination implementer is obtained based on the risk characteristic parameters of the high-risk individuals among the implementers of the physical examination. This process includes the following steps: The risk characteristic parameters include the current number of individuals at risk of examination in the physical examination implementation entity, the number of individuals in the same period in history, the overlap rate of risk symptoms, and the age and gender distribution characteristics of the at-risk population. The variation in the number of individuals under risk testing was determined by comparing the historical number of individuals in the same period with the current number. A comprehensive index of at-risk populations is obtained by combining the overlap rate of risk symptoms and the characteristics of age and gender distribution. The second evaluation coefficient for the physical examination implementer is obtained based on the magnitude of change and the comprehensive index of the risk population.

8. The method for multi-dimensional data analysis of physical examinations based on deep learning according to claim 7, characterized in that, The comprehensive evaluation index of the physical examination implementer is obtained based on the first evaluation coefficient and the second evaluation coefficient, which specifically includes the following steps: Set a first evaluation weight and a second evaluation weight; the first evaluation weight and the second evaluation weight are set according to the needs of physical examination service quality assessment; The comprehensive evaluation index corresponding to the physical examination implementation entity is obtained based on the first evaluation weight, the first evaluation coefficient, the second evaluation weight, and the second evaluation coefficient.

9. The method for multi-dimensional data analysis of physical examinations based on deep learning according to claim 8, characterized in that, The health assessment index of the risk-examined individual is obtained based on the risk information parameters and the risk correlation coefficient of the deviation data, specifically including the following steps: The risk information parameters include the number of items in the deviation data corresponding to the risk-tested individual, the duration of the deviation, the correlation with past medical history, and the influence coefficient of lifestyle habits; The individual cumulative risk coefficient of the risk subject is obtained based on the duration of deviation, correlation with past medical history, and influence coefficient of lifestyle habits; Based on the number of items that deviate from the data, the cumulative risk coefficient of the risk-examined individual is obtained by calculating the comprehensive risk coefficient corresponding to the deviation data in the risk-examined individual. The health assessment index of the risk-examined individual is obtained based on the individual risk cumulative coefficient and the risk cumulative coefficient.

10. A deep learning-based multi-dimensional data analysis system for physical examinations, applied to the deep learning-based multi-dimensional data analysis method for physical examinations as described in any one of claims 1-9, characterized in that, include: Data Acquisition Module: Collects raw data from multiple sources of physical examinations. The raw data from multiple sources of physical examinations is associated with the physical examination implementer and the examinee's identifier. The first processing module filters the health examination data of interest from the multi-source original health examination data, extracts the actual detection feature values ​​of the health examination data of interest, obtains the deviation data based on the actual detection feature values ​​and the dynamic reference standards of the corresponding health monitoring dimension type, and obtains the risk correlation coefficient based on the actual detection feature values ​​and the deviation data. The second processing module: obtains the first evaluation coefficient of the physical examination implementation entity based on the deviation parameters and risk correlation coefficient of the deviation data; Individuals whose data deviates from the target data are identified as high-risk individuals. The second evaluation coefficient of the physical examination implementer is obtained based on the risk characteristic parameters of the high-risk individuals. The comprehensive evaluation index of the physical examination implementer is obtained based on the first and second evaluation coefficients. The third processing module collects risk information parameters corresponding to the deviation data of the risk-tested individuals, and obtains the health evaluation index of the risk-tested individuals based on the risk information parameters and the risk correlation coefficient of the deviation data. The module develops a health management plan for individuals at risk based on the health evaluation index, determining the service optimization level of the physical examination implementation entity and the health management plan for individuals at risk based on the health evaluation index.

Images