Osteoporosis risk intelligent early warning method and system based on big database

Abstract

The invention relates to the technical field of medical information, in particular to an osteoporosis risk intelligent early warning method and system based on a big database. The method comprises the following steps: extracting user bone health data in a multi-source health database; evaluating a bone mass state based on the user bone health data; screening osteoporosis risk users in the user bone health data according to the bone mass state; inquiring a medical image of the osteoporosis risk user; extracting a bone mineral density value by using the medical image, and determining a bone mineral density change trend according to the bone mineral density value; detecting the degeneration degree of the bone microstructure by using a medical image; constructing an osteoporosis risk early warning model according to the bone mineral density change trend and the bone microstructure degeneration degree; and inputting the user bone health data into the osteoporosis risk early warning model to generate a risk early warning score. According to the invention, accurate identification and dynamic monitoring of the osteoporosis risk user are realized based on the medical information technology, and the accuracy and response rate of osteoporosis risk early warning are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical information technology, in particular to a big database-based osteoporosis risk intelligent early warning method and system. BACKGROUND

[0002] In the existing bone health monitoring and risk management technology, individual bone health status is usually analyzed by relying on a single type of health indicator or static evaluation method, such as identifying bone health risks based on only a single bone density detection result, fixed threshold grading rules or artificial experience judgment. In the application scenarios where the sample size is small or the individual bone health changes slowly, this kind of method can complete the basic risk screening and health evaluation, but in the scenarios of large-scale population, long-term health data accumulation and multi-source information parallel analysis, due to the significant individual differences, the diversity of bone health related indicators and the dynamic changes over time, the traditional method based on a single indicator or static rules is difficult to comprehensively and accurately reflect the change trend of bone health risk. Although the existing technology can evaluate the bone health status through bone density measurement or image analysis, the following problems generally exist: first, the existing evaluation methods mostly focus on static numerical judgment, lack of dynamic analysis capability for the change trend and structure feature evolution process of bone health indicators over time, and it is difficult to discover potential risk changes in time; second, it is difficult to effectively integrate bone density data, image structure features and historical health information and other multi-source data for comprehensive analysis, resulting in insufficient accuracy and pertinence of risk classification and early warning results; third, the existing bone health risk management process relies on artificial rules and experience parameters to a large extent, and the automation and intelligence level is limited, lacking the ability of continuous learning, intelligent analysis and risk early warning based on large-scale historical data. SUMMARY

[0003] Therefore, it is necessary to provide a big database-based osteoporosis risk intelligent early warning method and system to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, a big database-based osteoporosis risk intelligent early warning method comprises the following steps: Step S1: extracting user bone health data in a multi-source health database; evaluating bone mass status based on the user bone health data; screening osteoporosis risk users in the user bone health data according to the bone mass status; Step S2: querying medical images of the osteoporosis risk users; extracting bone density values from the medical images, and determining bone density change trends according to the bone density values; detecting bone microstructure degradation degrees by using the medical images; Step S3: constructing an osteoporosis risk early warning model according to the bone density change trends and the bone microstructure degradation degrees; inputting the user bone health data into the osteoporosis risk early warning model to generate a risk early warning score; Step S4: sending an early warning signal to the osteoporosis risk user to perform the osteoporosis risk intelligent early warning task.

[0005] Preferably, the present specification also provides an osteoporosis risk intelligent early warning system based on a large database for performing the osteoporosis risk intelligent early warning method based on a large database as described above, the osteoporosis risk intelligent early warning system based on a large database comprising: a bone mass assessment module for extracting user bone health data in a multi-source health database; assessing a bone mass state based on the user bone health data; screening an osteoporosis risk user in the user bone health data according to the bone mass state; a trend analysis module for querying medical images of the osteoporosis risk user; extracting bone density values using the medical images and determining a bone density change trend according to the bone density values; detecting a bone microstructure degradation degree using the medical images; a model construction module for constructing an osteoporosis risk early warning model according to the bone density change trend and the bone microstructure degradation degree; inputting the user bone health data into the osteoporosis risk early warning model to generate a risk early warning score; an intelligent early warning module for sending an early warning signal to the osteoporosis risk user to perform the osteoporosis risk intelligent early warning task.

[0006] The present application has the following beneficial effects: (1) In the bone health data collection and bone mass state assessment process, the user bone health data in the multi-source health database is systematically extracted and coded, including bone density values, bone resorption ratios, bone formation activities, and historical bone mass change trends, combined with bone metabolism imbalance analysis, bone microstructure degradation feature extraction, and age grading analysis, to realize accurate assessment of the user's bone mass state and screening of osteoporosis risk users, improving the systematicness and accuracy of bone health data processing.

[0007] (2) In the bone density change trend and bone microstructure degradation analysis process, based on bone density time series data trend fitting, bone trabecula thickness and spacing calculation, bone trabecula connectivity analysis, and bone microstructure integrity assessment, the quantification of osteoporosis risk related parameters is realized, and the bone density trend feature map and the bone microstructure degradation feature map are generated, and the accurate extraction of the comprehensive features of osteoporosis is realized through feature fusion, improving the data integrity and operability of osteoporosis risk assessment.

[0008] (3) In the process of constructing the osteoporosis risk warning model and generating the risk score, the bone density trend characteristics and bone microstructure degradation characteristics are weighted and fused, and the osteoporosis risk probability and grade are generated through the risk warning judgment layer to realize the automatic generation and accurate quantification of the osteoporosis risk warning, and through the historical report matching and signal writing into the multi-source health database, the complete closed loop of the risk warning information is realized, and the response speed and traceability of the osteoporosis risk warning are improved.

[0009] (4) In the osteoporosis risk user management and warning signal pushing link, based on the bone mass state coding threshold, age grading reference value and historical fracture data, an osteoporosis risk user index table is established to realize the accurate screening and grading management of the osteoporosis risk users, and through the warning grade signal pushing and database updating mechanism, the real-time distribution and visual management of the osteoporosis risk warning information are realized, and the fine level and intelligent level of the osteoporosis risk monitoring and management are improved. BRIEF DESCRIPTION OF DRAWINGS

[0010] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-restrictive embodiments made with reference to the accompanying drawings: Fig. 1 The step flowchart of the osteoporosis risk intelligent warning method based on a large database of the present application is shown in the figure. Fig. 2 The detailed step flowchart of step S14 in the present application is shown in the figure. Fig. 3 The detailed step flowchart of step S16 in the present application is shown in the figure. The implementation of the purpose of the present application, functional features and advantages will be further described with reference to the accompanying drawings. DETAILED DESCRIPTION

[0011] The technical method of the present application will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative efforts belong to the scope of protection of the present application.

[0012] In addition, the accompanying drawings are only schematic illustrations of the present application, and are not necessarily drawn to scale. The same reference signs in the drawings represent the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities, which do not necessarily have to correspond to physically or logically independent entities. The functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0013] It should be understood that, while terms such as "first," "second," and so on can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element without departing from the scope of the example embodiments. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0014] To achieve the above object, there is provided Figs. 1 to 3 The present application provides an intelligent early warning method for osteoporosis risk based on a large database, which comprises the following steps: Step S1: Extract user bone health data from a multi-source health database; evaluate bone mass status based on user bone health data; screen osteoporosis risk users from user bone health data according to bone mass status; In this embodiment, the multi-source health database is accessed in a structured manner, and the multi-source health database at least includes a physical examination database, an image examination database, a biochemical detection database, and a long-term health follow-up database. By setting a unified user unique identification code, the records in different databases are associated and matched to form a single user's bone health data set. The bone health data at least includes bone density detection records, bone metabolism related index records, age data, gender data, and historical timestamp information. In the data extraction process, for the bone density detection records, the bone density values corresponding to the detection sites are extracted, and the bone density values are stored in units of g / cm²; for the bone metabolism related indexes, the bone resorption marker values and the bone formation marker values are extracted and uniformly converted into a standardized value interval; for the time information, the detection dates are sorted to construct a time series data table. Based on the extracted bone density values and bone metabolism related indexes, the bone resorption ratio is calculated, which is obtained by the ratio of the bone resorption marker value to the bone formation marker value. In the bone mass status evaluation process, the bone density value is compared with the preset bone density threshold value, and the preset bone density threshold value is set according to age segmentation, for example, taking every 5 years as an age segment, and configuring a corresponding bone density reference interval for each age segment. If the user's bone density value is lower than the lower limit of the bone density reference interval of the corresponding age segment, it is marked as a low bone mass status. At the same time, the bone metabolism imbalance is judged in combination with the bone resorption ratio to form a bone mass status label. Finally, the osteoporosis risk users in the user bone health data are screened according to the bone mass status label, and a risk user list is generated for subsequent step calling.

[0015] Step S2: Query the medical image of the osteoporosis risk user; extract the bone density value from the medical image, and determine the bone density change trend according to the bone density value; detect the bone microstructure degradation degree by using the medical image; In this embodiment, for the osteoporosis risk users screened, the corresponding medical image data is queried according to the user unique identification code. The medical image data at least includes dual-energy X-ray image or other structure imaging data, the image data is stored in DICOM format, and contains spatial resolution parameter, gray depth parameter and acquisition time parameter.

[0016] In the bone density value extraction process, the medical image is subjected to gray scale correction processing, the correction parameters include gray scale offset and gray scale gain coefficient, the gray scale offset is set to 0, and the gray scale gain coefficient is set to the image device calibration value. A pre-set bone analysis region is located in the image, which is defined by a rectangular frame with fixed size, and the size of the rectangular frame is set in pixel units. The pixel gray scale values in the region are counted, and the gray scale values are converted into bone density values according to the image calibration curve to form bone density time series data.

[0017] In the bone density change trend determination process, the bone density time series data is arranged in time sequence, the bone density difference between adjacent time points is calculated, and the change rate sequence is obtained; further, a linear fitting method is used to fit the overall bone density data to obtain the trend slope parameter; at the same time, the standard deviation of the bone density value is calculated as the fluctuation amplitude parameter. According to the trend slope and the fluctuation amplitude parameter, the bone density change trend is quantitatively described.

[0018] In the bone microstructure degradation degree detection process, edge enhancement processing is performed on the medical image, and the enhancement parameters include a high-pass filter kernel with a kernel size of 3x3. The image is subjected to binaryzation processing, and the binaryzation threshold is set to the image gray scale mean value. The bone trabecula structure skeleton is extracted based on the binaryzation result, and the bone trabecula spacing is calculated, which is obtained by calculating the shortest pixel distance between adjacent skeleton branches. Further, the number of connected nodes and the number of broken nodes of the bone trabecula are counted, the connectivity index is calculated according to the node ratio, and finally the bone microstructure degradation degree parameter is formed.

[0019] Step S3: Construct an osteoporosis risk early warning model according to the bone density change trend and the bone microstructure degradation degree; input the user bone health data into the osteoporosis risk early warning model to generate a risk early warning score; In this embodiment, before constructing the osteoporosis risk early warning model, the bone density change trend parameter and the bone microstructure degradation degree parameter are subjected to unified dimension processing. The trend slope parameter is mapped to the 0-1 interval by maximum minimum normalization, the fluctuation amplitude parameter is standardized, and the bone microstructure degradation degree parameter is inversely mapped to the 0-1 interval according to the connectivity index.

[0020] In the model construction process, the processed bone density change trend parameter is taken as a first input vector, and the bone microstructure degradation degree parameter is taken as a second input vector. The model structure comprises a feature extraction unit, a feature fusion unit and a pre-warning judgment unit. The feature extraction unit performs convolution operation on the input vector, the convolution kernel size is set to 1x3, the step size is set to 1, and the continuous change feature is extracted; the feature fusion unit performs weighted summation on the two types of features, and the weighted coefficients are obtained by historical sample statistics, for example, the bone density trend weight is set to 0.6, and the bone microstructure weight is set to 0.4.

[0021] In the risk pre-warning score generation process, the fused comprehensive features are input into the pre-warning judgment unit, the judgment unit performs weighted accumulation operation and outputs the risk pre-warning score, and the risk pre-warning score is set to 0-100. The score is divided into multiple interval intervals according to the numerical interval, and each interval corresponds to a different pre-warning level code. Finally, the user bone health data is input into the pre-warning model to generate a corresponding risk pre-warning score, and stored in the user risk record table.

[0022] Step S4: sending the pre-warning signal to the osteoporosis risk user to perform the osteoporosis risk intelligent pre-warning task.

[0023] In this embodiment, according to the risk pre-warning score, the pre-set pre-warning level table is queried. The table is set to have multiple numerical intervals, for example, 0-30, 31-60, 61-100, and each interval corresponds to a unique pre-warning level identification code. According to the interval to which the risk pre-warning score belongs, the corresponding osteoporosis pre-warning level is determined.

[0024] The historical report data of the osteoporosis risk user is queried based on the unique identification code of the user, and the historical report at least includes the past risk record, the pre-warning level record and the time mark information. The current pre-warning level is matched and analyzed with the pre-warning level in the historical report to form a current pre-warning level signal data packet, and the data packet contains the user identification code, the pre-warning level identification code, the generation time stamp and the associated parameter index.

[0025] In the pre-warning signal sending process, the pre-warning level signal data packet is sent to the corresponding user terminal or management end through the pre-set information pushing interface, and the pre-warning record of this time is written in the multi-source health database synchronously. The writing content includes the risk pre-warning score, the pre-warning level identification code and the corresponding parameter source index, which is used for subsequent data tracing and historical analysis. Through the above operation, the execution process of the osteoporosis risk intelligent pre-warning task is completed.

[0026] Preferably, step S1 is specifically: Step S11: extracting the bone density parameter based on the user bone health data; In this embodiment, the data fields related to bone health in the multi-source health database are structurally parsed, and the multi-source health database at least includes an imaging examination database, a biochemical detection database, and a basic health information database. Through a unified data field mapping rule, the bone density detection records of different sources are aligned in the field to ensure the unit consistency of the bone density parameters, and g / cm2 is used as the storage unit. For each bone density detection record, the detection site identifier, the detection timestamp, and the corresponding bone density original value are extracted and written into the bone density parameter table.

[0027] In the bone density parameter extraction process, in the case that there are multiple records of the same user at the same detection time point, the device records matching the database calibration table are retained through the de-duplication processing according to the detection device number. If there are bone density data of the same user at multiple detection time points, the bone density time sequence data structure is formed in ascending order of the timestamp. The bone density parameter table at least includes four types of fields, such as a user unique identifier code, a detection site code, a bone density value, and a detection timestamp. Through the above-mentioned manner, the bone density parameter extraction operation based on the user bone health data is completed, and the extraction result is taken as the data input of the subsequent step.

[0028] Step S12: Comparing the bone density parameters with the preset bone density threshold value to screen the low bone density users; In this embodiment, the preset bone density threshold value is set in a hierarchical configuration manner, and the hierarchy at least includes two dimensions of age segmentation and gender grouping. The age segmentation is divided according to 5 years as an interval, such as 30-34 years old, 35-39 years old, 40-44 years old, etc.; the gender grouping is distinguished according to the gender field in the basic health information database. For each age segment and gender combination, the corresponding bone density lower threshold value is pre-stored in the threshold configuration table, which is derived from the mean value of historical statistical data minus a fixed offset, and the fixed offset is set to 0.15 g / cm2.

[0029] In the screening process, the extracted bone density parameters are associated and matched with the threshold configuration table, and the corresponding bone density threshold value is located according to the user age data and gender data. The bone density value and the corresponding threshold value are compared in value, if the bone density value is less than the lower threshold value, a low bone density mark field is written in the user bone health data, and the mark value is set to 1; if the bone density value is not less than the lower threshold value, the mark value is set to 0. Through the mark field, the screening of the low bone density users is completed, and the screening result is stored in the low bone density user list.

[0030] Step S13: Calculating the bone resorption ratio of the low bone density users; In this embodiment, for the low bone density users screened, the corresponding bone metabolism related index data is further extracted. The bone metabolism related index at least includes bone resorption marker value and bone formation marker value, both of which are derived from the biochemical detection database and are accompanied by detection time stamp. In order to ensure time consistency, only the bone metabolism detection records within 30 days of the bone density detection time interval are selected.

[0031] In the bone resorption ratio calculation process, the bone resorption marker value and the bone formation marker value are processed in units, and are converted to ng / mL. According to the same detection time point, the two types of indexes are paired, and if there are multiple records, they are matched according to the principle of closest time stamp. The bone resorption ratio is calculated by dividing the bone resorption marker value by the bone formation marker value, and the calculation result is rounded to two decimal places. After calculation, the bone resorption ratio, the user unique identification code and the detection time stamp are written into the bone resorption ratio data table as the basis data for subsequent bone metabolism analysis.

[0032] Step S14: bone metabolism imbalance analysis based on bone resorption ratio, obtaining bone metabolism imbalance data; In this embodiment, the bone metabolism imbalance analysis is based on the bone resorption ratio. A bone metabolism balance interval is preset in the system, and the upper and lower limits of the interval are determined by historical sample statistics, with the lower limit set to 0.80 and the upper limit set to 1.20. Each bone resorption ratio is compared and analyzed with the balance interval.

[0033] When the bone resorption ratio is higher than 1.20, the bone metabolism offset direction is recorded as absorption side offset, and the offset degree is calculated by subtracting 1.20 from the bone resorption ratio; when the bone resorption ratio is lower than 0.80, the bone metabolism offset direction is recorded as formation side offset, and the offset degree is obtained by subtracting the bone resorption ratio from 0.80; when the bone resorption ratio is between 0.80 and 1.20, the offset degree is recorded as 0. The offset direction and the offset degree are combined to form bone metabolism imbalance data, wherein the bone metabolism imbalance data at least includes offset direction identification code and offset degree value. The data is stored in the bone metabolism analysis table as a structured result of quantitatively describing the bone metabolism state.

[0034] Step S15: evaluating bone mass state according to bone metabolism imbalance data; In this embodiment, the bone mass state evaluation is based on the bone metabolism imbalance data. The offset degree value in the bone metabolism imbalance data is classified, and the classification threshold is set according to fixed intervals, for example, 0-0.20, 0.21-0.50, 0.51 and above three grade intervals. Each interval corresponds to a bone metabolism imbalance grade code.

[0035] The bone metabolism imbalance level code is combined with the extracted bone density value for joint analysis. A two-dimensional evaluation matrix is constructed to map the bone density value interval and the bone metabolism imbalance level code to generate a bone mass state code. The bone density value interval is divided by 0.05 g / cm2 as an interval, and the bone mass state code is represented in integer form. After the mapping is completed, the bone mass state code is written into the user bone health data record to form a standardized bone mass state data field, which provides a unified criterion for subsequent risk user screening.

[0036] Step S16: Screening the osteoporosis risk users in the user bone health data according to the bone mass state.

[0037] In this embodiment, a bone mass state screening threshold is preset in the system, and the threshold is set to be not less than 75 in the bone mass state code value, that is, the record with the code value ≥ 75 meets the screening condition. The threshold is derived from the distribution statistics of the user bone mass state code in the historical bone health database. All historical user bone mass state codes are arranged in ascending order, and the code value corresponding to the 75th percentile is taken as the screening threshold, and the threshold is written into the "bone mass state threshold" field of the screening rule table.

[0038] The bone mass state code field of all user bone health data is read one by one, and the bone mass state code is compared with the screening threshold 75. If the bone mass state code ≥ 75, the osteoporosis risk marker field is written in the user data, and the marker value is set to 1; if the bone mass state code < 75, the marker value is set to 0. After the marking is completed, the system screens the users with the osteoporosis risk marker field value of 1 to generate an osteoporosis risk user set, and writes the set into the risk user index table. The risk user index table includes user unique identification, bone mass state code, risk marker value and generation timestamp, which is used for data calling, risk score calculation and signal sending in the subsequent early warning process.

[0039] Preferably, step S14 is specifically: Step S141: determining the bone metabolism deviation degree based on the bone resorption ratio; In this embodiment, the bone resorption ratio data table is read in a structured manner, and each record includes at least user unique identification code, bone resorption ratio value and corresponding detection timestamp. A bone metabolism balance reference interval is preset in the system, which is obtained by statistically analyzing the bone resorption ratio of long-term stable samples in a large database. The median value 1.00 is taken as the reference value, and the upper and lower deviation tolerances are set, with the lower limit being 0.85 and the upper limit being 1.15.

[0040] During the offset calculation, each bone resorption ratio value is read and its difference from the baseline value of 1.00 is calculated using the formula: Bone Metabolism Offset Value = Bone Resorption Ratio - 1.00. The offset direction is determined based on the difference: a positive difference is recorded as "Resorption Side Offset"; a negative difference is recorded as "Forming Side Offset"; and a difference of 0 is recorded as "Equilibrium State". The offset degree value is the absolute value of the difference, uniformly rounded to three decimal places. Finally, the offset direction identifier, offset degree value, and original bone resorption ratio are written into the bone metabolism offset data table to form standardized bone metabolism offset degree data.

[0041] Step S142: Perform bone loss analysis based on the degree of bone metabolism shift to obtain bone loss data; In this embodiment, the system presets a reference range for bone loss analysis. This range is segmented according to the degree of deviation, specifically including four ranges: 0.000–0.050, 0.051–0.150, 0.151–0.300, and above 0.301. Each range corresponds to a bone loss intensity level code, with the code values ​​increasing sequentially.

[0042] During the analysis, bone metabolism shift data were read one by one, and bone loss analysis was performed only on records with a shift direction of "absorption side shift". The shift value was compared with the aforementioned interval threshold to determine the interval to which it belonged, and a corresponding bone loss intensity level code was assigned accordingly. Based on the shift value and the detection timestamp, the shift change rate per unit time was calculated. The change rate was calculated by dividing the current shift value by the number of days between two adjacent detection intervals, and the result was rounded to four decimal places. Bone loss data includes at least the bone loss intensity level code, the shift change rate value, and the detection timestamp, and is uniformly written into the bone loss data table as the basis for subsequent bone-related index assessment.

[0043] Step S143: Assess osteocalcin content based on bone loss data; In this embodiment, test records related to osteocalcin are extracted from a biochemical testing database. These records include at least the osteocalcin value, test timestamp, and testing unit. The osteocalcin values ​​are then uniformly converted to... As a standard storage unit, only osteocalcin test records with a timestamp no more than 20 days after the bone loss data are retained to ensure data consistency.

[0044] During the assessment process, the bone loss intensity level code is correlated with the osteocalcin value within the corresponding time period. The system has a preset osteocalcin reference range, determined based on long-term database statistics, with the lower limit set at [value missing]. The upper limit is set to Osteocalcin status codes are obtained by mapping osteocalcin values ​​to reference intervals. If the osteocalcin value is below the lower limit, it is recorded as a low interval code; if it is within the interval, it is recorded as a middle interval code; and if it is above the upper limit, it is recorded as a high interval code. The assessment results are written into the osteocalcin assessment data table in the form of osteocalcin status codes.

[0045] Step S144: Determine bone formation activity based on osteocalcin content to obtain bone formation activity data; In this embodiment, a bone formation activity mapping rule table is constructed in the system. This rule table takes osteocalcin status code as input and bone formation activity level code as output. The bone formation activity level is divided into three levels, corresponding to low, medium and high levels, and the code values ​​are set to 1, 2 and 3 respectively.

[0046] During processing, osteocalcin status codes are read one by one, and the corresponding bone formation activity level codes are directly matched according to the mapping rule table. Simultaneously, to introduce a time continuity factor, the bone formation activity level codes for three consecutive testing periods for the same user are weighted and calculated, with weight coefficients set sequentially to 0.2, 0.3, and 0.5. The weighted results are summed and rounded down to obtain the final bone formation activity data code. This code, along with the corresponding timestamp, is written into the bone formation activity data table to describe the bone formation-related data status.

[0047] Step S145: Perform bone metabolism imbalance analysis based on bone formation activity data to obtain bone metabolism imbalance data.

[0048] In this embodiment, bone formation activity data and bone loss intensity level codes are jointly read to construct a two-parameter analysis matrix. The system has a preset bone metabolism imbalance judgment matrix, where the horizontal axis represents the bone loss intensity level code and the vertical axis represents the bone formation activity level code, with each matrix unit corresponding to a bone metabolism imbalance level code.

[0049] During the analysis, corresponding cells in the matrix are located based on the bone loss intensity level code and the bone formation activity level code to retrieve the bone metabolism imbalance level code. The bone metabolism imbalance level code is represented as an integer, with the numerical value reflecting the degree of metabolic shift. This code, along with the detection timestamp and the user's unique identifier, is written into the bone metabolism imbalance data table to form structured bone metabolism imbalance data, providing a unified data input basis for subsequent bone mass status assessment and risk warning steps.

[0050] Preferably, step S16 specifically includes: Step S161: Perform age grading analysis based on bone mass status to obtain age grade data; In this embodiment, the user's basic information table is read from a multi-source health database, the date of birth field is extracted, and it is uniformly converted into a standard time format. The user's actual age is obtained by calculating the difference between the current system timestamp and the date of birth, with the age unit uniformly in "years" and rounded down. The age value is then range-mapped according to a pre-defined age classification rule table.

[0051] The age grading rule table is constructed through statistical analysis of user samples accumulated over a long period in the database. Age intervals are divided into 5-year units, specifically: 20-24 years old, 25-29 years old, 30-34 years old, 35-39 years old, 40-44 years old, 45-49 years old, 50-54 years old, 55-59 years old, 60-64 years old, 65-69 years old, and 70 years and above. Each age group corresponds to a unique age level code, with the code values ​​set sequentially in ascending order of age. The system compares the calculated age value with each of the above intervals. Upon successful matching, the corresponding age level data is generated, and the user's unique identifier, age value, and age level code are written into the age level data table as the basis for subsequent bone mass reference and actual value extraction.

[0052] Step S162: Determine bone mass reference values ​​for each age group based on age gradation data; In this embodiment, user samples from a multi-source health database that maintain stable bone mass over a long period are screened. Screening criteria include consistent bone mass markers for more than three consecutive years and the absence of abnormal fluctuations. Samples meeting these criteria are then grouped according to age-level coding.

[0053] Within each age group, corresponding bone mass parameters are extracted. These parameters are uniformly calculated using the standardized bone mass index derived from bone mineral density, with the unit being g / cm². The bone mass parameters within each group are sorted, and the median value for that age group is calculated, along with its upper and lower quartile ranges. The median value is used as the reference bone mass value for that age group, and the quartile range is recorded as the boundary of the reference interval. Each age group code corresponds to a specific bone mass reference value, which is stored in a fixed numerical form and does not change with individual user data. Finally, the age group codes and their corresponding bone mass reference values ​​are written into a bone mass reference value table, forming a standardized age-stratified bone mass reference data system.

[0054] Step S163: Extract the actual bone mass values ​​for each age group based on the age group data; In this embodiment, the bone mass status data table is sorted by time. For each user, only the bone mass status record from the most recent testing period is retained to avoid interference from historical data. Users are grouped according to age level codes. Within each age level group, the corresponding user's bone mass parameter value is extracted. The bone mass parameter maintains the same calculation method and unit of measurement as the bone mass index used in step S162. For a single user, the actual bone mass value is the bone mass index value from the user's most recent testing period. The system associates and stores the user's unique identifier, age level code, and actual bone mass value, and writes them into the actual bone mass value data table. This data table maintains a field correspondence with the bone mass reference value table in structure, providing a unified data foundation for subsequent numerical comparison and risk marking.

[0055] Step S164: If the actual bone mass value of each age group is lower than the bone mass reference value of each age group, then mark them as first-risk users; obtain historical fracture users and mark them as second-risk users; In this embodiment, the labeling of the first-risk user and the second-risk user is based on a comparison of actual bone mass values ​​and reference bone mass values, as well as historical data. The system reads and associates the actual bone mass value table and the reference bone mass value table, with the association field being the age group code. For each user record, the actual bone mass value is compared and calculated with the reference bone mass value for the corresponding age group.

[0056] When the actual bone mass is less than the reference bone mass, the system generates a first risk marker and records the first risk user identifier code in the risk marker field. Bone-related historical records are extracted from the historical event record table of the multi-source health database; these records are stored using a unified encoding method. The system filters users with historical fracture records by matching their unique user identifier codes and marks these users as second-risk users. The first-risk user marker data and the second-risk user marker data are stored in separate data fields, each with a clear timestamp to ensure the accuracy of subsequent intersection calculations.

[0057] Step S165: Perform an intersection operation based on the first risk user and the second risk user to determine the osteoporosis risk user.

[0058] In this embodiment, the system reads the first risk user tag set and the second risk user tag set, respectively. Both sets use the user's unique identifier as the set element. The two sets are deduplicated to ensure that each user identifier appears only once in its respective set.

[0059] The system performs a set intersection operation, with the rule that an intersection result is generated only if the same user's unique identifier exists in both the first and second risk user sets. The system then marks the user identifiers in the intersection result as osteoporosis-risk users and writes this marking result into the osteoporosis-risk user data table. This data table serves as the input data source for subsequent early warning signal generation and transmission steps, completing the process of determining osteoporosis-risk users based on a joint screening of bone mass status and historical data.

[0060] Preferably, the step S2 of determining the trend of bone mineral density change based on the bone mineral density value specifically involves: Extract time-series bone mineral density data based on bone mineral density values; In this embodiment, bone density test records associated with the user's unique identifier are extracted from the bone health data table. These test records include at least a test timestamp, a test site identifier, and the corresponding bone density value. To ensure temporal consistency, only bone density records from the same test site are retained, for example, uniformly limited to the lumbar spine or proximal femur, avoiding data mixing between different test sites.

[0061] The extracted bone mineral density (BMD) values ​​were standardized by converting all values ​​to g / cm² as the standard unit of measurement. Records with missing detection timestamps or abnormal values ​​were removed; records with BMD values ​​less than 0.300 g / cm² or greater than 2.000 g / cm² were excluded from subsequent calculations. The cleaned data was then sorted in ascending order by detection timestamp to form a continuous BMD time series. Each time series data entry contained at least three fields: the detection time point, the corresponding BMD value, and the time series index number. Finally, the structured data was written into a BMD time series data table as the basic input for trend analysis.

[0062] Trend fitting analysis was performed based on time-series bone mineral density data to obtain a bone mineral density trend curve; In this embodiment, continuous time series records of the same user are read from the bone density time series data table. The series length is required to be no less than three detection cycles, and the detection cycle interval is calculated in calendar days, with a minimum interval set to 90 days. The detection time points are numerically processed, with the first detection time as the time origin, and subsequent detection times converted into days from the origin.

[0063] In the trend fitting process, the least squares method is used for linear fitting. Specifically, a linear function expression is constructed using time values ​​as the independent variable and bone density values ​​as the dependent variable. The coefficients of the fitted line are then obtained through least squares calculation. No nonlinear terms are introduced during the fitting process; only linear terms and constant terms are retained to ensure consistency in the trend calculation rules. After calculation, the fitted values ​​corresponding to all time points are connected in chronological order to form a continuous bone density trend curve. The trend curve is stored in parametric form, including slope parameters, intercept parameters, and corresponding time ranges, for subsequent feature extraction steps.

[0064] Extract the trend slope and fluctuation amplitude based on the bone mineral density trend curve; In this embodiment, the linear fitting slope parameter is directly read from the trend curve data. This slope parameter represents the change in bone mineral density value per unit time, and the unit of the slope is uniformly g / cm² / year. To facilitate uniform comparison, the slope parameter is converted from a daily scale to an annual scale by multiplying the slope value by 365.

[0065] The fluctuation amplitude is calculated based on the deviation between the actual bone mineral density time-series data and the trend curve. Specifically, for each detection time point, the difference between the actual bone mineral density value and the corresponding fitted value of the trend curve is calculated, and the absolute value of the difference is taken. The absolute differences for all detection points are summed and divided by the number of detection points to obtain the average deviation value, which is used as the bone mineral density fluctuation amplitude value. The unit of fluctuation amplitude is consistent with the bone mineral density value, which is g / cm². Finally, the trend slope and fluctuation amplitude are written as independent parameters into the trend feature data table.

[0066] The trend of bone density change is determined by the slope of the trend and the amplitude of the fluctuation.

[0067] In this embodiment, a trend slope determination threshold and a fluctuation amplitude determination threshold are preset in the system. The trend slope threshold is set to ±0.005 g / cm² / year, and the fluctuation amplitude threshold is set to 0.030 g / cm². These thresholds are derived from long-term sample statistical results in a large database and are stored in the form of fixed parameters.

[0068] During the determination process, the trend slope value is compared with positive and negative thresholds: when the slope is less than the negative threshold, it is recorded as a downward trend; when the slope is between the positive and negative thresholds, it is recorded as a stable trend; and when the slope is greater than the positive threshold, it is recorded as an upward trend. Simultaneously, the fluctuation amplitude value is compared with a fluctuation amplitude threshold. If the fluctuation amplitude is greater than the threshold, a high fluctuation marker is added to the trend type. Finally, the system generates a bone density change trend code based on the slope determination result and the fluctuation amplitude marker, and writes this code and the corresponding time period into the bone density change trend data table, completing the process of determining the bone density change trend.

[0069] Preferably, in step S2, the use of medical imaging to detect the degree of bone microstructural degeneration specifically involves: Medical images are preprocessed to obtain enhanced images; In this embodiment, medical image preprocessing uses digital image formats (such as DICOM format) as input, aiming to improve the resolvability of trabecular bone structures. The images are standardized in grayscale, normalizing the original grayscale values ​​to the range of 0-255 to unify the pixel dynamic range of images acquired by different devices. A high-pass filter is then applied to the standardized image, using a 3×3 pixel Laplacian operator to enhance trabecular bone edge details while suppressing low-frequency background noise. After filtering, histogram equalization is performed to generate an enhanced image, making the trabecular bone significantly brighter and more contrasting than the surrounding cancellous bone region. The final output of the enhanced image is a high-resolution two-dimensional matrix with the same pixel spacing as the original image and pixel values ​​ranging from 0 to 255. This enhanced image serves as input data for subsequent trabecular bone structure feature recognition.

[0070] Identify the trabecular bone structure features in enhanced images; calculate the trabecular spacing based on the trabecular bone structure features; In this embodiment, trabecular bone structure feature recognition is performed based on enhanced images. Edge extraction is performed on the enhanced images using an edge detection operator (such as the Canny operator). The edge detection thresholds are set to a low threshold of 50 and a high threshold of 150 to ensure the extraction of fine trabecular bone edges. The extracted edge map is binarized, with pixel values ​​above the edge threshold set to 1 and those below the threshold set to 0, forming a trabecular bone binary map. Morphological operations are applied to the binary map, including opening operations (3×3 structuring elements) to remove isolated noise and closing operations (5×5 structuring elements) to fill small gaps, thereby obtaining cleaned skeleton data. The cleaned skeleton data is the structural feature of the trabecular bone, used to calculate trabecular bone spacing and connectivity analysis. The output of the trabecular bone structural feature is a two-dimensional matrix, with matrix elements representing the presence of trabecular bone. The trabecular bone binary map is converted into a representation of trabecular bone centerlines, with each trabecular bone represented by its pixel centroid coordinates. A nearest neighbor search algorithm is performed on the centroid coordinates of adjacent trabecular bone to calculate the Euclidean distance between centroids. Spacing calculations are expressed in pixels and converted to physical units (millimeters) based on pixel resolution, with each pixel corresponding to a physical spacing parameter of 0.1 mm. For each pixel row or column, the trabecular spacing is calculated separately, generating a trabecular spacing matrix. Matrix elements represent the center-to-center spacing of the trabecular bones in the corresponding region, ultimately forming a trabecular spacing data table that can be used for connectivity analysis.

[0071] Connectivity analysis was performed based on the trabecular spacing to obtain connectivity data; In this embodiment, trabecular connectivity analysis uses the trabecular spacing matrix as input. The trabecular spacing matrix is ​​converted into an adjacency matrix, where any two trabecular nodes are considered connected if the center-to-center distance is less than 2.5 mm. Connectivity component analysis is performed on the adjacency matrix, counting the number of trabecular nodes in each connected component and calculating the maximum and average chain lengths of each component. Connectivity data includes: the number of trabecular connected components, the maximum chain length, the average chain length, and the connectivity component density (the proportion of connected nodes to the total number of trabecular nodes). This data is stored in a structured table, providing fundamental parameters for trabecular integrity assessment.

[0072] The integrity of trabecular bone is assessed based on connectivity data; the degree of bone microstructural degradation is determined based on the integrity of trabecular bone.

[0073] In this embodiment, trabecular integrity assessment is based on connectivity data. A trabecular integrity threshold is set; a decline in trabecular integrity is defined as a maximum chain length of less than 20 trabecular nodes or a connected component density below 0.6. An integrity index is calculated for each trabecular connected component based on its length and density: Integrity Index = 0.7 × (connected component density) + 0.3 × (maximum chain length / maximum observed chain length), with an index range of 0 to 1. The average integrity index is calculated for all components and output as the trabecular integrity value. Ultimately, the degree of bone microstructure degradation is encoded through the integrity value; a lower value indicates a higher degree of trabecular sparsity and fragmentation. The integrity value is stored as a floating-point number with three decimal places for input into the subsequent osteoporosis risk warning model.

[0074] Preferably, the specific features for identifying trabecular bone structure in enhanced images are as follows: The cancellous bone region is determined based on the enhanced image, and a region bounding box is set for the cancellous bone region to obtain the bounding box data; In this embodiment, the enhanced image is input as a high-resolution two-dimensional matrix, where the pixel value ranges from 0 to 255. Global thresholding is performed on the image, and regions with pixel values ​​below 120 are marked as candidate regions for cancellous bone. Connectivity analysis is performed on the candidate regions to identify contiguous regions with an area greater than 200 pixels as preliminary cancellous bone regions. For each contiguous region, its minimum bounding box is calculated. The coordinates of the four vertices of the bounding box are recorded in pixels and converted to physical units (millimeters), with the pixel spacing parameter set to 0.1 mm. The bounding box data includes the coordinates of the top-left corner (x_min, y_min), the bottom-right corner (x_max, y_max), width, and height information of the rectangle. Overlapping or close bounding boxes (horizontal or vertical spacing less than 5 pixels) are merged to ensure complete coverage of the cancellous bone region. The final output bounding box data is stored in tabular form, with each row corresponding to the bounding box coordinates and area information of a cancellous bone region, providing a reference range for subsequent binarization segmentation.

[0075] Binarization segmentation is performed based on the bounding box data to obtain a binary image of the trabecular bone. In this embodiment, bounding box data is used to perform local binarization processing on the enhanced image. For each bounding box region, the pixel values ​​within that region are extracted to form a local sub-matrix. Local threshold binarization processing is performed on the pixels within the sub-matrix. The threshold is the average pixel value of that region plus a brightness deviation of 15, used to distinguish between trabeculae and cancellous bone voids. The binarization rule is that a pixel value greater than the threshold is set to 1, representing trabeculae; a pixel value less than or equal to the threshold is set to 0, representing bone voids. After processing all bounding box regions, the local binary image is embedded into the original image matrix to generate a complete trabeculae binary image. The elements of the binary image matrix are either 0 or 1, and the pixel spacing is consistent with the original image (0.1 mm / pixel) to ensure the accuracy of subsequent calculations of trabeculae geometry and connectivity.

[0076] Based on the removal of pseudo-branches and burrs from the binary images of trabeculae, purified skeletal data is obtained; In this embodiment, morphological cleansing is performed on the binary image of the trabecular bone to remove pseudo-branches and spurs. A 3×3 pixel cross-shaped erosion operator is applied to the binary image to eliminate isolated noise points with a single pixel width. A 3×3 pixel cross-shaped dilation operator is applied to restore the main width of the trabecular bone. A skeletonization algorithm is then performed on the image after erosion and dilation to unify the trabecular bone width to a single pixel skeleton while preserving the topological structure of the trabecular bone. The trabecular bone skeleton image is then filtered using connected component length to remove isolated pseudo-branches with a length less than 5 pixels. After the cleansing process, the output cleaned skeleton data is a two-dimensional matrix, where a value of 0 represents the background and a value of 1 represents the trabecular bone skeleton. The pixel spacing is maintained at 0.1 mm / pixel, and the topological relationships and spatial positions of the trabecular bone are completely preserved, making it suitable for trabecular bone spacing and connectivity analysis.

[0077] The trabecular structure characteristics were determined based on the purified skeleton data.

[0078] In this embodiment, trabecular bone structural feature extraction uses purified skeleton data as input. A node extraction algorithm is performed on the trabecular bone skeleton map, marking skeleton intersections and endpoints as trabecular bone nodes. Node coordinates are recorded in pixels and converted to physical units. The length of each trabecular bone is calculated, equal to the number of consecutive pixels in the skeleton multiplied by the pixel spacing (0.1 mm). Adjacency analysis is performed on the spatial distribution between trabeculae, calculating the center distance, intersection angle, and number of connected components of adjacent trabecular bone. Trabecular bone thickness is determined by the local maximum width corresponding to each skeleton pixel in the binary trabecular bone map. Finally, a trabecular bone structural feature data table is output, including parameters such as trabecular bone length, thickness, connectivity, node coordinates, and spatial distribution. All data is stored in a structured format, providing basic parameters for subsequent bone microstructure degradation analysis.

[0079] Preferably, in step S3, the osteoporosis risk early warning model is constructed based on the trend of bone density change and the degree of bone microstructure degeneration as follows: An osteoporosis risk warning model is constructed based on the trend of bone density change and the degree of bone microstructure degeneration. The osteoporosis risk warning model includes a bone density trend feature extraction layer, a bone microstructure degeneration feature extraction layer, a feature fusion layer, and a risk warning judgment layer. The bone density trend feature extraction layer is used to perform convolution operations and activation mapping on the bone density change trend to extract the bone density change rate and long-term trend features, and generate a bone density trend feature map. In this embodiment, the bone density trend feature extraction layer takes time-series bone density data as input. This time-series data includes bone density values ​​measured by the user at different time points, in g / cm², with time intervals in months. The time-series data is normalized to scale the bone density values ​​to the range of 0-1, using the following formula: ,in For the bone mineral density at the i-th time point, and These represent the minimum and maximum values ​​of the user's bone density sequence, respectively. The normalized sequence is then processed using a one-dimensional convolution operation to extract the rate of change features. The convolution kernel length is set to 3, the stride to 1, and the kernel weights are initialized to a uniform distribution. The convolution result is mapped using the Rectified Linear Activation Function (ReLU), which sets negative values ​​to zero and retains positive features. The convolution output is then processed using sliding window average pooling, with a window length of 3 and a stride of 1, to smooth short-term fluctuations and preserve long-term trends. The final output bone density trend feature map is a normalized numerical matrix, where rows correspond to the time series, columns correspond to the convolution feature channels, and each element represents the trend strength of bone density at that time point, used for subsequent feature fusion analysis.

[0080] The bone microstructure degradation feature extraction layer is used to perform convolution operations and feature mapping on the degree of bone microstructure degradation, extract degradation features such as changes in trabecular thickness, increased trabecular spacing, and decreased bone structure integrity, and generate a bone microstructure degradation feature map. In this embodiment, the bone microstructure degradation feature extraction layer takes bone microstructure degradation data as input, including parameters such as trabecular thickness (unit: mm), trabecular spacing (unit: mm), and trabecular connectivity score (unitless, value 0~1). Thickness, spacing, and connectivity are normalized to the range of 0~1, using the following normalization formula: Where Xi is the original parameter value, and This represents the minimum and maximum values ​​of the parameter across all users or in the historical database. The normalized parameter matrix is ​​used to extract local spatial features through a two-dimensional convolution operation. The convolution kernel size is set to 3×3, the stride is 1, and the kernel weights are initialized to a Gaussian distribution. The convolution output is then activated using the ReLU activation function, setting negative values ​​to zero. Max pooling is applied to the convolution output with a 3×3 pooling window and a stride of 1 to highlight local features indicating increased trabecular spacing and decreased thickness. The final result is a bone microstructure degradation feature map. The matrix rows represent the spatial distribution of bone microstructures, and the columns represent different feature channel values. Each element reflects the local intensity of trabecular degradation.

[0081] The feature fusion layer is used to weightedly fuse the bone density trend feature map with the bone microstructure degeneration feature map to determine the comprehensive characteristics of osteoporosis; In this embodiment, the bone density trend feature map and the bone microstructure degeneration feature map are adjusted to the same size matrix. A linear interpolation method is used to unify the number of rows and columns in the feature maps, with rows corresponding to time points or spatial locations and columns corresponding to feature channels. The fusion formula is as follows: ,in This is a bone mineral density trend feature map. Image showing the characteristics of bone microstructure degeneration , The weighting coefficients reflect the contribution of bone mineral density changes to short-term trends, and the weights are obtained through historical data statistics. The fusion matrix processes negative values ​​through another ReLU activation mapping to ensure that the feature strengths are non-negative. Each row of the fusion matrix represents the comprehensive features at a specific time point or spatial location, and each column corresponds to the fusion channel value used in the risk assessment layer calculation. The final output osteoporosis comprehensive feature matrix has normalized values, complete row and column information, and retains the comprehensive features of bone mineral density changes and bone microstructure degeneration.

[0082] The risk warning and judgment layer is used to generate the osteoporosis risk probability and output the corresponding osteoporosis risk warning result.

[0083] In this embodiment, each row of the comprehensive feature matrix is ​​standardized, and the formula is as follows: ,in For matrix elements, and These represent the mean and standard deviation of the row, respectively. Then, the standardized matrix is ​​weighted and summed along the feature channel direction, with weight coefficients... The importance of different feature channels is determined by weights derived from the correlation analysis of historical features with risk indicators. The weighted summation result is mapped to the 0-1 interval using a Sigmoid function to obtain the risk probability value for each time point or spatial location. All probability values ​​are then weighted by time or space, with the average weight determined by the time interval or spatial area, ultimately yielding the osteoporosis risk probability. The risk probability is judged according to a set threshold of 0.7, outputting a binary warning result: 0 indicates low risk, and 1 indicates high risk. The probability values ​​and judgment results are recorded in a database for further analysis and operation.

[0084] Preferably, step S4 specifically includes: Step S41: Determine the osteoporosis warning level based on the warning signal; In this embodiment, an osteoporosis risk probability value is extracted from the warning signal. This value is a real number between 0 and 1, derived from the calculation results of the risk warning judgment layer. The risk probability value is compared with preset level thresholds: a low-level threshold is set to 0.3, a medium-level threshold to 0.6, and a high-level threshold to 0.85. Warning signals with a risk probability less than 0.3 are marked as "low-level," those between 0.3 and 0.6 as "medium-level," those between 0.6 and 0.85 as "high-level," and those greater than 0.85 as "very high-level." For each warning signal, the corresponding level information is written into the warning signal table, including a unique user identifier, risk probability value, level number, and timestamp. The timestamp is accurate to the second to ensure complete signal recording. The signal table uses a database index to accelerate queries. The index fields include the user ID and timestamp for easy subsequent data matching and sending operations.

[0085] Step S42: Query the historical reports of users at risk of osteoporosis; In this embodiment, historical reports of users at risk of osteoporosis are queried, and user bone health records are retrieved from a multi-source health database using the user's unique identifier. The historical reports in the database include bone mineral density (BMD) values, bone microstructure degradation parameters, and historical warning signals. Each record includes a generation timestamp, parameter values, and source type fields. The search results are sorted by timestamp, retrieving data for the past 12 months from the most recent record to ensure coverage of trends over the past year. For each record, the units for BMD values ​​are standardized to g / cm², trabecular bone thickness to millimeters, and trabecular bone spacing to millimeters for ease of subsequent calculation and comparison. When historical report data is extracted from different sources, the data structure is standardized using a field matching algorithm to form a unified matrix format, where rows represent time points, columns represent different parameter fields, and missing values ​​in the matrix are filled using linear interpolation to maintain data continuity.

[0086] Step S43: Match the osteoporosis warning level with historical reports and send the corresponding osteoporosis warning level signal; In this embodiment, historical reports are used to match osteoporosis warning levels. The risk probability is calculated by analyzing bone mineral density values ​​and bone microstructure parameters at various time points in the historical reports. The calculated risk probability is compared with a preset level threshold to determine the level number for each historical record. Then, the historical levels are matched with the current warning signal level in chronological order. The matching rule is to use the most recent historical level as a reference, and simultaneously calculate the average level change over three consecutive months. An average value less than 0.3 is classified as low level, 0.3 to 0.6 as medium level, 0.6 to 0.85 as high level, and greater than 0.85 as very high level. After matching is complete, the matching result is used as the final warning level. A complete record table is formed by the user ID, current risk probability, historical average level, matched level, and generation timestamp, serving as the basis for sending the osteoporosis warning level signal.

[0087] Step S44: Write the osteoporosis early warning level signal into the multi-source health database to perform the intelligent early warning task for osteoporosis risk.

[0088] In this embodiment, osteoporosis warning level signals are written into a multi-source health database. A corresponding table is created for the warning level record table according to the database field structure, including user ID, risk probability, historical average level, matching level, level number, timestamp, and record status fields. A transaction mechanism is used to write the data to the database in batches, ensuring atomicity and preventing data loss or duplicate writing. During the writing process, a unique index is created for the level number, and the user ID is used as the primary key index for easy querying and updating. After the database writing is complete, a log recording operation is performed, recording the total number of records written, the number of failed records, and the operation time. To ensure data integrity, verification operations are performed on the written data, including field value type verification, timestamp order verification, and level number consistency verification, ensuring that each signal record meets preset standards and can be used for subsequent invocation and analysis of intelligent osteoporosis risk warning tasks.

[0089] Preferably, this specification also provides an intelligent early warning system for osteoporosis risk based on a large database, used to execute the intelligent early warning method for osteoporosis risk based on a large database as described above. The intelligent early warning system for osteoporosis risk based on a large database includes: The bone mass assessment module is used to extract user bone health data from a multi-source health database; assess bone mass status based on user bone health data; and filter users at risk of osteoporosis from the user bone health data based on bone mass status. The trend analysis module is used to query medical images of users at risk of osteoporosis; extract bone mineral density values ​​from medical images and determine the trend of bone mineral density changes based on the bone mineral density values; and detect the degree of bone microstructure degeneration using medical images. The model building module is used to construct an osteoporosis risk warning model based on the trend of bone density changes and the degree of bone microstructure degeneration; user bone health data is input into the osteoporosis risk warning model to generate a risk warning score; The intelligent early warning module is used to send early warning signals to users at risk of osteoporosis in order to perform intelligent early warning tasks for osteoporosis risk.

[0090] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0091] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A smart early warning method for osteoporosis risk based on big data, characterized in that, Includes the following steps: Step S1: Extract user bone health data from the multi-source health database; Bone mass status is assessed based on user bone health data; users at risk of osteoporosis are screened based on bone mass status from the user bone health data. Step S2: Query medical images of users at risk of osteoporosis; extract bone mineral density values ​​from the medical images and determine the trend of bone mineral density changes based on the bone mineral density values; detect the degree of bone microstructure degeneration using medical images; Step S3: Construct an osteoporosis risk early warning model based on the trend of bone density changes and the degree of bone microstructure degeneration; Input users' bone health data into the osteoporosis risk warning model to generate a risk warning score; Step S4: Send the warning signal to users at risk of osteoporosis to execute the intelligent warning task for osteoporosis risk.

2. The intelligent early warning method for osteoporosis risk based on big data as described in claim 1, characterized in that, Step S1 is as follows: Step S11: Extract bone mineral density parameters based on user bone health data; Step S12: Compare the bone mineral density parameters with the preset bone mineral density threshold to screen users with low bone mineral density; Step S13: Calculate the bone resorption ratio for users with low bone mineral density. Step S14: Perform bone metabolism imbalance analysis based on bone resorption ratio to obtain bone metabolism imbalance data; Step S15: Assess bone mass status based on bone metabolism imbalance data; Step S16: Filter users at risk of osteoporosis from the user bone health data based on bone mass status.

3. The intelligent early warning method for osteoporosis risk based on big data as described in claim 2, characterized in that, Step S14 is as follows: Step S141: Determine the degree of bone metabolism shift based on the bone resorption ratio; Step S142: Perform bone loss analysis based on the degree of bone metabolism shift to obtain bone loss data; Step S143: Assess osteocalcin content based on bone loss data; Step S144: Determine bone formation activity based on osteocalcin content to obtain bone formation activity data; Step S145: Perform bone metabolism imbalance analysis based on bone formation activity data to obtain bone metabolism imbalance data.

4. The intelligent early warning method for osteoporosis risk based on a large database according to claim 2, characterized in that, Step S16 is as follows: Step S161: Perform age grading analysis based on bone mass status to obtain age grade data; Step S162: Determine bone mass reference values ​​for each age group based on age gradation data; Step S163: Extract the actual bone mass values ​​for each age group based on the age group data; Step S164: If the actual bone mass value of each age group is lower than the reference value of bone mass for each age group, then the user is marked as the first risk user; Obtain users with a history of fractures and mark them as secondary risk users; Step S165: Perform an intersection operation based on the first risk user and the second risk user to determine the osteoporosis risk user.

5. The intelligent early warning method for osteoporosis risk based on big data as described in claim 1, characterized in that, Step S2, determining the trend of bone mineral density change based on bone mineral density values, specifically involves: Extract time-series bone mineral density data based on bone mineral density values; Trend fitting analysis was performed based on time-series bone mineral density data to obtain a bone mineral density trend curve; Extract the trend slope and fluctuation amplitude based on the bone mineral density trend curve; The trend of bone density change is determined by the slope of the trend and the amplitude of the fluctuation.

6. The intelligent early warning method for osteoporosis risk based on big data as described in claim 1, characterized in that, Step S2, which utilizes medical imaging to detect the degree of bone microstructural degeneration, specifically involves: Medical images are preprocessed to obtain enhanced images; Identify the trabecular bone structure features in enhanced images; calculate the trabecular spacing based on the trabecular bone structure features; Connectivity analysis was performed based on the trabecular spacing to obtain connectivity data; The integrity of trabecular bone is assessed based on connectivity data; the degree of bone microstructural degradation is determined based on the integrity of trabecular bone.

7. The intelligent early warning method for osteoporosis risk based on big data as described in claim 6, characterized in that, The specific features of trabecular bone structure identified in enhanced images are as follows: The cancellous bone region is determined based on the enhanced image, and a region bounding box is set for the cancellous bone region to obtain the bounding box data; Binarization segmentation is performed based on the bounding box data to obtain a binary image of the trabecular bone. Based on the removal of pseudo-branches and burrs from the binary images of trabeculae, purified skeletal data is obtained; The trabecular structure characteristics were determined based on the purified skeleton data.

8. The intelligent early warning method for osteoporosis risk based on big data as described in claim 1, characterized in that, Step S3 involves constructing an osteoporosis risk early warning model based on the trend of bone density changes and the degree of bone microstructure degeneration. An osteoporosis risk warning model is constructed based on the trend of bone density change and the degree of bone microstructure degeneration. The osteoporosis risk warning model includes a bone density trend feature extraction layer, a bone microstructure degeneration feature extraction layer, a feature fusion layer, and a risk warning judgment layer. The bone density trend feature extraction layer is used to perform convolution operations and activation mapping on the bone density change trend to extract the bone density change rate and long-term trend features, and generate a bone density trend feature map. The bone microstructure degradation feature extraction layer is used to perform convolution operations and feature mapping on the degree of bone microstructure degradation, extract degradation features such as changes in trabecular thickness, increased trabecular spacing, and decreased bone structure integrity, and generate a bone microstructure degradation feature map. The feature fusion layer is used to weightedly fuse the bone density trend feature map with the bone microstructure degeneration feature map to determine the comprehensive characteristics of osteoporosis; The risk warning and judgment layer is used to generate the osteoporosis risk probability and output the corresponding osteoporosis risk warning result.

9. The intelligent early warning method for osteoporosis risk based on big data as described in claim 1, characterized in that, Step S4 is as follows: Step S41: Determine the osteoporosis warning level based on the warning signal; Step S42: Query the historical reports of users at risk of osteoporosis; Step S43: Match the osteoporosis warning level with historical reports and send the corresponding osteoporosis warning level signal; Step S44: Write the osteoporosis early warning level signal into the multi-source health database to perform the intelligent early warning task for osteoporosis risk.

10. An intelligent early warning system for osteoporosis risk based on a large database, characterized in that, For executing the intelligent early warning method for osteoporosis risk based on a large database as described in claim 1, the intelligent early warning system for osteoporosis risk based on a large database includes: The bone mass assessment module is used to extract user bone health data from a multi-source health database; assess bone mass status based on user bone health data; and filter users at risk of osteoporosis from the user bone health data based on bone mass status. The trend analysis module is used to query medical images of users at risk of osteoporosis; extract bone mineral density values ​​from medical images and determine the trend of bone mineral density changes based on the bone mineral density values; and detect the degree of bone microstructure degeneration using medical images. The model building module is used to construct an osteoporosis risk warning model based on the trend of bone density changes and the degree of bone microstructure degeneration; user bone health data is input into the osteoporosis risk warning model to generate a risk warning score; The intelligent early warning module is used to send early warning signals to users at risk of osteoporosis in order to perform intelligent early warning tasks for osteoporosis risk.

Images