Intelligent assessment device, method, apparatus and medium for fracture risk in liver transplant recipients

Abstract

This invention provides an intelligent assessment device, method, and medium for fracture risk in liver transplant recipients, comprising: a data extraction module for obtaining the management stage identifier of the liver transplant recipient from a medical information system and extracting clinical data; a fracture risk assessment module for inputting clinical data into a fracture risk scoring model to obtain a fracture risk score, constructing a scoring model based on a neural network, and training it using the patient's historical data; a risk mapping module for calling the corresponding stage-specific calibration curve according to the management stage identifier and mapping the fracture risk score to a standardized risk probability value; and a profile report output module for generating and outputting a multi-dimensional risk profile report based on the risk probability value, highlighting at least one key feature that contributes most to the risk score, thereby addressing the problem of lacking a fracture risk assessment scheme that can integrate liver transplant-specific risks and dynamically calibrate in stages, making it difficult to provide accurate individualized management decision support for clinical practice.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of transplant medicine and artificial intelligence, and in particular to an intelligent assessment device, method, equipment and medium for assessing fracture risk in liver transplant recipients. Background Technology

[0002] Liver transplantation is an effective treatment for end-stage liver disease, and with improvements in surgical techniques and immunosuppressive regimens, recipient survival rates have significantly increased. However, osteoporotic fractures have become a major complication affecting the long-term prognosis of liver transplant recipients. Studies show that the incidence of vertebral fractures within one year after liver transplantation is as high as 20%-40%, with postoperative bone loss, osteoporosis, and fragility fracture rates of 34.53%, 11.68%, and 20.40%, respectively. Due to the combined effects of pre-existing hepatic osteodystrophy, the osteotoxicity of high-dose glucocorticoids and calcineurin inhibitors (CNIs) postoperatively, and renal impairment, bone metabolism management in liver transplant recipients is more complex than in the general osteoporosis population, necessitating precise risk assessment tools.

[0003] Currently, the most commonly used fracture risk assessment tool in clinical practice is FRAX®. This tool calculates an individual's 10-year fracture probability based on age, bone mineral density, body mass index, and some clinical risk factors. However, the application of FRAX® in liver transplant recipients has significant limitations: First, its model does not include risk factors specific to liver transplantation, such as cumulative exposure to CNI drugs, estimated glomerular Filtration Rate (eGFR), etiologies of liver disease such as primary biliary cholangitis (PBC), and bone turnover markers (beta-C-terminal telopeptide of type I collagen, β-CTX). Second, FRAX® is a static assessment model and cannot reflect the temporal pathophysiological evolution of liver transplant recipients from the preoperative and perioperative rapid bone loss phase to the long postoperative period, leading to a severe underestimation of perioperative fracture risk. In addition, although the osteoporosis fracture prediction patents disclosed in recent years (such as CN121393875A) have introduced dynamic prediction models, they are aimed at the general osteoporosis population, do not involve liver transplantation-specific risk factors, and do not solve the technical problem of dynamically adjusting risk thresholds at different management stages.

[0004] Therefore, there is an urgent need for an intelligent assessment device for fracture risk in liver transplant recipients to address the technical problem that existing technologies lack fracture risk assessment schemes that can integrate liver transplant-specific risk factors and achieve phased dynamic calibration, resulting in a large deviation between assessment results and actual risks, making it difficult to provide accurate individualized management decision support for clinical practice. Summary of the Invention

[0005] To overcome the problems existing in related technologies, this disclosure provides an intelligent assessment device, method, equipment, and medium for fracture risk in liver transplant recipients. This addresses the technical problem that the lack of a fracture risk assessment scheme that can integrate liver transplant-specific risk factors and achieve phased dynamic calibration in related technologies leads to a large deviation between the assessment results and the actual risk, making it difficult to provide accurate individualized management decision support for clinical practice.

[0006] This specification provides one or more embodiments of an intelligent assessment device for fracture risk in liver transplant recipients, including: The data extraction module is used to obtain the management stage identifier of the liver transplant recipient from the medical information system, and extract the corresponding clinical data according to the management stage identifier. The management stage identifier includes the preoperative stage, the perioperative stage, or the long-term postoperative stage. The fracture risk assessment module is used to input the clinical data into a pre-trained fracture risk scoring model to obtain a fracture risk score. The fracture risk scoring model is constructed based on a gradient boosting tree or neural network and trained using historical data of liver transplant patients. The input features include: calcineurin inhibitor dosage, cumulative glucocorticoid dosage, renal function indicators, liver disease etiology coding features, and bone turnover marker features. The risk mapping module is used to call the corresponding stage-specific calibration curve based on the management stage identifier and map the fracture risk score to a standardized risk probability value. The profile report output module is used to generate and output a multi-dimensional risk profile report based on the risk probability value, highlighting at least one key feature that contributes the most to the risk score.

[0007] Preferably, the fracture risk assessment module includes an input feature unit, configured as follows: Input features also include at least one of the following: bone mineral density value features, fragility fracture history coding features, glucocorticoid cumulative dose features, calcium and vitamin D deficiency marker features, and sarcopenia or low body mass index features; The input feature set of the fracture risk scoring model is dynamically adjusted based on the patient age field in the clinical dataset. When the value of the age field is less than the preset age, a first feature set is generated with bone mineral density value features, fragility fracture history coding features and glucocorticoid cumulative dose features as core inputs, and the weights of other preset features are reduced.

[0008] Preferably, the risk mapping module includes a calibration curve construction unit, configured as follows: Construct the phase-specific calibration curves according to the management phase identifiers; For the same fracture risk score, the calibration curve for the perioperative phase is configured to map to a higher risk probability value compared to the preoperative phase, and the calibration curve for the long-term postoperative phase is configured to map to a lower risk probability value compared to the perioperative phase.

[0009] Preferably, the profile report output module includes a prompt information generation unit, used to generate structured intervention prompt information corresponding to the risk probability value in the multidimensional risk profile report. The specific generation rules include the following steps: When the risk probability value is lower than the first threshold, a first type of prompt information is generated, which includes a basic nutrition supplementation direction indicator. When the risk probability value is higher than the first threshold and lower than the second threshold, a second type of prompt information is generated. The second type of prompt information includes a basic nutritional supplementation direction indicator and a bone density follow-up recommendation at the first frequency. When the risk probability value is higher than the second threshold, a third type of prompt information is generated. The third type of prompt information includes a drug intervention direction indicator, a second frequency of bone density follow-up suggestions, and a multidisciplinary consultation prompt indicator.

[0010] Preferably, the portrait report output module further includes a data recording unit and a model update unit; The data recording unit is used to record follow-up outcome data in the medical information system. The follow-up outcome data includes changes in bone mineral density during the follow-up period, markers of new fragility fractures, and markers of drug-related adverse reactions. The model update unit is used to incrementally train the fracture risk scoring model using the follow-up outcome data when the accumulated amount of follow-up outcome data meets a preset condition, and update the parameters of the fracture risk scoring model. The incremental training adopts Bayesian online learning or incremental random forest algorithm.

[0011] This specification provides one or more embodiments of an intelligent assessment method for fracture risk in liver transplant recipients, applied to the assessment device described above, characterized by comprising the following steps: The management stage identifier of the liver transplant recipient is obtained from the medical information system, and the corresponding clinical data is extracted based on the management stage identifier. The management stage identifier includes the preoperative stage, the perioperative stage, or the long-term postoperative stage. The clinical data is input into a pre-trained fracture risk scoring model to obtain a fracture risk score. The fracture risk scoring model is constructed based on a gradient boosting tree or neural network and trained using historical data of liver transplant patients. The input features include: calcineurin inhibitor dosage, renal function indicators, liver disease etiology coding features, and bone turnover marker features. Based on the management stage identifier, the corresponding stage-specific calibration curve is invoked to map the fracture risk score into a standardized risk probability value; Based on the risk probability value, a multi-dimensional risk profile report is generated and output, highlighting at least one key feature that contributes the most to the risk score.

[0012] Preferably, the input features also include at least one of the following: bone mineral density value features, fragility fracture history coding features, glucocorticoid cumulative dose features, calcium and vitamin D deficiency marker features, and sarcopenia or low body mass index features; The input feature set of the fracture risk scoring model is dynamically adjusted based on the patient age field in the clinical dataset. When the value of the age field is less than the preset age, a first feature set is generated with bone mineral density value features, fragility fracture history coding features and glucocorticoid cumulative dose features as core inputs, and the weights of other preset features are reduced.

[0013] Preferably, the stage-specific calibration curves are constructed according to the management stage identifiers, specifically including the following steps: For the same fracture risk score, the calibration curve for the perioperative phase is configured to map to a higher risk probability value compared to the preoperative phase, and the calibration curve for the long-term postoperative phase is configured to map to a lower risk probability value compared to the perioperative phase.

[0014] This specification provides one or more embodiments of a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the intelligent assessment method for fracture risk in liver transplant recipients as described above.

[0015] This specification provides one or more embodiments of a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the intelligent assessment method for fracture risk in liver transplant recipients described above.

[0016] This disclosure provides an intelligent assessment device, method, equipment, and medium for fracture risk in liver transplant recipients. Its advantages lie in its data extraction module, which identifies the management stage markers of the liver transplant recipient and automatically extracts clinical data matching that stage from the medical information system. This achieves precise targeted acquisition and structured integration of data from multiple periods, eliminating bias from manual screening and providing high-quality, standardized input for subsequent assessments. The fracture risk assessment module inputs the extracted clinical data into a fracture risk scoring model constructed based on a gradient boosting tree or neural network. The model integrates multidimensional features such as cumulative exposure to calcineurin inhibitors, renal function, etiology of liver disease, and bone turnover markers. After training with historical data, it outputs a quantitative fracture risk score, thereby deeply capturing... The unique complex risk patterns of liver transplant patients enhance the accuracy and objectivity of assessments. The risk mapping module calls the corresponding stage-specific calibration curve based on the management stage identifier, converting the risk score into a calibrated standardized risk probability value. This eliminates the quantitative bias caused by baseline differences between different management stages, allowing the output probability to directly reflect the true clinical risk level and facilitating unified interpretation and comparison across stages. The profile report output module generates a multi-dimensional risk profile report based on the risk probability value and automatically highlights at least one key feature that contributes the most to the score. This transforms abstract risk values ​​into an interpretable feature attribution view, helping medical staff quickly identify major risk factors to support individualized and precise fracture prevention decisions. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic diagram of a smart assessment device for fracture risk in liver transplant recipients provided in one or more embodiments of this specification; Figure 2 A flowchart illustrating an intelligent assessment method for fracture risk in liver transplant recipients, provided for one or more embodiments of this specification; Figure 3 This is a schematic diagram of the structure of a computer device provided for one or more embodiments of this specification. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this invention.

[0020] The present invention will now be described in detail with reference to specific embodiments and accompanying drawings.

[0021] Device Examples According to embodiments of the present invention, a smart assessment device for fracture risk in liver transplant recipients is provided, such as... Figure 1 The diagram shown is a structural schematic of the intelligent assessment device for fracture risk in liver transplant recipients provided in this embodiment. The intelligent assessment device for fracture risk in liver transplant recipients according to this embodiment includes: Data extraction module 11 is deployed within the liver transplant center's information system. This module connects to the hospital's electronic medical record system via an HL7 / FHIR interface to retrieve the liver transplant recipient's management stage identifier from the medical information system. Based on this identifier, corresponding clinical data is extracted. The management stage identifier includes preoperative, perioperative, or long-term postoperative stages. Specifically, based on the patient's transplant date entered by medical staff or the surgery date automatically extracted from the electronic medical record, the number of days since the transplant is completed is calculated. If no surgery date is recorded or the current date is within 30 days before the transplant, the management stage identifier is set to preoperative; if it is within 0 to 6 months postoperatively, it is set to perioperative; and if it is more than 6 months postoperatively, it is set to long-term postoperative. Then, the data acquisition template corresponding to that stage is retrieved. Preoperative stage: Collect DXA bone mineral density (lumbar spine, hip, distal radius), history of fragility fracture, etiology of liver disease (whether it is primary biliary cholangitis PBC), BMI, serum 25(OH)D, serum calcium, serum phosphorus, PTH.

[0022] Perioperative period: In addition to the above, prednisone equivalent daily dose, CNI type and dose, blood drug concentration (tacrolimus or cyclosporine), eGFR, and bone turnover marker β-CTX were collected at 1 month and 3 months postoperatively.

[0023] Long-term postoperative care: Annual DXA, current prednisone dose, CNI dose and blood concentration, eGFR, bone turnover markers, and whether new fragility fractures occur. Key points for the comprehensive management of osteoporosis in transplant recipients are shown in Table 1.

[0024] Table 1 Key Points for the Comprehensive Management of Osteoporosis in Liver Transplant Recipients

[0025] The fracture risk assessment module 12 is used to input clinical data into a pre-trained fracture risk scoring model to obtain a fracture risk score. The fracture risk scoring model is constructed based on XGBoost gradient boosting tree or neural network and is trained using historical data from 500 liver transplant patients, which can include 5-year follow-up data after surgery. Input features include: cumulative exposure to calcineurin inhibitors, renal function indicators, liver disease etiology coding features, and bone turnover marker features.

[0026] Example of input feature numericalization: The cumulative exposure characteristic of calcineurin inhibitors is calculated as: tacrolimus plasma concentration (ng / mL) × number of months of exposure, or cyclosporine dose (mg / day) × number of months of exposure. Higher values ​​are assigned when the plasma concentration is >10 ng / mL or the dose is >150 mg / day.

[0027] Renal function indicator characteristic = current eGFR value (mL / min). If eGFR < 45 and continues to decrease by > 20%, the characteristic value is doubled.

[0028] Etiological coding features for liver diseases: PBC=1, alcoholic liver disease=2, autoimmune hepatitis=3, others=0. The PBC feature has the highest weight.

[0029] Bone turnover marker characteristic = β-CTX value 1 month postoperatively / multiple of the upper limit of normal. If it is >2 times, the characteristic value is set to 1 (high bone resorption state).

[0030] The fracture risk assessment module 12 includes an input feature unit, configured to include at least one of the following input features: bone mineral density value feature, fragility fracture history coding feature, glucocorticoid cumulative dose feature, calcium and vitamin D deficiency marker feature, and sarcopenia or low body mass index feature. Specifically, the bone mineral density T-score is used to characterize the bone mineral density value feature for postmenopausal women and men over 50 years of age, while the bone mineral density Z-score is used to characterize the bone mineral density value feature for premenopausal women and men under 50 years of age, including men over 50 years of age.

[0031] The input feature set of the fracture risk scoring model is dynamically adjusted based on the patient age field in the clinical dataset.

[0032] When the value of the age field is less than the preset age, a first feature set is generated with bone mineral density value features, fragility fracture history coding features, and glucocorticoid cumulative dose features as core inputs, and the weights of other preset features are reduced.

[0033] For example, with a preset age of 40, for patients aged 40 and above, a full feature set is used, including four mandatory features: cumulative calcineurin inhibitor exposure, renal function indicators, liver disease etiology coding, and bone turnover markers, as well as optional additional features: bone mineral density (BMD), history of fragility fractures, cumulative glucocorticoid dose, calcium / vitamin D deficiency markers, and sarcopenia / low body mass index (BMI). The weights of each feature are obtained through model training. For patients under 40 years of age, a first feature set is generated, with core features including BMD Z-score (weight increased to 0.4), history of fragility fractures (weight 0.3), and cumulative glucocorticoid dose (weight 0.3); other features, such as CNI exposure, renal function, and PBC etiology, have their weights reduced to 0.05 or zero. When the model recalculates the risk score, only the core features are used.

[0034] For example, a 35-year-old liver transplant recipient had no history of fragility fractures preoperatively, but used high-dose hormones (prednisone 15 mg / day for more than 3 months) postoperatively, resulting in a bone mineral density Z-score of -2.1. The system automatically reduces the contribution of features such as CNI and PBC, primarily assessing risk based on hormone dosage and bone mineral density.

[0035] The risk mapping module 13 is used to call the corresponding stage-specific calibration curve based on the management stage identifier and map the fracture risk score to a standardized risk probability value.

[0036] The profile report output module 14 is used to generate and output a multi-dimensional risk profile report based on the risk probability value. The report is output as a visual dashboard, with a radar chart showing the contribution of each feature. At least one key feature that contributes the most to the risk score is highlighted to alert medical staff to pay attention.

[0037] The device provided in this embodiment identifies the management stage markers of liver transplant recipients through the data extraction module 11, automatically extracting clinical data matching the stage from the medical information system. This achieves precise targeted capture and structured integration of data from multiple periods, eliminating biases from manual screening and providing high-quality, standardized input for subsequent assessments. The fracture risk assessment module 12 inputs the extracted clinical data into a fracture risk scoring model constructed based on a gradient boosting tree or neural network. The model integrates multidimensional features such as cumulative exposure to calcineurin inhibitors, renal function, etiology of liver disease, and bone turnover markers. After training with historical data, it outputs a quantitative fracture risk score, thereby deeply capturing the unique complex risk patterns of liver transplant patients. This improves the accuracy and objectivity of the assessment; the risk mapping module 13 calls the corresponding stage-specific calibration curve according to the management stage identifier, and converts the risk score into a calibrated standardized risk probability value, thereby eliminating the quantitative bias caused by the difference in baselines between different management stages, so that the output probability directly reflects the true clinical risk level, which is convenient for unified interpretation and comparison across stages; the profile report output module 14 generates a multi-dimensional risk profile report based on the risk probability value, and automatically highlights at least one key feature that contributes the most to the score, transforming the abstract risk value into an interpretable feature attribution view, helping medical staff to quickly identify the main risk factors to support individualized and precise fracture prevention decisions.

[0038] In one embodiment, the risk mapping module includes a calibration curve construction unit configured as follows: Based on the management stage identifier, a stage-specific calibration curve was constructed by fitting a score-probability mapping function for each stage using historical cohort data. Specifically, data from liver transplant recipients who had completed at least 2 years of follow-up (n≥200) were collected. For each stage, the calibration curve was fitted using logistic regression or locally weighted regression (LOWESS) with the raw risk score output by the model as the independent variable and the observed fracture incidence rate within 12 months as the dependent variable.

[0039] The specific mapping rules are as follows: The calibration curve used in the preoperative stage employs a linear mapping, assuming that the original preoperative score ranges from 0 to 100, corresponding to a fracture probability of 0% to 15%. For example, an original score of 60 points maps to a probability of 8%.

[0040] For the same fracture risk score, the calibration curve for the perioperative phase is configured to map to a higher risk probability value compared to the preoperative phase, while the calibration curve for the long postoperative phase is configured to map to a lower risk probability value compared to the perioperative phase.

[0041] Perioperative calibration curve: Since the period from 0 to 6 months postoperatively is a period of rapid bone loss, the same original score is mapped to a higher probability using a logistic function during the perioperative period. For example, an original score of 60 points is mapped to a probability of 25%, which is about 3 times higher than in the preoperative stage.

[0042] Long-term postoperative calibration curve: The rate of bone loss slows significantly at 12 months postoperatively, with some patients entering the bone mineral density recovery phase. A smooth curve fitted based on real cohort data maps the original score to a stable-period risk probability. However, bone microstructural damage still exists, with a risk higher than preoperatively but lower than perioperatively. For example, an original score of 60 points maps to 15%.

[0043] The device provided in this embodiment constructs differentiated calibration curves for different management stages. For the same fracture risk score, the perioperative period maps a higher risk probability than the preoperative period, while the long-term postoperative period maps a lower risk probability than the perioperative period. This dynamically restores the evolution of fracture risk throughout the entire liver transplantation cycle, eliminating the systematic bias caused by fixed mapping and ensuring that the final output probability value accurately matches the actual clinical risk level at each stage.

[0044] In one embodiment, the profile report output module 14 includes a prompt information generation unit, which embeds a decision rule engine to generate structured intervention prompt information corresponding to the risk probability value in the multidimensional risk profile report. The specific generation rules include the following steps: When the risk probability value is below the first threshold (i.e., <10% (the boundary between low and medium risk)), a first-type prompt message is generated. This first-type prompt message includes a basic nutritional supplementation direction indicator. Example content: "Intervention direction: Basic nutritional supplementation. Supplement with elemental calcium 1000-1200mg / d, vitamin D 800-1000IU / d, maintain 25(OH)D >30μg / L."

[0045] When the risk probability value is higher than the first threshold but lower than the second threshold (i.e., 10% ≤ risk probability value < 20%), a second type of alert is generated. This second type of alert includes a basic nutritional supplementation direction indicator and a bone mineral density follow-up recommendation at the first frequency. Example content: "Intervention direction: Basic nutritional supplementation + bone mineral density monitoring. It is recommended to have a dual-energy X-ray absorptiometry (DXA) examination every 6 months." It also suggests "Consider oral bisphosphonates (alendronate sodium)."

[0046] When the risk probability value exceeds the second threshold (≥20% (the boundary between medium and high risk)), a third type of prompt information is generated. This third type of prompt information includes a drug intervention direction indicator, a second frequency of bone mineral density follow-up recommendations, and a multidisciplinary consultation prompt indicator. Example content: "Intervention direction: Drug intervention. Recommended intravenous zoledronic acid 4mg or denosumab 60mg every 6 months (denosumab preferred if eGFR<35). Bone mineral density monitoring frequency: Every 3 months. Triggering multidisciplinary consultation: Transplantation, Endocrinology, Rehabilitation."

[0047] The device provided in this embodiment automatically generates three-level structured intervention prompts by comparing the risk probability value with preset dual thresholds: low risk outputs basic nutritional supplementation direction, medium risk adds regular frequency bone density follow-up, and high risk triggers drug intervention direction, increased frequency of bone density follow-up, and multidisciplinary consultation indicators. In this way, the abstract risk score is directly transformed into a tiered clinical action list that is precisely matched with the risk level, realizing a seamless closed loop from assessment to decision-making.

[0048] In one embodiment, the portrait report output module 14 further includes a data recording unit and a model update unit; The data recording unit automatically records key information about each assessment in the background after a risk report is generated, such as the assessment time, risk probability value, and model version used, generating a unique assessment ID for each assessment. Follow-up outcome data is recorded in the medical information system. This data includes changes in bone mineral density during the follow-up period (the difference between the most recent DXA and the DXA at ​​the time of this assessment (absolute value or percentage), e.g., "lumbar vertebral bone mineral density increased by 3%"). New fragility fracture markers (whether a newly diagnosed vertebral or hip fracture was recorded (ICD-10 code)) and drug-related adverse reaction markers (whether hypocalcemia, renal function deterioration (eGFR decrease >30%), atypical femoral fracture, or jaw necrosis occurred, etc.). Based on the temporal relationship between the assessment timestamp and the follow-up occurrence timestamp, the follow-up outcome data is matched to the corresponding assessment.

[0049] The model update unit is used to incrementally train the fracture risk scoring model using follow-up outcome data when the accumulated amount of follow-up outcome data meets the preset conditions, and update the model parameters. The incremental training adopts Bayesian online learning or incremental random forest algorithm.

[0050] The device provided in this embodiment continuously writes back follow-up outcomes such as changes in bone density, new fragility fractures, and adverse drug reactions to the medical information system through a data recording unit. When the accumulated data reaches a preset condition, the model update unit uses Bayesian online learning or incremental random forest algorithms to incrementally train the fracture risk scoring model and update its parameters, thereby constructing a complete closed loop of assessment, intervention, follow-up, and evolution. This allows the model to seamlessly absorb real-world feedback and achieve online self-evolution without interrupting service or requiring full retraining, continuously improving prediction accuracy and clinical adaptability.

[0051] Method Implementation Examples According to embodiments of the present invention, a method for intelligently assessing fracture risk in liver transplant recipients is provided, such as... Figure 1 The diagram shown is a flowchart illustrating the intelligent assessment method for fracture risk in liver transplant recipients provided in this embodiment. The intelligent assessment method for fracture risk in liver transplant recipients according to this embodiment includes the following steps: S210. Obtain the management stage identifier of the liver transplant recipient from the medical information system, and extract the corresponding clinical data based on the management stage identifier. The management stage identifier includes the preoperative stage, perioperative stage, or long-term postoperative stage.

[0052] S220. Input clinical data into a pre-trained fracture risk scoring model to obtain a fracture risk score. The fracture risk scoring model is constructed based on a gradient boosting tree or neural network and trained using historical data from liver transplant patients. Input features include: cumulative exposure to calcineurin inhibitors, renal function indicators, liver disease etiology coding features, and bone turnover marker features.

[0053] The input features also include at least one of the following: bone mineral density value features, fragility fracture history coding features, glucocorticoid cumulative dose features, calcium and vitamin D deficiency marker features, and sarcopenia or low body mass index features. Based on the patient age field in the clinical dataset, the input feature set of the fracture risk scoring model is dynamically adjusted. When the value of the age field is less than the preset age, a first feature set is generated with bone mineral density value features, fragility fracture history coding features, and glucocorticoid cumulative dose features as the core inputs, and the weights of other preset features are reduced.

[0054] S230. Based on the management stage identifier, call the corresponding stage-specific calibration curve to map the fracture risk score into a standardized risk probability value.

[0055] S240. Based on the risk probability value, generate and output a multi-dimensional risk profile report, highlighting at least one key feature that contributes the most to the risk score to alert medical staff.

[0056] The method provided in this embodiment automatically extracts clinical data matching the management stage of a liver transplant recipient from the medical information system, achieving precise targeted extraction and structured integration of multi-stage data. This eliminates the bias of manual screening and provides high-quality, standardized input for subsequent assessment. The extracted clinical data is then input into a fracture risk scoring model constructed based on a gradient boosting tree or neural network. The model integrates multi-dimensional features such as cumulative exposure to calcineurin inhibitors, renal function, etiology of liver disease, and bone turnover markers. After training with historical data, it outputs a quantitative fracture risk score, thereby deeply capturing the unique complex risk patterns of liver transplant patients. This improves the accuracy and objectivity of assessments; by calling the corresponding stage-specific calibration curve based on the management stage identifier, the risk score is converted into a calibrated standardized risk probability value, thereby eliminating the quantitative bias caused by baseline differences between different management stages. This allows the output probability to directly reflect the true clinical risk level, facilitating unified interpretation and comparison across stages; a multi-dimensional risk profile report is generated based on the risk probability value, and at least one key feature that contributes most to the score is automatically highlighted, transforming abstract risk values ​​into an interpretable feature attribution view, helping medical staff quickly identify major risk factors to support individualized and precise fracture prevention decisions.

[0057] In one embodiment, stage-specific calibration curves are constructed based on management stage identifiers, specifically including the following steps: For the same fracture risk score, the calibration curve for the perioperative phase is configured to map to a higher risk probability value compared to the preoperative phase, while the calibration curve for the long postoperative phase is configured to map to a lower risk probability value compared to the perioperative phase.

[0058] The method provided in this embodiment constructs differentiated calibration curves for different management stages. For the same fracture risk score, the perioperative period maps a higher risk probability than the preoperative period, while the long-term postoperative period maps a lower risk probability than the perioperative period. This dynamically restores the evolution of fracture risk throughout the entire liver transplantation cycle, eliminating the systematic bias caused by fixed mapping and ensuring that the final output probability value accurately matches the actual clinical risk level at each stage.

[0059] The embodiments of the present invention are method embodiments corresponding to the above-described device embodiments. The specific operations of each module processing step can be understood by referring to the description of the method embodiments, and will not be repeated here.

[0060] like Figure 3 As shown, the present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements the intelligent assessment method for fracture risk of liver transplant recipients in the above embodiments, or when the computer program is executed by a processor, it implements the intelligent assessment method for fracture risk of liver transplant recipients in the above embodiments.

[0061] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0062] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for apparatus or system embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The apparatus and system 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 creative effort.

[0063] 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 or all of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and the contents not described in detail in the specification of the present invention are known to those skilled in the art.

Claims

1. A smart device for assessing fracture risk in liver transplant recipients, characterized in that, include: The data extraction module is used to obtain the management stage identifier of the liver transplant recipient from the medical information system, and extract the corresponding clinical data according to the management stage identifier. The management stage identifier includes the preoperative stage, the perioperative stage, or the long-term postoperative stage. The fracture risk assessment module is used to input the clinical data into a pre-trained fracture risk scoring model to obtain a fracture risk score. The fracture risk scoring model is constructed based on a gradient boosting tree or neural network and trained using historical data of liver transplant patients. The input features include: cumulative exposure to calcineurin inhibitors, renal function indicators, liver disease etiology coding features, and bone turnover marker features. The risk mapping module is used to call the corresponding stage-specific calibration curve based on the management stage identifier and map the fracture risk score into a standardized risk probability value. The profile report output module is used to generate and output a multi-dimensional risk profile report based on the risk probability value, highlighting at least one key feature that contributes the most to the risk score.

2. The intelligent assessment device for fracture risk in liver transplant recipients as described in claim 1, characterized in that, The fracture risk assessment module includes an input feature unit, configured as follows: Input features also include at least one of the following: bone mineral density value features, fragility fracture history coding features, glucocorticoid cumulative dose features, calcium and vitamin D deficiency marker features, and sarcopenia or low body mass index features; The input feature set of the fracture risk scoring model is dynamically adjusted based on the patient age field in the clinical dataset. When the value of the age field is less than the preset age, a first feature set is generated with bone mineral density value features, fragility fracture history coding features and glucocorticoid cumulative dose features as core inputs, and the weights of other preset features are reduced.

3. The intelligent assessment device for fracture risk in liver transplant recipients as described in claim 1, characterized in that, The risk mapping module includes a calibration curve construction unit, configured as follows: Construct the phase-specific calibration curves according to the management phase identifiers; For the same fracture risk score, the calibration curve for the perioperative phase is configured to map to a higher risk probability value compared to the preoperative phase, and the calibration curve for the long-term postoperative phase is configured to map to a lower risk probability value compared to the perioperative phase.

4. The intelligent assessment device for fracture risk in liver transplant recipients as described in claim 1, characterized in that, The profile report output module includes a prompt information generation unit, used to generate structured intervention prompt information corresponding to the risk probability value in the multidimensional risk profile report. The specific generation rules include the following steps: When the risk probability value is lower than the first threshold, a first type of prompt information is generated, which includes a basic nutrition supplementation direction indicator. When the risk probability value is higher than the first threshold and lower than the second threshold, a second type of prompt information is generated. The second type of prompt information includes a basic nutritional supplementation direction indicator and a bone mineral density follow-up recommendation at the first frequency. When the risk probability value is higher than the second threshold, a third type of prompt information is generated. The third type of prompt information includes a drug intervention direction indicator, a second frequency of bone density follow-up recommendations, and a multidisciplinary consultation prompt indicator.

5. The intelligent assessment device for fracture risk in liver transplant recipients as described in claim 1, characterized in that, The portrait report output module also includes a data recording unit and a model update unit; The data recording unit is used to record follow-up outcome data in the medical information system. The follow-up outcome data includes changes in bone mineral density during the follow-up period, markers of new fragility fractures, and markers of drug-related adverse reactions. The model update unit is used to incrementally train the fracture risk scoring model using the follow-up outcome data when the accumulated amount of follow-up outcome data meets a preset condition, and update the parameters of the fracture risk scoring model. The incremental training adopts Bayesian online learning or incremental random forest algorithm.

6. A method for intelligently assessing fracture risk in liver transplant recipients, applied to the assessment device as described in any one of claims 1-5, characterized in that, Includes the following steps: The management stage identifier of the liver transplant recipient is obtained from the medical information system, and the corresponding clinical data is extracted based on the management stage identifier. The management stage identifier includes the preoperative stage, the perioperative stage, or the long-term postoperative stage. The clinical data is input into a pre-trained fracture risk scoring model to obtain a fracture risk score. The fracture risk scoring model is constructed based on a gradient boosting tree or neural network and trained using historical data of liver transplant patients. The input features include: cumulative exposure to calcineurin inhibitors, renal function indicators, liver disease etiology coding features, and bone turnover marker features. Based on the management stage identifier, the corresponding stage-specific calibration curve is invoked to map the fracture risk score into a standardized risk probability value; Based on the risk probability value, a multi-dimensional risk profile report is generated and output, highlighting at least one key feature that contributes the most to the risk score.

7. The intelligent assessment method for fracture risk in liver transplant recipients as described in claim 1, characterized in that, The input features also include at least one of the following: bone mineral density value features, fragility fracture history coding features, glucocorticoid cumulative dose features, calcium and vitamin D deficiency marker features, and sarcopenia or low body mass index features; The input feature set of the fracture risk scoring model is dynamically adjusted based on the patient age field in the clinical dataset. When the value of the age field is less than the preset age, a first feature set is generated with bone mineral density value features, fragility fracture history coding features and glucocorticoid cumulative dose features as core inputs, and the weights of other preset features are reduced.

8. The intelligent assessment method for fracture risk in liver transplant recipients as described in claim 1, characterized in that, The stage-specific calibration curves are constructed based on the management stage identifiers, specifically including the following steps: For the same fracture risk score, the calibration curve for the perioperative phase is configured to map to a higher risk probability value compared to the preoperative phase, and the calibration curve for the long-term postoperative phase is configured to map to a lower risk probability value compared to the perioperative phase.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the intelligent assessment method for fracture risk in liver transplant recipients as described in any one of claims 6 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the intelligent assessment method for fracture risk in liver transplant recipients as described in any one of claims 6 to 8.

Images