Lung transplantation rehabilitation evaluation method, system and device based on ICF and medium

Abstract

The invention relates to the technical field of medical information, and discloses a lung transplantation rehabilitation evaluation method based on ICF entries and clinical data, comprising the following steps: step 1, collecting multi-modal data including ICF entries, clinical data, image data and behavior data; step 2, dynamically distributing ICF entry weights according to postoperative time and image data confidence; 3, based on dynamic weight and multi-modal data fusion, an evaluation result is output through a three-level progressive model, and in the step 1, ICF entries are selected from entries related to lung transplantation rehabilitation in a WHO international functional classification library; the clinical data comprises lung function indexes, plasma concentration and donor specific antibody titer; the image data includes an HRCT image. According to the method, a weight distribution technical scheme based on postoperative time decay and cross-modal verification coupling is adopted, the acute stage entry weight is dynamically reduced through the exponential decay model, and the technical effect of adapting to evaluation requirements of different rehabilitation stages is achieved.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to an ICF-based lung transplant rehabilitation assessment method, system, device and medium. Background Art

[0002] The lung transplant rehabilitation effect evaluation method based on ICF items and clinical data is a quantitative evaluation system that integrates the World Health Organization's International Classification of Functioning (ICF) framework with multimodal medical data. This method defines the evaluation dimensions through standardized ICF items (such as respiratory function and immunosuppression complications), and integrates lung function indicators (FEV1 / FVC), imaging features (HRCT ground-glass opacities) and laboratory data (DSA antibody titer). Its core goal is to provide clinicians with a decision-making basis for postoperative rehabilitation stage division, complication warning and personalized intervention through structured functional evaluation and quantitative data analysis. In the existing technology, ICF items are often statically mapped into fixed-dimensional assessment scales, for example, respiratory function (b440) is directly linked to FEV1 values, while social role adaptation (d710) is quantified through questionnaires.

[0003] In existing technologies, such methods typically employ a static weight allocation strategy, where fixed weights are pre-set for ICF items based on expert experience or historical data regression results, and a comprehensive score is calculated using linear weighting combined with clinical data. For example, respiratory function (b440) might be assigned a weight of 0.6, and social adaptability (d710) a weight of 0.1. A single score is then obtained through weighted summation, directly mapping the scores to "low risk," "medium risk," or "high risk." At the data fusion level, imaging data (such as the proportion of ground-glass opacities on HRCT) and clinical indicators (such as FEV1 values) are often analyzed independently: for example, image segmentation algorithms are used to quantify the area of ​​the pathological region and generate a separate rejection risk score; or the FEV1 value measured by a spirometer is used alone to assess lung function recovery. Furthermore, some technologies attempt to introduce a time dimension, such as dividing the stages by weeks after surgery and adjusting the weights in segments. However, these adjustment rules rely on manual presetting and lack data-driven dynamic adaptability.

[0004] However, this approach faces a significant technical bottleneck: static weighting is disconnected from the dynamic rehabilitation needs of postoperative patients. For example, in the early postoperative period (1-3 months), monitoring for acute complications (such as infection and acute rejection) should be prioritized, and respiratory function (b440) and immunosuppressive complications (b1305) should be given the primary weight. Meanwhile, in the postoperative stable period (after 6 months), the assessment shifts to social function recovery, and the weight of social role adaptation (d710) should be significantly increased. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a lung transplant rehabilitation assessment method based on ICF items and clinical data, which solves the problems of poor adaptability and insufficient assessment accuracy in the postoperative stage caused by static weight allocation, multimodal data conflicts and missing patient behavior data in the existing lung transplant rehabilitation assessment methods.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a lung transplant rehabilitation assessment method based on ICF items and clinical data, comprising the following steps: Step 1: Collect multimodal data, including ICF items, clinical data, imaging data, and behavioral data; Step 2: Dynamically assign ICF item weights based on postoperative time and imaging data confidence; Step 3: Based on dynamic weights and multimodal data fusion, the evaluation results are output through a three-level progressive model.

[0007] Preferably, in step 1: The ICF items were selected from the WHO International Classification of Functioning database related to lung transplant rehabilitation; The clinical data include lung function indicators, blood drug concentrations and donor-specific antibody titers; The imaging data includes HRCT images, which are used to quantify characteristics related to postoperative complications, including: The volume proportion of the atelectasis area is calculated using the lung lobe volume segmentation algorithm; Infectious lesion scoring, based on the density of the consolidation area and the presence of air bronchograms; The airway stenosis index was measured by three-dimensional reconstruction of the bronchial tree to measure the diameter deviation rate at the stenosis site.

[0008] Preferably, the step 2 includes the following sub-steps: The initial weight is adjusted for time decay according to the postoperative time t, and the time decay factor satisfies: W t (e i )=W0(e i )·e -λt Where W0(e i ) is the initial weight of the item, and λ is the attenuation coefficient based on the postoperative complication risk curve fitting; Calculate the confidence coefficient C through cross-modality verification of imaging data and clinical data img , and modify the weight: W(e i ,t)=W t (e i )·(1+α·C img ); Where α is the image verification weight coefficient.

[0009] Preferably, the image confidence coefficient C in step 2 is img The calculation formula is: Where: V Atelectasis : The ratio of atelectasis volume to total lung lobe volume (0-1); S Infection : Infectious lesion score, when the density of the consolidation area is greater than 50%, it is scored as 0.8, and when there is air bronchogram, it is scored as +0.2; I Stenosis : Airway stenosis index, when the main bronchial diameter is less than 60% of the normal value, it is 0.6; when the lobar bronchial stenosis occurs, it is 0.3.

[0010] Preferably, the three-level progressive model in step 3 includes: First-level assessment: Calculate the comprehensive risk score step 1 based on dynamic weights and clinical data. If step 1 ≥ θ1, mark the patient as a patient with high rehabilitation needs; Second-level assessment: For patients with high rehabilitation needs, HRCT imaging data and DSA titers are integrated and the classification result G is output through the classifier; Level 3 assessment: combined with self-monitoring adherence R 02 , dynamically optimize the final evaluation result Sfinal.

[0011] Preferably, the calculation formula for the comprehensive risk score S1 in step 3 is: where x norm (e i ) is the normalized clinical data, and the normalized range is the clinically reasonable value range.

[0012] Preferably, when the image confidence coefficient C img When <0.4, the manual review process is triggered; The classifier was used to classify the lung function FEV1 after verification adj =FEV1 meas ·C img , DSA titer, atelectasis volume ratio, and infectious lesion score were used for grading.

[0013] Preferably, the dynamic optimization in step 3 satisfies: S final =γ·G+(1-γ)·R O2 ; in: G is the grading result, with the coding bit value as mild = 0.2, moderate = 0.5, and severe = 0.8; R O2 is the oxygen therapy compliance rate, normalized to the interval [0,1]; γ is the grading weight coefficient, and its value range is 0.7≤γ≤0.9.

[0014] Preferably, the ICF entry includes: respiratory function; complications of immunosuppression; social role adaptation; body functions; Ability to take care of oneself; Environmental factors.

[0015] The present invention also provides a lung transplant rehabilitation assessment system based on ICF items and clinical data, comprising: Data acquisition module, used to obtain ICF entries, clinical data and imaging data; Dynamic weight allocation module, used to dynamically adjust item weights based on postoperative time and imaging data confidence; The hierarchical evaluation module is used to perform three-level evaluation and output the results. The three-level evaluation includes rapid initial screening, refined classification and dynamic optimization.

[0016] The present invention provides a lung transplant rehabilitation assessment method based on ICF items and clinical data. It has the following beneficial effects: 1. This invention utilizes a weight allocation technique based on postoperative time decay coupled with cross-modal validation. This approach dynamically reduces the weight of acute-phase items through an exponential decay model and modifies the weights through consistent analysis of imaging and clinical data, achieving a technical effect that adapts to the assessment needs of different rehabilitation stages. Compared to existing approaches that use fixed weights and result in rigid assessments, this approach addresses the inability to respond to changes in postoperative time and data reliability.

[0017] 2. This invention is based on a three-level progressive model. First, comprehensive scoring is used to rapidly screen patients with high rehabilitation needs. Second, multimodal fusion improves grading accuracy. Third, behavioral data optimizes long-term management strategies. This achieves a technical balance between computational efficiency and assessment accuracy. Compared to existing technologies that waste resources or lack precision through a single assessment process, this approach addresses the inability to balance rapid response in the acute phase with meticulous management during the stable phase.

[0018] 3. This invention uses an imaging confidence factor to verify the reliability of clinical data. By quantifying the difference between image-predicted pulmonary function and measured values, it dynamically suppresses the interference of low-quality data on the assessment results, thereby improving the robustness of the assessment results. Compared with existing solutions that analyze images and clinical indicators separately, this approach eliminates the risk of misjudgment due to data conflicts.

[0019] 4. This technical solution dynamically optimizes grading results by combining behavioral data such as oxygen therapy adherence rates. By weightedly integrating biomedical indicators with patient behavioral characteristics, this solution supports the development of personalized rehabilitation strategies. Compared to existing approaches that rely solely on laboratory test data, this approach addresses the shortcomings of prior art approaches that ignore the impact of patient compliance on long-term recovery. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] Please see the attached Figure 1 The embodiment of the present invention provides a lung transplant rehabilitation assessment method based on ICF items and clinical data, comprising the following steps: Step 1: Collect multimodal data, including ICF items, clinical data, imaging data, and behavioral data; Step 2: Dynamically assign ICF item weights based on postoperative time and imaging data confidence; Step 3: Based on dynamic weights and multimodal data fusion, the evaluation results are output through a three-level progressive model.

[0023] In step one: ICF items were selected from the WHO International Classification of Functioning database related to lung transplant rehabilitation; Clinical data included lung function indicators, blood drug concentrations, and donor-specific antibody titers; Imaging data included HRCT images to quantify characteristics related to postoperative complications, including: The volume proportion of the atelectasis area is calculated using the lung lobe volume segmentation algorithm; Infectious lesion scoring, based on the density of the consolidation area and the presence of air bronchograms; The airway stenosis index was measured by three-dimensional reconstruction of the bronchial tree to measure the diameter deviation rate at the stenosis site.

[0024] Step 2 includes the following sub-steps: The initial weight is adjusted for time decay according to the postoperative time t, and the time decay factor satisfies: W t (e i )=W0(e i )·e-λt Where W0(e i ) is the initial weight of the item, and λ is the attenuation coefficient based on the postoperative complication risk curve fitting; Calculate the confidence coefficient C through cross-modality verification of imaging data and clinical data img , and modify the weights: W(e i ,t)=W t (e i )·(1+α·C img ); Where α is the image verification weight coefficient.

[0025] Image confidence coefficient C in step 2 img The calculation formula is: Where: V Atelectasis : The ratio of atelectasis volume to total lung lobe volume (0-1); S Infection : Infectious lesion score, when the density of the consolidation area is greater than 50%, it is scored as 0.8, and when there is air bronchogram, it is scored as +0.2; I Stenosis : Airway stenosis index, when the main bronchial diameter is less than 60% of the normal value, it is 0.6; when the lobar bronchial stenosis occurs, it is 0.3.

[0026] The three-level progressive model in step three includes: First-level assessment: Calculate the comprehensive risk score step 1 based on dynamic weights and clinical data. If step 1 ≥ θ1, mark the patient as a patient with high rehabilitation needs; Second-level assessment: For patients with high rehabilitation needs, the proportion of atelectasis volume, infectious lesion score, airway stenosis index, and DSA titer are integrated to output the classification result G through random forest classifier; Level 3 assessment: combined with self-monitoring adherence R 02 , dynamically optimize the final evaluation result S final .

[0027] The calculation formula for the comprehensive risk score S1 in step 3 is: where x norm (e i ) is the normalized clinical data, and the normalized range is the clinically reasonable value range.

[0028] The refined classification in step three includes: When the image confidence coefficient C img When <0.4, the manual review process is triggered; The classifier was used to classify the lung function FEV1 after verification adj=FEV1 meas ·C img , DSA titer, atelectasis volume ratio, and infectious lesion score were used for grading.

[0029] In step 3, dynamic optimization satisfies: S final =γ·G+(1-γ)·R O2 ; in: G is the grading result, with the coding bit value as mild = 0.2, moderate = 0.5, and severe = 0.8; R O2 is the oxygen therapy compliance rate, normalized to the interval [0,1]; γ is the grading weight coefficient, and its value range is 0.7≤γ≤0.9.

[0030] ICF items include: respiratory function; complications of immunosuppression; social role adaptation; body functions; Ability to take care of oneself; Environmental factors.

[0031] In this embodiment, step 1 is used to implement the collection and preprocessing of multimodal data. Its technical solution includes the definition of data sources, screening rules and normalization processing. The specific implementation method is as follows: As an option, multimodal data include items selected from the WHO International Classification of Functioning (ICF) library, clinical monitoring data, medical imaging data, and patient behavior data. Specifically, ICF items are functional assessment items directly related to post-lung transplant rehabilitation, such as respiratory function (b440), immunosuppression complications (b1305), and social role adaptation (d710). It should be noted that the screening of ICF items needs to be combined with the rehabilitation needs of different stages after surgery, and clinical experts should select them according to predefined correlation scoring criteria to ensure that the item set covers acute phase physiological indicators and stable phase social function assessment.

[0032] In one possible implementation, clinical data includes pulmonary function indicators and laboratory test data. For example, pulmonary function indicators are collected via a spirometer, including FEV1 (Fever in one second) and FVC (Final capacity); laboratory test data include tacrolimus blood concentration and donor-specific antibody (DSA) titer. It is understood that blood concentration is used to monitor the effectiveness of immunosuppressive therapy, while DSA titer reflects the risk of rejection, both of which are key parameters for assessing postoperative complications.

[0033] In terms of data normalization, this embodiment adopts a differentiated mapping strategy for heterogeneous data characteristics. Specifically: Pulmonary function index: normalized according to the reasonable value range defined by medical guidelines. For example, the normalization formula for FEV1 is: where x min with x max They are the lower and upper limits of the normal range of postoperative FEV1; Blood drug concentration: Truncation and normalization were performed based on the therapeutic window threshold (e.g., tacrolimus concentration 5-15 ng / mL), and the concentration was forced to be mapped to 0 or 1 when it exceeded the range.

[0034] It should be noted that the purpose of normalization is to eliminate dimensional differences so that the eigenvalues ​​of different data sources can be weighted and integrated. Preferably, the selection of the normalization range should be consistent with clinical medical consensus to avoid subjective settings that lead to evaluation bias.

[0035] In this embodiment, the behavioral data includes oxygen therapy-related parameters, and the collection, quantification, and fusion methods are as follows: 1. Oxygen therapy behavior data collection and quantification As an option, oxygen therapy compliance rate R O2 The quantification is based on the usage log of the smart oxygen therapy device and the patient's physiological indicators. Specifically: - Usage log data: The daily activation time T of the oxygen therapy device is collected through the Internet of Things (IoT) module use , single use interval Δt and oxygen flow setting value F O2 - Physiological indicator association: Simultaneously record blood oxygen saturation (SpO2) monitoring data and calculate the proportion of effective oxygen therapy time: where [x] + =max(x,0), which means that the effective duration is only counted when SpO2 increases after oxygen therapy. It should be noted that the data is transmitted to the central server via Bluetooth or NB-IoT protocol and stored after denoising and outlier filtering. For example, the oxygen flow setting value F O2 When the flow rate exceeds the prescribed range (e.g. >5L / min), an alarm is triggered and invalid data is excluded.

[0036] 2. Integration of oxygen therapy behavior data and the three-level evaluation model In the third-level evaluation of claim 5, the dynamic optimization formula is adjusted to: S final =γ·G+(1-γ)·(ω1·R O2 +ω2·C O2 ); Where: -C O2 is the oxygen therapy behavior consistency index, and the calculation formula is: Among them: ω1, ω2 are weight coefficients, satisfying ω1+ω2=1; -γ is the grading weight coefficient, and the value range is 0.7≤γ≤0.9.

[0037] For example, when the oxygen therapy behavior consistency index C O2 When the value is <0.5, it is determined that the patient has not followed the oxygen therapy plan in a standardized manner, triggering the nursing staff to provide on-site guidance.

[0038] 3. Oxygen therapy data-driven rehabilitation recommendation generation Association rules based on oxygen therapy behavior data and evaluation results: When R O2 <0.6 and C O2 When the value is <0.4, it is judged as "inadequate oxygen therapy implementation" and it is recommended to adjust the type of oxygen therapy equipment (such as switching from nasal cannula to high-flow humidifier) ​​and shorten the follow-up period; When R O2 ≥0.8 but S final If the expected result is not achieved: Combined with imaging data (such as atelectasis volume V Atelectasis >15%), it is recommended to add airway clearance physical therapy; When the oxygen flow rate setting value F O2 Persistently exceeding the doctor's orders: triggers a respiratory consultation to assess for undiagnosed diffusion dysfunction.

[0039] It is understandable that the above rules are implemented by a decision tree model, and the node splitting conditions are obtained based on the correlation between oxygen therapy parameters and clinical outcomes in historical data. Preferably, the model input features include R O2 、C O2 、F O2 and SpO2 coefficient of variation.

[0040] In this embodiment, step 2 is used to implement dynamic allocation of ICF item weights. Its technical solution includes time attenuation adjustment based on postoperative time and cross-modal data verification correction. The specific implementation method is as follows: As an option, dynamic weight allocation is based on the multimodal data preprocessed in step 1, including postoperative time t, HRCT imaging data and clinical indicators. Specifically, the dynamic weight W(e i The calculation of the time decay weight is divided into two sub-processes: time decay and image verification and correction, which are described below. In one possible implementation, the calculation of the time decay weight is implemented by an exponential decay model. For example, for the initial weight W0(e i ), its attenuation formula with time t is: W t (e i )=W0(e i )·e -λt ; It should be noted that the determination of the attenuation coefficient λ depends on the time distribution characteristics of the risk of postoperative complications. Preferably, by fitting the probability curves of acute rejection and infection events in the 3-year follow-up data after surgery, it is found that the value range of λ is 0.08 to 0.12. It is understandable that this design makes the weights of acute phase-related items (such as respiratory function b440) gradually decrease with the passage of time after surgery, while the weights of stable phase items (such as social role adaptation d710) are relatively increased, thereby dynamically matching the assessment needs of different rehabilitation stages.

[0041] In this embodiment, the image confidence coefficient C img The calculation method includes the quantification and fusion process of multi-parameter image features. Specifically, the image data is digitally processed through HRCT images to extract the characteristics related to postoperative complications, including the volume proportion of the atelectasis area, infectious lesion score and airway stenosis index, and dynamic verification is performed in combination with the lung function prediction error. As an option, the volume proportion of the atelectasis area V Atelectasis The volume of the lung lobe is calculated by a lung lobe volume segmentation algorithm. For example, a threshold segmentation method is used to perform three-dimensional reconstruction of the lung parenchyma area of ​​the HRCT image, and a Hounsfield unit (HU) threshold range is set to distinguish normal lung tissue from atelectasis areas. In one possible implementation, the volume percentage of the atelectasis area is calculated to meet the following requirements: It should be noted that the segmentation algorithm can be combined with morphological opening operations to remove noise interference and ensure the continuity of the boundaries of the non-expanded areas.

[0042] For infectious lesions, score S Infection , its calculation depends on the joint determination of the distribution density of the consolidation area and the bronchial inflation sign. Specifically, the texture features of the consolidation area (such as the grayscale co-occurrence matrix energy value) are quantified by radiomics feature extraction technology, and the density threshold is set to judge the consolidation range. Preferably, when the density of the consolidation area exceeds the preset threshold, a basic score is assigned; if there is a bronchial inflation sign, an additional score is added. Exemplarily, the score calculation satisfies: It is understood that the determination of air bronchogram is achieved by a convolutional neural network (CNN) model, which is trained based on annotated HRCT image datasets and is used to automatically identify the image features of air bronchogram. Stenosis , which measures the diameter deviation rate at the stenosis site through 3D reconstruction of the bronchial tree. Specifically, a region growing algorithm is used to track the direction of bronchial branches, extract the center lines of the main bronchi and lobar bronchi, and measure the diameter change along the center lines. In one possible implementation, the stenosis index calculation satisfies: It should be noted that the normal reference diameter is set based on the statistical average of the ipsilateral healthy bronchial segment. img The fusion calculation satisfies: For example, FEV1 pred The FEV1 value is predicted based on HRCT image features using a random forest regression model. Specifically, the input features include the percentage of atelectasis volume, the mean HU value of the consolidation area, and the bronchial stenosis index, and the output is the predicted FEV1 value. Preferably, the model training data is derived from historical HRCT images and concurrent pulmonary function test results.

[0043] It is understandable that the confidence coefficient C img Used to dynamically modify the ICF item weights. img When the value falls below a preset threshold, it indicates a significant conflict between the imaging data and clinical indicators, triggering a manual review process. In one possible implementation, the threshold is set based on ROC curve analysis to balance the risks of false positives and false negatives.

[0044] It should be noted that the quantification and fusion process of the above-mentioned multi-parameter image features is implemented through medical image processing software (such as MITK and 3D Slicer) combined with a custom algorithm module to ensure the repeatability and clinical applicability of the processing flow.

[0045] It should be noted that the initial weight W0(e i ) allocation rules are strongly correlated with the postoperative period. Optimally, the initial weights for respiratory function (b440) and immunosuppressive complications (b1305) within one month after surgery are set at 0.6, and 0.4, respectively. Three months after surgery, the initial weight for social role adaptation (d710) is increased to 0.3. This allocation strategy is based on clinical expert experience and consensus on postoperative rehabilitation pathways, ensuring that weight allocation conforms to medical principles.

[0046] In this embodiment, step three is used to output the evaluation results of lung transplant rehabilitation effects through a three-level progressive model. The technical solution includes three progressive stages: rapid initial screening, refined grading, and dynamic optimization. The specific implementation method is as follows: Alternatively, a three-tiered progressive model uses the dynamic weights calculated in step 2 and the multimodal data preprocessed in step 1 to sequentially perform risk assessment, grading, and outcome optimization. This tiered design aims to balance assessment efficiency and accuracy, prioritizing patients with high rehabilitation needs for in-depth analysis, thereby optimizing the allocation of computing resources.

[0047] In one possible implementation, the first-level assessment (rapid initial screening) generates a comprehensive risk score by dynamically weighting and fusing clinical data. Specifically, the calculation formula for the comprehensive risk score S1 is: Among them, W(e i ,t) is the dynamic weight output in step 2, x norm (e i ) is the normalized clinical data in step one. Exemplarily, if the weight of respiratory function (b440) is 0.415, and its corresponding FEV1 normalized value is 0.58, then the contribution value of this entry is 0.415×0.58≈0.241. It can be understood that the scoring mechanism strengthens the contribution of key indicators through dynamic weights, while suppressing the influence of low-confidence data. It should be noted that the output results of the first-level assessment are used for screening patients with high rehabilitation needs. Preferably, a threshold θ1 is set, and the second-level assessment is triggered when S1≥θ1. Exemplarily, if θ1=0.5, when the comprehensive score exceeds the threshold, the system automatically marks the patient as high risk and starts a refined grading process. For the second-level assessment (refined grading), this embodiment fuses imaging data with biomarkers for multimodal analysis. Specifically, the input features include: Lung function after verification: FEV1 adj =FEV1 meas ·C img , where C img is the image confidence coefficient calculated in step 2; DSA titer: converted to a [0,1] interval value through the normalization process in step 1; Ground glass opacity ratio: It is directly obtained from the image segmentation result of step 1.

[0048] In one possible implementation, support vector machine (SVM) is used as a classifier for grading. Preferably, the training data of the classifier includes the imaging features, laboratory indicators and clinical diagnosis results of historical cases, and the output grading result G∈{mild=0.2, moderate=0.5, severe=0.8}. It should be noted that when the image confidence C img When the value is <0.3, the system automatically sends an alert to the doctor for manual review to avoid misjudgment caused by low-quality imaging data.

[0049] For the third level evaluation (dynamic optimization), this embodiment introduces patient behavior data to correct the grading results. Specifically, the final score S final The calculation formula is: S final =γ·G+(1-γ)·R O2 ; Among them, R O2is the oxygen therapy compliance rate collected in step 1, normalized to the interval [0,1]; γ is the grading weight coefficient, which is used to balance the contribution of biomedical indicators and behavioral data. Preferably, the value range of γ is 0.7 to 0.9 to reflect the dominance of clinical indicators. For example, if the grading result G = 0.5 and the oxygen therapy compliance rate R O2 =0.6, γ=0.8, then: S final =0.8×0.5+0.2×0.6=0.52; In this embodiment, the dynamic optimization result S final It is used to generate differentiated rehabilitation recommendations. Its core logic is to evaluate the degree of deviation of the patient's rehabilitation status from the baseline of the healthy population, and to formulate recovery-oriented intervention strategies accordingly. Specifically, by comparing the patient's multi-dimensional indicators with the statistical distribution of the healthy population, the functional recovery gap is determined, and then targeted recommendations are output. As an option, the baseline data of the healthy population is obtained through a large-sample health database, covering lung function, exercise tolerance and environmental exposure parameters matched by age, gender and region. For example, the Z-score standardization method is used to quantify the degree of patient deviation: Where X is the patient's current indicator value, μ is the mean of the healthy population, and σ is the standard deviation. It should be noted that the Z-score calculation is performed independently for each ICF item, and the final S final is the weighted comprehensive Z value.

[0050] In one possible implementation, the rehabilitation suggestion generation rule is adjusted as follows: When S final <0.3: It is judged as "significantly deviating from the normal state" and it is recommended to increase the frequency of high-frequency monitoring (such as daily lung function self-test) and adjust the intensity of the rehabilitation program. For example, if the respiratory function (b440) Z-score is <-2, the respiratory muscle strengthening training program is triggered; When 0.3≤S final <0.6: It is judged as "partial functional impairment" and it is recommended to implement precise intervention based on behavioral data. For example, when the social role adaptation (d710) Z-score is <-1, the frequency of vocational rehabilitation psychological counseling should be increased; When S final When the Z-score of the environmental factor (e260) is greater than 1, the home environment optimization guide is output to maintain stability.

[0051] It is understandable that the recommendations are closely related to the comparison with the healthy baseline. Specifically, in the assessment of respiratory function recovery, dynamic thresholds are set based on the FEV1 / FVC ratio distribution of healthy people: When the recovery progress is <30%, it is determined that intensive intervention is needed; 30%-70% is the adjustment period; >70% is the maintenance period.

[0052] It should be noted that the generation of rehabilitation recommendations relies on multi-source data fusion. For example, when both exercise tolerance (b455) and dietary self-care ability (d550) deviate from the normal range, a combined intervention recommendation (such as nutritional intake optimization combined with step-by-step exercise training) is generated. Preferably, the recommendation library is constructed based on clinical guidelines and expert experience, and the priority of recommendations is dynamically optimized through a reinforcement learning model.

[0053] Regarding the recovery assessment of environmental factors (e260), the regional health exposure index is introduced as a benchmark. Specifically, the deviation of the air quality index (AQI) of the patient's residence from the national average is calculated, and the room for environmental improvement is evaluated in combination with the utilization rate of indoor purification equipment. For example, when outdoor PM2.5 continues to exceed the standard and the utilization rate of purification equipment is <50%, it is recommended to upgrade the air purification system and adjust the time of going out.

[0054] It is understandable that the above rules achieve closed-loop management through a three-level evaluation model: Rapid initial screening: Identify significant deviations (e.g., Z-score < -2); Refined grading: Locating the root cause of functional impairment (such as complications of immunosuppression or environmental exposure); Dynamic optimization: Rolling updates of recommendations based on real-time data ensure synchronization with recovery progress.

[0055] In this example, the ICF item screening and quantification method is based on the multidimensional assessment needs of post-lung transplant rehabilitation, covering three categories: physical function, activity participation, and environmental factors. Specifically, these include respiratory function, immunosuppressive complications, social role adaptation, physical function, dietary self-care ability, and environmental factors. Through multi-source data fusion and a dynamic weighting mechanism, a comprehensive assessment of rehabilitation outcomes is achieved.

[0056] As an option, respiratory function (corresponding to ICF item b440) is quantified by combining lung function indicators with image features. Specifically, lung function parameters such as forced expiratory volume (FEV1) and vital capacity (FVC) are collected, and the ventilation heterogeneity index is calculated in combination with the pulmonary lobe ventilation distribution map of the HRCT image. For example, the ventilation heterogeneity index is quantified by the entropy value of the image grayscale histogram, and the calculation formula is: where p i H represents the percentage of each grayscale level in the lung lobe area. Higher H values ​​indicate more significant ventilation malfunction. It should be noted that pulmonary function data were collected using a portable spirometer, and the imaging data were parsed and processed in DICOM format.

[0057] For immunosuppressive complications (corresponding to ICF entry b1305), their quantification relies on cross-modal verification of blood drug concentration and donor-specific antibody (DSA) titer. In one possible implementation, high-performance liquid chromatography (HPLC) is used to detect tacrolimus blood drug concentration, and DSA titer is quantified by flow cytometry. Preferably, the data is normalized to a clinically reasonable value range and then input into a dynamic weight allocation model. It is understandable that too low a blood drug concentration may lead to rejection, while too high a blood drug concentration increases the risk of infection. Therefore, the weight needs to be dynamically adjusted in combination with the postoperative time.

[0058] Regarding social role adaptation (corresponding to ICF item d710), its assessment is achieved through the integration of questionnaire surveys and wearable device data. Specifically, a standardized scale is designed to assess the patient's family role, occupational participation and social activity frequency, while the duration of daily activities and the frequency of social interactions are collected through smart bracelets. For example, the frequency of social interactions is based on GPS positioning data and communication record analysis, and the calculation formula is: It should be noted that the baseline value is set based on statistical data of healthy people of the same age group.

[0059] For the physical function of exercise tolerance (corresponding to ICF item b455), its quantification method includes the fusion of the 6-minute walk test (6MWT) and cardiopulmonary function monitoring data. In one possible implementation, wearable devices are used to collect heart rate variability (HRV) and blood oxygen saturation (SpO2) during walking, and the exercise tolerance index is calculated: Preferably, the predicted value is dynamically adjusted based on the patient's age, height, and preoperative lung function history data.

[0060] The assessment of dietary self-care ability (corresponding to ICF item d550) is combined with the dietitian's assessment and the patient's behavior log. For example, the patient's meal intake and food types are recorded using image recognition technology, and the nutritional balance index is calculated: where w i is the weight coefficient of nutrient categories such as protein and carbohydrates. It should be noted that the image recognition technology is based on the convolutional neural network (CNN) model, which can automatically identify the food ingredients and quantity on the plate.

[0061] It should be noted that the dynamic weight assignment of the aforementioned ICF items follows the time decay and image verification rules in claim 2. For example, during high-infection seasons, the weight coefficient of the environmental factor (e260) can be dynamically increased based on historical infection rate data. Preferably, the weight adjustment is optimized using a gradient descent algorithm to minimize the deviation between the assessment results and clinical outcomes.

[0062] The lung transplant rehabilitation assessment system based on ICF items and clinical data based on frequency data and phase data described below and the lung transplant rehabilitation assessment method based on ICF items and clinical data described above can correspond to each other.

[0063] A lung transplant rehabilitation assessment system based on ICF items and clinical data, including: Data acquisition module, used to obtain ICF entries, clinical data and imaging data; Dynamic weight allocation module, used to dynamically adjust item weights based on postoperative time and imaging data confidence; The hierarchical assessment module is used to perform three-level assessment and output the results. The three-level assessment includes rapid initial screening, refined grading and dynamic optimization.

[0064] The device of this embodiment can be used to execute the above method embodiment, and its principles and technical effects are similar, so they will not be repeated here.

[0065] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A lung transplant rehabilitation assessment method based on ICF items and clinical data, characterized by: The following steps are involved: Step 1: Collect multimodal data, including ICF items, clinical data, imaging data, and behavioral data; Step 2: Dynamically assign ICF item weights based on postoperative time and imaging data confidence; Step 3: Based on dynamic weights and multimodal data fusion, the evaluation results are output through a three-level progressive model.

2. A lung transplant rehabilitation assessment method based on ICF items and clinical data according to claim 1, characterized in that: In the step 1: The ICF items were selected from the WHO International Classification of Functioning database related to lung transplant rehabilitation; The clinical data include lung function indicators, blood drug concentrations and donor-specific antibody titers; The imaging data includes HRCT images, which are used to quantify characteristics related to postoperative complications, including: The volume proportion of the atelectasis area is calculated using the lung lobe volume segmentation algorithm; Infectious lesion scoring, based on the density of the consolidation area and the presence of air bronchograms; The airway stenosis index was measured by three-dimensional reconstruction of the bronchial tree to measure the diameter deviation rate at the stenosis site.

3. A lung transplant rehabilitation assessment method based on ICF items and clinical data according to claim 1, characterized in that: The second step includes the following sub-steps: The initial weight is adjusted for time decay according to the postoperative time t, and the time decay factor satisfies: W t (e i )=W0(e i )·e -λt Where W0(e i ) is the initial weight of the item, and λ is the attenuation coefficient based on the postoperative complication risk curve fitting; Calculate the confidence coefficient C through cross-modality verification of imaging data and clinical data img , and modify the weights: W(e i ,t)=W t (e i )·(1+α·C img ); Where α is the image verification weight coefficient.

4. A lung transplant rehabilitation assessment method based on ICF items and clinical data according to claim 1, characterized in that: The image confidence coefficient C in step 2 img The calculation formula is: Where: V Atelectasis : The ratio of atelectasis volume to total lung lobe volume (0-1); S Infection : Infectious lesion score: when the density of the consolidation area is greater than 50%, it is scored as 0.8, and the presence of air bronchogram is +0.2; I Stenosis : Airway stenosis index, when the main bronchial diameter is less than 60% of the normal value, it is 0.6; when the lobar bronchial stenosis occurs, it is 0.

3.

5. The lung transplant rehabilitation assessment method based on ICF items and clinical data according to claim 1, characterized in that: The three-level progressive model in step 3 includes: First-level assessment: Calculate the comprehensive risk score step 1 based on dynamic weights and clinical data. If step 1 ≥ θ1, mark the patient as a patient with high rehabilitation needs; Second-level assessment: For patients with high rehabilitation needs, the proportion of atelectasis volume, infectious lesion score, airway stenosis index, and DSA titer are integrated to output the classification result G through random forest classifier; Level 3 assessment: combined with self-monitoring adherence R 02 , dynamically optimize the final evaluation result S final .

6. The lung transplant rehabilitation assessment method based on ICF items and clinical data according to claim 1, characterized in that: The calculation formula of the comprehensive risk score S1 in step 3 is: where x norm (e i ) is the normalized clinical data, and the normalized range is the clinically reasonable value range.

7. The lung transplant rehabilitation assessment method based on ICF items and clinical data according to claim 1, characterized in that: The refined classification in step 3 includes: When the image confidence coefficient C img When <0.4, the manual review process is triggered; The classifier was used to classify the lung function FEV1 after verification adj =FEV1 meas ·C img , DSA titer, atelectasis volume ratio, and infectious lesion score were used for grading.

8. The lung transplant rehabilitation assessment method based on ICF items and clinical data according to claim 1, characterized in that: The dynamic optimization in step 3 satisfies: S final =γ·G+(1-γ)·R O2 ; in: G is the grading result, with the coding bit value as mild = 0.2, moderate = 0.5, and severe = 0.8; R O2 is the oxygen therapy compliance rate, normalized to the interval [0,1]; γ is the grading weight coefficient, and its value range is 0.7≤γ≤0.

9.

9. The lung transplant rehabilitation assessment method based on ICF items and clinical data according to claim 1, characterized in that: The ICF entries include: respiratory function; complications of immunosuppression; social role adaptation; body functions; Ability to take care of oneself; Environmental factors.

10. A lung transplant rehabilitation assessment system based on ICF items and clinical data, according to a lung transplant rehabilitation assessment method based on ICF items and clinical data according to any one of claims 1 to 9, characterized in that: include: Data acquisition module, used to obtain ICF entries, clinical data and imaging data; Dynamic weight allocation module, used to dynamically adjust item weights based on postoperative time and imaging data confidence; The hierarchical evaluation module is used to perform three-level evaluation and output the results. The three-level evaluation includes rapid initial screening, refined classification and dynamic optimization.

Images