A personalized rehabilitation shared decision making system, method, and medium based on digital humans

Abstract

The application discloses a kind of personalized rehabilitation sharing decision system, method and medium based on digital person, system includes data acquisition and monitoring module, digital person state generation module, shared decision auxiliary module, interaction and incentive module, health literacy adaptation module and adaptive interactive interface module;Digital person state generation module obtains comprehensive state index and generates rehabilitation digital person;Interaction and incentive module are used for the interaction of rehabilitation digital person and patient;Health literacy adaptation module is used to evaluate electronic health literacy score and generate classification label;Adaptive interactive interface module is used to select the interactive interface corresponding to classification label.The application converts abstract rehabilitation data into digital person image through comprehensive state index, provides quantitative, visual decision basis and improves patient compliance;Also use weight algorithm and differentiating interaction mechanism, perceive and adapt to patient cognitive ability, realize the dual personalization of rehabilitation training and interactive experience, effectively solve rehabilitation training interaction problem.

Description

Technical Field

[0001] This invention relates to the fields of medical information technology and intelligent health management technology, specifically to a personalized rehabilitation sharing decision-making system, method and medium based on digital human, which assists patients in home rehabilitation training through a visualized digital human image, adaptive quantitative assessment algorithm and differentiated interaction mechanism. Background Technology

[0002] Rehabilitation training is a core medical approach to promote functional recovery and improve quality of life for patients. However, the traditional hospital-centered rehabilitation model is limited by the uneven distribution of medical resources and the shortage of professional personnel. Patients generally face the dilemma of lacking continuous professional guidance, supervision and motivation after discharge, resulting in low compliance with home rehabilitation, poor rehabilitation effects and even relapse of the condition.

[0003] Current research indicates that in remote rehabilitation programs without any decision-making support tools, patient adherence at 12 weeks is only around 50%, and this rate drops further to less than 40% among older adults with low electronic health literacy. This further confirms the inadequacy of existing technologies in addressing differences in user cognition and decision-making participation. While the emerging remote rehabilitation technologies have alleviated geographical and time limitations to some extent, enabling remote online communication and guidance, they still reveal the following common shortcomings in practical applications: (1) The information presentation is abstract, and the cognitive burden on patients is heavy: Existing remote rehabilitation systems mostly use numerical values, charts and other forms to provide patients with feedback on their rehabilitation progress. For elderly patients or patients with low health literacy, such abstract information is difficult to understand and use effectively. Patients cannot establish an intuitive connection between their daily behavior and rehabilitation goals, which weakens their intrinsic motivation for behavior change. (2) Lack of effective doctor-patient interaction and decision-making participation mechanisms: Most existing systems only support one-way rehabilitation plan push functions and fail to effectively support the patient-centered medical model of "Shared Decision Making (SDM)". Patients passively accept the pre-set plan, and their personal values, rehabilitation preferences and actual concerns are not fully expressed and respected in the decision-making process, resulting in low psychological acceptance of the rehabilitation plan and difficulty in maintaining long-term compliance; (3) Lack of emotional support and dynamic encouragement during the rehabilitation process: The home rehabilitation environment naturally lacks real-time companionship and emotional support from professional medical personnel. Existing technology has not yet provided a humanized interactive role that combines supervision, companionship and encouragement. Patients are very likely to feel tired and lonely due to the boring process, delayed feedback or lack of a sense of accomplishment, and thus interrupt their rehabilitation training; (4) The system design does not take into account the heterogeneity of users' health literacy: The user interface and interaction logic of existing remote rehabilitation systems are mostly designed in a "one-size-fits-all" manner, without differentiated adaptation for users with different levels of eHealth Literacy (eHL). Some patients are often marginalized due to the complexity of operation, information overload, or difficulty in interpretation, which exacerbates the "digital divide" problem in the utilization of rehabilitation services.

[0004] In summary, there is an urgent need for a new type of intelligent rehabilitation support system that can transform abstract rehabilitation data into intuitive visual feedback, support efficient collaborative decision-making between doctors and patients, provide personalized companionship and incentives, and adaptively match different users' cognitive abilities. Summary of the Invention

[0005] To address the aforementioned issues, the present invention aims to propose a personalized rehabilitation shared decision-making system, method, and medium based on digital humans. This system transforms abstract rehabilitation data into a dynamic digital human image through a comprehensive state index, providing quantitative and visual decision-making support and improving patient adherence to long-term rehabilitation training. Simultaneously, it employs a weighted algorithm and a differentiated interaction mechanism to perceive and adapt to the patient's cognitive abilities, achieving dual personalization in rehabilitation training and interactive experience, thus bridging the interaction problems faced by the elderly and those with low health literacy scores.

[0006] This was achieved through the following technical solutions: A personalized rehabilitation sharing decision-making system based on digital humans includes a data acquisition and monitoring module, a digital human status generation module, a sharing decision support module, an interaction and incentive module, a health literacy adaptation module, and an adaptive interactive interface module. The data acquisition and monitoring module is used to collect multi-source data from patients. The health literacy adaptation module is used to assess patients' electronic health literacy scores (E_score) and generate corresponding grade labels. The shared decision support module provides a visual comprehensive assessment report based on a rehabilitation digital human, explanations of rehabilitation option trade-offs, goal setting tools, and program simulation functions. The interaction and motivation module is used for voice and text interaction between the rehabilitation digital human and the patient, as well as to provide action demonstrations, schedule reminders and instant motivational feedback; The adaptive interactive interface module is used to select the corresponding interactive interface based on the hierarchical labels. The digital human status generation module includes: a normalization processor, used to map multi-source data to the [0,1] interval to obtain a normalized index set; a microprocessor, which integrates a weight calculation engine and a weighted summation unit. The weight calculation engine is used to calculate the Comprehensive Status Index (CSI) based on the weight coefficients corresponding to the grading labels output by the health literacy adaptation module, and the weighted summation unit is used to perform weighted summation operations; and an image processor, used to convert the CSI into visual status parameters of the rehabilitation digital human according to preset mapping rules and output them to the display interface. The microprocessor is communicatively connected to the normalization processor and the image processor.

[0007] Optionally, the weighted summation unit first dynamically determines the clinical weight adoption factor α and the preference weight adoption factor β based on the electronic health literacy score E_score, where α is positively correlated with E_score and β is negatively correlated with E_score; then, based on α, β, and the preset clinical weights and patient preference weights, it calculates the comprehensive weight Wi for each indicator, where i refers to the i-th comprehensive weight; finally, it calculates the CSI based on the weighted sum of the comprehensive weight Wi and the normalized indicators; the weighted summation unit dynamically adjusts the contribution ratio of clinical weights and patient preference weights in the comprehensive weight calculation based on the grading labels output by the health literacy adaptation module; when the electronic health literacy score E_score is higher, the contribution ratio of clinical weights is greater; when the E_score is lower, the contribution ratio of patient preference weights is greater.

[0008] Optionally, the formula for calculating the overall weight Wi is: ,in, ; α base , β base All are preset constants, and γ is a preset amplification factor.

[0009] Optionally, the digital human status generation module also includes a safety constraint correction unit: when any preset key safety indicator normalized value x'_k is lower than the corresponding safety threshold S_k, the comprehensive status index CSI is penalized and the corrected CSI′=CSI×min(1,x'_k / S_k) is obtained; CSI′ is used to replace CSI and converted into visual status parameters of the rehabilitation digital human; the safety threshold S_k is dynamically set according to the patient's age, underlying diseases and clinical guidelines, or preset by medical staff.

[0010] Optionally, a state evolution prediction engine, a rehabilitation option value clarification tool, and a decision consistency evaluation unit are provided in the shared decision-making assistance module; the state evolution prediction engine is used to predict the predicted value of the CSI in the future period based on the current value and historical data of the comprehensive state index CSI, and generate a corresponding preview image of the rehabilitated digital human; the rehabilitation option value clarification tool is used to guide the patient to quantify the preference weights for different rehabilitation goals in the form of an interactive questionnaire or graphical cards; the decision consistency evaluation unit is used to record the decision-making process of the patient and medical staff and generate a traceable decision-making file.

[0011] Optionally, the adaptive interaction interface module switches between the first interaction interface and the second interaction interface according to the hierarchical label; wherein, the second interaction interface uses the image change of the rehabilitated digital human as the core information carrier, and is supplemented by voice broadcast for explanation; the first interaction interface provides a text list, data charts, and parameter adjustment controls.

[0012] Optionally, when the health literacy adaptation module generates the hierarchical label, it compares the electronic health literacy score E_score with the preset hierarchical threshold T_eHL; when E_score≥T_eHL, it is determined as the first interaction mode; when E_score<T_eHL, it is determined as the second interaction mode; when the adaptive interaction interface module selects the interaction interface, if in the first interaction mode, it displays the interface for each core information with a text list, data charts, and trend curves, and provides parameter adjustment controls; if in the second interaction mode, it is based on the dimensionality reduction transformation of each core information into the image change of the rehabilitated digital human, supplemented by voice broadcast for corresponding explanation, and at the same time adopts a binary comparison selection interface or the default simplified interaction interface.

[0013] A personalized rehabilitation shared decision-making method is also proposed, which operates using the personalized rehabilitation shared decision-making system as described above, including an initial stage and a periodic operation stage; Among them, the initial stage includes: S1: Evaluate the patient's electronic health literacy score E_score through the health literacy adaptation module to determine the interaction mode; S2: Set the functional relationship between the clinical weight adoption factor α and the preference weight adoption factor β according to the interaction mode; α is positively correlated with E_score, and β is negatively correlated with E_score; ; α base 、 β base Both are preset constants, and γ is a preset amplification factor; The periodic operation stage is used to repeat the following steps in each rehabilitation training cycle: S3: Collect multi-source data from patients through the data acquisition and monitoring module, and obtain an indicator set through normalization processing; S4: Based on the functional relationship determined in step S2, calculate the comprehensive weight of each indicator, and then calculate the Comprehensive State Index (CSI). S5: Based on the preset mapping rules, the CSI is mapped to the visual image of the rehabilitation digital human, and a smooth interpolation algorithm is used for continuous gradual changes; the preset mapping rules include at least three intervals: when CSI>0.7, the rehabilitation digital human presents a good image with a rosy complexion and a pleasant expression; when 0.4≤CSI≤0.7, the rehabilitation digital human presents a warning image with a pale complexion and a slight frown; when CSI<0.4, the rehabilitation digital human presents a risky image with cyanotic lips and a painful expression. S6: Based on the interaction mode determined in step S1, execute differentiated information presentation and interaction strategies through the adaptive interactive interface module.

[0014] S7: Through the shared decision support module and interaction and incentive module, it enables the interaction between the rehabilitation digital human and the patient and records a visual comprehensive assessment report.

[0015] Secondly, a computer-readable storage medium is proposed, on which a computer program is stored, which, when executed by a processor, implements the personalized rehabilitation sharing decision-making method described above.

[0016] Optionally, a computer program is stored in a memory to support a rehabilitation digital human terminal device. The rehabilitation digital human terminal device includes a display screen, a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it generates and outputs a rehabilitation digital human image as described in the above-mentioned personalized rehabilitation sharing decision-making system based on digital humans, and displays it on the display screen.

[0017] The beneficial effects of this invention compared to the prior art are: 1) By adding specific hardware modules such as normalization processors, microprocessors and image processors, abstract rehabilitation data is transformed into dynamic digital human images, overcoming the shortcomings of existing technologies that rely solely on abstract feedback based on numerical values ​​and charts. This significantly lowers the cognitive threshold for patients, enabling them to intuitively observe rehabilitation progress and actively participate in decision-making.

[0018] 2) The system adopts a weighted algorithm that couples health literacy, enabling it to automatically adjust the contribution ratio of clinical objective indicators and patient subjective preferences based on the patient's electronic health literacy score. At the same time, a safety constraint correction unit is introduced to proactively reduce the comprehensive status index and trigger an early warning when key safety indicators are abnormal, thus ensuring the safety of home rehabilitation from a technical perspective.

[0019] 3) The adaptive interactive interface module enables automatic switching between the two interaction modes, which solves the digital divide problem caused by the existing "one-size-fits-all" design and ensures that elderly patients and patients with low health literacy can also effectively use the system.

[0020] 4) The digital human integrates companionship, motivation, and early warning functions to form a closed-loop feedback, significantly improving long-term rehabilitation compliance; the system architecture is universal and can be easily adapted to various fields such as pulmonary rehabilitation, orthopedic rehabilitation, cardiac rehabilitation, and neurological rehabilitation, and has broad clinical application prospects. Attached Figure Description

[0021] Figure 1 A schematic diagram of a module for a personalized rehabilitation shared decision-making system based on digital humans; Figure 2 This is a flowchart of a personalized rehabilitation sharing decision-making method based on digital humans. Detailed Implementation

[0022] The technical solutions in the embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0023] like Figure 1 The diagram shows a modular design of a personalized rehabilitation shared decision-making system based on a digital human. The system comprises multiple modules, including a data acquisition and monitoring module, a digital human status generation module, a shared decision support module, an interaction and incentive module, a health literacy adaptation module, and an adaptive interactive interface module. Through the collaborative efforts of these modules, the system transforms the patient's abstract rehabilitation training data into a dynamic digital human image. Furthermore, by employing a weighted algorithm and a differentiated interaction mechanism, it perceives and adapts to the patient's cognitive abilities, achieving dual personalization in both rehabilitation training and interactive experience. This effectively overcomes the interaction and compliance issues inherent in traditional rehabilitation training.

[0024] In this embodiment, the data acquisition and monitoring module is used to collect multi-source data related to the patient's rehabilitation training, including but not limited to: physiological parameters (such as heart rate and blood oxygen), functional assessment indicators (such as 6-minute walking distance, joint range of motion, and muscle strength), symptom scores (such as dyspnea, pain, and fatigue), rehabilitation training completion rate, and medication records.

[0025] In this embodiment, the health literacy adaptation module is used to evaluate the patient's e - health literacy score E_score and generate corresponding classification labels. The health literacy adaptation module can guide the patient to complete an online assessment of a standardized e - health literacy scale (such as eHEALS) to obtain the original score; then normalize this original score and map it to the interval [0, 1] to obtain the e - health literacy score E_score. When generating the classification label, the health literacy adaptation module compares the e - health literacy score E_score with a preset classification threshold T_eHL; when E_score≥T_eHL, it is determined as the first interaction mode; when E_score < T_eHL, it is determined as the second interaction mode. The classification mechanism realizes a rapid and standardized stratification of the patient's cognitive ability, provides a clear control signal for subsequent adaptive interaction, thus facilitating the patient's use and achieving a differentiated user experience.

[0026] In another embodiment, the e - health literacy score E_score can be replaced by other parameters characterizing the patient's cognitive ability, including but not limited to: user age, education level, mobile device operation proficiency score, or the digital health literacy index inferred by a historical interaction behavior machine learning model. These alternative parameters are also used to dynamically adjust the proportional relationship between the clinical weight adoption factor α and the preference weight adoption factor β to achieve a similar cognitive ability adaptation effect.

[0027] In this embodiment, the digital human state generation module is used to receive multi - source data, obtain a comprehensive state index CSI (Comprehensive Status Index) based on a multi - dimensional state fusion algorithm, and then generate a corresponding rehabilitation digital human in real - time in terms of visual image according to a preset mapping rule.

[0028] The digital human state generation module includes a normalization processor, a microprocessor, and an image processor. The microprocessor is communicatively connected to the normalization processor and the image processor respectively. The normalization processor is used to map the multi - source data to the interval [0, 1] to obtain a normalized index set. The microprocessor internally integrates a weight calculation engine and a weighted summation unit. The weight calculation engine is used to calculate the comprehensive state index CSI according to the weight coefficients corresponding to the classification labels output by the health literacy adaptation module; the weighted summation unit is used to perform a weighted sum operation. The image processor is used to convert the CSI into the visual state parameters of the rehabilitation digital human according to a preset mapping rule and output them to the display interface.

[0029] The process by which the weighted summation unit obtains the comprehensive state index CSI is as follows: 1) First, based on the electronic health literacy score E_score, dynamically calculate the clinical weight adoption factor α and the preference weight adoption factor β for each indicator; where α is positively correlated with E_score and β is negatively correlated with E_score. α and β are used to dynamically balance the contribution ratio of objective clinical standards and subjective patient preferences in the overall weight. 2) Then, based on α, β, and the preset clinical weights and patient preference weights, the overall weight Wi for each indicator can be calculated. ,in, , α base , β base All are preset constants, γ is a preset amplification coefficient, and i is used to refer to the comprehensive weight of the i-th indicator. 3) Then, based on the comprehensive weight Wi and the weighted sum of the normalized indicators, the corresponding Comprehensive Status Index (CSI) is calculated. Through the weight adjustment mechanism, the clinical objective standards and the patient's subjective preferences can be adaptively integrated according to the patient's own electronic health literacy score, so that the CSI more realistically reflects the patient's actual recovery status.

[0030] The weighted summation unit can dynamically adjust the contribution ratio of clinical weight and patient preference weight in the comprehensive weight calculation based on the graded labels output by the health literacy adaptation module. When the electronic health literacy score E_score is higher, the contribution ratio of clinical weight is greater; when the E_score is lower, the contribution ratio of patient preference weight is greater.

[0031] The digital human status generation module also incorporates a safety constraint correction factor when calculating the Comprehensive Status Index (CSI): when any preset key safety indicator (such as blood oxygen saturation, heart rate, etc.) among multiple indicators is detected to be below the corresponding safety threshold, a punitive correction is applied to the CSI. This correction enables timely detection of abnormal physiological states in patients, significantly improving system safety.

[0032] In this embodiment, the shared decision support module is equipped with a state evolution prediction engine, a rehabilitation option value clarification tool, and a decision consistency assessment unit. This module can be used to provide a visual comprehensive assessment report based on a rehabilitation digital human, a rehabilitation option trade-off explanation, a goal setting tool, and a program simulation function.

[0033] The state evolution prediction engine is used to simulate and predict the CSI prediction value within a preset future period based on the current value of the Comprehensive State Index (CSI) and the corresponding historical data, using time series prediction models (such as ARIMA models or linear regression models), and generate a corresponding preview image of the rehabilitation digital human prediction. This allows patients or remote medical staff to intuitively observe the expected future effects of rehabilitation training.

[0034] The rehabilitation option value clarification tool is used to guide patients to express and quantify their preference weights for different rehabilitation goals. For example, it can guide patients to express the relative importance of different rehabilitation goals (such as "improving dyspnea" and "increasing walking distance") through interactive questionnaires or graphical card classification, and quantify the ranking results as input for preference weights.

[0035] The decision consistency assessment unit is used to review and confirm the decision process and create a traceable record after patients and healthcare professionals have shared decision-making discussions or after patients interact with the rehabilitation digital human.

[0036] By visually previewing future states, the expected effects of different training programs can be simulated intuitively, greatly improving the comprehensibility and willingness to participate in shared decision-making. Value clarification tools transform patients' subjective feelings into quantifiable weight parameters, while decision files provide a traceable basis for subsequent program adjustments, ensuring the scientific nature and continuity of training and decision-making.

[0037] In this embodiment, the interaction and incentive module is used for voice and text interaction between the rehabilitation digital human and the patient, as well as for providing standardized action demonstrations for rehabilitation training, schedule reminders, real-time incentive feedback based on the completion of rehabilitation training, and automatic early warning for abnormal situations.

[0038] In this embodiment, the adaptive interactive interface module is used to switch between a first interactive interface and a second interactive interface based on hierarchical labels. When selecting an interactive interface, if in the first interactive mode, the adaptive interactive interface module displays each core piece of information as a text list, data chart, and trend curve, and provides parameter adjustment controls; if in the second interactive mode, it reduces the dimensionality of each core piece of information to the image changes of a rehabilitation digital human, supplemented by voice broadcast for corresponding explanations, and simultaneously uses a binary comparison selection interface or a default simplified interactive interface.

[0039] Secondly, combining Figure 2 The flowchart of the personalized rehabilitation shared decision-making method shown in this embodiment also proposes a personalized rehabilitation shared decision-making method, which is operated using the above-mentioned personalized rehabilitation shared decision-making system based on digital humans. The method includes an initial stage and a periodic operation stage.

[0040] The initial stage includes the following steps: S1: Assess the patient's electronic health literacy score E_score through the health literacy adaptation module to determine the interaction mode.

[0041] When obtaining the e-health literacy score, first guide the patient to complete an online assessment of the standardized e-health literacy scale (such as eHEALS) to obtain the original score. The system normalizes this original score and maps it to the interval [0, 1] to obtain the e-health literacy score (also called the e-health literacy index) E_score.

[0042] To prevent the imbalance of subsequent weight calculation caused by extreme evaluation values, boundary truncation processing can also be performed on E_score: when E_score < E_min, let E_score = E_min; when E_score > E_max, let E_score = E_max. Here, E_min can be set to 0.1 and E_max can be set to 0.95.

[0043] The system presets at least one grading threshold T_eHL. When E_score ≥ T_eHL, it is determined that the patient is in the first interaction mode (standard mode); when E_score < T_eHL, it is determined that the patient is in the second interaction mode (enhanced assistance mode).

[0044] S2: Set the functional relationship between the clinical weight adoption factor α and the preference weight adoption factor β according to the interaction mode.

[0045] The periodic operation stage is used to repeat the following steps in each rehabilitation training cycle: S3: Collect the multi-source data of the patient through the data collection and monitoring module and obtain the index set through normalization. For the specific rehabilitation field where the patient is located, a set of multiple core indicators with clinical evaluation significance (which can be determined according to the clinical practice guidelines of related diseases, etc.) can be selected from the data collection and monitoring module to form the index set X = {x1, x2,..., xn}, where n corresponds to the number of multi-core core indicators. The system normalizes the original values of each collected indicator according to the preset clinical reference range and maps them to the interval [0, 1], where the value 1 represents the optimal state of the indicator in terms of clinical significance and the value 0 represents the worst state, to obtain the normalized index set X' = {x'1, x'2,..., x'n}.

[0046] S4: Based on the functional relationship determined in step S2, calculate the comprehensive weight of each indicator, and then calculate the comprehensive status index CSI.

[0047] Specifically, when calculating the comprehensive weight of each indicator using the multidimensional state fusion algorithm, the comprehensive weight Wi for each normalized arbitrary indicator x'i is first defined as Wi = (α × W_clinical_i + β × W_preference_i) / (α + β). Here, α is the clinical weight adoption factor, β is the preference weight adoption factor, and α and β are used to dynamically balance the contribution ratio of objective clinical standards and patient subjective preferences in the comprehensive weight. W_clinical_i is the preset clinical weight corresponding to indicator x'i (which can be preset according to clinical treatment guidelines for relevant diseases), and W_preference_i is the preset preference weight corresponding to indicator x'i (which can be determined by medical staff based on the patient's personal rehabilitation goal preferences). α and β are not fixed constants, but rather dynamic functions associated with the electronic health literacy score E_score: α = α base ×(1+γ×E_score), β= β base ×(2-E_score), i=1, 2, ..., n; α base , β base γ is a preset baseline constant used to ensure the numerical stability of weight adjustments; γ is a preset amplification factor used to adjust the sensitivity of E_score to the influence of clinical weight adoption factors.

[0048] Taking a patient with E_score=0.55 as an example. Setting... α base =1.0, β base =1.0, γ=2.0. Step 1: α=1.0×(1+2.0×0.55)=2.10, β=1.0×(2-0.55)=1.45. Step 2: Let the clinical weight of a certain indicator be 0.25 and the preference weight be 0.35, then Wi=(2.10×0.25+1.45×0.35) / (2.10+1.45)=0.291. Step 3: If the normalized values ​​of the four indicators are all 0.70, then CSI=0.291×0.70×4≈0.815. Step 4: If blood oxygen saturation x'SpO2=0.88 and S_SpO2=0.90, then CSI'=0.815×0.978≈0.797.

[0049] Using a multidimensional state fusion algorithm, the following adaptive effects can be achieved: 1) When a patient has high electronic health literacy (E_score close to 1), the α value increases accordingly, and the β value decreases accordingly. When calculating the overall state, the system will adopt objective clinical weights to a greater extent. These patients can correctly understand the meaning of clinical indicators, thus the rehabilitation plan developed accordingly is more scientific and challenging. 2) When a patient has low electronic health literacy (E_score close to 0), the α value returns to the baseline level, while the β value increases significantly. The system will adaptively favor the patient's subjectively expressed preferences. For example, if a patient with low health literacy repeatedly emphasizes their fear of a certain symptom (such as shortness of breath or pain), the system will assign a higher overall weight to the symptom-related indicators during calculation, making the quantitative results closer to the patient's personal feelings and core concerns. This effectively maintains their motivation to participate and their confidence in rehabilitation training, even when their cognitive abilities are limited.

[0050] After obtaining the comprehensive weight of each indicator, the comprehensive status index CSI=Σ(Wi×x'i) can be calculated. At the same time, in order to ensure the safety of patients during rehabilitation training, when any of the preset key safety indicators x'k (such as blood oxygen saturation, heart rate, etc.) is lower than the corresponding safety threshold S_k, a punitive correction formula CSI'=CSI×min(1,x'k / S_k) is used for correction.

[0051] Through punitive correction, it can be ensured that when a patient exhibits an abnormal condition posing a clinical risk, the CSI value will be significantly lowered regardless of other indicators, thus ensuring that the digital avatar accurately reflects the urgent health risk. Simultaneously, the calculated CSI value (or the safety-corrected CSI' value) is a dimensionless value within the range [0,1], comprehensively reflecting the patient's overall recovery status across both objective clinical indicators and subjective personal feelings.

[0052] In this embodiment, the rehabilitation digital human is a multi-dimensional, dynamic, and visual model, whose state is determined by multiple sub-modules related to rehabilitation goals: Core Physiological / Functional Status Submodule: This is the basic module. It dynamically changes based on indicators reflecting the patient's core functional impairments (such as blood oxygen saturation and dyspnea in patients with cardiopulmonary diseases; pain level and joint range of motion in orthopedic patients; and muscle strength and balance in patients with neurological diseases). For example, in a pulmonary rehabilitation scenario, this module can be represented as "Respiratory Status" (Good, Warning, Risk); in an orthopedic rehabilitation scenario, it can be represented as "Limb Movement Status" (Free Movement, Slightly Limited, Severely Limited).

[0053] The compliance and behavior submodule links to the patient's rehabilitation training completion status, medication records, etc. The digital human can visualize compliance through methods such as wearing badges and displaying status icons above their head.

[0054] The Emotion and Psychological State submodule: Based on the patient's reported emotion scores or related scales, the dynamic changes in the digital human's facial expressions indicate their psychological state.

[0055] Quality of life related submodule: Symbolic cues can be provided through the digital human's facial expressions, posture, or background elements based on other quality of life indicators reported by the patient (such as sleep quality and fatigue level).

[0056] S5: Based on preset mapping rules, the Comprehensive State Index (CSI) is mapped to the visual image of the rehabilitation digital human, and the visual image is continuously and gradually changed through a smooth interpolation algorithm. The smooth interpolation algorithm defines the key visual parameters of the rehabilitation digital human as a continuous function of the CSI, including at least facial rosiness, the upward curve of the corners of the mouth, and the forward tilt angle of the body. The digital human is preset with multiple visual state templates, each corresponding to a range of CSI values ​​and possessing differentiated visual features (such as facial skin color, facial expression, body posture, auxiliary symbols, etc.). The digital human state generation module dynamically switches and renders the corresponding digital human visual image on the patient's end based on the real-time calculated CSI values, according to the above mapping rules and smoothing algorithm. This quantifies and dynamically maps abstract, multi-dimensional rehabilitation data into visual feedback that the patient can intuitively perceive.

[0057] The preset mapping rules include at least three intervals: when CSI > 0.7, the rehabilitation android presents a first state, such as a rosy complexion and a pleasant expression, indicating a good condition; when 0.4 ≤ CSI ≤ 0.7, the rehabilitation android presents a second state, such as a pale complexion and a slightly furrowed brow, indicating a warning condition; when CSI < 0.4, the rehabilitation android presents a third state, such as cyanosis of the lips and a pained expression, indicating a risk condition. Each different state can be preset in the system, and then the corresponding state is selected and presented based on the corresponding CSI value.

[0058] When using smooth interpolation algorithms, key visual parameters of the digital human (such as RGB values ​​of facial rosiness, the upward curve of the corners of the mouth, and the angle of body leaning forward) are defined as continuous functions of CSI values. For example, the facial rosiness R-channel value R(CSI) = R_min + (R_max - R_min) × f(CSI), where f(CSI) is a preset smoothing mapping function (such as the sigmoid function or a piecewise linear function). This makes the changes in the digital human's state appear as a natural, continuous, and gradual process, further enhancing the patient's immersion and trust.

[0059] S6: Based on the interaction mode determined in step S1, the adaptive interaction interface module executes differentiated information presentation and interaction strategies. The adaptive interaction interface module has adaptive information carrier capabilities and adaptive interaction complexity capabilities. Regarding the adaptive information carrier capability, it includes: 1) When the first interaction mode is determined, the patient-side interface uses text lists, data charts, and trend curves as the main information carriers, providing a detailed data view; 2) When the second interaction mode is determined, the system automatically reduces the dimensionality of the core information and converts it into visual state changes of a digital human image, supplemented by speech synthesis for explanation. In other words, the system actively hides or folds complex values ​​and charts to reduce the cognitive load on the patient.

[0060] Regarding the adaptive capability for interaction complexity, the system includes: 1) In the first interaction mode, the shared decision interface provides a complete list of rehabilitation options and refined parameter adjustment controls (such as sliders and input boxes); 2) In the second interaction mode, the system activates a simplified decision navigation protocol, which means that a binary comparison selection or default preference recommendation mechanism is adopted on the interface. Simple and intuitive graphical buttons and guided questions replace the complex parameter setting interface, and a safe default option is preset based on the clinical pathway. If the patient does not make a selection within the preset time, the system automatically adopts the default option to ensure the continuity and safety of nursing intervention.

[0061] S7: In each rehabilitation training cycle, the interaction between the rehabilitation digital human and the patient is executed through the shared decision support module and interaction and incentive module, and the corresponding visual comprehensive assessment report is recorded.

[0062] For patients' long-term rehabilitation training cycle, when the patient first uses the personalized rehabilitation shared decision-making system, the system completes data collection, eHL assessment, and initial CSI calculation, generating a corresponding initial digital avatar. This initial digital avatar explains the patient's current overall health status in a way that adapts to the patient's interaction mode (voice / animation / simplified charts). Then, the patient and remote medical staff can conduct the first shared decision-making meeting: the medical staff uses the state evolution prediction engine of the shared decision-making support module to show the simulation results of the digital avatar's future appearance under different rehabilitation plans. The medical staff and the patient jointly agree on an initial personalized rehabilitation plan that includes specific and quantifiable goals (such as "the CSI value will increase to above 0.6 after 4 weeks" or "the digital avatar's complexion will be rosier") using visual goal-setting tools and rehabilitation option value clarification tools. After the meeting, the decision consistency assessment unit records the decision process in the corresponding file. Secondly, patients undergo home-based rehabilitation training according to the initial personalized rehabilitation plan. The digital human provides standardized movement demonstrations, personalized reminders, and incentives. The CSI value changes with the patient's daily / weekly data updates and is mapped in real time to subtle changes in the digital human's image (such as a gradually rosier complexion and increased range of motion) through a smooth transition algorithm, forming an immediate and intuitive positive feedback loop. Simultaneously, when the monitored data triggers a preset safety threshold, the digital human proactively changes its image and issues a voice warning. Finally, a mid-term evaluation and plan re-decision are conducted: at the preset mid-term evaluation time point, the system automatically generates a comprehensive visual report including the CSI change curve and key indicator trends. Both doctors and patients can use this report to evaluate the effectiveness of the initial rehabilitation training and, again through a shared decision-making process assisted by the digital human, jointly determine the direction of adjustments to the next stage of the rehabilitation training plan.

[0063] The system of this invention can be divided into three modules: a patient terminal, a remote medical care terminal, and a cloud server. The patient terminal includes a data acquisition and monitoring module, an adaptive interactive interface module, and an interaction and incentive module. The cloud server includes a health literacy adaptation module and a digital human status generation module. The remote medical care terminal includes a shared decision support module. The cloud server can use a general-purpose cloud computing server (such as a cloud virtual host based on x86 architecture) running a Linux operating system. The mini-program used on the patient terminal is built based on the WeChat mini-program development framework and supports iOS and Android mobile operating systems. The web management backend of the medical care terminal adopts a B / S architecture and is accessed through mainstream browsers. Data storage adopts a hybrid architecture of relational databases (such as MySQL) and time-series databases (such as InfluxDB) to accommodate the storage needs of structured scale data and high-frequency physiological monitoring data.

[0064] This solution also proposes a computer-readable storage medium on which a computer program is stored, which, when executed by a processor, implements the aforementioned personalized rehabilitation sharing decision-making method.

[0065] In addition, a rehabilitation digital human terminal device is proposed, including: a display screen, a memory, a processor, and a computer program stored in the memory and capable of running on the processor; when the processor executes the program, it generates and outputs a rehabilitation digital human image, such as in a personalized rehabilitation shared decision-making system, and displays it on the display screen.

[0066] Secondly, a personalized rehabilitation shared decision server based on digital humans is proposed, comprising: a data receiving module for receiving multi-source data collected from the patient via a network; a health literacy adaptation module for evaluating the patient's electronic health literacy score (E_score) and generating corresponding grading labels; a digital human status generation module, including a normalization processor, a microprocessor, and an image processor, for calculating the Comprehensive Status Index (CSI) based on multi-source data and grading labels, and generating visual status parameters of the rehabilitation digital human; and a communication module for sending the visual status parameters to the patient, enabling the patient to render and display the rehabilitation digital human image based on the visual status parameters.

[0067] In addition, a personalized rehabilitation sharing decision-making user terminal based on digital humans is proposed, including: a data acquisition and monitoring module for collecting multi-source data from patients and uploading it to a cloud server; an adaptive interactive interface module for selecting the corresponding interactive interface according to the graded labels issued by the cloud; and a display and interaction module for receiving the visual state parameters of the rehabilitation digital human issued by the cloud, rendering the image of the rehabilitation digital human on the display screen, and interacting with the patient through voice and text.

[0068] The following two general examples illustrate the application of this invention: Demonstration I: Application in Remote Pulmonary Rehabilitation for Patients with Chronic Obstructive Pulmonary Disease (COPD) First, system deployment and initialization: The patient end uses a mini-program integrated into the WeChat platform, the medical staff end is a browser-based web management backend, and the cloud server is responsible for data storage, logical operations and model management.

[0069] Secondly, eHL assessment and adaptive mode settings: Upon initial login, the system guides patients to complete the eHEALS scale assessment. Assume a patient's raw score, after normalization and boundary truncation, is E_score = 0.55. The system's preset grading threshold T_eHL = 0.5. Since 0.55 ≥ 0.5, this patient is determined to enter the first interactive mode (standard mode). Another patient's assessment score is E_score = 0.35, which is below the threshold, and is determined to enter the second interactive mode (enhanced auxiliary mode).

[0070] Next, data collection and preprocessing are performed: For the COPD pulmonary rehabilitation field, the system selects an indicator set according to the relevant treatment guidelines: X = {Predicted FEV1%, 6-minute walk distance (6MWD), mMRC dyspnea score, CAT quality of life score}. The system collects the above data regularly through Bluetooth medical devices and online scales, and normalizes it according to the clinical reference ranges specified in the GOLD (Global Initiative for Chronic Obstructive Lung Disease) guidelines. For example, FEV1% ≥ 80% is mapped to 1, < 30% to 0; 6MWD ≥ 450 meters is mapped to 1, 6MWD < 150 meters to 0; mMRC level 0 is mapped to 1, level 4 to 0; CAT score 0 is mapped to 1, 40 to 0. The key safety indicator x'k is set as blood oxygen saturation (SpO2), and its safety threshold S_k is 90%.

[0071] To further illustrate the working process of the safety constraint correction unit, a numerical example is given below: Assume a patient's current Comprehensive Status Index (CSI) is 0.60 (on the boundary of the "Status Warning" interval), and the measured blood oxygen saturation is 88% (normalized value x'_k = 0.88, since the lower limit of normal 92% is mapped to 1, 88% corresponds to 0.88). The safety threshold S_k is set to 0.90 (i.e., 90%). Then the penalty correction factor is min(1, 0.88 / 0.90) = 0.978, and the corrected CSI' = 0.60 × 0.978 = 0.5868. This value is still within the "Status Warning" interval, but the digital human's image will be slightly biased towards the risk direction (e.g., facial rosiness decreases by 5%). If blood oxygen saturation further drops to 82% (x'_k=0.82), the correction factor = 0.82 / 0.90≈0.911, CSI' = 0.60 × 0.911 = 0.5466. The digital avatar will shift from "warning" to the edge of "risk," triggering a voice alert: "Your blood oxygen is low today. Please pause your activity and contact a doctor." This correction mechanism ensures timely and intuitive visual feedback for dangerous situations such as low blood oxygen.

[0072] Regarding personalized settings for safety thresholds: For COPD patients with heart failure, clinical guidelines recommend maintaining SpO2 above 92%, therefore S_k can be preset to 0.92 in the system; while for ordinary stable COPD patients, S_k can be set to 0.90. The system allows healthcare professionals to modify the S_k value for each patient through the healthcare professional interface and records the modification log.

[0073] Next, adaptive CSI calculation and digital human generation are performed: the preset clinical weight W_clinical={0.25,0.25,0.25,0.25}. In the initial shared decision-making meeting, both patients expressed primary concern about the symptom of "dyspnea" using the value clarification tool; therefore, the preference weight W_preference_mMRC for the mMRC index is increased in the system. 1) For patients in the first interaction mode (E_score=0.55), the system calculates that α is relatively large and β is relatively small (e.g., β>2α). The CSI calculation results will appropriately balance objective clinical indicators with patient subjective preferences; 2) For patients in the second interaction mode (E_score=0.35), the system calculated that β was significantly greater than α (e.g., β>2α). The CSI calculation results will be strongly biased towards the mMRC score, which reflects the patient's subjective feeling of dyspnea, making the digital human image extremely sensitive to changes in respiratory status.

[0074] Suppose that at a certain moment, other indicators of the patient in the second interaction mode are stable, but the mMRC score fluctuates slightly due to weather changes, and their CSI value will decrease significantly accordingly. Simultaneously, if the patient's uploaded SpO2 value is 88%, below the safety threshold of 90%, the system automatically triggers safety constraint correction, further lowering the CSI value. The digital avatar will quickly and smoothly transition from "good condition" to "status warning" and even "status risk," proactively triggering voice reminders: "Your blood oxygen is a little low today, and your breathing is a little rapid. Shall we pause for a moment and rest?" This immediate and highly sensitive feedback, guided by the patient's core pain points and combined with safety constraints, is crucial for maintaining the participation and sense of security of patients with low health literacy.

[0075] Finally, the adaptive interactive and shared decision-making process is completed: 1) Patient in the first interactive mode: The mini-program interface displays a CSI trend chart, a detailed list of values ​​for each indicator, and a slider control for finely adjusting training intensity and frequency. The state evolution prediction engine displays the predicted CSI trend for the next 4 weeks in the form of a line graph. 2) Patient in the second interactive mode: The interface is mainly occupied by an enlarged digital human figure. The system announces the daily / weekly core conclusions in a voice synthesis manner (such as "Old Wang, you have been walking very well this week, and I feel much better breathing"). The state evolution prediction is presented as two digital human comparison charts ("Maintain the status quo" vs. "Increase training" after 4 weeks). When the patient confirms that the rehabilitation plan needs to be adjusted, the interface only presents two large graphic buttons: "A. Maintain the current pace and proceed steadily" and "B. Try walking a little longer", and the recommended safe option A is highlighted by default to simplify the decision-making process. After the meeting, the system pops up a simple confirmation interface "Are you sure you want to follow plan A next week?", recording the patient's choice and confirmation.

[0076] Demonstration II: Application in remote rehabilitation of patients after knee replacement surgery.

[0077] First, the system collects indicators. For the rehabilitation scenario after knee replacement surgery, the indicator set X is configured as follows: 1) Active range of motion (ROM) of the knee joint: the optimal value is 120° (mapped to 1), and the worst value is 0° (mapped to 0); 2) Visual analog scale (VAS) score for pain: the optimal value is 0 (no pain) (mapped to 1), and the worst value is 10 (severe pain) (mapped to 0); 3) Manual muscle strength grading (MMT) of the quadriceps femoris: the optimal value is 5 (normal muscle strength) (mapped to 1), and the worst value is 0 (no contraction) (mapped to 0); 4) Modified stand-up-walk time test (TUG): the optimal value is ≤10 seconds (mapped to 1), and the worst value is ≥30 seconds (mapped to 0). The key safety indicator is the pain VAS score, and the safety threshold S_k is 7 points (i.e., when VAS > 7 points, safety constraint correction is triggered).

[0078] Secondly, digital human status mapping is performed, and the digital human status can be represented as follows: 1) Good status (CSI>0.7): the digital human's knee joints move freely, the gait is stable, and the expression is relaxed; 2) Status warning (0.4≤CSI≤0.7): the digital human's knee joints are slightly bent, the digital human walks slowly with crutches, and the expression is slightly frowning; 3) Status risk (CSI<0.4): the digital human's legs are stiff, the digital human needs assistance to stand, and the expression is painful.

[0079] The system then adjusts according to adaptive logic. Based on the patient's electronic health literacy score, the system adaptively adjusts the complexity of information presentation and the depth of decision-making engagement. For example, for patients with low eHL, the system prioritizes the patient's fear of pain during shared decision-making sessions, assigns higher preference weights to VAS indicators, makes the digital avatar more sensitive to changes in pain, and simplifies rehabilitation options to a binary choice between "more rest" and "moderate activity," supplemented by a safe default recommendation.

[0080] The above description is merely a preferred embodiment of the present invention and does not limit the scope of patent protection of the present invention. Any equivalent structural or procedural modifications made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

[0081] In summary, this invention transforms abstract rehabilitation data into a dynamic digital human image through the Comprehensive Status Index (CSI), significantly lowering the cognitive threshold and enabling patients to intuitively observe changes in rehabilitation training and actively participate in decision-making. Furthermore, by employing a weighted algorithm and a differentiated interaction mechanism, it perceives and adapts to the patient's cognitive abilities, achieving dual personalization in rehabilitation training and interactive experience. This bridges the interaction problems for the elderly and those with low health literacy scores, ensuring the technology's accessibility and fairness. In addition, the digital human integrates companionship and motivation, forming a closed-loop feedback loop that significantly improves long-term rehabilitation adherence. The system is highly versatile and can be easily adapted to various rehabilitation training fields, demonstrating significant advancements.

[0082] The above embodiments are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. A personalized rehabilitation shared decision-making system based on digital humans, characterized in that, It includes a data acquisition and monitoring module, a digital human status generation module, a shared decision support module, an interaction and incentive module, a health literacy adaptation module, and an adaptive interactive interface module; The data acquisition and monitoring module is used to collect multi-source data from patients. The health literacy adaptation module is used to assess patients' electronic health literacy scores (E_score) and generate corresponding grade labels. The shared decision support module provides a visual comprehensive assessment report based on a rehabilitation digital human, explanations of rehabilitation option trade-offs, goal setting tools, and program simulation functions. The interaction and motivation module is used for voice and text interaction between the rehabilitation digital human and the patient, as well as to provide action demonstrations, schedule reminders and instant motivational feedback; The adaptive interactive interface module is used to select the corresponding interactive interface based on the hierarchical labels. The digital human status generation module includes: a normalization processor, used to map the multi-source data to the [0,1] interval to obtain a normalized index set; a microprocessor, which integrates a weight calculation engine and a weighted summation unit, wherein the weight calculation engine is used to calculate the Comprehensive Status Index (CSI) based on the weight coefficients corresponding to the grading labels output by the health literacy adaptation module, and the weighted summation unit is used to perform weighted summation operations; and an image processor, used to convert the CSI into visual status parameters of the rehabilitation digital human according to a preset mapping rule and output them to the display interface; wherein the microprocessor is communicatively connected to the normalization processor and the image processor.

2. The personalized rehabilitation shared decision-making system based on digital humans according to claim 1, characterized in that, The weighted summation unit first dynamically determines the clinical weight adoption factor α and the preference weight adoption factor β based on the electronic health literacy score E_score, where α is positively correlated with E_score and β is negatively correlated with E_score; then, based on α, β, and the preset clinical weight and patient preference weight, it calculates the comprehensive weight Wi for each indicator, where i refers to the i-th comprehensive weight. Finally, the CSI is calculated based on the weighted sum of the comprehensive weight Wi and the normalized index. The weighted summation unit dynamically adjusts the contribution ratio of clinical weight and patient preference weight in the comprehensive weight calculation based on the grading labels output by the health literacy adaptation module. When the electronic health literacy score E_score is higher, the contribution ratio of the clinical weight is greater; when the E_score is lower, the contribution ratio of the patient preference weight is greater.

3. The personalized rehabilitation shared decision-making system based on digital humans according to claim 2, characterized in that, The formula for calculating the comprehensive weight Wi is: ,in, ; α base , β base All are preset constants, and γ is a preset amplification factor.

4. The personalized rehabilitation shared decision-making system based on digital humans according to claim 1, characterized in that, The digital human state generation module also includes a safety constraint correction unit: when any preset key safety indicator normalized value x'_k is lower than the corresponding safety threshold S_k, the comprehensive state index CSI is penalized and the corrected CSI′=CSI×min(1,x'_k / S_k) is obtained; CSI′ is used to replace CSI and converted into visual state parameters of the rehabilitation digital human; the safety threshold S_k is dynamically set according to the patient's age, underlying diseases and clinical guidelines, or preset by medical staff.

5. A personalized rehabilitation sharing decision-making system based on digital humans according to claim 1, characterized in that, The shared decision-making assistance module is provided with a state evolution prediction engine, a rehabilitation option value clarification tool, and a decision consistency evaluation unit; the state evolution prediction engine is used to predict the predicted value of the CSI within a future period based on the current value and historical data of the comprehensive state index CSI, and generate a corresponding preview image of the rehabilitated digital human prediction image; the rehabilitation option value clarification tool is used to guide the patient to quantify the preference weights for different rehabilitation goals in the form of an interactive questionnaire or graphical cards. The decision consistency evaluation unit is used to record the decision-making process of the patient and the medical staff and generate a traceable decision-making file.

6. A personalized rehabilitation sharing decision-making system based on digital humans according to claim 1, characterized in that, The adaptive interaction interface module switches between the first interaction interface and the second interaction interface according to the classification label; wherein, the second interaction interface uses the image change of the rehabilitated digital human as the core information carrier and is supplemented by voice broadcast for explanation; the first interaction interface provides a text list, a data chart, and a parameter adjustment control.

7. A personalized rehabilitation sharing decision-making system based on digital humans according to claim 1, characterized in that, When the health literacy adaptation module generates the classification label, it compares the electronic health literacy score E_score with the preset classification threshold T_eHL; when E_score≥T_eHL, it is determined as the first interaction mode; when E_score<T_eHL, it is determined as the second interaction mode. When the adaptive interaction interface module selects the interaction interface, if it is in the first interaction mode, it displays the interface for each core information with a text list, a data chart, and a trend curve, and provides a parameter adjustment control. If it is in the second interaction mode, it is based on the dimensionality reduction transformation of each core information into the image change of the rehabilitated digital human, supplemented by voice broadcast for corresponding explanation, and at the same time adopts a binary comparison selection interface or the default simplified interaction interface.

8. A personalized rehabilitation shared decision-making method, operating using the personalized rehabilitation shared decision-making system as described in any one of claims 1 to 7, characterized in that, It includes an initial stage and a periodic operation stage. Among them, the initial stage includes: S1: Evaluate the patient's electronic health literacy score E_score through the health literacy adaptation module to determine the interaction mode. S2: Based on the interaction mode, set the functional relationship between the clinical weight adoption factor α and the preference weight adoption factor β; α is positively correlated with E_score, and β is negatively correlated with E_score; ; α base , β base All are preset constants, and γ is a preset amplification factor; The periodic operation stage is used to repeat the following steps in each rehabilitation training cycle: S3: Collect multi-source data of the patient through the data collection and monitoring module and obtain an index set through normalization processing. S4: Based on the functional relationship determined in step S2, calculate the comprehensive weight of each index, and then calculate the comprehensive state index CSI. S5: Map the CSI to the visual image of the rehabilitated digital human according to the preset mapping rule and perform continuous gradient through the smooth interpolation algorithm; the preset mapping rule includes at least three intervals: when CSI>0.7, the rehabilitated digital human presents a good image with a rosy complexion and a pleasant expression; when 0.4≤CSI≤0.7, the rehabilitated digital human presents a warning image with a pale complexion and a slight frown; when CSI<0.4, the rehabilitated digital human presents a risk image with cyanosis of the lips and a painful expression. S6: According to the interaction mode determined in step S1, execute a differentiated information presentation and interaction strategy through the adaptive interaction interface module. S7: Through the shared decision-making assistance module and the interaction and incentive module, execute the interaction between the rehabilitated digital human and the patient and record the visual comprehensive evaluation report.

9. A computer-readable storage medium having a computer program stored thereon, said computer program, when executed by a processor, implementing the personalized rehabilitation shared decision-making method as described in any one of claims 8 to 9.

10. A computer-readable storage medium according to claim 9, wherein a computer program is stored in a memory for supporting a rehabilitation digital human terminal device, the rehabilitation digital human terminal device comprising: The display screen, memory, processor, and computer program stored in the memory and executable on the processor are characterized in that, when the processor executes the computer program, it generates and outputs a rehabilitation digital human image in the personalized rehabilitation sharing decision-making system based on digital humans as described in claim 1, and presents it on the display screen.

Images