Intelligent evaluation and monitoring method for muscle mass of senile sarcopenia
By introducing a physical information neural network model of metabolic energy equivalent, the problem of physiological factors interfering with BIA in the assessment of sarcopenia in the elderly was solved, achieving stable assessment and long-term monitoring under non-ideal conditions, thus improving the accuracy of the assessment and the user experience.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- AFFILIATED HOSPITAL OF SHAOXING UNIV OF ARTS & SCI
- Filing Date
- 2025-12-29
- Publication Date
- 2026-04-21
AI Technical Summary
Existing bioelectrical impedance analysis (BIA) is easily affected by short-term physiological factors such as hydration, food intake, and exercise when assessing sarcopenia in the elderly, resulting in unstable measurement results. It is difficult to accurately assess and monitor muscle mass in the elderly under non-ideal conditions, and the reliability of long-term monitoring is low.
A physical information neural network model with metabolic energy equivalent as a latent variable is adopted. This model is combined with multi-frequency bioelectrical impedance data and user state information. The model is trained through physical constraints and a stability loss function to generate muscle mass assessment results. This isolates noise caused by extracellular fluid changes and ensures the stability and reliability of the assessment results.
It achieves stable and accurate assessment of muscle mass in older adults under non-ideal conditions, improves user experience compliance and measurement feasibility, enhances the reliability of long-term monitoring, can automatically correct for interference and generate reliable reports on muscle mass change trends, and supports personalized muscle health management.
Smart Images
Figure CN121890977A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent monitoring technology, and in particular to a method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia. Background Technology
[0002] Sarcopenia is an age-related, progressive, systemic muscle disease characterized by decreased muscle mass, strength, and function. It is a significant risk factor for falls, fractures, disability, and even death in older adults, severely impacting their quality of life and imposing a substantial social and medical burden. Therefore, accurate and convenient assessment and long-term monitoring of sarcopenia in the elderly are crucial for early intervention and effective management.
[0003] Currently, methods for assessing muscle mass mainly include imaging examinations, clinical measurements, and bioelectrical impedance analysis (BIA). Dual-energy X-ray absorptiometry (DXA) and computed tomography (CT) are considered the "gold standard" for muscle mass assessment, offering high accuracy. However, these devices are expensive, bulky, require specialized operation, and emit radiation (DXA uses low doses), making them difficult to widely implement for routine screening and frequent monitoring in communities, homes, or primary healthcare facilities. Bioelectrical impedance analysis (BIA), due to its ease of operation, low cost, non-invasiveness, and safety, has become the most widely used tool for assessing body composition, including muscle mass, in various scenarios. Its basic principle involves applying a weak alternating current to the body, measuring the body's impedance to the current (including resistance and reactance), and using a regression model to estimate parameters such as body water and lean mass. Traditional BIA devices, even multi-frequency BIA, typically rely on the ideal assumption that the body is in a stable hydration state.
[0004] However, BIA technology has a long-standing and unresolved inherent flaw in the practical assessment and monitoring of sarcopenia in the elderly: its measurement results are highly susceptible to significant interference from short-term physiological factors such as body hydration status, food intake, exercise, and diurnal cycle fluctuations. Specifically: 1) Water fluctuation interference: When elderly individuals drink water, take medications (such as diuretics), or experience mild edema / dehydration, their extracellular fluid undergoes drastic changes, which significantly alters electrical impedance values, leading to a substantial error in the estimated muscle mass, far exceeding the actual physiological changes. A single instance of water intake may artificially inflate the measurement results, while mild dehydration may underestimate muscle mass. 2) Stringent measurement conditions: To avoid the above interferences, traditional BIA protocols require subjects to be in a standard state of fasting, bladder emptying, and no strenuous exercise before measurement. However, for the elderly population with poor compliance and variable lifestyles, strictly adhering to this protocol is extremely difficult, resulting in inconsistent measurement conditions and a significant reduction in the reliability and comparability of the data. 3) Low reliability of long-term monitoring: Since the accuracy of a single measurement is already questionable, the reliability of conclusions drawn from long-term trend analysis based on a series of highly fluctuating BIA data obtained under different body fluid conditions is naturally reduced. Clinicians or researchers find it difficult to determine whether changes in the measured values represent actual muscle growth or loss, or merely "noise" caused by water fluctuations.
[0005] While existing technologies have attempted to alleviate this problem by fixing measurement times, strictly controlling pre-measurement behavior, or using statistical averaging methods, none have fundamentally overcome it at the measurement principle and mathematical model level. Therefore, developing a BIA (Biological Analysis) technology that can effectively resist short-term physiological fluctuations and stably and accurately assess and monitor muscle mass in the elderly under non-ideal conditions has become a critical technological bottleneck that urgently needs to be overcome. To address this, a method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia is proposed. Summary of the Invention
[0006] The main objective of this invention is to provide an intelligent assessment and monitoring method for muscle mass in elderly patients with sarcopenia. By introducing advanced mathematical models and neural network models, the robustness and reliability of BIA in practical applications are fundamentally improved, which can effectively solve the problems in the background technology.
[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia, including the following steps: Acquire multi-frequency bioelectrical impedance data and user status information from users; The multi-frequency bioelectrical impedance data and user status information are input into a pre-trained muscle mass assessment model; The muscle mass assessment model calculates an assessment result characterizing the user's muscle mass based on the multi-frequency bioelectrical impedance data and user status information; wherein, the muscle mass assessment model is a physical information neural network model that introduces metabolic energy equivalent as a latent variable, the metabolic energy equivalent is used to characterize the energy metabolism activity of the user's body tissues, and is related to fat removal quality and is not sensitive to short-term body fluid fluctuations. Based on the assessment results, a report on sarcopenia risk level or muscle mass change trend is generated and output.
[0008] A smart assessment and monitoring system for muscle mass in elderly patients with sarcopenia, comprising: The data acquisition module is configured to acquire the user's multi-frequency bioelectrical impedance data and user status information; The data processing and analysis module, which integrates the pre-trained muscle mass assessment model, is configured to calculate muscle mass assessment results based on the input data. The results generation and output module is configured to generate a report on sarcopenia risk level or muscle mass change trend based on the assessment results, and output the report to the user or medical staff.
[0009] The data acquisition module includes a multi-frequency bioelectrical impedance analysis device and a mobile terminal application that communicates with the device. The user status information is input through the human-computer interaction interface of the mobile terminal application.
[0010] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, enables a method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia.
[0011] Furthermore, the muscle mass assessment model is trained using a physical constraint loss function, which includes at least: The mass conservation constraint is set to constrain the predicted total body water to be equal to the predicted sum of extracellular fluid and intracellular fluid. An energy-mass correlation constraint is set to constrain the metabolic energy equivalent to be linearly correlated with the predicted intracellular fluid. Impedance-fluid relationship constraints are set to constrain the predicted electrical properties of extracellular and intracellular fluids that can be derived from the biophysical model in accordance with the input multi-frequency bioelectrical impedance data.
[0012] Furthermore, the muscle mass assessment model is trained using a stability loss function, which constrains the predicted value of the metabolic energy equivalent to remain stable in consecutive measurements taken by the same user over short time intervals.
[0013] Furthermore, the muscle mass assessment model is trained using a fidelity loss function, which measures the difference between the muscle mass predicted by the model and the "gold standard" measurement.
[0014] Furthermore, the muscle mass assessment model is a dual-channel network structure, including: A structural encoder is configured to extract body structural feature vectors based on the multi-frequency bioelectrical impedance data; A state encoder is configured to extract a current physiological state feature vector based on the user state information. A physical constraint fusion decoder is configured to fuse the body structure feature vector with the current physiological state feature vector, and output the metabolic energy equivalent, extracellular fluid, and intracellular fluid under the constraint of the physical constraint loss function. The output of the structural encoder is correlated with parameters obtained by fitting the multi-frequency bioelectrical impedance data using a Cole-Cole model.
[0015] Furthermore, the total loss of the muscle mass assessment model includes fidelity loss, physical constraint loss, and stability loss. The fidelity loss is used to measure the difference between the muscle mass predicted by the model and the "gold standard" measurement value. The physical constraint loss is used to constrain the model to comply with basic biophysical laws. The stability loss is used to constrain the stability of the metabolic energy equivalent measured continuously by an individual over a short period of time. The total loss function of the muscle mass assessment model is expressed as: = + + ; in This is the total loss function; The fidelity loss function is calculated as follows: = , Degreasing quality data measured using gold standard measuring equipment. This is a predicted value for metabolic energy equivalent. Learnable parameters are used to transform unitless latent variables. Degreasing quality data mapped to actual mass units It is obtained through model training; The physical constraint loss function is obtained by weighting the mass conservation constraint, energy-mass correlation constraint, and impedance-fluid relationship constraint. The calculation method is as follows: = + + , The mass conservation constraint term is calculated as follows: = , The total body water content can be roughly estimated using empirical formulas and used as a reference. The model predicts values for extracellular fluid. The model predicts the value of intracellular fluid; The energy-mass correlation constraint term is calculated as follows: = , This is the model prediction value for metabolic energy equivalent. , All are learnable parameters; The impedance-fluid relationship constraint term is calculated based on the Cole-Cole model. The calculation method is as follows: = , The multi-frequency resistor is obtained from actual measurements by the user. The reactance is the actual reactance measured by the user. , For predictions based on the model and The user's theoretical multi-frequency resistance and reactance are derived by reverse calculation; , , All are weighted coefficients within the interval (0,1), and + + =1; The stability loss function is obtained by calculating the variance of the metabolic energy equivalent predicted by the model over a short period of time. = ,in This indicates the calculation of metabolic energy equivalent. variance =1,2,..., ; , These are all hyperparameters of the model, which are either manually set before training begins or determined through a hyperparameter optimization algorithm.
[0016] Furthermore, the user status information includes at least one of the following: the time since the last meal, the time and intensity since the last exercise, the daily cycle time, and the subjective thirst score.
[0017] Furthermore, after acquiring the user's multi-frequency bioelectrical impedance data, the method further includes: Based on the multi-frequency bioelectrical impedance data, the phase angle and the ratio of extracellular fluid to body fluid were calculated; The phase angle and the ratio are used as additional features and are input into the muscle mass assessment model along with the user state information.
[0018] The present invention has the following beneficial effects: Compared with existing technologies, this scheme maps the raw BIA signal, which is susceptible to moisture, to metabolic energy equivalent data that is insensitive to short-term fluctuations. This fundamentally isolates the noise caused by changes in extracellular fluid, ensuring that the muscle mass assessment results remain stable and reliable after drinking water, eating, and exercising. Furthermore, the embedded physical constraints such as mass conservation and impedance-fluid relationship ensure that the output results not only fit the data mathematically, but also enhance the rationality and interpretability of the scheme in terms of biophysics, avoiding the gross errors that may be generated by traditional pure data-driven models.
[0019] Compared with existing technologies, this solution breaks through the strict limitation that traditional BIA must be measured under the standard conditions of "fasting and resting". The elderly can take measurements in various normal daily situations (such as after meals or after walking). The system can automatically correct for interference, which greatly improves user experience compliance and measurement feasibility.
[0020] Compared to existing technologies, this solution ensures the reliability and sensitivity of long-term monitoring. Due to the enhanced anti-interference capability of single measurements, the signal-to-noise ratio of time-series muscle mass change trend graphs is significantly improved. Medical staff can clearly and reliably identify the true trends of muscle mass growth or loss, and make more timely and sensitive assessments of the effectiveness of interventions, avoiding the dilemma of being unable to draw conclusions due to large data fluctuations.
[0021] Compared to existing technologies, this solution encapsulates complex signal processing, physical modeling, and artificial intelligence inference into automated services. Users only need to complete simple measurements and questionnaires, and the system can automatically generate reports containing risk assessments and trend analyses, lowering the professional barrier for operators.
[0022] Compared with existing technologies, this solution shifts from "single assessment" to "whole-process management," supporting the establishment of personal health records and continuous data tracking. It realizes the transformation from one-time sarcopenia screening to personalized, dynamic, whole-process muscle health management, providing an effective technical tool for preventing disability in the elderly. Attached Figure Description
[0023] Figure 1 This is a flowchart illustrating the intelligent assessment and monitoring method for muscle mass in elderly patients with sarcopenia according to the present invention. Figure 2 This is a schematic diagram of the implementation scheme of the intelligent assessment and monitoring method for muscle mass in elderly patients with sarcopenia according to the present invention; Figure 3This is a schematic diagram of the intelligent assessment and monitoring system for muscle mass in elderly patients with sarcopenia, as described in this invention. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0025] Example 1: See Figure 1 The flowchart shown is a method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia according to the present invention, which includes the following steps: Step 1: Obtain the user's multi-frequency bioelectrical impedance data and user status information; the user status information includes: time since last meal, time and intensity since last exercise, daily cycle time, and subjective thirst score. Step 2: Input multi-frequency bioelectrical impedance data and user status information into a pre-trained muscle mass assessment model; wherein, the muscle mass assessment model is a two-channel network structure, including: A structural encoder is configured to extract body structural feature vectors from multi-frequency bioelectrical impedance data, and its output is correlated with parameters obtained by fitting the multi-frequency bioelectrical impedance data through a Cole-Cole model. A state encoder is configured to extract a current physiological state feature vector based on user state information. The physical constraint fusion decoder is configured to fuse the body structure feature vector with the current physiological state feature vector, and output metabolic energy equivalent, extracellular fluid, and intracellular fluid under the constraint of the physical constraint loss function.
[0026] The muscle mass assessment model is trained using a physical constraint loss function, which includes at least the following: The mass conservation constraint is set to constrain the predicted total body water to be equal to the predicted sum of extracellular fluid and intracellular fluid. Energy-mass correlation constraints are set to constrain the linear correlation between metabolic energy equivalent and predicted intracellular fluid. Impedance-fluid relationship constraints are set to constrain the predicted electrical properties of extracellular and intracellular fluids that can be derived from the biophysical model in accordance with the input multi-frequency bioelectrical impedance data.
[0027] The muscle mass assessment model is trained using a stability loss function, which constrains the predicted value of metabolic energy equivalent to remain stable in consecutive measurements taken by the same user over short time intervals.
[0028] Step 3: The muscle mass assessment model calculates the assessment result representing the user's muscle mass based on multi-frequency bioelectrical impedance data and user status information. The muscle mass assessment model is a physical information neural network model that introduces metabolic energy equivalent as a latent variable. Metabolic energy equivalent is used to represent the energy metabolism activity of body tissues and is related to fat-free mass and is not sensitive to short-term body fluid fluctuations. After obtaining the user's multi-frequency bioelectrical impedance data, the phase angle and the ratio of extracellular fluid to body fluid are calculated based on the multi-frequency bioelectrical impedance data. The phase angle and ratio are used as additional features and are input into the muscle mass assessment model along with the user status information.
[0029] Step 4: Based on the assessment results, generate and output a report on sarcopenia risk level or muscle mass change trend.
[0030] Based on the above technical process, this embodiment provides a specific implementation plan step, see [link to implementation plan]. Figure 2 As shown, the details are as follows: Phase 1: System Initialization and Model Deployment Step S101: Model Building and Pre-training Build the network structure: On a server or in the cloud, build a dual-channel physical information neural network, including a structural encoder, a state encoder, and a physical constraint fusion decoder.
[0031] Define the loss function: Total loss function = + + ; in For the total loss function, This represents the loss of fidelity; it is used to measure the difference between the model's predicted muscle mass and the "gold standard" measurement. = , Degreasing quality data measured using gold standard measuring equipment. This is a predicted value for metabolic energy equivalent. Learnable parameters are used to transform unitless latent variables. Degreasing quality data mapped to actual mass units It is obtained through model training; It includes constraints related to mass conservation, energy-mass correlation, and impedance-fluid relationship, which are obtained by weighting each constraint term, specifically: = + + , This is a mass conservation constraint term used to force the sum of the predicted body fluid components to equal the total body water volume. The calculation method is as follows: = , The total body water content can be initially estimated using a well-established empirical formula based on the BIA (Body Water Index) and used as a reference benchmark. The model predicts values for extracellular fluid. The model predicts the value of intracellular fluid; This is an energy-mass correlation constraint term, used to constrain the linear relationship between metabolic energy equivalent and predicted intracellular fluid. The calculation method is as follows: = , This is the model prediction value for metabolic energy equivalent. , All are learnable parameters; The impedance-fluid relationship constraint term is used to constrain the predicted extracellular and intracellular fluid electrical properties, which can be derived from the biophysical model to be consistent with the input multi-frequency bioelectrical impedance data. The calculation is based on the Cole-Cole model. = , The multi-frequency resistor is obtained from actual measurements by the user. The reactance is the actual reactance measured by the user. , For predictions based on the model and The user's theoretical multi-frequency resistance and reactance are derived by reverse calculation; , , All are weighted coefficients within the interval (0,1), and + + =1; The stability loss is obtained by calculating the variance of the metabolic energy equivalent predicted by the model over a short period of time. = ,in This indicates the calculation of metabolic energy equivalent. variance =1,2,..., ; , These are all hyperparameters of the model, which are either manually set before training begins or determined through a hyperparameter optimization algorithm.
[0032] Step S102: System Integration and Deployment The trained muscle mass assessment model is integrated into the data processing and analysis module.
[0033] Develop user terminals (such as smartphone apps, tablet applications, or dedicated device firmware) that include data acquisition modules and result generation and output modules, and establish communication connections with cloud servers or local processors where the model is deployed.
[0034] Phase Two: User-side Measurement and Data Acquisition Step S201: Guide users in standardized preparation The system provides user-friendly prompts (especially for the elderly) through the app interface, guiding them to make standardized preparations before measurement, such as: try to measure in the early morning, on an empty stomach, and after emptying the bladder; avoid measuring immediately after strenuous exercise or drinking a lot of water.
[0035] Step S202: Collect bioelectrical impedance data Users take measurements using multi-frequency BIA devices that are compatible with the system (such as smart scales with multi-frequency BIA functionality, handheld devices, or foot-operated electrode pads).
[0036] The device automatically collects raw resistance and reactance data at multiple frequencies (such as 5kHz, 50kHz, 100kHz, 200kHz).
[0037] The terminal app receives this raw data from the BIA device via Bluetooth or Wi-Fi.
[0038] Step S203: Collect user status information Before and after the measurement, the app guides users to input or select their current user status information through a questionnaire, which may include the following questions: Question 1: "When was the last time you ate?" (Enter the number of hours) Question 2: "Have you engaged in moderate-intensity or higher exercise in the last 2 hours?" (Yes / No, or select intensity) Question 3: "Are you feeling thirsty right now?" (Rating 1-5) The system automatically records the current time (daily cycle time).
[0039] Phase 3: Data Processing and Intelligent Analysis Step S301: Data Preprocessing and Feature Extraction The terminal packages the received raw BIA data and user status information and sends it to the cloud-based data processing and analysis module.
[0040] In the cloud, the structural encoder processes multi-frequency BIA data to extract body structure feature vectors.
[0041] At the same time, the state encoder processes the user's state information and extracts state feature vectors.
[0042] Step S302: Fusion Reasoning under Physical Constraints Physical constraints are fused to the decoder's receiver.
[0043] The decoder operates internally, and its output layer simultaneously generates estimates of three core variables: metabolic energy equivalent M, extracellular fluid ECW, and intracellular fluid ICW.
[0044] Step S303: Calculate final muscle mass and evaluation indicators The system uses the inferred metabolic energy equivalent M to calculate the user's lean mass and skeletal muscle mass through a simple linear relationship: muscle mass ≈ K × M (K is a known parameter obtained from training). Skeletal muscle mass is used to reflect the user's sarcopenia risk level, while lean mass can be used to reflect the energy metabolism activity of the user's body tissues.
[0045] Furthermore, based on the user's height, the skeletal muscle mass index (SMI) is calculated: SMI = muscle mass (kg) / height (m) 2 .
[0046] Phase 4: Results Generation, Feedback, and Long-Term Monitoring Step S401: Generate assessment report and risk warning The results generation and output module compares the calculated skeletal muscle mass index with pre-stored sarcopenia diagnostic cutoff points (such as the AWGS criteria). It then generates a user report, including: Current muscle mass level and skeletal muscle mass index value.
[0047] Sarcopenia risk level (e.g., normal, low, sarcopenia).
[0048] The trend of change compared to the previous measurement (increasing, decreasing, or remaining stable).
[0049] If a rapid decline in muscle mass is detected or the criteria for sarcopenia are met, the system will send an alert to the user and / or their designated healthcare provider.
[0050] Step S402: Establish personal health records and trend analysis The system stores all raw data from previous measurements, the inferred M-value, the calculated muscle mass, and the user's status in the individual health record.
[0051] Visual charts (such as line graphs) can be used to show users the long-term trends of their muscle mass, M-value, and body water (ECW, ICW). This helps users see their progress intuitively and validate the effectiveness of the model (the trends of M-value and muscle mass should be smoother and more reliable than simple weight or traditional BIA results).
[0052] Example 2: This invention also provides an intelligent assessment and monitoring system for muscle mass in elderly patients with sarcopenia, see [link to relevant documentation]. Figure 3 The system architecture diagram shown includes: The data acquisition module is configured to acquire the user's multi-frequency bioelectrical impedance data and user status information; The data processing and analysis module integrates a pre-trained muscle mass assessment model, which is set up to calculate muscle mass assessment results based on the input data. The results generation and output module is configured to generate reports on sarcopenia risk levels or muscle mass change trends based on assessment results, and output the reports to users or healthcare professionals.
[0053] The data acquisition module includes a multi-frequency bioelectrical impedance analysis device and a mobile terminal application that communicates with the device. User status information is input through the human-computer interaction interface of the mobile terminal application.
[0054] Example 2: The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, enables a method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia.
[0055] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.
Claims
1. A method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia, characterized in that, Includes the following steps: Acquire multi-frequency bioelectrical impedance data and user status information from users; The multi-frequency bioelectrical impedance data and user status information are input into a pre-trained muscle mass assessment model; The muscle mass assessment model calculates an assessment result characterizing the user's muscle mass based on the multi-frequency bioelectrical impedance data and user status information; wherein, the muscle mass assessment model is a physical information neural network model that introduces metabolic energy equivalent as a latent variable, the metabolic energy equivalent is used to characterize the energy metabolism activity of the user's body tissues, and is related to fat removal quality and is not sensitive to short-term body fluid fluctuations. Based on the assessment results, generate and output a report on sarcopenia risk level or muscle mass change trend; The muscle mass assessment model is a two-channel network structure, including: A structural encoder is configured to extract body structural feature vectors based on the multi-frequency bioelectrical impedance data; A state encoder is configured to extract a current physiological state feature vector based on the user state information. The physical constraint fusion decoder is configured to fuse the body structure feature vector with the current physiological state feature vector, and output the metabolic energy equivalent, extracellular fluid, and intracellular fluid under the constraint of the physical constraint loss function.
2. The method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia according to claim 1, characterized in that, The muscle mass assessment model is trained using a physical constraint loss function, which includes at least the following: The mass conservation constraint is set to constrain the predicted total body water to be equal to the predicted sum of extracellular fluid and intracellular fluid. An energy-mass correlation constraint is set to constrain the metabolic energy equivalent to be linearly correlated with the predicted intracellular fluid. Impedance-fluid relationship constraints are set to constrain the predicted electrical properties of extracellular and intracellular fluids that can be derived from the biophysical model in accordance with the input multi-frequency bioelectrical impedance data.
3. The method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia according to claim 2, characterized in that, The muscle mass assessment model is trained using a stability loss function, which constrains the predicted value of the metabolic energy equivalent to remain stable in consecutive measurements taken by the same user over short time intervals.
4. The method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia according to claim 3, characterized in that, The muscle mass assessment model is trained using a fidelity loss function, which measures the difference between the model's predicted muscle mass and the "gold standard" measurement.
5. The method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia according to claim 4, characterized in that, The total loss of the muscle mass assessment model is a weighted average of the fidelity loss, the physical constraint loss, and the stability loss, expressed as: = + + ;in For the total loss function, For the fidelity loss function, The physical constraint loss function, For stability loss function, , These are all hyperparameters of the model.
6. The method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia according to any one of claims 5, characterized in that, The output of the structural encoder is correlated with parameters obtained by fitting the multi-frequency bioelectrical impedance data using a Cole-Cole model.
7. The method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia according to claim 1, characterized in that, The user status information includes at least one of the following: the time since the last meal, the time and intensity of the last exercise, the daily cycle time, and the subjective thirst score.
8. The method for intelligent assessment and monitoring of muscle mass in elderly patients with sarcopenia according to claim 1, characterized in that, After acquiring the user's multi-frequency bioelectrical impedance data, the method further includes: Based on the multi-frequency bioelectrical impedance data, the phase angle and the ratio of extracellular fluid to body fluid were calculated; The phase angle and the ratio are used as additional features and are input into the muscle mass assessment model along with the user state information.
9. A smart assessment and monitoring system for muscle mass in elderly patients with sarcopenia, used to implement the method described in any one of claims 1-8, characterized in that, include: The data acquisition module is configured to acquire the user's multi-frequency bioelectrical impedance data and user status information; The data processing and analysis module, which integrates the pre-trained muscle mass assessment model, is configured to calculate muscle mass assessment results based on the input data. The results generation and output module is configured to generate a report on sarcopenia risk level or muscle mass change trend based on the assessment results, and output the report to the user or medical staff. The data acquisition module includes a multi-frequency bioelectrical impedance analysis device and a mobile terminal application that communicates with the device. The user status information is input through the human-computer interaction interface of the mobile terminal application.
10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent assessment and monitoring method for muscle mass in elderly patients with sarcopenia as described in any one of claims 1-8.