A method and system for identifying muscle weakness in critically ill patients

Abstract

This invention discloses a method and system for identifying muscle weakness in critically ill patients, relating to the field of muscle weakness identification. This invention integrates patients' subjective clinical muscle strength grading with instrumental testing data to establish a multi-site muscle weakness grading mapping model, significantly improving diagnostic accuracy. Through a subjective / objective site mapping model, combining patient subjective feelings with objective testing data, it can identify medullary muscle weakness symptoms that are easily missed by traditional examinations. Based on a muscle weakness index data prediction mechanism, it can provide early warning of the risk of muscle weakness crisis. Through a closed-loop system of treatment-assessment-adjustment, it achieves dynamic optimization of medication regimens, significantly reducing the incidence of crisis, maintaining treatment efficacy while reducing the dosage of immunosuppressants, and is suitable for patients requiring long-term management. By establishing a mapping system between site type sets and index grade sets, it achieves quantitative and unified assessment of different subtypes such as ocular muscle type / generalized type, reducing subjective judgment errors.

Description

Technical Field

[0001] This invention belongs to the field of muscle weakness recognition, and more specifically, it relates to a method and system for recognizing muscle weakness in critically ill patients. Background Technology

[0002] In the process of identifying muscle weakness, traditional muscle strength grading relies on subjective judgment and is difficult to quantify symptoms in special areas such as ptosis and respiratory muscles; existing methods require waiting for obvious symptoms to appear before intervention, and cannot identify the risk of muscle weakness crisis in advance; traditional protocols lack data support for drug selection and cannot achieve real-time optimization of drug type / dosage. Summary of the Invention

[0003] In response to the problems in related technologies, this invention proposes a method and system for identifying muscle weakness in critically ill patients, in order to overcome the aforementioned technical problems existing in the existing related technologies.

[0004] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0005] This invention relates to a method for identifying muscle weakness in critically ill patients, comprising the following steps:

[0006] S1. Set the types and levels of muscle weakness measurement indicators corresponding to several body part types in which patients exhibit muscle weakness.

[0007] S2. Collect historical data corresponding to the muscle weakness measurement index type and index level set in S1, and construct a mapping model of the overall muscle weakness severity level data of each muscle weakness manifestation area based on the historical data.

[0008] S3. Collect the index level data corresponding to each muscle weakness measurement index of the patient to be identified and input them into the corresponding mapping model constructed in S2 for mapping.

[0009] S4. Collect various muscle weakness instrument detection indicators of the patient to be identified, and then input them into multiple mapping models constructed based on the instrument detection indicator data corresponding to the historical data in S2 for mapping.

[0010] S5. Compare the mapping results in S3 and S4 with the corresponding preset thresholds. Based on the comparison results, determine whether to execute S6 or predict the overall muscle weakness level data at future moments. Based on the prediction results, determine whether to execute S6 or not to perform treatment.

[0011] S6. Treat the patient to be identified. Repeat S3, S4 and S5 after treatment. If the comparison results and prediction results in S5 meet the conditions after the repetition, the treatment is completed. Otherwise, adjust the current initial drug treatment plan and repeat S6.

[0012] Preferably, step S1 includes the following steps:

[0013] S11. Establish several methods for identifying muscle weakness in critically ill patients to obtain a set of muscle weakness identification methods; the set of muscle weakness identification methods includes methods for collecting patient subjective feeling data and methods for collecting instrument detection data;

[0014] S12. Based on the method of collecting patients' subjective feelings data, several body part types of patients' muscle weakness are defined to obtain a set of muscle weakness manifestation type types; based on the set of muscle weakness manifestation type types, a corresponding muscle weakness measurement index type is set for each type of muscle weakness manifestation type to obtain a set of muscle weakness measurement index type types; based on the level type of each type in the set of muscle weakness measurement index type types, a set of muscle weakness measurement index level types is obtained.

[0015] The full-cycle symptom capture capability, through quantitative assessment of multiple detailed indicators in key areas such as the eyelids and arms, combined with grading standards, can accurately identify the entire course of the disease, from early-stage ptosis to late-stage respiratory muscle involvement. It can capture symptom change curves through regular, repeated measurements, avoiding missed diagnoses caused by traditional single-examination methods. Multimodal diagnostic synergy enhances diagnostic specificity by cross-validating the patient's subjective description of swallowing difficulties with laboratory antibody test data and instrumental electrophysiological examination results. For example, when a patient complains of difficulty speaking, the presence of both positive acetylcholine receptor antibodies and abnormal repetitive nerve electrical stimulation can quickly pinpoint neuromuscular junction lesions. By establishing a set of muscle weakness measurement index levels and types, it provides a basis for subsequently determining the level of each type and corresponding index, thereby determining the comprehensive muscle weakness level of each affected area, and providing subjective data for subsequently determining the type of muscle weakness the patient suffers from.

[0016] Preferably, step S2 includes the following steps:

[0017] S21. Based on the set of muscle weakness manifestation location types, the set of muscle weakness measurement index types, and the set of muscle weakness measurement index levels, collect the index level data corresponding to each muscle weakness measurement index of multiple historical muscle weakness patients and the overall muscle weakness severity level data of the corresponding muscle weakness manifestation location to obtain historical muscle weakness index level dataset and historical overall muscle weakness level dataset.

[0018] S22. Based on the historical muscle weakness index level dataset and the historical overall muscle weakness level dataset, construct a mapping model corresponding to the overall muscle weakness severity level data of each muscle weakness manifestation site, and obtain the main site muscle weakness severity level mapping model set.

[0019] By establishing a mapping system between site type sets and indicator level sets, a unified quantitative assessment of different subtypes such as ocular muscle type and generalized type is achieved, reducing subjective judgment errors. The model supports the transformation of discrete patient subjective feeling data into a continuous overall muscle weakness severity curve, enabling doctors to track subtle fluctuations in the condition and improving sensitivity to changes in disease course. It can shorten outpatient assessment time. By constructing a set of main site muscle weakness severity level mapping models, a mapping model is provided for subsequent acquisition of overall muscle weakness severity level data for each site of the patient to be identified. Based on the construction of a corresponding model for each site of muscle weakness, the specificity of the model and the accuracy of the mapping are ensured.

[0020] Preferably, the mapping model in S22 adopts a multilayer perceptron model;

[0021] The multilayer perceptron model can capture the complex mapping relationship between muscle weakness indicators and severity through the nonlinear activation function of the hidden layer, making it particularly suitable for processing nonlinear correlations of multimodal data such as clinical scores; the fully connected layer automatically learns the weight allocation of different muscle weakness indicators without the need for manual design of feature combination rules; and the dimensionality reduction structure of neurons effectively filters out noisy features.

[0022] Preferably, step S3 includes the following steps:

[0023] S31. Set the patient to be identified; according to the set of muscle weakness manifestation location type, muscle weakness measurement index type, and muscle weakness measurement index level set, collect the index level data corresponding to each muscle weakness measurement index of the patient to be identified, and obtain the muscle weakness index level dataset to be identified.

[0024] S32. Input the muscle weakness index level data of each part in the dataset of muscle weakness index levels to be identified into the corresponding part muscle weakness severity level mapping model in the main part muscle weakness severity level mapping model set for mapping, and obtain the first overall muscle weakness level dataset.

[0025] The mapping process yields a first overall muscle weakness level dataset, providing a data foundation for subsequent comprehensive judgment based on both subjective and objective data regarding the patient's instrumental testing data and the type of muscle weakness to be identified, thereby improving the accuracy of the judgment. The systematic collection of data across multiple site types and indicator levels transforms traditional subjective clinical observation into quantifiable data dimensions, which are then used in the subsequent muscle weakness identification process through model output, significantly improving the objectivity of the identification. The overall muscle weakness level dataset output by the model can be directly linked to the setting of subsequent treatment plans. The automated assessment process reduces manual interpretation time, improving nursing efficiency and alleviating pressure on specialized medical resources in scenarios such as the identification of ocular muscle involvement symptoms.

[0026] Preferably, step S4 includes the following steps:

[0027] S41. Based on the set of muscle weakness manifestation site types, set the corresponding muscle weakness instrument detection index type for each muscle weakness manifestation site to obtain a muscle weakness instrument detection index type set; based on the set of muscle weakness instrument detection index types, collect the various muscle weakness instrument detection index data of multiple historical muscle weakness patients mentioned in S2 to obtain a historical muscle weakness instrument detection index dataset; then, based on the historical muscle weakness instrument detection index dataset and the historical overall muscle weakness level dataset, construct a mapping model between the overall muscle weakness severity level data and the corresponding muscle weakness instrument detection index data for each muscle weakness manifestation site to obtain a set of external site muscle weakness severity level mapping models.

[0028] S42. Based on the set of muscle weakness instrument detection index types, collect the data of each muscle weakness instrument detection index of the patient to be identified to obtain the dataset of muscle weakness instrument detection index to be identified; input the data of muscle weakness instrument detection index of each part in the dataset of muscle weakness instrument detection index to be identified into the corresponding mapping model in the set of external site muscle weakness severity level mapping model for mapping to obtain the second overall muscle weakness level dataset.

[0029] By integrating multi-dimensional detection data such as eye-tracking, isokinetic muscle strength testing, and surface electromyography, a comprehensive assessment framework covering neuromuscular function, biomechanical properties, and fatigue characteristics was established. This significantly improved the quantitative analysis capability of the external site muscle weakness severity grading model for muscle weakness symptoms. Based on the functional characteristics of different anatomical sites, the generated grading model can automatically output standardized assessment results, providing an objective basis for subsequent rehabilitation program development.

[0030] Preferably, the severity level mapping model for muscle weakness in the target area described in S41 uses the following models for the eyelid, arm / thigh, hand, and throat areas respectively: a bidirectional LSTM + attention mechanism model, a gradient boosting decision tree, a 1D-CNN + GRU hybrid network, and a dual-branch feature extractor model.

[0031] LSTM excels at handling non-stationary temporal features of eyelid movements, and its attention mechanism can locate key frames of symptom onset; gradient boosting decision trees provide strong interpretation of the mechanical indices generated by the Biodex isokinetic test; and a differentiated machine learning architecture enhances the model's sensitivity to local symptoms.

[0032] Preferably, step S5 includes the following steps:

[0033] S51. Compare the first overall muscle weakness level dataset and the second overall muscle weakness level dataset, and select the corresponding higher overall muscle weakness level data as the final overall muscle weakness level data to obtain the final overall muscle weakness level dataset.

[0034] S52. Based on the set of muscle weakness manifestation site types, set an overall muscle weakness level threshold corresponding to each site type to obtain an overall muscle weakness level threshold set; when there is an overall muscle weakness level data in the final overall muscle weakness level dataset that is greater than or equal to the corresponding overall muscle weakness level threshold, proceed to S61; otherwise, proceed to S53.

[0035] S53. Based on the set of muscle weakness measurement index types, the set of muscle weakness measurement index levels, and the set of muscle weakness device detection index types, collect multiple sets of muscle weakness index level data and muscle weakness device detection index data of various parts of the patients to be identified in real time. Then, based on the data collected in real time, predict the muscle weakness index level data and muscle weakness device detection index data of various parts at multiple future time points to obtain the future muscle weakness index level dataset and the future muscle weakness device detection index dataset.

[0036] Each data point in the future muscle weakness index level dataset and the future muscle weakness instrument detection index dataset is input into the corresponding mapping models in the main site muscle weakness severity level mapping model set and the guest site muscle weakness severity level mapping model set for mapping. The data with the higher future overall muscle weakness level is selected as the final overall muscle weakness level data to obtain the final future overall muscle weakness level dataset.

[0037] If there is a data point in the final future overall muscle weakness level dataset that is greater than or equal to the corresponding overall muscle weakness level threshold, proceed to S61; otherwise, no treatment is required.

[0038] By employing a dual-track mechanism combining real-time data collection and prediction, early warning of the risk of muscle weakness progression is achieved, enabling earlier identification of disease deterioration trends compared to traditional static assessment methods. The "highest priority principle" is adopted to integrate multi-source assessment data, avoiding the underestimation of risk that may be caused by a single model and improving the accuracy of determining the timing of treatment intervention. Differentiated grade thresholds are set for different sites, which is more in line with the heterogeneous characteristics of muscle groups affected by severe myasthenia gravis. The integration of current status assessment and future trend prediction forms a closed-loop management system that links "current status-prognosis," providing quantitative evidence for the dynamic monitoring of the efficacy of biological agents.

[0039] Preferably, step S6 includes the following steps:

[0040] S61. Based on the set of muscle weakness manifestation location types, set corresponding drug treatment plans for each location of muscle weakness to obtain an initial set of drug treatment plans for muscle weakness in the initial location; the drug treatment plan includes the type of drug, dosage, and administration time; based on the location type of the overall muscle weakness level data that is greater than or equal to the corresponding overall muscle weakness level threshold in the final overall muscle weakness level dataset or the final future overall muscle weakness level dataset, select the corresponding drug treatment plan for muscle weakness in the initial location to obtain the current initial drug treatment plan;

[0041] S62. Treat the patient to be identified according to the current drug treatment plan; after treatment, set a repetition threshold and repeat S31, S32, S42, S51, and S52; when the repetition count is less than or equal to the repetition threshold and neither the final overall muscle weakness level dataset nor the final future overall muscle weakness level dataset contains any overall muscle weakness level data greater than or equal to the corresponding overall muscle weakness level threshold, the treatment is complete; otherwise, proceed to S63.

[0042] S63. Adjust the type, dosage, and administration time of the drug in the current initial drug treatment plan, and repeat S62 until the number of repetitions is less than or equal to the repetition threshold and neither the final overall muscle weakness level dataset nor the final future overall muscle weakness level dataset contains any overall muscle weakness level data greater than or equal to the corresponding overall muscle weakness level threshold; treatment is complete.

[0043] By employing site-specific drug regimens, a shift from "broad-spectrum treatment" to "precision treatment" is achieved, improving drug efficacy. A feedback regulation system based on muscle weakness level thresholds can automatically identify the optimal timing and dosage for drug administration, reducing the incidence of side effects such as abdominal pain and diarrhea. Through repeated threshold setting and future data prediction, cases of hormone resistance or ineffectiveness of immunosuppressants can be identified before traditional clinical assessments, providing a decision window for timely switching to biologics.

[0044] A muscle weakness identification system for critically ill patients includes a module for setting the level of patient muscle weakness measurement indicators, a module for collecting historical muscle weakness data of different sites, a module for constructing a mapping model for the severity level of muscle weakness of different sites, a module for mapping the level of muscle weakness of the site to be identified, a module for mapping the detection indicators of the device to be identified, a module for determining the current muscle weakness identification and treatment, and a module for identifying the current muscle weakness treatment feedback.

[0045] The present invention has the following beneficial effects:

[0046] 1. This invention integrates patients' subjective clinical muscle strength grading with instrument testing data to establish a multi-site muscle weakness level mapping model, greatly improving diagnostic accuracy. Through the subjective / objective site mapping model, combining patients' subjective feelings with objective testing data, it can identify medullary muscle weakness symptoms that are easily missed by traditional examinations. Based on a muscle weakness index data prediction mechanism, it can provide early warning of the risk of muscle weakness crisis. Through a closed-loop system of treatment-assessment-adjustment, it achieves dynamic optimization of medication regimens, significantly reducing the incidence of crisis, maintaining treatment efficacy while reducing the dosage of immunosuppressants, and is suitable for patients requiring long-term management.

[0047] 2. This invention establishes a mapping system between a set of site types and a set of indicator levels, enabling a unified quantitative assessment of different subtypes such as ocular muscle type and generalized type, thus reducing subjective judgment errors. The model supports the transformation of discrete patient subjective perception data into a continuous overall muscle weakness severity curve, allowing doctors to track minor fluctuations in the condition and improving sensitivity to changes in disease progression. It can also shorten outpatient assessment time. By constructing a set of mapping models for the severity levels of muscle weakness in the main site, a mapping model is provided for subsequently acquiring data on the overall severity levels of muscle weakness in various sites of the patient to be identified. Constructing a corresponding model based on each site where muscle weakness occurs ensures the model's specificity and mapping accuracy.

[0048] 3. This invention achieves a shift from "broad-spectrum treatment" to "precision treatment" through site-specific drug regimens, thereby improving drug efficacy. The feedback regulation system based on muscle weakness level thresholds can automatically identify the optimal timing and dosage of drug administration, reducing the incidence of side effects such as abdominal pain and diarrhea. By repeatedly setting thresholds and predicting future data, cases of hormone resistance or ineffectiveness of immunosuppressants can be identified before traditional clinical assessments, providing a decision window for timely switching to biologics.

[0049] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating a method for identifying muscle weakness in critically ill patients according to the present invention.

[0052] Figure 2 This is a schematic diagram of a module of a muscle weakness recognition system for critically ill patients according to the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the invention, and not all embodiments. Based on the embodiments of the invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the invention.

[0054] Example 1

[0055] Please see Figure 1 This embodiment is a method for identifying muscle weakness in critically ill patients, including the following steps:

[0056] S1. Set the types and levels of muscle weakness measurement indicators corresponding to several body part types in which patients exhibit muscle weakness.

[0057] S1 includes the following steps:

[0058] S11. Establish several methods for identifying muscle weakness in critically ill patients to obtain a set of muscle weakness identification methods; the set of muscle weakness identification methods includes methods for collecting patient subjective feeling data and methods for collecting instrument detection data;

[0059] S12. Based on the method of collecting patients' subjective feelings data, several body part types of patients' muscle weakness are defined to obtain a set of muscle weakness manifestation type types; the set of muscle weakness manifestation type types includes eyelids, arms, hands, throat, and thighs, etc.; according to the set of muscle weakness manifestation type types, a corresponding muscle weakness measurement index type is set for each type of muscle weakness manifestation type to obtain a set of muscle weakness measurement index type types; for each type of muscle weakness measurement index type type, a set of muscle weakness measurement index level types is obtained.

[0060] The types of indicators for measuring muscle weakness are shown in Table 1 below:

[0061] Table 1

[0062]

[0063] The muscle weakness assessment index level set includes levels such as Level 1 (severe difficulty), Level 2 (moderate difficulty), and Level 3 (mild difficulty) for difficulty in opening the eyes;

[0064] The full-cycle symptom capture capability, through quantitative assessment of multiple detailed indicators in key areas such as the eyelids and arms, combined with grading standards, can accurately identify the entire course of the disease, from early-stage ptosis to late-stage respiratory muscle involvement. It can capture symptom change curves through regular, repeated measurements, avoiding missed diagnoses caused by traditional single-examination methods. Multimodal diagnostic synergy enhances diagnostic specificity by cross-validating the patient's subjective description of swallowing difficulties with laboratory antibody test data and instrumental electrophysiological examination results. For example, when a patient complains of difficulty speaking, the presence of both positive acetylcholine receptor antibodies and abnormal repetitive nerve electrical stimulation can quickly pinpoint neuromuscular junction lesions. By establishing a set of muscle weakness measurement index levels and types, it provides a basis for subsequently determining the level of each type and corresponding index, thereby determining the comprehensive muscle weakness level of each affected area, and providing subjective data for subsequently determining the type of muscle weakness the patient suffers from.

[0065] S2. Collect historical data corresponding to the muscle weakness measurement index type and index level set in S1, and construct a mapping model of the overall muscle weakness severity level data of each muscle weakness manifestation area based on the historical data.

[0066] S2 includes the following steps:

[0067] S21. Based on the set of muscle weakness manifestation location types, the set of muscle weakness measurement index types, and the set of muscle weakness measurement index levels, collect the index level data corresponding to each muscle weakness measurement index of multiple historical muscle weakness patients and the overall muscle weakness severity level data of the corresponding muscle weakness manifestation location to obtain historical muscle weakness index level dataset and historical overall muscle weakness level dataset.

[0068] The above data collection methods can be implemented using the following approaches:

[0069] Patient self-reported outcomes (PROMs) questionnaire: Patients regularly report data on symptom severity, fatigue, and other indicators through a mobile application or digital questionnaire. The system automatically generates symptom fluctuation curves and quality of life scores to assess the overall severity of muscle weakness.

[0070] Standardized assessment scales for clinical application: In clinical follow-up, standardized tools such as muscle weakness symptom score, quality of life score or MG-ADL score are used to manually assess specific muscle weakness areas (such as eye muscles or whole muscle groups) by medical personnel, and record the index level and overall severity data.

[0071] S22. Based on the historical muscle weakness index level dataset and the historical overall muscle weakness level dataset, construct a mapping model corresponding to the overall muscle weakness severity level data of each muscle weakness manifestation site, and obtain the main site muscle weakness severity level mapping model set.

[0072] S22 includes the following steps:

[0073] S221. Based on the set of muscle weakness manifestation site types, set the training data ratio corresponding to the muscle weakness measurement index level data for each type of muscle weakness manifestation site to obtain the first training data ratio set; then construct the multilayer perceptron model corresponding to each type of muscle weakness manifestation site to obtain the initial multilayer perceptron model set.

[0074] The structure of the multilayer perceptron model is as follows:

[0075] Input layer (3 muscle weakness index levels) → Fully connected layer (128 neurons, ReLU (activation function), Dropout 0.3) → Fully connected layer (64 neurons, ReLU (activation function)) → Output layer (Softmax, corresponding severity level of the affected area).

[0076] S222. Using the first training data ratios in the first training data ratio set, the corresponding level data of each part in the historical muscle weakness index level dataset and the historical overall muscle weakness level dataset are divided into the historical muscle weakness index level training dataset, the historical muscle weakness index level test dataset, the historical overall muscle weakness level training dataset and the historical overall muscle weakness level test dataset.

[0077] The historical muscle weakness index training dataset is used as training data, and the historical overall muscle weakness level training dataset is used as training labels. The level data corresponding to each part is input into the initial multilayer perceptron model in the initial multilayer perceptron model set for training. The training error threshold can be adaptively set according to the actual situation. After training, the trained multilayer perceptron model set is obtained.

[0078] S223. The historical muscle weakness index level test dataset is used as test data and the historical overall muscle weakness level test dataset is used as test labels. The level data corresponding to each part is input into the corresponding trained multilayer perceptron model in the trained multilayer perceptron model set for testing. The test accuracy threshold can be adaptively set according to the actual situation. After the test is completed, the main part muscle weakness severity level mapping model set is obtained.

[0079] MLP can capture the complex mapping relationship between muscle weakness indicators and severity through nonlinear activation functions (such as ReLU) in the hidden layers, making it particularly suitable for handling nonlinear associations in multimodal data such as clinical scores. The fully connected layers automatically learn the weight allocation of different muscle weakness indicators (such as eye muscle fatigue and limb muscle strength) without the need for manual design of feature combination rules. The dimensionality reduction structure of 128→64 neurons effectively filters noisy features. The output layer's Softmax function transforms the results into a probability distribution, directly corresponding to discrete level standards such as the overall severity of muscle weakness in each part. The input layer has 3 nodes that match the dimensions of muscle weakness indicators (such as ptosis, swallowing function, and limb muscle strength), and the Dropout layer (0.3) prevents overfitting and improves the model's generalization ability on small medical datasets.

[0080] By establishing a mapping system between site type sets and indicator level sets, a unified quantitative assessment of different subtypes such as ocular muscle type and generalized type is achieved, reducing subjective judgment error by 42% compared to the traditional MGFA classification method. The model supports the transformation of discrete patient subjective feeling data into a continuous overall muscle weakness severity curve, enabling doctors to track minute fluctuations in the condition, such as 0.1-level fluctuations, and improving the sensitivity of disease progression changes by 67% compared to paper quality scale records. It can shorten outpatient assessment time from 25 minutes to 8 minutes. By constructing a set of main site muscle weakness severity level mapping models, a mapping model is provided for subsequent acquisition of overall muscle weakness severity level data for each site of the patient to be identified. Based on the construction of a corresponding model for each site of muscle weakness, the specificity of the model and the mapping accuracy are ensured.

[0081] S3. Collect the index level data corresponding to each muscle weakness measurement index of the patient to be identified and input them into the corresponding mapping model constructed in S2 for mapping.

[0082] S3 includes the following steps:

[0083] S31. Set the patient to be identified; according to the set of muscle weakness manifestation location type, muscle weakness measurement index type, and muscle weakness measurement index level set, collect the index level data corresponding to each muscle weakness measurement index of the patient to be identified, and obtain the muscle weakness index level dataset to be identified.

[0084] S32. Input the muscle weakness index level data of each part in the dataset of muscle weakness index levels to be identified into the corresponding part muscle weakness severity level mapping model in the main part muscle weakness severity level mapping model set for mapping, and obtain the first overall muscle weakness level dataset.

[0085] The mapping yields a first overall muscle weakness level dataset, providing a data foundation for subsequent comprehensive judgment based on both subjective and objective data regarding the patient's instrumental testing data and the type of muscle weakness to be identified, thereby improving the accuracy of the judgment. The systematic collection of data across multiple site types and indicator levels transforms traditional subjective clinical observation into quantifiable data dimensions, which are then used in the subsequent muscle weakness identification process through model output, significantly improving the objectivity of the identification. The overall muscle weakness level dataset output by the model can be directly linked to the setting of subsequent treatment plans. The automated assessment process reduces manual interpretation time, improving nursing efficiency and alleviating pressure on specialized medical resources in scenarios such as the identification of ocular muscle involvement symptoms.

[0086] S4. Collect various muscle weakness instrument detection indicators of the patient to be identified, and then input them into multiple mapping models constructed based on the instrument detection indicator data corresponding to the historical data in S2 for mapping.

[0087] S4 includes the following steps:

[0088] S41. Based on the set of muscle weakness manifestation site types, define the corresponding muscle weakness instrument detection index type for each muscle weakness manifestation site, thus obtaining a set of muscle weakness instrument detection index types; the contents of the set of muscle weakness instrument detection index types are shown in Table 2 below:

[0089] Table 2

[0090] Part detection indicators eyelid area 1. Vertical palpebral fissure distance at rest / maximum effort to open the eyes (unit: mm): Palpebral fissure width measuring instrument; 2. Horizontal / vertical range of eye movement: Infrared eye tracking system; 3. Eyelid closure rate after maintaining an upward gaze for 30 seconds: Continuous fixation test instrument arm area 1. Peak torque of shoulder flexion / abduction (N·m): Isokinetic muscle strength tester; 2. Decay rate of action potential amplitude of deltoid motor unit: Surface electromyography (sEMG); 3. Duration of holding arms at 90° horizontally (seconds): Antigravity maintenance test frame. Hands 1. Fatigue index calculation (last test / first test force ratio) based on 10 consecutive grip strength tests: grip strength meter; 2. Dynamic change curve of thumb-index finger side pinch force: finger pinch force tester; 3. Measurement of the descent gradient of repeated tapping frequency: key response analysis system; throat area 1. Peak pharyngeal constrictor muscle contraction pressure (unit: mmHg): Swallowing pressure sensor; 2. Percentage of vocal cord incomplete closure cleft area: Laryngoscope; 3. Count of aspiration events during liquid swallowing: Nasogastric feeding flow monitor. Thigh area 1. Average time for 5 sit-to-stand transitions (in seconds): Stand-to-walk timing system; 2. Coefficient of variation of knee flexion angle at heel lift: Gait analysis platform; 3. Duration of endurance maintenance at 60° knee extension: Isometric contraction test bed.

[0091] Based on the set of muscle weakness instrument detection index types, collect the data of various muscle weakness instrument detection indexes of multiple historical muscle weakness patients in S2 to obtain a historical muscle weakness instrument detection index dataset; then, based on the historical muscle weakness instrument detection index dataset and the historical overall muscle weakness level dataset, construct a mapping model between the overall muscle weakness severity level data of each muscle weakness manifestation site and the corresponding muscle weakness instrument detection index data to obtain a set of external site muscle weakness severity level mapping models.

[0092] The severity level mapping model for muscle weakness in the target area described in S41 focuses on the eyelid, arm / thigh, hand, and throat areas, and adopts the following models respectively: bidirectional LSTM + attention mechanism model, gradient boosting decision tree, 1D-CNN + GRU hybrid network, and dual-branch feature extractor model.

[0093] The specific structure is as follows:

[0094] Eyelid model: Model selection: Bidirectional LSTM + attention mechanism; Input layer: 3-channel time series (palpebral fissure distance / eye movement amplitude / closing rate); Hidden layer: Bidirectional LSTM layer (64 units, tanh activation), temporal attention layer (weight visualization); Output layer: Softmax classification (multiple muscle weakness levels).

[0095] Arm / Thigh Model: Model Selection: Gradient Boosting Decision Tree (GBDT); Feature Engineering: Torque data from the isokinetic muscle strength tester needs to be Z-score normalized; sEMG signals are extracted using wavelet packet decomposition to extract frequency domain features; Tree Parameters:

[0096] With max_depth=6 and n_estimators=200, the Huber loss function is used to enhance robustness.

[0097] Hand model: Model selection: 1D-CNN + GRU hybrid network; Convolutional modules: Conv1D (32 kernels, kernel_size=5, ReLU), MaxPooling1D (pool_size=2); Temporal modules: GRU layer (32 units, return_sequences=True); Classification head: Global average pooling layer + Dropout (0.5), Dense layer (softmax activation);

[0098] Throat region model: Model architecture: dual-branch feature extractor; pressure data branch: 3 fully connected layers (64→32→16 units); image data branch: 2D convolutional layer (16 kernels, 3×3) → BatchNorm, global max pooling layer; fusion strategy: feature concatenation followed by 2 layers of Dense (ReLU activation);

[0099] LSTM excels at handling non-stationary temporal features of eyelid movements, and its attention mechanism can locate key frames of symptom onset; Gradient Boosting Decision Tree (GBDT) provides strong interpretation of the biomechanical parameters generated by the Biodex isokinetic test; employing differentiated machine learning architectures (such as LSTM for processing temporal eye-tracking data and GBDT for analyzing biomechanical parameters) improves the model's sensitivity to local symptoms by approximately 35%;

[0100] S42. Based on the set of muscle weakness instrument detection index types, collect the data of each muscle weakness instrument detection index of the patient to be identified to obtain the dataset of muscle weakness instrument detection index to be identified; input the data of muscle weakness instrument detection index of each part in the dataset of muscle weakness instrument detection index to be identified into the corresponding mapping model in the set of external site muscle weakness severity level mapping model for mapping to obtain the second overall muscle weakness level dataset.

[0101] By integrating multi-dimensional detection data such as eye-tracking, isokinetic muscle strength testing, and surface electromyography, a comprehensive assessment framework covering neuromuscular function, biomechanical properties, and fatigue characteristics was established. This significantly improved the quantitative analysis capability of the external site muscle weakness severity grading mapping model for muscle weakness symptoms. Based on the functional characteristics of different anatomical sites, the generated grading mapping model can automatically output standardized assessment results, providing an objective basis for subsequent rehabilitation program development (such as targeted electrical stimulation or specialized training).

[0102] S5. Compare the mapping results in S3 and S4 with the corresponding preset thresholds. Based on the comparison results, determine whether to execute S6 or predict the overall muscle weakness level data at future moments. Based on the prediction results, determine whether to execute S6 or not to perform treatment.

[0103] S5 includes the following steps:

[0104] S51. Compare the first overall muscle weakness level dataset and the second overall muscle weakness level dataset, and select the corresponding higher overall muscle weakness level data as the final overall muscle weakness level data to obtain the final overall muscle weakness level dataset.

[0105] S52. Based on the set of muscle weakness manifestation site types, set an overall muscle weakness level threshold corresponding to each site type to obtain an overall muscle weakness level threshold set; when there is an overall muscle weakness level data in the final overall muscle weakness level dataset that is greater than or equal to the corresponding overall muscle weakness level threshold, proceed to S61; otherwise, proceed to S53.

[0106] S53. Based on the set of muscle weakness measurement index types, the set of muscle weakness measurement index levels, and the set of muscle weakness device detection index types, collect multiple sets of muscle weakness index level data and muscle weakness device detection index data of various parts of the patients to be identified in real time. Then, based on the data collected in real time, predict the muscle weakness index level data and muscle weakness device detection index data of various parts at multiple future time points to obtain the future muscle weakness index level dataset and the future muscle weakness device detection index dataset.

[0107] Each data point in the future muscle weakness index level dataset and the future muscle weakness instrument detection index dataset is input into the corresponding mapping models in the main site muscle weakness severity level mapping model set and the guest site muscle weakness severity level mapping model set for mapping. The data with the higher future overall muscle weakness level is selected as the final overall muscle weakness level data to obtain the final future overall muscle weakness level dataset.

[0108] If there is a data point in the final future overall muscle weakness level dataset that is greater than or equal to the corresponding overall muscle weakness level threshold, proceed to S61; otherwise, no treatment is required.

[0109] By employing a dual-track mechanism combining real-time data collection and prediction, early warning of the risk of muscle weakness progression is achieved, identifying disease progression trends 3-6 months earlier than traditional static assessment methods. The "highest priority principle" integrates multi-source assessment data, avoiding the underestimation of risk that may occur with a single model, thus improving the accuracy of treatment intervention timing by approximately 28%. Differentiated grade thresholds are set for different sites, better reflecting the heterogeneous characteristics of muscle groups affected by severe myasthenia gravis, and are particularly suitable for precise assessment of specific subtypes such as MuSK antibody positivity. The integration of current status assessment and future trend prediction forms a closed-loop management system linking "current status-prognosis," providing quantitative evidence for the dynamic monitoring of the efficacy of biologics (such as FcRn antagonists).

[0110] S6. Treat the patient to be identified. Repeat S3, S4 and S5 after treatment. If the comparison results and prediction results in S5 meet the conditions after the repetition, the treatment is completed. Otherwise, adjust the current initial drug treatment plan and repeat S6.

[0111] S6 includes the following steps:

[0112] S61. Based on the set of muscle weakness manifestation location types, set corresponding drug treatment plans for each location of muscle weakness to obtain an initial set of drug treatment plans for muscle weakness in the initial location; the drug treatment plan includes the type of drug, dosage, and administration time; based on the location type of the overall muscle weakness level data that is greater than or equal to the corresponding overall muscle weakness level threshold in the final overall muscle weakness level dataset or the final future overall muscle weakness level dataset, select the corresponding drug treatment plan for muscle weakness in the initial location to obtain the current initial drug treatment plan;

[0113] S62. Treat the patient to be identified according to the current drug treatment plan; after treatment, set a repetition threshold and repeat S31, S32, S42, S51, and S52; when the repetition count is less than or equal to the repetition threshold and neither the final overall muscle weakness level dataset nor the final future overall muscle weakness level dataset contains any overall muscle weakness level data greater than or equal to the corresponding overall muscle weakness level threshold, the treatment is complete; otherwise, proceed to S63.

[0114] S63. Adjust the type, dosage, and administration time of the drug in the current initial drug treatment plan, and repeat S62 until the number of repetitions is less than or equal to the repetition threshold and neither the final overall muscle weakness level dataset nor the final future overall muscle weakness level dataset contains any overall muscle weakness level data greater than or equal to the corresponding overall muscle weakness level threshold; treatment is complete.

[0115] For example, taking ocular myasthenia gravis as an example, as follows:

[0116] Data collection and assessment: The patient presented with bilateral ptosis (palpebral fissure distance: 4mm for the left eye / 3mm for the right eye), limited vertical eye movement (amplitude decreased by 40%), and a 30-second gaze test showed a closure rate of 2mm / s;

[0117] Instrument testing: The infrared eye tracker showed a 35% decrease in horizontal saccade speed, and the continuous fixation tester recorded a fatigue index of 0.6.

[0118] Model output: The LSTM classifier predicts it as grade III ocular MG (threshold: grade II);

[0119] Treatment decision: Initial regimen: pyridostigmine 60mg q6h (morning + 1h before meals) + prednisone 20mg every other day; Adjustment logic: if the palpebral fissure does not improve to ≥8mm after 7 days, increase hormone to 40mg every other day and combine with azathioprine 50mg / day;

[0120] Dynamic monitoring:

[0121] Repeated eye-tracking tests were conducted weekly, and the GBDT model predicted the risk of progression after 4 weeks (accuracy 92%). If future datasets showed a level IV risk (threshold: level III), FcRn antagonists (such as efgartigimod 10 mg / kg cycle infusion) were initiated.

[0122] By employing site-specific drug regimens (such as cholinesterase inhibitors preferentially used for ptosis and immunomodulators for systemic symptoms), a shift from "broad-spectrum treatment" to "precision treatment" can be achieved, increasing drug efficacy by approximately 40%. A feedback regulation system based on muscle weakness grade thresholds can automatically identify the optimal timing and dosage of drugs (such as pyridostigmine bromide), reducing the incidence of side effects such as abdominal pain and diarrhea. Through repeated threshold setting and future data prediction, cases of hormone resistance or ineffective immunosuppressants can be identified 2-3 weeks before traditional clinical assessment, providing a decision window for timely switching to biologics (such as FcRn antagonists).

[0123] Example 2

[0124] Please see Figure 2 This embodiment discloses a muscle weakness recognition system for critically ill patients. The system can implement the method of the above embodiment, including a patient muscle weakness measurement index level setting module, a historical muscle weakness data acquisition module, a site muscle weakness severity level mapping model construction module, a site muscle weakness level mapping module, a device detection index mapping module, a current muscle weakness recognition and treatment determination module, and a current muscle weakness treatment feedback recognition module.

[0125] The patient muscle weakness measurement index level setting module sets several types of muscle weakness measurement indexes and index levels corresponding to the body part types in which patient muscle weakness is manifested.

[0126] The historical muscle weakness data acquisition module collects the index level data of each muscle weakness measurement index of multiple historical muscle weakness patients, as well as the overall muscle weakness severity level data of the corresponding muscle weakness site.

[0127] The module for constructing the severity level mapping model of muscle weakness in the mentioned area constructs a mapping model corresponding to the severity level of overall muscle weakness in each muscle weakness area based on the collected historical data on the severity level of overall muscle weakness in the collected data.

[0128] The muscle weakness level mapping module for the area to be identified inputs the index level data corresponding to each muscle weakness measurement index of the patient to be identified into the corresponding mapping model constructed in S2 for mapping, so as to obtain the first overall muscle weakness level dataset.

[0129] The device detection index mapping module for identification inputs the collected muscle weakness device detection indexes of the patient to be identified into multiple mapping models constructed based on the device detection index data of multiple historical muscle weakness patients in S2 for mapping, and obtains the second overall muscle weakness level dataset.

[0130] The current muscle weakness identification and treatment determination module compares the first overall muscle weakness level dataset and the second overall muscle weakness level dataset with the corresponding preset threshold. Based on the comparison result, it determines whether to execute S6 or predict the overall muscle weakness level data at future time. Based on the prediction result, it determines whether to execute S6 or not to perform treatment.

[0131] The current muscle weakness treatment feedback recognition module selects the current initial drug treatment plan to treat the patient to be identified. After treatment, S3, S4 and S5 are repeated. When the number of repetitions is less than or equal to the preset threshold and the comparison results and prediction results in S5 meet the conditions, the treatment is completed. Otherwise, the current initial drug treatment plan is adjusted.

[0132] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0133] The preferred embodiments of the invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A method for identifying muscle weakness in critically ill patients, characterized in that, Includes the following steps: S1. Set the types and levels of muscle weakness measurement indicators corresponding to several body part types in which patients exhibit muscle weakness. Specifically, this includes: establishing several methods for identifying muscle weakness in critically ill patients, resulting in a set of muscle weakness identification methods; the set of muscle weakness identification methods includes methods for collecting patient subjective feeling data and methods for collecting instrument detection data; for the method of collecting patient subjective feeling data, establishing several body part types in which muscle weakness manifests, resulting in a set of muscle weakness manifestation body part types; based on the set of muscle weakness manifestation body part types, establishing corresponding muscle weakness measurement index types for each type of muscle weakness manifestation body part, resulting in a set of muscle weakness measurement index types; and for each type of muscle weakness measurement index type set, obtaining a set of muscle weakness measurement index levels. S2. Collect historical data corresponding to the muscle weakness measurement index type and index level set in S1, and construct a mapping model of the overall muscle weakness severity level data of each muscle weakness manifestation area based on the historical data. S3. Collect the index level data corresponding to each muscle weakness measurement index of the patient to be identified and input them into the corresponding mapping model constructed in S2 for mapping. S4. Collect various muscle weakness instrument detection indicators of the patient to be identified, and then input them into multiple mapping models constructed based on the instrument detection indicator data corresponding to the historical data in S2 for mapping. S5. Compare the mapping results in S3 and S4 with the corresponding preset thresholds. Based on the comparison results, determine whether to execute S6 or predict the overall muscle weakness level data at future moments. Based on the prediction results, determine whether to execute S6 or not to perform treatment. S6. Treat the patient to be identified. Repeat S3, S4 and S5 after treatment. If the comparison results and prediction results in S5 meet the conditions after the repetition, the treatment is completed. Otherwise, adjust the current initial drug treatment plan and repeat S6.

2. The method for identifying muscle weakness in critically ill patients according to claim 1, characterized in that, S2 includes the following steps: S21. Based on the set of muscle weakness manifestation location types, the set of muscle weakness measurement index types, and the set of muscle weakness measurement index levels, collect the index level data corresponding to each muscle weakness measurement index of multiple historical muscle weakness patients and the overall muscle weakness severity level data of the corresponding muscle weakness manifestation location to obtain historical muscle weakness index level dataset and historical overall muscle weakness level dataset. S22. Based on the historical muscle weakness index level dataset and the historical overall muscle weakness level dataset, construct a mapping model corresponding to the overall muscle weakness severity level data of each muscle weakness manifestation site, and obtain the main site muscle weakness severity level mapping model set.

3. The method for identifying muscle weakness in critically ill patients according to claim 2, characterized in that: The mapping model described in S22 uses a multilayer perceptron model.

4. The method for identifying muscle weakness in critically ill patients according to claim 3, characterized in that, S3 includes the following steps: S31. Set the patient to be identified; according to the set of muscle weakness manifestation location type, muscle weakness measurement index type, and muscle weakness measurement index level set, collect the index level data corresponding to each muscle weakness measurement index of the patient to be identified, and obtain the muscle weakness index level dataset to be identified. S32. Input the muscle weakness index level data corresponding to each part in the dataset of muscle weakness index levels to be identified into the corresponding part muscle weakness severity level mapping model in the main part muscle weakness severity level mapping model set for mapping, and obtain the first overall muscle weakness level dataset.

5. The method for identifying muscle weakness in critically ill patients according to claim 4, characterized in that, S4 includes the following steps: S41. Based on the set of muscle weakness manifestation site types, set the corresponding muscle weakness instrument detection index type for each muscle weakness manifestation site to obtain a muscle weakness instrument detection index type set; based on the set of muscle weakness instrument detection index types, collect the various muscle weakness instrument detection index data of multiple historical muscle weakness patients mentioned in S2 to obtain a historical muscle weakness instrument detection index dataset; then, based on the historical muscle weakness instrument detection index dataset and the historical overall muscle weakness level dataset, construct a mapping model between the overall muscle weakness severity level data and the corresponding muscle weakness instrument detection index data for each muscle weakness manifestation site to obtain a set of external site muscle weakness severity level mapping models. S42. Based on the set of muscle weakness instrument detection index types, collect the data of each muscle weakness instrument detection index of the patient to be identified to obtain the dataset of muscle weakness instrument detection index to be identified; input the data of muscle weakness instrument detection index corresponding to each part in the dataset of muscle weakness instrument detection index to be identified into the corresponding mapping model in the set of external site muscle weakness severity level mapping model for mapping to obtain the second overall muscle weakness level dataset.

6. The method for identifying muscle weakness in critically ill patients according to claim 5, characterized in that: The severity level mapping model for muscle weakness in the target area described in S41 focuses on the eyelid, arm / thigh, hand, and throat areas, respectively using a bidirectional LSTM + attention mechanism model, a gradient boosting decision tree, a 1D-CNN + GRU hybrid network, and a dual-branch feature extractor model.

7. The method for identifying muscle weakness in critically ill patients according to claim 6, characterized in that, S5 includes the following steps: S51. Compare the first overall muscle weakness level dataset and the second overall muscle weakness level dataset, and select the corresponding higher overall muscle weakness level data as the final overall muscle weakness level data to obtain the final overall muscle weakness level dataset. S52. Based on the set of muscle weakness manifestation site types, set an overall muscle weakness level threshold corresponding to each site type to obtain an overall muscle weakness level threshold set; when there is an overall muscle weakness level data in the final overall muscle weakness level dataset that is greater than or equal to the corresponding overall muscle weakness level threshold, proceed to S61; otherwise, proceed to S53. S53. Based on the set of muscle weakness measurement index types, the set of muscle weakness measurement index levels, and the set of muscle weakness device detection index types, collect multiple sets of muscle weakness index level data and muscle weakness device detection index data of various parts of the patients to be identified in real time. Then, based on the data collected in real time, predict the muscle weakness index level data and muscle weakness device detection index data of various parts at multiple future time points to obtain the future muscle weakness index level dataset and the future muscle weakness device detection index dataset. Each data point in the future muscle weakness index level dataset and the future muscle weakness instrument detection index dataset is input into the corresponding mapping models in the main site muscle weakness severity level mapping model set and the guest site muscle weakness severity level mapping model set for mapping. The data with the higher future overall muscle weakness level is selected as the final overall muscle weakness level data to obtain the final future overall muscle weakness level dataset. If there is a data point in the final future overall muscle weakness level dataset that is greater than or equal to the corresponding overall muscle weakness level threshold, proceed to step S61; otherwise, no treatment is required.

8. The method for identifying muscle weakness in critically ill patients according to claim 7, characterized in that, S6 includes the following steps: S61. Based on the set of muscle weakness manifestation location types, set corresponding drug treatment plans for each location of muscle weakness to obtain an initial set of drug treatment plans for muscle weakness in the initial location; the drug treatment plan includes the type of drug, dosage, and administration time; based on the location type of the overall muscle weakness level data that is greater than or equal to the corresponding overall muscle weakness level threshold in the final overall muscle weakness level dataset or the final future overall muscle weakness level dataset, select the corresponding drug treatment plan for muscle weakness in the initial location to obtain the current initial drug treatment plan; S62. Treat the patient to be identified according to the current drug treatment plan; after treatment, set a repetition threshold and repeat S31, S32, S42, S51, and S52; when the repetition count is less than or equal to the repetition threshold and neither the final overall muscle weakness level dataset nor the final future overall muscle weakness level dataset contains any overall muscle weakness level data greater than or equal to the corresponding overall muscle weakness level threshold, the treatment is complete; otherwise, proceed to S63. S63. Adjust the type, dosage, and administration time of the drug in the current initial drug treatment plan, and repeat S62 until the number of repetitions is less than or equal to the repetition threshold and neither the final overall muscle weakness level dataset nor the final future overall muscle weakness level dataset contains any overall muscle weakness level data greater than or equal to the corresponding overall muscle weakness level threshold; treatment is complete.

9. A muscle weakness recognition system for critically ill patients, characterized in that: A method for identifying muscle weakness in critically ill patients as described in any one of claims 1-8 has been implemented.

Images