A detection method and system for digital multi-modal isometric muscle training

Abstract

The application discloses a kind of digital multi-modal muscle isometric training detection method and system.The detection method includes: collecting the skeletal key point data and muscle tissue hardness data when patient executes muscle isometric training;Judge whether the training posture of patient and effort is up to standard;Record the training data of patient in up to standard state;Training data is transmitted to cloud platform, based on pre-trained GAMLSS model, corresponding percentile standard curve of the attribute of patient is matched, to convert the muscle tissue hardness data of patient in up to standard state into percentile score;Based on the training performance of patient's percentile score evaluation, and provide training feedback.Using the present application can automatically monitor patient muscle isometric rehabilitation training and provide training feedback in home scene, save the time cost of patient.And, the present application can adopt different detection standards for different users, to adapt to the training situation of different users.

Description

Technical Field

[0001] This invention relates to a method for detecting digital multimodal isometric muscle training, and also to a corresponding detection system, belonging to the field of healthcare informatics technology. Background Technology

[0002] Patients with cervical spondylosis typically require systematic isometric muscle training during the rehabilitation phase to enhance neck stability and promote functional recovery. Extensive clinical practice has shown that standardized rehabilitation training plays a crucial role in promoting functional recovery and improving prognosis. However, due to the long rehabilitation period, patients often find it difficult to consistently attend professional training sessions, making home-based rehabilitation a mainstream option. How to effectively monitor the training process and assess its quality in a home environment has become one of the most pressing issues in the field of rehabilitation medicine.

[0003] Isometric muscle training is a static force exertion training method with no significant limb displacement, widely used in rehabilitation programs for patients after cervical spine surgery or with chronic cervical spondylosis. Because the joint angles do not change during the movement, traditional techniques based on posture recognition or motion capture are insufficient to effectively assess the intensity of force exertion and the correctness of the movement. Especially in typical movements such as isometric extension contraction, it is often difficult to objectively assess remotely whether the patient is truly exerting force, whether the force exertion reaches the therapeutic threshold, and whether the posture is standard. This leads to unreliable training effectiveness and may even delay the rehabilitation process.

[0004] Currently, home-based rehabilitation training mainly relies on patient self-execution or family-assisted supervision, lacking professional guidance and real-time feedback mechanisms. Rehabilitation therapists struggle to remotely monitor patients' training progress, making it difficult to correct errors or motivate patients to achieve their goals in a timely manner. This "black box" state during training not only affects rehabilitation outcomes but also increases the cost of doctor-patient communication and hinders the efficient use of rehabilitation resources. Therefore, there is an urgent need for a technological means to accurately identify, quantify, and provide real-time feedback on isometric muscle training in a home setting, filling the gaps in existing monitoring methods. Summary of the Invention

[0005] The primary technical problem to be solved by this invention is to provide a method for detecting digital multimodal isometric muscle training.

[0006] Another technical problem to be solved by the present invention is to provide a digital multimodal muscle isometric training detection system.

[0007] To achieve the above-mentioned technical objectives, the present invention adopts the following technical solution: According to a first aspect of the present invention, a method for detecting digital multimodal isometric muscle training is provided, comprising the following steps: Collect skeletal key point data and muscle tissue stiffness data when patients perform isometric muscle training; Based on the skeletal key point data, it is determined whether the patient's training posture conforms to the standard posture, and based on the muscle tissue stiffness data, it is determined whether the patient's force exertion reaches the preset threshold. When the patient's training posture meets the standard and the force exertion reaches the threshold, it is determined to be in a qualified state, and the training data in the qualified state is recorded; wherein, the training data in the qualified state includes the muscle tissue stiffness data; The training data is transmitted to the cloud platform; On the cloud platform, based on the pre-trained GAMLSS model, percentile standard curves corresponding to the patient's attributes are matched; The GAMLSS model is constructed based on historical training data. Each sample in the historical training data consists of a response variable and multiple explanatory variables. The response variable is the measured value of muscle tissue stiffness, and the multiple explanatory variables include at least: postoperative time, gender, age group, specific cervical spine disease type, and specific surgical procedure. Furthermore, the GAMLSS model is used to predict the conditional probability distribution of the response variable for a specific combination of explanatory variables, thereby generating the corresponding percentile standard curve. Based on the percentile standard curve corresponding to the patient, the muscle tissue stiffness data of the patient under the target state is converted into percentile scores. The patient's training performance was evaluated based on their percentile score, and feedback was provided to the patient.

[0008] Preferably, the skeletal key points include C7-T1 and the acromion key points; Determining whether the patient's training posture conforms to the standard posture includes: calculating the neck tilt angle or rotation angle and comparing it with the standard angle. When the error is less than a preset degree, it is determined to conform to the standard posture.

[0009] Preferably, in the GAMLSS model, the conditional probability distribution of the response variable is defined by multiple distribution parameters. Each of the distribution parameters is mapped to a linear predictor through a link function; Each of the linear predictors consists of an intercept term and one or more smoothing or factor terms corresponding to the explanatory variables.

[0010] Preferably, the muscle tissue stiffness data of the patient in the target state is converted into percentile scores, specifically including: The muscle tissue stiffness value corresponding to each instance of achieving the target force was compared with the patient's corresponding percentile standard curve to determine the corresponding percentile P. i ; Calculate the percentile score S for a single training session: S = (P1 + P2 + ... + P) i ) / n, i∈(1,n), where n is the number of training iterations for this group.

[0011] Preferably, the detection method further includes: When the amount of new patient training data accumulated on the cloud platform reaches a preset threshold, the GAMLSS model is retrained, and the muscle tissue stiffness percentile standard curve is updated based on the retrained model.

[0012] Preferably, when it is determined that the patient's posture does not meet the standard and / or the force exerted does not reach the threshold, a real-time correction prompt message is generated and output.

[0013] According to a second aspect of the present invention, a detection system for digital multimodal isometric muscle training is provided for implementing the above-described detection method, wherein the detection system includes a patient terminal, a cloud platform, and a doctor terminal; wherein, The patient terminal includes: The visual acquisition module is used to collect skeletal key point data when patients perform isometric muscle training; The tissue stiffness monitoring module is used to collect muscle tissue stiffness data when patients perform isometric muscle training. An edge computing device is connected to the visual acquisition module and the tissue stiffness monitoring module, and is configured to perform standard judgments on posture and force exertion; A patient feedback module is connected to the edge computing device and configured to provide training feedback to the patient; The cloud platform connects to the edge computing device and is configured to store the GAMLSS model and various types of percentile standard curves generated by the GAMLSS model, in order to perform matching of patient data with percentile standard curves and calculation of the patient's percentile score. The doctor's terminal is connected to the patient feedback module, allowing the doctor to view all of the patient's training data in real time and provide auxiliary suggestions to the patient.

[0014] Preferably, the visual acquisition module is a camera integrated into a tablet device or smartphone; The tissue stiffness monitoring module is a wearable device that supports wireless communication.

[0015] Preferably, the cloud platform further includes a model update unit, which is configured to monitor the accumulation of new data and automatically trigger the retraining of the model and the updating of the standard curve when preset conditions are met.

[0016] Compared with the prior art, the present invention has the following technical effects: (1) Effective quantification and precise monitoring of isometric muscle training in home settings have been achieved.

[0017] This invention solves the technical problem of traditional methods being unable to quantify force intensity due to the lack of limb displacement by integrating visual skeletal key points and multimodal data from wearable devices. It can objectively judge whether the posture standard and force intensity of training movements meet the standards.

[0018] (2) A highly personalized rehabilitation assessment system has been constructed.

[0019] This invention, through the construction of percentile standard curves based on the GAMLSS model, encompasses multidimensional explanatory variables such as postoperative time, gender, age group, disease type, and surgical procedure. The system can match personalized assessment standards for patients with different characteristics, achieving a leap from a "one-size-fits-all" approach to precise assessment that is "personalized and time-dependent," significantly improving the scientific nature and effectiveness of rehabilitation guidance.

[0020] (3) A real-time feedback and closed-loop optimization mechanism for cloud-edge collaboration has been formed.

[0021] This invention achieves real-time posture and force assessment and immediate correction feedback through edge computing on the patient's end, ensuring low latency and high timeliness in training interactions. Simultaneously, the cloud platform handles complex model calculations and data persistence, automatically updating the model and standard curves as data accumulates, enabling the detection system to self-evolve and gradually improve detection accuracy.

[0022] (4) Significantly improved the quality of rehabilitation training and patient compliance.

[0023] The detection system of this invention provides patients with clear training goals and positive immediate incentives through real-time feedback and visualized percentile scores. At the same time, it synchronizes the patient's training data to the doctor's end, enabling the doctor to conduct remote supervision and precise guidance. This effectively solves the problems of uncontrolled home training quality and reliance on manual supervision, and frees up medical resources.

[0024] (5) It has pioneered a new path for objective evaluation based on the fusion of statistical models and multimodal data.

[0025] This invention proposes for the first time to use muscle tissue stiffness as the core quantitative indicator of isometric muscle exertion, and combines it with computer vision technology to generate dynamic standards through a rigorous GAMLSS statistical model. This provides a verifiable and replicable technical solution for the field of digital rehabilitation and has significant industry demonstration value. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a method for detecting digital multimodal isometric muscle training according to the first embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the working principle of the GAMLSS model in the first embodiment of the present invention. Figure 3 A schematic diagram of the structure of a digital multimodal isometric muscle training detection system provided in the second embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a digital multimodal isometric muscle training detection system provided in the third embodiment of the present invention. Detailed Implementation

[0027] The technical content of the present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0028] To address the problems of existing technologies, such as difficulty in quantifying the effectiveness of training movements, inability to ensure training quality, and over-reliance on manual supervision, this invention provides a digital detection scheme based on multimodal data fusion and statistical modeling. First, based on historical patient training data, a GAMLSS (Generalized Additive Model Position-Scale-Shape) model is constructed using deep learning. This model uses measured muscle stiffness as the response variable and postoperative time, gender, age group, disease type, and surgical procedure as explanatory variables, generating personalized percentile standard curves for different patient characteristics. In the application phase, the detection system collects visual skeletal keypoint data (for posture judgment) and muscle stiffness data (for force assessment) in real time during patient training. Once training is deemed satisfactory, the measured stiffness value is compared with the corresponding standard curve and converted into a percentile score, thereby achieving a quantitative assessment of training performance. This assessment result reflects the standardization of training in real time, allowing patients to receive professional feedback without on-site guidance from a rehabilitation therapist. Simultaneously, all training data that meets the standards can be uploaded to the cloud for continuous optimization and updating of the GAMLSS model, forming a data-driven self-evolutionary closed loop that continuously improves assessment accuracy.

[0029] First Embodiment like Figure 1 As shown, the first embodiment of the present invention provides a method for detecting digital multimodal isometric muscle training, which includes at least the following steps: S1: Collect multimodal data.

[0030] Specifically, when patients perform isometric muscle training, a tablet device with a camera detects key skeletal points (e.g., C7-T1, acromion) and calculates the neck tilt / rotation angle. If the deviation from the standard angle is less than 5 degrees, it is considered a standard posture. If the error is greater than or equal to 5 degrees, it indicates that the training posture does not conform to the standard posture, and the patient needs to be reminded.

[0031] Similarly, wearable devices supporting wireless communication can detect muscle tissue stiffness in real time to determine whether the patient's exertion is sufficient. Specifically, a tablet device provides voice prompts to the patient to exert as much force as possible. When the detected muscle tissue stiffness value rises and reaches its peak, it is determined that the target has been met, and this peak value is recorded as the patient's maximum muscle tissue stiffness. Conversely, if the value falls below the peak, it indicates that the patient's exertion is insufficient.

[0032] It is important to emphasize that this invention uses muscle tissue stiffness as the core quantitative indicator of isometric training force intensity, rather than simply adopting ultrasound elastography technology used in the medical field for passive palpation or single-shot measurements. Compared to electromyography (EMG) signals (susceptible to interference from skin impedance and electrode position) or force sensors (requiring fixed contact and unsuitable for home settings), muscle tissue stiffness can be dynamically monitored in real-time, continuously, and non-invasively through wearable devices, and it has a direct physical correlation with muscle contraction intensity. More importantly, this invention fuses muscle stiffness data with visual skeletal keypoint data for determination: only when the posture meets the standard and the stiffness reaches the threshold is it considered effective training. This dual-condition fusion mechanism avoids misjudgment based on a single modality (e.g., stiffness alone cannot distinguish whether posture compensation exists during force exertion; vision alone cannot determine force intensity), thereby achieving precise control over the quality of isometric training in a home environment.

[0033] S2: Record training data when patients achieve the target status.

[0034] When it is determined from step S1 above that the patient's training posture meets the standard and the force exertion reaches the threshold, it indicates that the patient's training is in a qualified state. At this time, it is necessary to record the patient's training data in the qualified state, including: the current training movement, the number of training sessions, the training duration, and the current muscle tissue stiffness, etc.

[0035] Furthermore, it should be noted that in step S2, training data is only recorded when the patient's training is in a target state. If the posture is not standard or the force exertion does not reach the threshold, the specific cause (such as postural deviation or insufficient force exertion) is determined based on multimodal data, and corresponding real-time correction prompts are generated.

[0036] For example, if the training posture is not in accordance with the standard posture, the patient will be prompted on the patient's end (e.g., on a tablet) to adjust their posture; if the force applied is insufficient, the patient will be prompted to increase the force. The specific content of the corrective prompts can be adaptively set as needed and is not specifically limited here.

[0037] S3: Transmit training data to the cloud platform to calculate the percentile score of the patient's training.

[0038] In one embodiment of the present invention, the cloud platform needs to pre-construct a GAMLSS model based on historically achieved training data. Each sample in the historically achieved training data consists of a response variable and multiple explanatory variables. The response variable is the measured value of muscle tissue stiffness, and the multiple explanatory variables include at least: postoperative time, gender, age group, specific cervical spine disease type, and specific surgical procedure.

[0039] The working principle of the GAMLSS model will be explained in detail below. Figure 2 As shown, the GAMLSS model is divided into a training phase and an application phase. The training phase uses historical training data as input and outputs a personalized GAMLSS model after training. The application phase takes a combination of explanatory variables from new patients (such as gender, age, surgical procedure, etc., determined through patient information) as input and outputs a percentile standard curve for that patient.

[0040] Understandably, after the aforementioned training phase, the GAMLSS model has learned how the distribution of the response variable (muscle stiffness) is influenced by the various explanatory variables. During model training, this structured data format (a combination of a response variable and multiple explanatory variables) enables the GAMLSS model to learn the normal range of muscle stiffness under various specific conditions, thus laying the foundation for subsequently generating personalized percentile standard curves.

[0041] Understandably, once the model training is complete, the GAMLSS model will predict the conditional probability distribution of the response variable based on all possible combinations of explanatory variables (such as male / female + age group + disease + surgical procedure), and then generate the corresponding percentile standard curves in batches.

[0042] Therefore, in the subsequent model application phase, the corresponding explanatory variables can be determined based on the new patient's personal information. Specifically, the patient's gender, age group, disease type, and specific surgical procedure (i.e., standardized surgical methods with significant differences in surgical path, operation technique, fixation method, and impact on muscle tissue in the treatment of cervical spine diseases) can be determined using the specific combination of explanatory variables identified for this patient. Then, the percentile standard curve corresponding to this patient can be matched from the various generated percentile standard curves.

[0043] In one embodiment of the present invention, the conditional probability distribution of the response variable of the GAMLSS model is defined by multiple distribution parameters; each distribution parameter is mapped to a linear predictor through a link function; each linear predictor consists of an intercept term and one or more smoothing terms or factor terms corresponding to the explanatory variables.

[0044] Specifically, the GAMLSS model is defined as follows: Where y is the response variable, which is usually a vector containing the observed values; D represents the probability distribution of the response variable, which has K parameters; θ1, θ2.....θ K The parameters that represent the distribution D are, for example: θ1 represents the location parameter (such as the mean). g k This represents the link function, used to link θ k Mapping to linear predictor η k ; η k This represents the k-th linear predictor; β 0k The intercept term representing the k-th distribution parameter is a scalar constant; 1 n The unit represents a vector of length n; This indicates that the design matrix or basis matrix is ​​used for the j-th smoothing term of the k-th distribution parameter; This represents the coefficient vector of the j-th smoothing term of the k-th distribution parameter; This represents the smoothing parameter, which controls the smoothing degree of the j-th smoothing term; This represents the penalty matrix, used to define the penalty term for the smoothing function, thereby ensuring the coefficient vector... The estimate is smooth.

[0045] In step S3, after the training data obtained in step S2 is transmitted to the cloud platform, a corresponding percentile standard curve is first matched for the patient based on the GAMLSS model. Then, the muscle tissue stiffness data of the patient in the target state is converted into a percentile score. The specific calculation process is as follows: ① During each training session, the muscle tissue stiffness value was recorded once for each time the target force was achieved. The value was then compared with the standard curve of the corresponding percentile for each patient to obtain the corresponding percentile Pi.

[0046] Specifically, the patient's measured muscle stiffness value is compared with the standard curve of their respective population to determine their relative position within that population. Then, linear interpolation can be used to calculate the precise percentile.

[0047] For example: The patient's measured muscle tissue stiffness value is 3.1. Using the percentile standard curve for this patient, the corresponding intervals are 3.08 (P40) and 3.12 (P50). Since the measured value of 3.1 falls between 3.08 and 3.12, the result calculated using linear interpolation is: Therefore, the muscle tissue stiffness value of this training session was 3.1, corresponding to the percentile of P45.

[0048] ② Calculate the percentile score S for a single training session.

[0049] After calculating the percentile for each training session using the above method, the percentile score S for a single training session is obtained by averaging the results. S = (P1 + P2 + ... + P...) i ) / n, i∈(1,n), where n is the number of training iterations for this group.

[0050] S4: Provide training feedback to patients.

[0051] Once the percentile score S corresponding to the patient is determined based on the above step S3, the percentile level that the patient has reached in training can be determined to evaluate the patient's training performance and display the current training status to the patient through the patient's end (e.g., a tablet).

[0052] In the above embodiments, preferably, the detection method further includes: S5: Model optimization.

[0053] Specifically, as new patients continue to train, the cloud platform will gradually accumulate training data on the achievement of new patient standards. When the amount of training data on the achievement of new patient standards reaches a preset threshold, the GAMLSS model will be retrained, and the standard curve of muscle tissue stiffness percentiles will be updated based on the retrained model.

[0054] Understandably, as the GAMLSS model is continuously optimized, the training and detection accuracy for new patients will continue to improve, thus forming a perfect closed loop of new patient training → model update → training feedback → model update again → new patient training.

[0055] This invention introduces the GAMLSS model into the field of rehabilitation medicine to construct a dynamic percentile standard curve for muscle tissue stiffness, which is not a simple application of known statistical methods. Unlike traditional fixed threshold assessments or simple regression analyses, the GAMLSS model uses postoperative time as the core dynamic explanatory variable, while incorporating multidimensional factors such as gender, age group, specific cervical spine disease type, and specific surgical procedure to predict the conditional probability distribution of muscle tissue stiffness at different rehabilitation stages. The resulting percentile standard curve can dynamically change with the patient's postoperative recovery process, achieving truly personalized assessment that varies from person to person and from time to time. In existing rehabilitation assessment technologies, neither absolute value comparisons based on force sensors nor threshold judgments based on surface electromyography can provide such a dynamically updated individualized reference system based on a large-sample statistical model. This invention is the first to apply the GAMLSS model to isometric rehabilitation training assessment, providing a new statistical benchmark for digital rehabilitation.

[0056] Second Embodiment like Figure 3 As shown, based on the first embodiment described above, the second embodiment of the present invention provides a digital multimodal muscle isometric training detection system, including a patient terminal 10, a cloud platform 20, and a doctor terminal 30.

[0057] Specifically, the patient terminal 10 is used to perform multimodal data acquisition, edge computing, and patient feedback operations. The patient terminal 10 includes a visual acquisition module 1, a tissue stiffness monitoring module 2, an edge computing device 3, and a patient feedback module 4. The visual acquisition module 1 is a camera integrated into the edge computing device 3 (e.g., a tablet or smartphone) and is used to acquire skeletal key point data when the patient performs isometric muscle training. The tissue stiffness monitoring module 2 is a wearable device supporting wireless communication and is used to acquire muscle tissue stiffness data when the patient performs isometric muscle training. The edge computing device 3 connects to the visual acquisition module 1 and the tissue stiffness monitoring module 2 and is configured to perform posture and force assessment. The patient feedback module 4 connects to the edge computing device 3 and is configured to provide training feedback to the patient.

[0058] The cloud platform 20 connects to the edge computing device 3 and is configured to store the GAMLSS model and various types of percentile standard curves generated by the GAMLSS model, in order to perform matching of patient data with percentile standard curves and calculation of patient percentile scores. Preferably, the cloud platform 20 also includes a model update unit 21, configured to monitor the accumulation of new data and automatically trigger model retraining and standard curve updates when preset conditions are met.

[0059] The doctor's terminal 30 is connected to the patient's terminal 10, which allows the doctor to view the patient's training data for the day, overall training data, patient compliance data and other information in real time, and thus provide auxiliary suggestions to the patient.

[0060] It is understood that the functions and connections of the modules in this embodiment are merely one specific implementation of the detection method described in the first embodiment. In other embodiments, the functions and connections of the modules can be adaptively adjusted as needed, and no specific limitations are made here.

[0061] Third Embodiment like Figure 4 As shown, based on the above-described method for detecting digital multimodal isometric muscle training, the third embodiment of the present invention further provides a system for detecting digital multimodal isometric muscle training. This detection system includes one or more processors and a memory. The memory is coupled to the processor and is used to store one or more programs. When the programs are executed by the processor, the processor implements the method for detecting digital multimodal isometric muscle training as described in the above embodiment.

[0062] The processor controls the overall operation of the detection system to complete all or part of the steps of the aforementioned digital multimodal isometric muscle training detection method. The processor can be a central processing unit (CPU), graphics processing unit (GPU), field-programmable gate array (FPGA), application-specific integrated circuit (ASIC), digital signal processing (DSP) chip, etc. The memory stores various types of data to support the operation of the detection system. This data may include, for example, instructions for any application or method operating on the detection system, as well as application-related data. The memory can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, etc.

[0063] In one exemplary embodiment, the detection system may be implemented by a computer chip or physical entity, or by a product with certain functions, for performing the aforementioned detection method for digital multimodal isometric muscle training and achieving the same technical effect as described above. A typical embodiment is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, an in-vehicle human-machine interface device, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0064] In another exemplary embodiment, the present invention also provides a computer-readable storage medium including program instructions, which, when executed by a processor, implement the steps of the detection method for digital multimodal isometric muscle training in any of the above embodiments. For example, the computer-readable storage medium may be the memory including the program instructions described above, which may be executed by the processor of the detection system to complete the detection method for digital multimodal isometric muscle training described above, and achieve the same technical effects as the method described above.

[0065] It should be noted that the above embodiments are merely illustrative examples. The technical solutions of each embodiment can be combined, and all are within the protection scope of this invention.

[0066] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0067] The above provides a detailed description of the detection method and system for digital multimodal isometric muscle training provided by this invention. Any obvious modifications made by those skilled in the art without departing from the essence of this invention will constitute an infringement of the patent rights of this invention and will incur corresponding legal liability.

Claims

1. A method for detecting digital multimodal isometric muscle training, characterized in that... Includes the following steps: Collect skeletal key point data and muscle tissue stiffness data when patients perform isometric muscle training; Based on the skeletal key point data, it is determined whether the patient's training posture conforms to the standard posture, and based on the muscle tissue stiffness data, it is determined whether the patient's force exertion reaches the preset threshold. When the patient's training posture meets the standard and the force exertion reaches the threshold, it is determined to be in a qualified state, and the training data in the qualified state is recorded; wherein, the training data in the qualified state includes the muscle tissue stiffness data; The training data is transmitted to the cloud platform; On the cloud platform, based on the pre-trained GAMLSS model, percentile standard curves corresponding to the patient's attributes are matched; The GAMLSS model is constructed based on historical training data. Each sample in the historical training data consists of a response variable and multiple explanatory variables. The response variable is the measured value of muscle tissue stiffness, and the multiple explanatory variables include at least: postoperative time, gender, age group, specific cervical spine disease type, and specific surgical procedure. Furthermore, the GAMLSS model is used to predict the conditional probability distribution of the response variable for a specific combination of explanatory variables, thereby generating the corresponding percentile standard curve. Based on the percentile standard curve corresponding to the patient, the muscle tissue stiffness data of the patient under the target state is converted into percentile scores. The patient's training performance was evaluated based on their percentile score, and feedback was provided to the patient.

2. The detection method as described in claim 1, characterized in that: The key skeletal points include C7-T1 and the acromion key points; Determining whether the patient's training posture conforms to the standard posture includes: calculating the neck tilt angle or rotation angle and comparing it with the standard angle. When the error is less than a preset degree, it is determined to conform to the standard posture.

3. The detection method as described in claim 1, characterized in that... In the GAMLSS model: The conditional probability distribution of the response variable is defined by multiple distribution parameters; Each of the distribution parameters is mapped to a linear predictor through a link function; Each of the linear predictors consists of an intercept term and one or more smoothing or factor terms corresponding to the explanatory variables.

4. The detection method as described in claim 1, characterized in that... The muscle tissue stiffness data of the patients under the target condition were converted into percentile scores, specifically including: The muscle tissue stiffness value corresponding to each instance of achieving the target force was compared with the patient's corresponding percentile standard curve to determine the corresponding percentile P. i ; Calculate the percentile score S for a single training session: S = (P1 + P2 + ... + P) i ) / n, i∈(1,n), where n is the number of training iterations for this group.

5. The detection method as described in claim 1, characterized in that... Also includes: When the amount of new patient training data accumulated on the cloud platform reaches a preset threshold, the GAMLSS model is retrained, and the muscle tissue stiffness percentile standard curve is updated based on the retrained model.

6. The detection method as described in claim 1, characterized in that: When it is determined that the patient's posture does not meet the standard and / or the force exerted does not reach the threshold, a real-time correction prompt message is generated and output.

7. A digital multimodal isometric muscle training detection system, used to implement the detection method according to any one of claims 1 to 6, characterized in that... This includes the patient's end, the cloud platform, and the doctor's end; among them, The patient terminal includes: The visual acquisition module is used to collect skeletal key point data when patients perform isometric muscle training; The tissue stiffness monitoring module is used to collect muscle tissue stiffness data when patients perform isometric muscle training. An edge computing device is connected to the visual acquisition module and the tissue stiffness monitoring module, and is configured to perform standard judgments on posture and force exertion; A patient feedback module is connected to the edge computing device and configured to provide training feedback to the patient; The cloud platform connects to the edge computing device and is configured to store the GAMLSS model and various types of percentile standard curves generated by the GAMLSS model, in order to perform matching of patient data with percentile standard curves and calculation of the patient's percentile score. The doctor's terminal is connected to the patient feedback module, allowing the doctor to view all of the patient's training data in real time and provide auxiliary suggestions to the patient.

8. The detection system as described in claim 7, characterized in that: The visual acquisition module is a camera integrated into a tablet device or smartphone; The tissue stiffness monitoring module is a wearable device that supports wireless communication.

9. The detection system as described in claim 7, characterized in that: The cloud platform also includes a model update unit; the model update unit is configured to monitor the accumulation of new data and automatically trigger the retraining of the model and the updating of the standard curve when preset conditions are met.

Images