Pelvic floor muscle repair instrument, pelvic floor muscle training indication method, equipment and medium

Abstract

The invention discloses a pelvic floor muscle repairing instrument, a pelvic floor muscle training indication method, pelvic floor muscle training indication equipment and a medium, and the pelvic floor muscle repairing instrument comprises a main body part which is used for being inserted into or placed in a part corresponding to pelvic floor muscle; the first electrode and the second electrode are arranged on the main body part and are used for electrically stimulating pelvic floor muscles and detecting pelvic floor muscle surface electromyogram signals after the pelvic floor muscles are electrically stimulated; the third electrode is attached to the peripheral muscle of the pelvic floor muscle during use and is used for detecting an electromyographic signal of the peripheral muscle of the pelvic floor muscle; the posture detection unit is arranged on the trunk part and is used for detecting a posture time sequence signal after the action of the trunk part is changed; the controller is electrically connected with the first electrode and the second electrode and is used for performing feature extraction on the three signals in the same time sequence, generating a fusion feature vector, determining a pelvic floor muscle training evaluation result and generating a training feedback instruction according to the pelvic floor muscle training evaluation result; the training feedback instruction is used for instructing the user to train.

Description

Technical Field

[0001] This application relates to the field of postpartum rehabilitation equipment technology, specifically to a pelvic floor muscle repair instrument, a pelvic floor muscle training instruction method, equipment, and medium. Background Technology

[0002] Pelvic floor muscle rehabilitation training is an important means of improving pelvic floor function in clinical settings such as postpartum rehabilitation for modern women and urinary continence training for the elderly. The accurate evaluation of its training effect is of key significance for improving the quality of rehabilitation.

[0003] Currently, existing assessment methods mainly rely on manual observation or single-channel electromyography (EMG) signal detection. However, existing technologies have several drawbacks. For example, manual observation is easily influenced by subjective factors, leading to inaccurate and subjective assessment results; and single-channel EMG signals can only reflect the activity of a single muscle group, failing to comprehensively assess the synergistic effects of multiple muscle groups during training, resulting in distorted assessments of training effectiveness. Therefore, existing technologies have the problem of failing to accurately identify the quality of pelvic floor muscle training, thus affecting the quality and efficiency of rehabilitation training.

[0004] The preceding description is intended to provide general background information and does not necessarily constitute prior art. Summary of the Invention

[0005] This application provides a pelvic floor muscle repair device, a pelvic floor muscle training instruction method, equipment, and medium, which can solve the problem that the existing technology cannot accurately identify the quality of pelvic floor muscle training, thus affecting the quality and efficiency of rehabilitation training.

[0006] In a first aspect, embodiments of this application provide a pelvic floor muscle repair device, the pelvic floor muscle repair device comprising:

[0007] The main body is used for insertion into or into the area corresponding to the pelvic floor muscles;

[0008] The first electrode and the second electrode are disposed on the main body and are used to electrically stimulate the pelvic floor muscles and to detect the electromyographic signal on the surface of the pelvic floor muscles after they have been electrically stimulated.

[0009] The third electrode is attached to the muscles surrounding the pelvic floor muscles during use to detect electromyographic signals of the muscles surrounding the pelvic floor muscles.

[0010] The posture detection unit is used to detect the posture timing signal after changes in torso movement; and

[0011] The controller, electrically connected to the first electrode, the second electrode, the third electrode, and the posture detection unit, is used to extract features from the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture temporal signals at the same time sequence, and generate a fused feature vector; it is also used to determine the pelvic floor muscle training evaluation result based on the fused feature vector and generate a training feedback instruction based on the pelvic floor muscle training evaluation result; the training feedback instruction is used to instruct the user to perform training.

[0012] Furthermore, in some embodiments of this application, the pelvic floor muscle repair device further includes a communication unit; the communication unit is electrically connected to the controller and is used to receive the training feedback instruction and send the training feedback instruction to the user terminal; the user terminal is used to generate an action reminder message according to the training feedback instruction; the action reminder message includes at least one of a prompt sound and a prompt image.

[0013] Furthermore, in some embodiments of this application, the pelvic floor muscle repair device further includes an output unit;

[0014] The output unit is electrically connected to the controller and is used to receive the training feedback instruction and instruct the user to perform training according to the training feedback instruction.

[0015] Furthermore, in some embodiments of this application, the pelvic floor muscle training assessment results include effectiveness assessment results and compensatory interference assessment results.

[0016] Furthermore, in some embodiments of this application, the controller is specifically configured to: output a pelvic floor muscle training evaluation result indicating proper movement when the effectiveness evaluation result is greater than or equal to a first threshold and the compensation interference evaluation result is less than or equal to a second threshold; output a pelvic floor muscle training evaluation result indicating excessive compensation by the muscles surrounding the pelvic floor muscles when the effectiveness evaluation result is greater than or equal to the first threshold and the compensation interference evaluation result is greater than the second threshold; output a pelvic floor muscle training evaluation result indicating insufficient pelvic floor muscle contraction strength when the effectiveness evaluation result is less than the third threshold and the compensation interference evaluation result is greater than the second threshold; and output a pelvic floor muscle training evaluation result indicating ineffective pelvic floor muscle contraction accompanied by significant compensation by the muscles surrounding the pelvic floor muscles when the effectiveness evaluation result is less than the third threshold and the compensation interference evaluation result is greater than the second threshold.

[0017] Secondly, embodiments of this application provide a method for instructing pelvic floor muscle training, including:

[0018] Simultaneously acquire electromyographic signals of the pelvic floor muscles, electromyographic signals of the muscles surrounding the pelvic floor muscles, and posture and temporal signals of the trunk.

[0019] Feature extraction is performed on the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture temporal signals at the same time sequence to generate a fused feature vector;

[0020] The fused feature vector determines the pelvic floor muscle training evaluation result through a pre-trained evaluation model and generates training feedback instructions based on the pelvic floor muscle training evaluation result; the training feedback instructions are used to instruct the user to perform training.

[0021] Furthermore, in some embodiments of this application, the step of extracting features from the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture temporal signals in the same time sequence to generate a fused feature vector includes:

[0022] The first feature extraction network is used to extract features from the electromyographic signals on the surface of the pelvic floor muscles to generate a first electromyographic feature vector;

[0023] A second feature extraction network is used to extract features from the electromyographic signals of the muscles surrounding the pelvic floor muscles, generating a second electromyographic feature vector.

[0024] A temporal coding network is used to extract features from the attitude temporal signal to generate an attitude feature vector.

[0025] The first electromyographic feature vector, the second electromyographic feature vector, and the posture feature vector are concatenated to form the fused feature vector.

[0026] Furthermore, in some embodiments of this application, the pelvic floor muscle training assessment results include effectiveness assessment results and compensatory interference assessment results.

[0027] Furthermore, in some embodiments of this application, the method further includes:

[0028] Based on the combined relationship between the effectiveness evaluation results and the compensation interference evaluation results, a training feedback instruction is generated;

[0029] An action reminder message is generated based on the training feedback instructions.

[0030] Furthermore, in some embodiments of this application, generating training feedback instructions based on the combined relationship between the effectiveness evaluation results and the compensation interference evaluation results includes:

[0031] When the effectiveness assessment result is greater than or equal to the first threshold and the compensation interference assessment result is less than or equal to the second threshold, an action specification instruction is output.

[0032] When the effectiveness assessment result is greater than or equal to the first threshold, and the compensation interference assessment result is greater than the second threshold, a compensation warning instruction is output.

[0033] When the effectiveness assessment result is less than the third threshold and the compensation interference assessment result is less than or equal to the second threshold, an intensity increase command is output.

[0034] When the effectiveness evaluation result is less than the third threshold and the compensation interference evaluation result is greater than the second threshold, an action abnormality interruption command is output.

[0035] Furthermore, in some embodiments of this application, the method further includes:

[0036] Real-time monitoring of the coordination coefficient between the electromyographic signals on the surface of the pelvic floor muscles and the electromyographic signals of the muscles surrounding the pelvic floor muscles;

[0037] When the coordination coefficient is detected to exceed the preset range, the compensation interference assessment result is dynamically corrected.

[0038] Thirdly, embodiments of this application provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the pelvic floor muscle training instruction method as described in the second aspect.

[0039] Fourthly, embodiments of this application provide a storage medium storing a computer program capable of being loaded by a processor and executing the pelvic floor muscle training instruction method as described in the second aspect.

[0040] This application provides a pelvic floor muscle repair device, a pelvic floor muscle training instruction method, equipment, and medium. The pelvic floor muscle repair device, through a specific electrode layout (a first electrode and a second electrode are set on the main body, and a third electrode is attached to the muscles surrounding the pelvic floor muscles during use) and a posture detection unit fixed to the trunk, achieves synchronous acquisition of multimodal physiological signals. It then acquires electromyographic signals from the pelvic floor muscle surface and surrounding muscles through the first, second, and third electrodes, and obtains posture temporal signals by capturing dynamic changes in the trunk's posture through the posture detection unit. Finally, it extracts features from the pelvic floor muscle surface electromyographic signals, surrounding muscles electromyographic signals, and posture temporal signals at the same temporal sequence and generates a fused feature vector. It is also used to determine the pelvic floor muscle training evaluation result based on the fused feature vector and generate training feedback instructions based on the evaluation result, thereby achieving a quantitative evaluation of the pelvic floor muscle training effect and guiding the user's training process based on the evaluation result. As can be seen, this application, by simultaneously collecting electromyographic signals from the surface of the pelvic floor muscles, the electromyographic signals from the muscles surrounding the pelvic floor muscles, and postural temporal signals, and combining them with a deep learning model, fully utilizes the complementarity of multiple physiological signals to overcome the limitations of single-signal evaluation. This improves the accuracy and comprehensiveness of pelvic floor muscle training assessment, provides strong support for optimizing training programs and improving rehabilitation quality, and effectively solves the problem that existing technologies cannot accurately identify the quality of pelvic floor muscle training, thus affecting the quality and efficiency of rehabilitation training. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a schematic diagram of the main body of the pelvic floor muscle repair device provided in the embodiments of this application;

[0043] Figure 2 This is a schematic diagram of the pelvic floor muscle repair device provided in the embodiments of this application;

[0044] Figure 3 This is another structural schematic diagram of the pelvic floor muscle repair device provided in the embodiments of this application;

[0045] Figure 4 This is a flowchart illustrating the pelvic floor muscle training instruction method provided in the embodiments of this application;

[0046] Figure 5 This is another schematic flowchart of the pelvic floor muscle training instruction method provided in the embodiments of this application;

[0047] Figure 6 This is another flowchart illustrating the pelvic floor muscle training instruction method provided in the embodiments of this application;

[0048] Figure 7 This is a schematic diagram of the pelvic floor muscle training and assessment device provided in the embodiments of this application;

[0049] Figure 8 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0050] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of systems and methods consistent with those detailed in the appended claims or with some aspects of this application.

[0051] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover descriptions such as non-exclusive inclusion, so that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, components, features, and elements with the same names in different embodiments of this application may have the same meaning or different meanings, the specific meaning of which must be determined by its interpretation in that specific embodiment or further in conjunction with the context of that specific embodiment.

[0052] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0053] In the following description, the use of suffixes such as "module," "part," or "unit" to denote elements is solely for the purpose of illustrative purposes and has no specific meaning in itself. Therefore, "module," "part," or "unit" may be used interchangeably.

[0054] To address the aforementioned technical problems and overcome the shortcomings of existing technologies, this application provides a pelvic floor muscle repair device, a pelvic floor muscle training instruction method, apparatus, equipment, and medium, which can quantitatively assess the effectiveness of pelvic floor muscle training and compensatory interference, and guide training based on the assessment results, thereby improving the quality and efficiency of pelvic floor muscle rehabilitation training.

[0055] like Figure 1 , 2 As shown, Figure 1 This is a schematic diagram of the main body 101 in a pelvic floor muscle repair device according to one embodiment. Figure 2 This is a schematic diagram of the structure of a pelvic floor muscle repair device in one embodiment. The pelvic floor muscle repair device in this application embodiment may include: a main body 101, a first electrode 102, a second electrode 103, a third electrode 104, a posture detection unit 105, and a controller 106.

[0056] The main body 101 is used to insert or place into the area corresponding to the pelvic floor muscles, which may refer to the vagina or rectum.

[0057] It should be noted that, based on the position and method of action of the main body 101, pelvic floor muscle repair devices can be mainly divided into two product forms: in-body and non-in-body. This application embodiment uses the in-body type as an example for explanation. In the in-body type, the main body 101 is inserted into the corresponding part of the pelvic floor muscles during use. It is worth noting that the solution of this application also applies to the non-in-body type, and this application does not limit it in this regard. For non-in-body pelvic floor muscle repair devices, the main body of this type of product is placed on the pelvic floor muscles. The specific product form can be a pad, a cushion, or other non-in-body product.

[0058] The main body 101 may be made of one of the following materials, including but not limited to medical-grade ABS plastic, medical-grade polycarbonate (PC), and medical-grade silicone composite materials. Medical-grade ABS plastic has good biocompatibility, is non-allergenic, and can come into direct contact with human skin; it also has high mechanical strength, can stably fix various functional modules, and has a smooth surface that is easy to clean and disinfect, making it suitable for the frequent cleaning needs in medical settings. Medical-grade polycarbonate (PC) has good light transmittance, excellent impact resistance, and a wide range of high and low temperature resistance, making it adaptable to temperature changes in different clinical environments, and also meets medical-grade biocompatibility standards.

[0059] like Figure 1 As shown, the main body 101 is rod-shaped, tapering at both ends and thickening in the middle, gradually tapering at the front end to form a smooth, blunt or ellipsoidal head. This ensures smooth insertion and allows easy passage through narrow passages, minimizing user discomfort and potential tissue damage. The middle section is the thickest part of the main body, forming the functional main area. Its increased diameter ensures that the electrodes on its surface can generate sufficient and uniform contact area with the target area of ​​the pelvic floor muscles, providing a physical basis for effective electrical stimulation and signal acquisition. The rear end gradually tapers compared to the middle section and is usually designed with an anti-slip structure or an interface for connecting to the handle, facilitating user grip or device fixation.

[0060] The first electrode 102 and the second electrode 103 are disposed on the main body 101 for electrically stimulating the pelvic floor muscles and for detecting the electromyographic signals on the surface of the pelvic floor muscles after electrical stimulation. For example, the first electrode 102 and the second electrode 103 are respectively disposed on opposite sides of the outer peripheral surface of the main body 101. The first electrode 102 and the second electrode 103 can be flexible electrodes or metal electrodes coated with conductive silicone, etc., and this application embodiment does not limit this.

[0061] One of the first electrode 102 and the second electrode 103 is a positive electrode, and the other is a negative electrode. Current will flow between the first electrode 102 and the second electrode 103 on the main body 101 to form a local, closed current loop, so as to electrically stimulate the pelvic floor muscles. The number of electrodes can be increased according to clinical needs (such as multi-channel acquisition, improving signal redundancy) to adapt to the rehabilitation training needs in different scenarios.

[0062] The third electrode 104 is attached to the muscles surrounding the pelvic floor muscles during use to detect the electromyographic signals of the muscles surrounding the pelvic floor muscles, which include the rectus abdominis or the lower abdomen.

[0063] The rectus abdominis muscle is located on both sides of the midline of the abdomen, extending from the pubic symphysis to the xiphoid process. It is the main target for collecting electromyographic signals from the muscles surrounding the pelvic floor muscles.

[0064] The lower abdomen refers to the lower side of the abdomen, with the following specific boundaries: the upper boundary is the level of the umbilicus; the lower boundary is the inguinal ligament and the upper edge of the pubic symphysis; and the lateral boundaries are the anterior abdominal line (a virtual line extending downward from the midclavicular line to the inguinal ligament).

[0065] After the pelvic floor muscles are electrically stimulated by the first electrode 102 and the second electrode 103, they will first cause the deep transverse abdominis, multifidus, and diaphragm to contract synergistically to complete the function of core stability; secondly, they will affect the gluteal and adductor muscles through biomechanical effects. The involvement of the rectus abdominis or lower abdomen requires more precise control to avoid compensation.

[0066] The posture detection unit 105 is used to detect the posture timing signal after the torso movement changes;

[0067] The attitude detection unit 105 is directly embedded in or fixed to the end or a specific part of the main body 101 inserted into the body.

[0068] Alternatively, in another embodiment, the posture detection unit 105 is an attachable module connected to the controller 106 via a flexible medical-grade wire. It is then fixed to the skin surface of the muscles surrounding the pelvic floor muscles, such as specific locations on the lower abdomen, buttocks, or groin, using medical adhesive or similar methods.

[0069] The controller 106 is electrically connected to the first electrode 102, the second electrode 103, the third electrode 104, and the posture detection unit 105. It is used to extract features from the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture temporal signals at the same time sequence, and generate a fused feature vector. It is also used to determine the pelvic floor muscle training evaluation result based on the fused feature vector and generate training feedback instructions based on the pelvic floor muscle training evaluation result. The training feedback instructions are used to instruct the user to perform training. The pelvic floor muscle training evaluation result includes an effectiveness evaluation result and a compensation interference evaluation result.

[0070] This application achieves simultaneous acquisition of multimodal physiological signals through a specific electrode layout (first electrode 102 and second electrode 103 are disposed on the main body 101, and third electrode 104 is attached to the muscles surrounding the pelvic floor muscles) and a posture detection unit 105 fixed to the trunk. Then, it acquires electromyographic signals of the pelvic floor muscle surface and the muscles surrounding the pelvic floor muscles through the first electrode 102, second electrode 103 and third electrode 104, and obtains posture temporal signals by capturing the dynamic changes in the posture of the trunk through the posture detection unit 105. Finally, it extracts features from the pelvic floor muscle surface electromyographic signals, the muscles surrounding the pelvic floor muscles and the posture temporal signals of the same time sequence and generates a fused feature vector. It is also used to determine the pelvic floor muscle training evaluation result based on the fused feature vector and generate training feedback instructions based on the pelvic floor muscle training evaluation result, thereby realizing a quantitative evaluation of the pelvic floor muscle training effect, and guiding the user's training process based on the evaluation result.

[0071] The two methods, inserted and non-inserted, each have their own characteristics in terms of applicable scenarios, technical implementation, and user experience, together forming a complete pelvic floor muscle rehabilitation treatment system. The following is a detailed explanation:

[0072] In the insertive position, after the main body 101 is inserted into the vagina, the first electrode 101 and the second electrode 102 directly contact the vaginal wall area corresponding to the pelvic floor muscles. Muscle contraction is triggered by close-range electrical stimulation (intensity adjustable range 5-100mA). Simultaneously, the same electrode synchronously collects surface electromyographic signals (sEMG_pfd) of the pelvic floor muscles, achieving integrated stimulation and monitoring. It is suitable for patients with moderate to severe muscle weakness who require direct stimulation of the deep fibers of the pelvic floor muscles.

[0073] In the non-invasive type, the first electrode 102 and the second electrode 103 are made of flexible silicone patches and are attached to the skin around the vaginal opening (such as the perineal body and the outer side of the labia majora).

[0074] The first electrode 102 indirectly activates the pelvic floor muscles through transcutaneous electrical stimulation (TES). Although the stimulation depth is relatively shallow, it can cover a wider range of muscle groups (such as the superficial transverse perineal muscle and the bulbospongiosus muscle); it simultaneously acquires electromyographic signals from the superficial parts of the pelvic floor muscles. It is suitable for sensitive individuals or those who are psychologically averse to internal probes, thus avoiding invasive procedures.

[0075] Furthermore, in some embodiments, the first electrode 102 and the second electrode 103 are distributedly encapsulated on the main body 101 of the pelvic floor muscle repair device for direct insertion or attachment to the corresponding part of the pelvic floor muscle, such as the perineum or vagina.

[0076] The first electrode 102 and the second electrode 103 are arranged symmetrically in the pelvic floor muscle region to ensure the balance and accuracy of signal acquisition.

[0077] The sEMG_pfd signals of the pelvic floor muscles acquired by the first electrode 102 and the second electrode 103 can directly reflect the contractile function of the pelvic floor muscles themselves, including:

[0078] Contraction intensity: Signal amplitude (such as root mean square value RMS) quantifies the magnitude of pelvic floor muscle contraction force; a high score indicates strong contraction.

[0079] Contraction endurance: The duration of the signal reflects the degree of fatigue in the muscle to maintain contraction;

[0080] Activation patterns: Signal timing characteristics (such as initiation speed and rhythmicity) reveal the efficiency of neuromuscular control.

[0081] Electromyographic signals on the surface of the pelvic floor muscles, as the direct output of the target muscle group, are a core indicator for evaluating training effectiveness.

[0082] High signal quality of electromyographic signals on the surface of the pelvic floor muscles indicates that the pelvic floor muscles can independently and efficiently complete the contraction task. Weakened signals suggest insufficient muscle strength or nerve activation defects, requiring targeted training. Disordered signals reflect poor control ability, requiring improvement in force coordination.

[0083] Furthermore, in some embodiments, the third electrode 104 is disposed in a symmetrical area of ​​the rectus abdominis muscle or the lower abdomen. In order to ensure the quality of the electromyographic signal of the muscles around the pelvic floor muscles, the third electrode 104 is attached to a location that avoids bony structures and areas with excessive fat.

[0084] Electromyographic signals (sEMG_abd) of the muscles surrounding the pelvic floor muscles acquired by the third electrode 104 can reflect the involvement patterns of muscles such as the rectus abdominis and are used for identification:

[0085] Compensatory behavior: Abnormally enhanced signals (compared to baseline) indicate overactivation of abdominal muscles, which substitute for pelvic floor muscle exertion;

[0086] Synergistic contraction: The temporal correlation with pelvic floor muscle signals (such as the Pearson coefficient) reflects the coordination between muscles;

[0087] Interference intensity: The severity of compensation quantified by the signal amplitude and duration.

[0088] Electromyographic signals from the muscles surrounding the pelvic floor are a key input for assessing compensatory interference, revealing “false efficiency” training, i.e. seemingly effective movements that actually rely on non-target muscle groups.

[0089] A persistently high electromyographic signal in the muscles surrounding the pelvic floor muscles suggests that the user has an incorrect force exertion pattern and needs to learn isolated contraction through biofeedback training; a lack of signal indicates a low risk of compensation.

[0090] Furthermore, in some embodiments, the posture detection unit 105 is positioned at the waist and indirectly reflects the compensatory behavior of non-target muscle groups such as the muscles around the pelvic floor muscles by detecting abnormal movement patterns of the pelvis (such as excessive anterior tilting or rotation of the pelvis during training).

[0091] The posture timing signal acquired by the posture detection unit 105 can reflect the overall motion state of the trunk and pelvis, including:

[0092] Movement standardization: Pelvic orientation (Euler angle) and displacement trajectory descriptions to determine whether the movement conforms to physiological standards;

[0093] Stability: The acceleration / angular velocity variance index quantifies the degree of motion jitter or loss of control;

[0094] Compensatory postures: such as anterior pelvic tilt, rotation, and other patterns that are highly correlated with compensatory behaviors.

[0095] The posture detection unit 105 provides objective evidence of the quality of motion execution, compensating for the inability of electromyographic signals to capture macroscopic motion defects. For example, abnormal anterior pelvic tilt during pelvic floor muscle contraction is often accompanied by abdominal compensation.

[0096] Furthermore, in some embodiments, such as Figure 3 As shown, the pelvic floor muscle repair device may also include a filter unit 107;

[0097] The first electrode 102, the second electrode 103, and the third electrode 104 are connected to the controller 106 through the filter unit 107. The electromyographic signals of the pelvic floor muscle surface and the electromyographic signals of the muscles surrounding the pelvic floor muscle are removed by bandpass filters to remove out-of-band noise, and then the 50Hz power frequency interference is eliminated by notch filters. Finally, the clean electromyographic signals of the pelvic floor muscle surface and the electromyographic signals of the muscles surrounding the pelvic floor muscle are output to the controller 106 for feature extraction.

[0098] For example, the filter unit 107 adopts a pre-bandpass filter (20-450Hz) design, which is implemented by a multi-stage operational amplifier and an RC network. It is used to retain the effective frequency band of the electromyographic signals on the surface of the pelvic floor muscles and the electromyographic signals of the muscles around the pelvic floor muscles (the main energy is concentrated in 20-200Hz), while suppressing low-frequency motion artifacts and high-frequency noise.

[0099] The filter unit 107 can also be integrated into the controller 106.

[0100] Furthermore, in some embodiments of this application, when the pelvic floor muscle training assessment result is "the movement is correct", the training feedback instruction generated by the training feedback unit is "the training movement is correct, continue to maintain it";

[0101] When the pelvic floor muscle training assessment result indicates excessive compensation by the muscles surrounding the pelvic floor muscles, the training feedback unit generates the training feedback instruction "There is obvious compensation".

[0102] When the pelvic floor muscle training assessment result indicates insufficient pelvic floor muscle contraction strength, the training feedback instruction generated by the training feedback unit is "insufficient activation, possibly insufficient contraction";

[0103] When the pelvic floor muscle training assessment result shows that the pelvic floor muscle contraction is ineffective and accompanied by significant compensation from the surrounding muscles, the training feedback unit generates the training feedback instruction as "abnormal movement, inefficient compensation exists".

[0104] Furthermore, in some embodiments, such as Figure 3 As shown, the pelvic floor muscle repair device may also include a communication unit 108;

[0105] The communication unit 108 is electrically connected to the controller 106 and is used to receive training feedback instructions and send the training feedback instructions to the user terminal.

[0106] User terminal 109 is used to generate action reminder messages based on training feedback instructions;

[0107] Action notification messages include at least one of the following: a notification sound or a notification video.

[0108] The training feedback instruction generated by the controller 106 is "Training action is correct, keep it up". The communication unit 108 encapsulates the instruction data into JSON format and sends it to the user terminal 109 via Bluetooth connection. The user terminal 109 generates a praising voice, such as "Great job! Keep up the current posture and contraction method", or an image containing praise or commendation to prompt the user.

[0109] The training feedback instruction generated by the controller 106 is "significant compensation exists". The communication unit 108 encapsulates the instruction data into JSON format and sends it to the user terminal 109 via Bluetooth connection. The user terminal 109 generates prompting voice messages based on these messages, such as "Please note that the abdominal muscles are over-engaged, please reduce the force", or warning images, such as displaying "a diagram of abdominal and pelvic floor muscle separation contraction" or "animation comparing incorrect movements", to visually guide the user to adjust the movements.

[0110] The training feedback instruction generated by the controller 106 is "Insufficient activation, possibly insufficient contraction". The communication unit 108 encapsulates the instruction data into JSON format and sends it to the user terminal 109 via Bluetooth connection. The user terminal 109 generates encouraging voice messages based on these messages, such as "Please increase the contraction force, exert force like interrupting the flow of urine", with a positive tone to motivate the user, or encourages the user with encouraging image resources, such as displaying "progress bar animation" or "star reward icon", accompanied by the text "Give it your all!"

[0111] The training feedback command generated by the controller 106 is "abnormal movement, inefficient compensation exists". The communication unit 108 encapsulates the command data into JSON format and sends it to the user terminal 109 via Bluetooth connection. The user terminal 109 generates an emergency prompt voice based on it, such as "Error movement detected! Please pause training", or "Common error movement diagram" or "Step-by-step correction video", highlighting the contrast between abdominal compensation and pelvic floor muscle weakness.

[0112] Furthermore, in some embodiments, such as Figure 3 As shown, the pelvic floor muscle repair device may also include an output unit 110;

[0113] The output unit 110 is electrically connected to the controller 106 and is used to receive training feedback instructions and instruct the user to perform training according to the training feedback instructions.

[0114] The training feedback instruction generated by the controller 106 is "Training action is correct, continue to maintain". The output unit 110 directly receives the training feedback instruction (through an internal electrical interface, such as SPI or I2C bus) and generates a praising voice, such as "Great job! Keep up the current posture and contraction method", or an image containing praise or commendation to prompt the user.

[0115] The training feedback command generated by the controller 106 is "significant compensation exists". The output unit 110 directly receives the training feedback command (through an internal electrical interface, such as SPI or I2C bus) and generates prompting voice, such as "Please note that the abdominal muscles are over-engaged, please reduce the force", or warning images, such as displaying "a diagram of abdominal and pelvic floor muscle separation contraction" or "animation comparing incorrect movements", to visually guide the user to adjust the movement.

[0116] The training feedback instruction generated by the controller 106 is "Insufficient activation, possible inadequate contraction". The output unit 110 directly receives the training feedback instruction (through an internal electrical interface, such as SPI or I2C bus) and generates encouraging voice messages, such as "Please increase the contraction force, exert force like interrupting the flow of urine", with a positive tone to motivate the user, or encouraging image resources, such as displaying "progress bar animation" or "star reward icon", accompanied by the text "Give it your all!" to motivate the user.

[0117] The training feedback command generated by the controller 106 is "abnormal action, inefficient compensation exists". The output unit 110 directly receives the training feedback command (through an internal electrical interface, such as SPI or I2C bus) and generates an emergency prompt voice based on it, such as "Error action detected! Please pause training", or "Common error action diagram" or "Step-by-step correction video", highlighting the contrast between abdominal compensation and pelvic floor muscle weakness.

[0118] Furthermore, in some embodiments, the controller 106 is configured to output a pelvic floor muscle training evaluation result with proper movement when the effectiveness evaluation result is greater than or equal to a first threshold and the compensation interference evaluation result is less than or equal to a second threshold. This indicates that the user can accurately isolate and activate the pelvic floor muscles, with sufficient contraction strength and no interference from non-target muscle groups. The training movement fully conforms to physiological norms, which represents an ideal training state and indicates that the user has mastered the correct force exertion technique and is making good progress in rehabilitation.

[0119] When the effectiveness assessment result is greater than or equal to the first threshold and the compensation interference assessment result is greater than the second threshold, the output is a pelvic floor muscle training assessment result with a compensation warning of excessive compensation by the muscles surrounding the pelvic floor muscles. This indicates that although the pelvic floor muscle contraction is effective, the movement is completed by excessive compensation by the muscles surrounding the pelvic floor muscles.

[0120] When the effectiveness assessment result is less than the third threshold and the compensation interference assessment result is less than or equal to the second threshold, the output is a pelvic floor muscle training assessment result indicating insufficient pelvic floor muscle contraction strength. This indicates that the user reports weak pelvic floor muscle contraction ability, but the force exertion pattern is correct and the control ability is insufficient, requiring a focus on strengthening intensity training.

[0121] When the effectiveness assessment result is less than the third threshold and the compensation interference assessment result is greater than the second threshold, the output is a pelvic floor muscle training assessment result of ineffective pelvic floor muscle contraction and significant compensation by the surrounding muscles of the pelvic floor muscles. This indicates that the user's pelvic floor muscles are not activated enough and are accompanied by significant compensation. The user lacks both correct force exertion techniques and has insufficient muscle strength. Training should be stopped immediately and the user should relearn the basic movements.

[0122] The controller 106 can comprehensively judge the effect of pelvic floor muscle training by combining the effectiveness evaluation results (y) and the compensation interference evaluation results (c). This is because it is based on multimodal physiological signal fusion and deep learning quantitative evaluation, which reveals the physiological nature of the training movement from two dimensions: the activation quality of the target muscle group and the degree of interference of the non-target muscle group.

[0123] The effectiveness assessment result (y) directly reflects the efficacy of independent contraction of the pelvic floor muscles, and is quantified by the temporal characteristics (such as root mean square value RMS, integrated electromyography value iEMG) and frequency domain characteristics (median frequency MF) of the pelvic floor muscle surface electromyography signal (sEMG_pfd). A high score (y≥80) indicates that the pelvic floor muscles can generate sufficient and sustained contractile force, which meets the intensity requirements of rehabilitation training.

[0124] The compensation interference assessment results (c) quantify the abnormal involvement of non-target muscle groups (i.e., the electromyographic signals of muscles surrounding the pelvic floor muscles) through synergistic analysis of the electromyographic signals (sEMG_abd) of muscles surrounding the pelvic floor muscles and pelvic floor muscle signals (such as Pearson correlation coefficient and mutual information content), as well as abnormal movement patterns identified by IMU posture signals. Low scores (c≤20) indicate that muscles such as the rectus abdominis are not overcompensating, and the training movement is well isolated.

[0125] Furthermore, in some embodiments, a portion of the main body 101 is inserted into the vagina or rectum, while another portion of the main body 101 is located outside the vagina or rectum. This portion may employ a liquid crystal display (LCD) or OLED display screen to display a graphical user interface (GUI), thereby displaying corresponding information to instruct the user to perform training based on training feedback instructions. A miniature speaker or buzzer may also be built-in to play voice prompts, sound effects, or alarm sounds, thereby instructing the user to perform training through prompts based on training feedback instructions. A linear motor may also be built-in to generate vibrations of different modes, thereby instructing the user to perform training through vibrations of different modes based on training feedback instructions.

[0126] Please see Figure 4 , Figure 4 This is a flowchart illustrating a pelvic floor muscle training instruction method according to an embodiment of this application. This embodiment mainly uses the application of this pelvic floor muscle training instruction method to a pelvic floor muscle repair device as an example for illustration. The pelvic floor muscle training instruction method provided in an embodiment of this application may specifically include the following steps:

[0127] S1. Simultaneously acquire electromyographic signals of the pelvic floor muscles, electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture and temporal signals of the trunk.

[0128] Specifically, for step S1, during pelvic floor muscle training, the surface electromyographic (EMG) signals of the pelvic floor muscles directly reflect their contraction and are a key indicator for evaluating training effectiveness. The EMG signals of muscles surrounding the pelvic floor muscles can be used to identify compensatory behaviors of non-target muscle groups, as other muscles such as the rectus abdominis may involuntarily participate when the pelvic floor muscles contract, thus affecting training effectiveness. Postural timing signals are acquired via an inertial measurement unit (IMU), containing information such as acceleration and angular velocity, reflecting the smoothness of movement and postural stability during training. By simultaneously acquiring these multi-source physiological signals, key information during training can be comprehensively captured, providing a rich data foundation for subsequent accurate evaluation. For example, when acquiring surface EMG signals of the pelvic floor and rectus abdominis muscles, it is necessary to rationally select electrode placement locations, avoiding bony structures and areas with excessive fat, and ensuring a sufficiently high sampling rate (e.g., ≥1000Hz) to ensure signal quality and preservation of detailed information.

[0129] S2. Extract features from the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture temporal signals at the same time sequence, and generate a fused feature vector;

[0130] Specifically, in step S2, the collected surface electromyographic (EMG) signals of the pelvic floor muscles, the EMG signals of the muscles surrounding the pelvic floor muscles, and the postural timing signals typically contain a large amount of data and noise, requiring feature extraction to extract representative and discriminative features. For the surface EMG signals of the pelvic floor muscles and rectus abdominis muscles, time-domain, frequency-domain, or time-frequency domain analysis methods can be used to extract features, such as root mean square (RMS) values, waveform length, and frequency center. These features can reflect the activity intensity and contraction characteristics of the muscle groups. Postural timing signals can be used to extract features through time-series analysis methods, such as the rate of change of acceleration and angular velocity, and stability indices, to reflect the continuity of movement and the stability of posture. After feature extraction, features from different modalities are fused to form a fused feature vector, enabling data from different sources to be integrated and analyzed in a unified feature space, providing a more comprehensive and effective input for subsequent evaluation models.

[0131] S3. The fusion feature vector determines the pelvic floor muscle training evaluation results through a pre-trained evaluation model and generates training feedback instructions based on the pelvic floor muscle training evaluation results. The training feedback instructions are used to instruct the user to perform training. The pelvic floor muscle training evaluation results include effectiveness evaluation results and compensation interference evaluation results.

[0132] Specifically, in step S3, the constructed evaluation model is first trained. The trained model can accurately determine the effectiveness of pelvic floor muscle training and the presence of compensatory interference based on the input fused feature vector. The evaluation model can be constructed using various machine learning or deep learning algorithms, such as multilayer perceptrons, convolutional neural networks, and recurrent neural networks. During model training, by using labeled training data, such as effectiveness evaluation result labels and compensation score labels, the model can learn the mapping relationship between different features and training effects. When the fused feature vector is input, the model can output effectiveness evaluation results and compensation interference evaluation results, providing quantitative indicators for the effectiveness of training movements and the presence of compensatory interference, respectively. This provides trainers and rehabilitation therapists with intuitive evaluation results, helping them adjust training plans in a timely manner and providing personalized feedback suggestions.

[0133] As can be seen, this embodiment comprehensively acquires key physiological information during the training process by simultaneously collecting electromyographic signals from the surface of the pelvic floor muscles, electromyographic signals from the muscles surrounding the pelvic floor muscles, and postural temporal signals. Then, feature extraction is performed on the multi-source physiological signals to generate a fused feature vector, effectively integrating the features of different modal signals, enabling the subsequent evaluation model to more accurately identify the training effect. Finally, the fused feature vector is input into the pre-trained evaluation model to obtain a quantitative evaluation result that includes the effectiveness evaluation result and the compensation interference evaluation result, achieving a precise and comprehensive evaluation of the pelvic floor muscle training effect. This provides strong technical support for optimizing pelvic floor muscle rehabilitation training and helps improve the targeting and rehabilitation effect of training.

[0134] Furthermore, such as Figure 5 As shown, in some embodiments, step S1, "synchronously acquiring electromyographic signals of the pelvic floor muscles, electromyographic signals of the muscles surrounding the pelvic floor muscles, and posture timing signals of the trunk," may specifically include:

[0135] S11. Acquire electromyographic signals on the surface of the pelvic floor muscles of the target object through the first electrode;

[0136] Specifically, for step S11, the electromyographic signals on the surface of the pelvic floor muscles directly reflect the contraction of the pelvic floor muscles and are one of the key indicators for evaluating training effectiveness. During the data acquisition process, it is necessary to rationally select the electrode placement location, avoiding bony structures and areas with excessive fat, to ensure signal quality and preservation of detailed information. For example, electrodes can be placed in areas where the pelvic floor muscles are relatively concentrated, such as the perineum. At the same time, attention should be paid to electrode fixation and good contact to reduce interference such as motion artifacts and electrode slippage.

[0137] S12. Electromyographic signals of the muscles surrounding the pelvic floor muscles are collected in the rectus abdominis region of the target subject using the second electrode;

[0138] Specifically, in step S12, the electromyographic (EMG) signals of the muscles surrounding the pelvic floor muscles can be used to identify compensatory behaviors of non-target muscle groups. This is because during pelvic floor muscle training, other muscles such as the rectus abdominis may involuntarily participate, thus affecting the training effect. Similar to the EMG signals on the surface of the pelvic floor muscles, the acquisition of EMG signals from the muscles surrounding the pelvic floor muscles also requires appropriate electrode placement. Electrodes are typically placed in a symmetrical area in the middle or lower part of the rectus abdominis, avoiding interfering areas such as the navel, and ensuring a sufficiently high sampling rate, such as ≥1000Hz, to capture detailed EMG information. Simultaneously, a pre-bandpass filter (20-450Hz) can be used, combined with a notch filter to suppress 50Hz power frequency interference, to improve the signal-to-noise ratio.

[0139] S13. Acceleration and angular velocity data are collected in the pelvic region of the target object through the inertial measurement unit to generate attitude timing signals;

[0140] Specifically, for step S13, the attitude timing signal reflects the smoothness and stability of movements during training, providing crucial information for evaluating the standardization and continuity of training movements. An inertial measurement unit (IMU) typically includes an accelerometer and a gyroscope, capable of measuring acceleration and angular velocity respectively. When placing the IMU in the pelvic region, attention must be paid to the stability and comfort of the fixed position to avoid signal distortion caused by device displacement. The acquired acceleration and angular velocity data can be processed through filtering, data fusion, etc., to generate an attitude timing signal for subsequent motion analysis.

[0141] S14. Perform time-series synchronization processing on the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture timing signals.

[0142] Specifically, for step S14, since the acquisition devices and sampling rates of different signals may vary, timing synchronization is required to ensure the consistency of each signal on the time axis, facilitating subsequent comprehensive analysis and evaluation. Timing synchronization can be achieved through hardware or software synchronization. Hardware synchronization can be achieved during the data acquisition phase using a synchronized clock or trigger signal; software synchronization can be performed after data acquisition using algorithms such as timestamp alignment and interpolation. For example, using the electromyographic signals on the surface of the pelvic floor muscles as a benchmark, the electromyographic signals and posture timing signals of the muscles surrounding the pelvic floor muscles can be time-aligned to ensure that the data of each signal at the same time point can accurately correspond, providing a reliable data foundation for subsequent feature extraction and fusion.

[0143] This embodiment provides different methods for acquiring physiological signals to ensure that the acquired signals are of high quality and reliability. At the same time, the time-synchronized processing provides an accurate data foundation for subsequent feature extraction and fusion, which helps to improve the accuracy and effectiveness of pelvic floor muscle training assessment.

[0144] Furthermore, such as Figure 6 As shown, in some embodiments, step S2, "extracting features from the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture temporal signals at the same time sequence, and generating a fused feature vector," may specifically include:

[0145] S21. The first feature extraction network is used to extract features from the electromyographic signals on the surface of the pelvic floor muscles to generate the first electromyographic feature vector;

[0146] Specifically, in step S21, the electromyographic (EMG) signal on the surface of the pelvic floor muscles contains rich time-domain, frequency-domain, and time-frequency-domain information. This information reflects key features such as the contraction intensity, duration, and rhythm of the pelvic floor muscles. The first feature extraction network can employ various deep learning architectures, such as a one-dimensional convolutional neural network (1D-CNN), which uses convolutional kernels to slide along the time axis to automatically learn local features, such as the waveform pattern and frequency components of the EMG signal. Through multi-layer convolution and pooling operations, deep-level features in the signal can be gradually extracted, ultimately generating the first EMG feature vector. This feature vector condenses the key information of the EMG signal on the surface of the pelvic floor muscles, providing a foundation for subsequent comprehensive evaluation.

[0147] S22. A second feature extraction network is used to extract features from the electromyographic signals of the muscles surrounding the pelvic floor muscles, generating a second electromyographic feature vector;

[0148] Specifically, in step S22, the electromyographic signals of the muscles surrounding the pelvic floor muscles also contain rich feature information, and their activity can help identify compensatory behaviors of non-target muscle groups. The architecture of the second feature extraction network can be similar to that of the first feature extraction network, using 1D-CNN or other network structures suitable for processing temporal signals. By extracting features from the electromyographic signals of the muscles surrounding the pelvic floor muscles, feature vectors reflecting the contraction intensity and temporal characteristics of the rectus abdominis can be obtained. These features are important for determining whether the rectus abdominis is overly involved in pelvic floor muscle training, because compensatory behaviors often disrupt training effects and affect the proper rehabilitation of the pelvic floor muscles.

[0149] S23. Use a time-series coding network to extract features from the attitude time-series signal and generate an attitude feature vector;

[0150] Specifically, for step S23, the posture temporal signal contains information on the continuity of movement and posture stability during training, which is crucial for evaluating the standardization of training movements. The temporal coding network can employ recurrent neural networks (RNNs) and their variants, such as Long Short-Term Memory (LSTM) networks or gated recurrent units (GRUs). Temporal coding networks can effectively process time-dependent sequence data. LSTMs, through their unique gating mechanism, can memorize key information from long-term sequences while forgetting irrelevant information, thereby extracting dynamic features from the posture temporal signal, such as the smoothness of movement, the rate of change of posture, and stability indices. The generated posture feature vector reflects the trainee's movement execution, providing a basis for judging whether the training movements are standardized.

[0151] S24. The first electromyographic feature vector, the second electromyographic feature vector, and the posture feature vector are concatenated to form a fused feature vector;

[0152] Specifically, in step S24, after extracting the feature vectors of the electromyographic signals on the surface of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the postural temporal signals, these feature vectors from different modalities need to be fused to comprehensively evaluate the training effect of the pelvic floor muscles. Concatenation is a common feature fusion method, which involves connecting the various feature vectors along a certain dimension (such as the channel dimension) to form a higher-dimensional fused feature vector. In this way, the fused feature vector includes both the contraction features of the pelvic floor muscles and the rectus abdominis muscles, as well as the motion features in the postural temporal signals, allowing the subsequent evaluation model to consider multiple aspects of information simultaneously, thus providing a more comprehensive and accurate evaluation of pelvic floor muscle training. For example, the fused feature vector can be input into a multi-task regression model to predict the evaluation results of training effectiveness and compensation interference, respectively.

[0153] This embodiment fully utilizes the feature extraction capabilities of deep learning networks for signals from different modalities, enabling the extraction of key information from each signal and achieving a comprehensive representation of multi-source physiological signals through fusion. The fused feature vector provides a comprehensive and highly representative input for the subsequent pelvic floor muscle training evaluation model, helping to improve the accuracy of the evaluation results and thus more effectively quantifying the effectiveness of pelvic floor muscle training and the degree of compensatory interference.

[0154] Furthermore, in some embodiments, step S2, "extracting features from the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture temporal signals at the same time sequence, and generating a fused feature vector," may specifically include:

[0155] S25. During the feature extraction process, the reconstructed root mean square curves corresponding to the electromyographic signals on the surface of the pelvic floor muscles and the electromyographic signals of the muscles surrounding the pelvic floor muscles are generated simultaneously.

[0156] Specifically, in step S25, while extracting features from the electromyographic (EMG) signals on the surface of the pelvic floor muscles and the EMG signals of the muscles surrounding the pelvic floor muscles, a reconstructed root mean square (RMS) curve is generated. The RMS value is an important indicator for measuring the intensity of EMG signals and can reflect the contraction strength and activity level of the muscle group. Generating the reconstructed RMS curve through the feature extraction network allows for a more intuitive evaluation of the model's effectiveness in extracting EMG signal features.

[0157] S26. Calculate the structural similarity loss between the reconstructed root mean square curve and the true root mean square. The structural similarity loss is used to constrain the training of the first feature extraction network and the second feature extraction network.

[0158] Specifically, for step S26, to improve the accuracy and robustness of feature extraction, a structural similarity loss (SSIM) is introduced. SSIM measures the similarity between the reconstructed root mean square curve (RMS_pred) and the true root mean square curve (RMS_true). This loss function ensures that the feature extraction network not only focuses on feature vector extraction during training but also guarantees that the extracted features can accurately reconstruct a root mean square curve similar to the real data, thereby enhancing the model's ability to recognize training rhythms and compensatory fluctuations.

[0159] This embodiment improves the accuracy and robustness of the feature extraction network in extracting electromyographic (EMG) signals from the pelvic floor muscles and surrounding muscles by simultaneously generating reconstructed root mean square (RMS) curves during feature extraction and calculating the structural similarity loss between the reconstructed and real RMS curves. This loss is then constrained.

[0160] Furthermore, in some embodiments, the training method for evaluating the model may specifically include:

[0161] A multi-task regression network is constructed. The input layer of the multi-task regression network is a shared feature extraction layer that receives fused feature vectors and generates shared embedded features. The output layer is a parallel two-branch regression layer that includes an effectiveness evaluation result branch and a compensation score branch.

[0162] Construct a training dataset, which includes sample fusion feature vectors and supervision signals. The supervision signals are sample validity labels and sample compensation labels corresponding to the sample fusion feature vectors, which are annotated manually.

[0163] After inputting the training dataset into the multi-task regression network, the corresponding joint loss function is calculated;

[0164] With the goal of minimizing the joint loss function, the network parameters of the multi-task regression network are updated to obtain the trained evaluation model;

[0165] Specifically, this embodiment provides a training method for the evaluation model, the specific process of which is as follows:

[0166] Before training, a multi-task regression network is first constructed. This network is designed to simultaneously predict the effectiveness evaluation results and compensation interference evaluation results of pelvic floor muscle training. The network's input layer receives a fused feature vector, which includes comprehensive features of surface electromyography (EMG) signals of the pelvic floor muscles, EMG signals of surrounding muscles, and postural temporal signals. A shared feature extraction layer extracts shared embedded features from the fused feature vector. These embedded features reflect key information during training, such as muscle contraction intensity and movement standardization. Parallel two-branch regression layers output the effectiveness evaluation results and compensation scores respectively, allowing the network to provide two important evaluation metrics simultaneously in a single forward propagation. Constructing the training dataset is fundamental to model training. The training dataset needs to contain sample fused feature vectors and corresponding supervision signals. Supervision signals are obtained through manual annotation, including sample effectiveness labels and sample compensation labels. Effectiveness labels reflect whether the training action achieved the expected effect, while compensation labels indicate whether there is excessive involvement of non-target muscle groups. To ensure the model's generalization ability, the training dataset should cover samples from different individuals, different training stages, and different levels of movement standardization to fully cover various possible training scenarios.

[0167] During training, the constructed training dataset is input into the multi-task regression network, which outputs corresponding predictions. The calculation of the joint loss function involves the difference between the outputs of the effectiveness evaluation branch and the compensation score branch and the true labels. The design of the joint loss function usually considers the losses of both branches. For example, a weighted summation method can be used to combine the mean squared error loss and the structural similarity loss to balance the training weights of the two tasks. By minimizing the joint loss function, the network can gradually learn how to accurately predict the effectiveness evaluation results and compensation scores. The core objective of model training is to minimize the joint loss function. To achieve this goal, optimization algorithms, such as stochastic gradient descent (SGD) or the Adam optimizer, are typically used to update the network parameters. In each iteration, the parameter values ​​are adjusted according to the gradient of the loss function with respect to the network parameters, so that the network's predictions gradually approach the true labels. After multiple iterations of training, the network parameters gradually converge, ultimately resulting in an evaluation model that can accurately assess the effectiveness of pelvic floor muscle training.

[0168] This embodiment, by designing a structure with a shared feature extraction layer and a parallel dual-branch regression layer, can simultaneously output the effectiveness evaluation results and compensation interference evaluation results of pelvic floor muscle training. By combining manually labeled training datasets and optimization strategies for joint loss functions, it effectively improves the model's ability to quantitatively evaluate training effects, not only improving the accuracy of evaluation but also providing solid technical support for personalized training feedback and optimization suggestions.

[0169] Furthermore, in some embodiments, after inputting the training dataset into the multi-task regression network, the corresponding joint loss function is calculated, which may specifically include:

[0170] The training dataset is input into a multi-task regression network, and the analysis results are output, including the output results of the effectiveness evaluation branch and the output results of the compensation score branch.

[0171] Calculate the first mean square error between the output of the validity assessment result branch and the sample validity label;

[0172] Calculate the second mean square error between the output of the compensation score branch and the sample compensation label;

[0173] The reconstructed root mean square curve is generated based on the intermediate features of the shared feature extraction layer, and the structural similarity loss between the reconstructed root mean square curve and the true root mean square curve is calculated.

[0174] Specifically, the calculation process for the joint loss function is as follows:

[0175] The constructed training dataset is input into a multi-task regression network. The network processes the fused feature vector of each sample through a forward propagation process, ultimately outputting the analysis results, including the output of the effectiveness evaluation branch and the output of the compensation score branch. Specifically, the network's shared feature extraction layer first extracts shared embedding features from the input data. These shared embedding features capture key information from the electromyographic signals on the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the temporal signals of posture. Then, the shared features are passed to two parallel branch regression layers, generating predicted values ​​for the effectiveness evaluation result and the compensation score, respectively. This process essentially maps the information from multi-source physiological signals after feature extraction and fusion onto two different evaluation dimensions: the effectiveness of the training and the degree of compensation interference.

[0176] Mean squared error (MSE) is a common metric used to measure the difference between predicted and true values. For the validity assessment branch, calculating the first mean squared error between its output and the sample validity label quantifies the model's accuracy in predicting validity assessment results. Specifically, for each training sample, the predicted validity assessment result from the branch is compared with the manually labeled true validity label, the square of the difference between the two is calculated, and the average of the squared differences over all samples is obtained to obtain the first mean squared error. This error value reflects the model's fit to the validity assessment result prediction task; the smaller the error, the closer the model's prediction is to the true label, and the better the model's performance on this task.

[0177] Similarly, for the compensation scoring branch, the second mean squared error (MSE) between its output and the sample compensation label is calculated to measure the model's accuracy in predicting compensation scores. The process is similar to calculating the first MSE, except that here the comparison is between the output of the compensation scoring branch and the sample compensation label. This metric assesses the model's ability to identify the degree of compensation interference, thus understanding its performance on the compensation score prediction task. A smaller second MSE means the model can predict compensation scores more accurately, contributing to improved reliability of the overall evaluation system.

[0178] In addition to the mean squared error loss, this embodiment introduces structural similarity loss (SSIM) to further optimize the model. SSIM focuses on the structural similarity between the reconstructed root mean square curve and the true root mean square curve. First, a reconstructed root mean square curve is generated starting from the intermediate features of the shared feature extraction layer. This curve reconstructs the root mean square value of the original electromyography (EMG) signal, reflecting the model's ability to capture and reconstruct EMG signal features. Then, the reconstructed root mean square curve is compared with the true root mean square curve, and the structural similarity loss between them is calculated. SSIM loss considers not only the differences in mean and variance between the two curves but also their covariance relationship, thus providing a more comprehensive measure of the similarity between the curves. By minimizing the structural similarity loss, the model can learn more accurate signal features, enhancing its ability to recognize training rhythmicity and compensatory fluctuations.

[0179] Finally, based on the first mean squared error, the second mean squared error, and the structural similarity loss, and their respective weight coefficients, the joint loss function output by the training dataset after being input into the multi-task regression network is calculated.

[0180] The multi-task regression network in this embodiment can simultaneously predict the effectiveness evaluation results and compensation interference evaluation results of pelvic floor muscle training. The calculation and optimization of the joint loss function ensures the accuracy of the model on both evaluation tasks. Specifically, the mean squared error loss enables the model to accurately fit the effectiveness and compensation labels in the training data, while the structural similarity loss further enhances the model's ability to capture the temporal features of electromyographic signals. This helps improve the overall performance of the pelvic floor muscle training evaluation model, enabling it to more accurately quantify training effects and provide more reliable technical support for subsequent personalized training feedback and optimization suggestions.

[0181] Furthermore, in some embodiments, step S3, "fusing feature vectors to determine pelvic floor muscle training evaluation results through a pre-trained evaluation model," may specifically include:

[0182] S31. Input the fused feature vector into the shared feature extraction layer of the evaluation model to generate fused embedded features;

[0183] Specifically, for step S31, the fused feature vector includes comprehensive features extracted from electromyography (EMG) signals on the surface of the pelvic floor muscles, EMG signals from muscles surrounding the pelvic floor muscles, and posture temporal signals. The shared feature extraction layer, as the front end of the evaluation model, is responsible for further feature processing and abstraction of the input fused feature vector, extracting higher-level fused embedding features. This process enables deeper integration of feature information from different modalities, allowing subsequent evaluation branches to make more accurate predictions based on these embedding features. For example, the shared feature extraction layer can consist of multiple fully connected layers or convolutional layers, capturing complex relationships between features through nonlinear transformations.

[0184] S32. Based on the fusion embedding features, the training effectiveness evaluation results and compensation interference evaluation results are output synchronously through the parallel dual-branch regression layer of the evaluation model;

[0185] Specifically, in step S32, the parallel dual-branch regression layer of the evaluation model is a key part of achieving multi-task evaluation. The effectiveness evaluation branch focuses on assessing whether the training movements have achieved the expected training effect, while the compensation interference evaluation branch is used to identify whether there is excessive involvement of non-target muscle groups during training. These two branches share the same feature extraction layer but each has an independent regression network to ensure that each task can focus on its own objective. By inputting the fused embedded features into these two branches respectively, the model can output two scores simultaneously. This parallel processing approach not only improves the efficiency of the model but also ensures that the two evaluation metrics do not interfere with each other, thereby improving the accuracy and reliability of the evaluation.

[0186] The shared feature extraction layer provided in this embodiment ensures efficient use of features, while the parallel dual-branch regression layer enables the model to provide two key evaluation metrics simultaneously during a single inference process, greatly improving the efficiency and accuracy of the evaluation.

[0187] Furthermore, in some embodiments, the pelvic floor muscle training instruction method further includes:

[0188] S41. Generate training feedback instructions based on the combination of effectiveness evaluation results and compensation interference evaluation results;

[0189] Specifically, for step S41, based on the effectiveness assessment results and compensation interference assessment results output by the evaluation model, targeted training feedback instructions can be generated. These feedback instructions aim to help trainees understand their current training status and provide suggestions for improvement. For example, when the effectiveness assessment result is high and the compensation interference assessment result is low, it indicates that the training movements are performed correctly, and the trainee can be prompted to continue maintaining this level. Conversely, if the effectiveness assessment result is low and the compensation interference assessment result is high, it indicates that there are abnormal movements, and it is recommended to adjust the training method or pause training for correction.

[0190] S42. Based on the training feedback instructions, invoke the corresponding device control strategy;

[0191] Specifically, in step S42, based on the generated training feedback instructions, the pelvic floor muscle repair device can invoke pre-set device control strategies. For example, when it is necessary to prompt the trainee to increase the contraction intensity, the device can guide the trainee through vibration or sound prompts; when severe compensatory interference is detected, the device can automatically reduce the training intensity or guide the trainee to use the correct force exertion method, thereby achieving intelligent training guidance and improving training effectiveness.

[0192] S43. Adjust the working status of the pelvic floor muscle repair device based on the equipment control strategy;

[0193] Specifically, in step S43, the pelvic floor muscle repair device adjusts its working state according to the invoked device control strategy, including changing the intensity of electrical stimulation, adjusting the biofeedback mode, and providing visual or auditory cues. Through these adjustments, the device can respond to changes in the trainee's state in real time, providing personalized training support and thus improving the targeting and efficiency of the training.

[0194] This embodiment generates training feedback instructions based on the effectiveness assessment results and the compensation interference assessment results, and accordingly calls the device control strategy to adjust the working state of the pelvic floor muscle repair device, so as to realize real-time and personalized feedback and guidance on the pelvic floor muscle training process, effectively improve the training quality of trainees, and reduce the decline in training effect caused by compensation interference.

[0195] Furthermore, in some embodiments, step S41, "generating training feedback instructions based on the combined relationship between the effectiveness evaluation results and the compensation interference evaluation results," may specifically include:

[0196] S41 a. When the effectiveness assessment result is greater than or equal to the first threshold and the compensation interference assessment result is less than or equal to the second threshold, output the action specification instruction;

[0197] Specifically, when the effectiveness assessment result is high (≥ the first threshold) and the compensation interference assessment result is low (≤ the second threshold), it indicates that the training movements are performed correctly, the pelvic floor muscle contraction is effective, and there is no obvious compensatory behavior. At this time, the system outputs a movement correction instruction to encourage the trainee to continue the current training method. The first and second thresholds can be set based on a large amount of experimental data, for example, by determining a reasonable range for effectiveness and compensation scores through statistical analysis. The device can also combine voice prompts, screen displays, and other methods to provide real-time feedback on the trainee's movement status.

[0198] S41 b. When the effectiveness assessment result is greater than or equal to the first threshold and the compensation interference assessment result is greater than the second threshold, output a compensation warning instruction;

[0199] Specifically, when the effectiveness assessment result is high but the compensation interference assessment result is also high, it indicates that although the trainee has performed effective pelvic floor muscle contraction, there is obvious compensatory behavior (such as excessive abdominal exertion). At this time, the system outputs a compensation warning instruction, prompting the trainee to reduce the involvement of non-target muscle groups. The device can also guide the trainee to adjust the exertion method through voice prompts or vibration feedback, such as prompting "reduce abdominal exertion and focus attention on pelvic floor muscle contraction".

[0200] S41c. When the effectiveness assessment result is less than the third threshold and the compensation interference assessment result is less than or equal to the second threshold, output the intensity increase instruction;

[0201] Specifically, when the effectiveness assessment result is low (< the third threshold) but the compensation interference assessment result is low, it indicates that the trainee's pelvic floor muscle contraction strength is insufficient, but there is no obvious compensatory behavior. At this time, the system outputs an intensity increase command, suggesting that the trainee increase the contraction strength or prompting the trainee to adjust their posture. Biofeedback technology can also be combined, such as displaying the pelvic floor muscle contraction strength curve on the screen, to guide the trainee to gradually increase the contraction strength.

[0202] S41d. When the effectiveness evaluation result is less than the third threshold and the compensation interference evaluation result is greater than the second threshold, output an action abnormality interruption command;

[0203] Specifically, when the effectiveness assessment result is low and the compensation interference assessment result is high, it indicates that the trainee's movements are significantly abnormal, with insufficient pelvic floor muscle contraction and significant compensatory behavior. In this case, the system outputs a movement abnormality interruption command, suggesting that the trainee pause the current training and relearn the correct movements. Detailed movement guidance videos or illustrated explanations can also be provided to help the trainee correct incorrect movements. Simultaneously, the frequency and type of abnormal movements can be recorded, providing reference data for rehabilitation therapists.

[0204] This embodiment generates corresponding training feedback instructions based on different combinations of effectiveness assessment results and compensation interference assessment results, enabling real-time and accurate feedback on the pelvic floor muscle training process. This helps trainees adjust their training methods in a timely manner, reduce compensatory behaviors, and improve training effectiveness.

[0205] Furthermore, in some embodiments, the pelvic floor muscle training instruction method further includes:

[0206] S61. Real-time monitoring of the coordination coefficient between the electromyographic signals on the surface of the pelvic floor muscles and the electromyographic signals of the muscles surrounding the pelvic floor muscles;

[0207] Specifically, for step S61, during pelvic floor muscle training, the coordinated contraction of the pelvic floor muscles and rectus abdominis muscles is crucial for assessing training effectiveness and identifying compensatory behaviors. By monitoring the coordination coefficient between the electromyographic signals of these two muscle groups in real time, their interaction can be dynamically captured. The coordination coefficient can be obtained by calculating the Pearson correlation coefficient or mutual information of the two electromyographic signals; these indicators can reflect the linear or nonlinear correlation between the signals.

[0208] S62. When the synergy coefficient is detected to exceed the preset range, the compensation interference assessment results are dynamically corrected;

[0209] Specifically, for step S62, a normal range for the synergy coefficient is defined, based on extensive experimental data and clinical experience. When the synergy coefficient monitored in real time exceeds this range, it indicates possible abnormal compensatory behavior, requiring dynamic correction of the compensatory interference assessment results. This correction can be achieved by adjusting relevant parameters in the assessment model, such as increasing the weight of the compensatory score, or directly adjusting the score through a specific compensation algorithm to more accurately reflect the degree of compensatory interference in the current training state.

[0210] This embodiment uses a dynamic correction mechanism to reflect compensatory changes during training in a timely manner, providing trainees with more accurate feedback, helping them adjust their training methods promptly, reduce compensatory behaviors, and thus improve training effectiveness.

[0211] In summary, the pelvic floor muscle training instruction method provided in this embodiment first comprehensively captures physiological information during training by simultaneously collecting electromyographic signals from the surface of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and postural temporal signals. Then, it extracts features from the multi-source physiological signals and generates a fused feature vector, integrating key features from different signal modalities to provide a rich data foundation for subsequent evaluation. Finally, the fused feature vector is input into a pre-trained evaluation model, which outputs an evaluation result including effectiveness assessment and compensation interference assessment results. Based on the pelvic floor muscle training evaluation results, it generates training feedback instructions, achieving a quantitative evaluation of the pelvic floor muscle training effect and guiding the user's training process based on the evaluation results. It is evident that this embodiment, by simultaneously collecting multi-source physiological signals and combining them with a deep learning model, fully utilizes the complementarity of multi-source physiological signals to overcome the limitations of single-signal evaluation, thereby improving the accuracy and comprehensiveness of pelvic floor muscle training evaluation. This provides strong support for optimizing training programs and improving rehabilitation quality, effectively solving the problem in existing technologies where the inability to accurately identify the quality of pelvic floor muscle training leads to a decline in the quality and efficiency of rehabilitation training.

[0212] It should be understood that, although Figure 4 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 4 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0213] To facilitate better implementation of the pelvic floor muscle training instruction method of this application, this application also provides a pelvic floor muscle training assessment device based on the above-described pelvic floor muscle training instruction method. The meanings of the terms used are the same as in the above-described pelvic floor muscle training instruction method, and specific implementation details can be found in the description of the method embodiments.

[0214] Please see Figure 7 , Figure 7 This is a schematic diagram of the pelvic floor muscle training and assessment device provided in the embodiments of this application. The pelvic floor muscle training and assessment device may specifically include:

[0215] The signal acquisition module 201 is used to synchronously acquire multi-source physiological signals of the target object. The multi-source physiological signals include electromyographic signals on the surface of the pelvic floor muscles, electromyographic signals of the muscles surrounding the pelvic floor muscles, and posture and timing signals.

[0216] Feature extraction module 202 is used to extract features from multi-source physiological signals and generate fused feature vectors;

[0217] The evaluation module 203 is used to input the fused feature vector into the pre-trained evaluation model and output the pelvic floor muscle training evaluation results of the target object. The pelvic floor muscle training evaluation results include the effectiveness evaluation results and the compensation interference evaluation results.

[0218] Furthermore, in some embodiments, the signal acquisition module 201 includes:

[0219] The first acquisition unit is used to acquire electromyographic signals on the surface of the pelvic floor muscles in the pelvic floor muscle region of the target object through the first electrode.

[0220] The second pair of acquisition units is used to acquire electromyographic signals of the muscles surrounding the pelvic floor muscles in the rectus abdominis region of the target object through the second electrode.

[0221] The third acquisition unit is used to acquire corresponding acceleration and angular velocity data in the pelvic region of the target object through the inertial measurement unit, and generate attitude timing signals.

[0222] The synchronization unit is used to perform time synchronization processing on the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture timing signals.

[0223] Furthermore, in some embodiments, the feature extraction module 202 includes:

[0224] The first extraction unit is used to extract features from the electromyographic signals on the surface of the pelvic floor muscles using a first feature extraction network, and generate a first electromyographic feature vector.

[0225] The second extraction unit is used to extract features from the electromyographic signals of the muscles surrounding the pelvic floor muscles using the second feature extraction network, and generate a second electromyographic feature vector.

[0226] The third extraction unit is used to extract features from the attitude time-series signal using a time-series coding network to generate an attitude feature vector;

[0227] The feature splicing unit is used to splice the first electromyographic feature vector, the second electromyographic feature vector, and the posture feature vector to form a fused feature vector.

[0228] Furthermore, in some embodiments, the feature extraction module 202 further includes:

[0229] The reconstructed root mean square curve unit is used to simultaneously generate the reconstructed root mean square curves corresponding to the electromyographic signals on the surface of the pelvic floor muscles and the electromyographic signals of the muscles surrounding the pelvic floor muscles during the feature extraction process.

[0230] The structural similarity loss unit is used to calculate the structural similarity loss between the reconstructed root mean square curve and the true root mean square curve. The structural similarity loss is used to constrain the training of the first feature extraction network and the second feature extraction network.

[0231] Furthermore, in some embodiments, the device further includes a model training module, specifically used for:

[0232] A multi-task regression network is constructed. The input layer of the multi-task regression network is a shared feature extraction layer that receives fused feature vectors and generates shared embedded features. The output layer is a parallel two-branch regression layer that includes an effectiveness evaluation result branch and a compensation score branch.

[0233] Construct a training dataset, which includes sample fusion feature vectors and supervision signals. The supervision signals are sample validity labels and sample compensation labels corresponding to the sample fusion feature vectors, which are annotated manually.

[0234] After inputting the training dataset into the multi-task regression network, the corresponding joint loss function is calculated;

[0235] With the goal of minimizing the joint loss function, the network parameters of the multi-task regression network are updated to obtain the trained evaluation model.

[0236] Furthermore, in some embodiments, after inputting the training dataset into the multi-task regression network, the corresponding joint loss function is calculated, including:

[0237] The training dataset is input into a multi-task regression network, and the analysis results are output, including the output results of the effectiveness evaluation branch and the output results of the compensation score branch.

[0238] Calculate the first mean square error between the output of the validity assessment result branch and the sample validity label;

[0239] Calculate the second mean square error between the output of the compensation score branch and the sample compensation label;

[0240] The reconstructed root mean square curve is generated based on the intermediate features of the shared feature extraction layer, and the structural similarity loss between the reconstructed root mean square curve and the true root mean square curve is calculated.

[0241] Furthermore, in some embodiments, the evaluation module 203 includes:

[0242] The input unit is used to input the fused feature vector into the shared feature extraction layer of the evaluation model to generate fused embedded features;

[0243] The output unit is used to synchronously output the training effectiveness evaluation results and the compensation interference evaluation results based on the fused embedding features through the parallel bi-branch regression layer of the evaluation model.

[0244] Furthermore, in some embodiments, the device further includes a feedback module, specifically used for:

[0245] Based on the combined relationship between the effectiveness evaluation results and the compensation interference evaluation results, training feedback instructions are generated.

[0246] Based on training feedback instructions, the corresponding device control strategy is invoked;

[0247] Adjust the working status of the pelvic floor muscle repair device based on the equipment control strategy.

[0248] Furthermore, in some embodiments, training feedback instructions are generated based on the combined relationship between the effectiveness evaluation results and the compensation interference evaluation results, including:

[0249] When the effectiveness assessment result is greater than or equal to the first threshold and the compensation interference assessment result is less than or equal to the second threshold, the action specification instruction is output.

[0250] When the effectiveness assessment result is greater than or equal to the first threshold and the compensation interference assessment result is greater than the second threshold, a compensation warning instruction is output.

[0251] When the effectiveness assessment result is less than the third threshold and the compensation interference assessment result is less than or equal to the second threshold, an intensity increase instruction is output.

[0252] When the effectiveness evaluation result is less than the third threshold and the compensation interference evaluation result is greater than the second threshold, an action abnormality interruption command is output.

[0253] Furthermore, in some embodiments of this application, the apparatus further includes a dynamic correction module, specifically used for:

[0254] Real-time monitoring of the coordination coefficient between the electromyographic signals on the surface of the pelvic floor muscles and the electromyographic signals of the muscles surrounding the pelvic floor muscles;

[0255] When the synergy coefficient is detected to exceed the preset range, the compensation interference assessment results are dynamically corrected.

[0256] Specific limitations regarding the pelvic floor muscle training assessment device can be found in the above description of the limitations on pelvic floor muscle training instruction methods, and will not be repeated here. Each module in the aforementioned pelvic floor muscle training assessment device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0257] The pelvic floor muscle training assessment device provided in this embodiment synchronously acquires multi-source physiological signals from the target subject through a signal acquisition module 201. These multi-source physiological signals include electromyographic signals from the pelvic floor muscle surface, electromyographic signals from surrounding muscles, and postural temporal signals. A feature extraction module 202 extracts features from these multi-source physiological signals to generate a fused feature vector. An assessment module 203 inputs the fused feature vector into a pre-trained assessment model and outputs the pelvic floor muscle training assessment results for the target subject. These results include effectiveness assessment results and compensation interference assessment results. This embodiment, by synchronously acquiring multi-source physiological signals and combining them with a deep learning model, fully utilizes the complementarity of these signals to overcome the limitations of single-signal assessment, thereby improving the accuracy and comprehensiveness of pelvic floor muscle training assessment. This provides strong support for optimizing training programs and improving rehabilitation quality, effectively solving the problem in existing technologies where the inability to accurately identify the quality of pelvic floor muscle training leads to reduced rehabilitation training quality and efficiency.

[0258] Furthermore, embodiments of this application also provide an electronic device, such as... Figure 8 As shown, it illustrates a structural schematic diagram of the electronic device involved in the embodiments of this application, specifically:

[0259] The electronic device may include components such as a processor 301 with one or more processing cores, a memory 302 with one or more computer-readable storage media, a power supply 303, and an input unit 304. Those skilled in the art will understand that... Figure 7 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0260] The processor 301 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines, and performs various functions and processes data by running or executing software programs and / or modules stored in the memory 302, and by calling data stored in the memory 302, thereby providing overall monitoring of the electronic device. Optionally, the processor 301 may include one or more processing cores; preferably, the processor 301 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 301.

[0261] The memory 302 can be used to store software programs and modules. The processor 301 executes various functional applications and pelvic floor muscle training instructions by running the software programs and modules stored in the memory 302. The memory 302 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 302 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 302 may also include a memory controller to provide the processor 301 with access to the memory 302.

[0262] The electronic device also includes a power supply 303 that supplies power to various components. Preferably, the power supply 303 can be logically connected to the processor 301 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 303 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0263] The electronic device may also include an input unit 304, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0264] Although not shown, the electronic device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 301 in the electronic device loads the executable files corresponding to the processes of one or more applications into the memory 302 according to the following instructions, and the processor 301 runs the applications stored in the memory 302 to realize various functions, as follows:

[0265] Simultaneously collect multi-source physiological signals from the target object, including electromyographic signals on the surface of the pelvic floor muscles, electromyographic signals of the muscles surrounding the pelvic floor muscles, and posture and temporal signals; extract features from the multi-source physiological signals to generate a fused feature vector; input the fused feature vector into a pre-trained evaluation model to output the pelvic floor muscle training evaluation results of the target object, including effectiveness evaluation results and compensation interference evaluation results.

[0266] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0267] This application embodiment, by synchronously collecting multi-source physiological signals and combining them with a deep learning model, fully utilizes the complementarity of multi-source physiological signals to overcome the limitations of single-signal evaluation, thereby improving the accuracy and comprehensiveness of pelvic floor muscle training evaluation. This provides strong support for optimizing training programs and improving rehabilitation quality, effectively solving the problem that existing technologies cannot accurately identify the quality of pelvic floor muscle training, thus affecting the quality and efficiency of rehabilitation training.

[0268] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0269] Therefore, embodiments of this application provide a storage medium storing a plurality of instructions that can be loaded by a processor to execute steps in any of the pelvic floor muscle training instruction methods provided in embodiments of this application. For example, the instructions can execute the following steps:

[0270] Simultaneously collect multi-source physiological signals from the target object, including electromyographic signals on the surface of the pelvic floor muscles, electromyographic signals of the muscles surrounding the pelvic floor muscles, and posture and temporal signals; extract features from the multi-source physiological signals to generate a fused feature vector; input the fused feature vector into a pre-trained evaluation model to output the pelvic floor muscle training evaluation results of the target object, including effectiveness evaluation results and compensation interference evaluation results.

[0271] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0272] The storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0273] Since the instructions stored in the storage medium can execute the steps in any of the pelvic floor muscle training instruction methods provided in the embodiments of this application, the beneficial effects that any of the pelvic floor muscle training instruction methods provided in the embodiments of this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0274] The foregoing has provided a detailed description of a pelvic floor muscle repair device, pelvic floor muscle training instruction method, apparatus, equipment, and medium provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A pelvic floor muscle repair device, characterized in that, The pelvic floor muscle repair device includes: The main body is used for insertion or placement into the corresponding area of ​​the pelvic floor muscles; The first electrode and the second electrode are disposed on the main body and are used to electrically stimulate the pelvic floor muscles and to detect the electromyographic signal on the surface of the pelvic floor muscles after they have been electrically stimulated. The third electrode is attached to the muscles surrounding the pelvic floor muscles during use to detect electromyographic signals of the muscles surrounding the pelvic floor muscles. The posture detection unit is used to detect the posture timing signal after changes in torso movement; and The controller, electrically connected to the first electrode, the second electrode, the third electrode, and the posture detection unit, is used to extract features from the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture temporal signals at the same time sequence, and generate a fused feature vector; it is also used to determine the pelvic floor muscle training evaluation result based on the fused feature vector and generate a training feedback instruction based on the pelvic floor muscle training evaluation result; the training feedback instruction is used to instruct the user to perform training.

2. The pelvic floor muscle repair device according to claim 1, characterized in that, The pelvic floor muscle repair device also includes a communication unit; The communication unit is electrically connected to the controller and is used to receive the training feedback instruction and send the training feedback instruction to the user terminal. The user terminal is used to generate an action reminder message based on the training feedback instruction; The action reminder message includes at least one of a prompt sound or a prompt video.

3. The pelvic floor muscle repair device according to claim 1, characterized in that, The pelvic floor muscle repair device also includes an output unit; The output unit is electrically connected to the controller and is used to receive the training feedback instruction and instruct the user to perform training according to the training feedback instruction.

4. The pelvic floor muscle repair device according to claim 1, characterized in that, The pelvic floor muscle training assessment results include effectiveness assessment results and compensation interference assessment results.

5. The pelvic floor muscle repair device according to claim 4, characterized in that, The controller is specifically configured to: output a pelvic floor muscle training evaluation result indicating proper movement when the effectiveness evaluation result is greater than or equal to a first threshold and the compensation interference evaluation result is less than or equal to a second threshold; output a pelvic floor muscle training evaluation result indicating excessive compensation by the surrounding muscles when the effectiveness evaluation result is greater than or equal to the first threshold and the compensation interference evaluation result is greater than the second threshold; output a pelvic floor muscle training evaluation result indicating insufficient pelvic floor muscle contraction strength when the effectiveness evaluation result is less than a third threshold and the compensation interference evaluation result is less than or equal to the second threshold; and output a pelvic floor muscle training evaluation result indicating ineffective pelvic floor muscle contraction accompanied by significant compensation by the surrounding muscles when the effectiveness evaluation result is less than the third threshold and the compensation interference evaluation result is greater than the second threshold.

6. A method for instructing pelvic floor muscle training, characterized in that, include: Simultaneously acquire electromyographic signals of the pelvic floor muscles, electromyographic signals of the muscles surrounding the pelvic floor muscles, and posture and temporal signals of the trunk. Feature extraction is performed on the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture temporal signals at the same time sequence to generate a fused feature vector; The fused feature vector determines the pelvic floor muscle training evaluation result through a pre-trained evaluation model and generates training feedback instructions based on the pelvic floor muscle training evaluation result; the training feedback instructions are used to instruct the user to perform training.

7. The pelvic floor muscle training instruction method according to claim 6, characterized in that, The step involves extracting features from the electromyographic signals of the pelvic floor muscles, the electromyographic signals of the muscles surrounding the pelvic floor muscles, and the posture temporal signals at the same time sequence to generate a fused feature vector, including: The first feature extraction network is used to extract features from the electromyographic signals on the surface of the pelvic floor muscles to generate a first electromyographic feature vector; A second feature extraction network is used to extract features from the electromyographic signals of the muscles surrounding the pelvic floor muscles, generating a second electromyographic feature vector. A temporal coding network is used to extract features from the attitude temporal signal to generate an attitude feature vector. The first electromyographic feature vector, the second electromyographic feature vector, and the posture feature vector are concatenated to form the fused feature vector.

8. The pelvic floor muscle repair device according to claim 6, characterized in that, The pelvic floor muscle training assessment results include effectiveness assessment results and compensation interference assessment results.

9. The pelvic floor muscle training instruction method according to claim 8, characterized in that, The method further includes: Based on the combined relationship between the effectiveness evaluation results and the compensation interference evaluation results, a training feedback instruction is generated; An action reminder message is generated based on the training feedback instructions.

10. The pelvic floor muscle training instruction method according to claim 8, characterized in that, The step of generating training feedback instructions based on the combination of the effectiveness evaluation results and the compensation interference evaluation results includes: When the effectiveness assessment result is greater than or equal to the first threshold and the compensation interference assessment result is less than or equal to the second threshold, an action specification instruction is output. When the effectiveness assessment result is greater than or equal to the first threshold, and the compensation interference assessment result is greater than the second threshold, a compensation warning instruction is output. When the effectiveness assessment result is less than the third threshold and the compensation interference assessment result is less than or equal to the second threshold, an intensity increase command is output. When the effectiveness evaluation result is less than the third threshold and the compensation interference evaluation result is greater than the second threshold, an action abnormality interruption command is output.

11. The pelvic floor muscle training instruction method according to claim 7, characterized in that, The method further includes: Real-time monitoring of the coordination coefficient between the electromyographic signals on the surface of the pelvic floor muscles and the electromyographic signals of the muscles surrounding the pelvic floor muscles; When the coordination coefficient is detected to exceed the preset range, the compensation interference assessment result is dynamically corrected.

12. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the pelvic floor muscle training instruction method as described in any one of claims 6-11.

13. A storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed as described in any one of claims 6-11.

Images