Remote kinesio treatment guidance and quality control system

Abstract

The invention relates to the field of remote quality control of kinesio, and provides a remote kinesio treatment guidance and quality control system. The image acquisition module is used for acquiring high-definition image data of a to-be-treated area through intelligent terminal equipment and marking anatomical feature points; wherein the high-definition image data comprises a static image and a dynamic muscle movement video stream; the AI analysis engine is used for carrying out feature extraction on the high-definition image data based on a muscle dynamics model of a to-be-treated region constructed by a convolutional neural network, identifying a muscle tension abnormal region and grading a nerve injury degree, and generating an initial sticking path recommendation scheme; the remote auditing interface transmits the recommended scheme generated by the AI to the doctor end interaction interface, obtains a target treatment scheme after the doctor end performs scheme correction based on the medical history data of the patient, and confirms a final treatment instruction through a digital signature; and the block chain quality control module is used for encrypting key operation data in the treatment process and then writing the key operation data into a distributed account book to generate a treatment process traceability chain which cannot be tampered.

Description

Technical Field

[0001] The present invention relates to the technical field of effect monitoring of intramuscular taping, and in particular to a remote intramuscular taping treatment guidance and quality control system. Background Art

[0002] Kinesio taping, a non-invasive rehabilitation therapy, has been widely used in recent years for sports injuries, neurological rehabilitation, and chronic pain. It works by mechanically stimulating the skin and subcutaneous tissue with an elastic tape, regulating muscle tone, promoting blood circulation, and improving neurofeedback. However, current clinical practice faces significant technical bottlenecks in taping protocol design, implementation, and treatment quality control, severely hindering its standardized efficacy and large-scale application.

[0003] Traditional kinesiology taping treatment is highly dependent on the physician's anatomical knowledge and clinical experience. The taping path design is mostly based on static palpation and patient complaints, and lacks real-time monitoring of the dynamic contraction of muscles and the sliding state of fascia (such as the tension change during compensatory contraction of the trapezius muscle). Studies have shown that the operating error rate of inexperienced physicians can reach 35% to 40%, resulting in insufficient tension or excessive stretching of the tape, causing skin allergies or reduced efficacy. In addition, existing technologies are unable to grade the degree of nerve damage (such as the grading and positioning of common peroneal nerve entrapment), resulting in a lack of precise indication matching for taping schemes.

[0004] In home rehabilitation and primary healthcare scenarios, existing remote guidance mostly uses two-dimensional video demonstrations or graphic descriptions, which makes it difficult to match the calibration of anatomical landmarks with the patient's individualized body shape. For example, the taping of patients with rotator cuff injuries requires adjusting the direction of the tape according to the dynamic trajectory of the scapula, but existing video tools cannot capture muscle kinematic parameters (such as angular velocity and electromyographic signals), resulting in a tape misplacement rate of more than 50% when patients operate on their own. At the same time, the lack of a real-time feedback mechanism means that doctors cannot verify the actual tension parameters of the tape (such as the standard stretching rate of 12%-15%) through remote means.

[0005] Existing clinical operation records are still mainly paper documents or simple electronic medical records, which pose the risk of data tampering and cannot be linked to multimodal data during the treatment process (such as pressure sensor data during taping and changes in the patient's joint range of motion). A multicenter study found that 72% of adverse reactions to Kinesio taping were difficult to attribute due to incomplete operation records. In addition, the lack of a mechanism to record the process of modifying treatment plans makes the collaborative decision-making process between physicians and AI systems impossible to audit, hindering the continuous improvement of medical quality.

[0006] Although some studies have attempted to use computer vision for taping positioning (such as OpenCV-based skin marker recognition), these algorithms are limited to static image processing and do not integrate muscle dynamics models, making them unable to predict changes in the mechanical distribution of the tape during dynamic activities. For example, taping for iliotibial band syndrome requires maintaining balanced tape tension during knee flexion and extension. However, existing AI tools only provide fixed path suggestions and are not integrated with biomechanical simulation, resulting in an error rate of up to 28% in dynamic scenarios.

[0007] The current clinical application of Kinesio taping is polarized. While standardized in tertiary hospitals, where rehabilitation physicians lead the procedure, it remains difficult to popularize due to labor costs. In primary healthcare settings and at home, patients often rely on online tutorials or assistance from non-professionals, resulting in high rates of procedural errors. A 2022 report from the World Health Organization (WHO) indicated that misuse of Kinesio taping has become the second-leading preventable medical risk factor in sports medicine. In particular, in cases of lymphatic obstruction caused by high-tension taping, the complication rate is 4.2 times higher than in the standard procedure group.

[0008] In summary, existing technologies have not yet solved the three core problems of Kinesio taping treatment: the lack of dynamic assessment, inaccurate remote execution, and broken quality control chains. There is an urgent need for a closed-loop management system that integrates artificial intelligence, the Internet of Things, and blockchain to achieve standardized reconstruction of the entire process from diagnosis to execution. Summary of the Invention

[0009] Firstly, the present application proposes a remote kinesiology tape treatment guidance and quality control system, which is used to solve the problems in the prior art where the industrial control computer of the system often has heat dissipation problems, insufficient multi-tasking processing capabilities, poor real-time performance and other problems.

[0010] This application proposes a remote Kinesio taping treatment guidance and quality control system, including:

[0011] An image acquisition module is used to obtain high-definition image data of the taping area through a smart terminal device and mark anatomical feature points; the high-definition image data includes static images and dynamic muscle movement video streams;

[0012] The AI ​​analysis engine uses a convolutional neural network to build a muscle dynamics model of the taping area, extract features from high-definition image data, identify areas of abnormal muscle tension and grade the degree of nerve damage, and generate initial taping path recommendations;

[0013] The remote review interface transmits the AI-generated recommendation plan to the physician's interactive interface, obtains the target taping recommendation items after the physician modifies the plan based on the patient's medical history data, and confirms the final taping instruction through a digital signature;

[0014] The blockchain quality control module is used to encrypt key operational data during the taping process and write it into the distributed ledger, generating an unalterable traceability chain for the taping process.

[0015] The patient-side execution feedback module obtains the taping path through augmented reality and superimposes it on the patient's real-time image screen. It also collects actual taping parameters based on the pressure sensor patch, compares the deviation with the target taping recommendation, and triggers abnormal warnings or automatic calibration instructions.

[0016] In combination with the first aspect, the image acquisition module includes:

[0017] The preprocessing submodule uses a preset OpenCV library to perform illumination equalization on the original image of the taping area, and aligns the original image after illumination equalization with the anteroposterior and lateral images of the taping area of ​​the patient using a non-rigid registration algorithm to generate a first image.

[0018] The dynamic feature extraction submodule tracks the muscle motion trajectory of the taping area in the first image based on the optical flow method and constructs a three-dimensional muscle motion vector field;

[0019] The anatomical landmark database is used to perform pixel enhancement on each acupuncture point in the taping area according to the three-dimensional muscle motion vector field to generate high-definition image data.

[0020] In conjunction with the first aspect, the AI ​​engine includes a multimodal fusion training module including:

[0021] Prediction submodule: used to determine the electromyographic characteristics of the taping area based on the marked anatomical feature points, and form a therapeutic effect prediction model;

[0022] The real-time optimization submodule uses a time-series analysis of historical taping cases based on an efficacy prediction model and an LSTM network to determine the tape stretching parameters for the target taping recommendation. These parameters are positively correlated with the progressive nature of neurological recovery within the patient's taping cycle.

[0023] The interpretable output unit determines the muscle state characteristics in the muscle dynamics model based on the tape stretch parameters.

[0024] In combination with the first aspect, the blockchain quality control module includes:

[0025] The tagging supervision submodule stores high-definition image data based on the IPFS protocol and writes the file hash value into the Hyperledger Fabric chain code to ensure data integrity and cross-institutional retrieval efficiency;

[0026] A smart contract-driven quality control rule base defines automated compliance checks for patch replacement cycle thresholds, tension deviation tolerance, and physician response time. Illegal operations trigger on-chain alerts and quality control score downgrades.

[0027] The privacy computing sub-module uses homomorphic encryption technology to desensitize patients' sensitive information, performs zero-knowledge proof verification when sharing data on the chain, and generates compliance values.

[0028] In combination with the first aspect, the patient-side execution feedback module includes:

[0029] Smart patch hardware integrates a flexible pressure sensor array and a micro inertial measurement unit to monitor the pressure distribution, tensile deformation, and angular offset data of the patch-skin contact surface in real time;

[0030] Identification submodule: Based on an adaptive calibration algorithm, when the deviation between the actual taping parameters and the recommended parameters exceeds a preset threshold, a correction guidance arrow is dynamically rendered through the AR interface and a vibration prompt message is generated;

[0031] The efficacy tracking sub-module triggers the patient's movement self-inspection process in the taping area on a daily basis, calculates the HB score improvement rate through video analysis, generates a taping progress report and synchronizes it to the physician end.

[0032] In conjunction with the first aspect, the remote audit interface 300 is further used to:

[0033] Establish a federated learning framework across medical institutions. Each node hospital locally stores patient taping data, uploads model gradients to a central server via homomorphic encryption and secure aggregation protocols, and updates the global efficacy prediction model in real time.

[0034] An XGBoost model was trained based on multi-center data to identify key factor combinations that affect efficacy and generate a priority list of combined taping regimens for different disease stages.

[0035] And according to the task thread after the scheduler allocates the signal sequence, the task counting information is generated; wherein, the technical information of the synchronization task has a synchronization identifier.

[0036] In conjunction with the first aspect, the AI ​​analysis engine includes:

[0037] A reinforcement learning agent module is used to obtain real-time image data of the taping area of ​​the patient, determine the real-time input state of the taping area, and generate a first set of recommended items based on the displacement of the taping anchor point and the tension adjustment amplitude;

[0038] Stagnation judgment module: used to automatically trigger the PPO algorithm to recalculate the taping parameters in the first set of recommended items and generate a second set of recommended items when the level of abnormal muscle tension area and nerve damage degree in real-time image data does not change within the preset supervision time.

[0039] In combination with the first aspect, the execution feedback module is further configured to:

[0040] Construct a virtual social environment, determine the movement tasks of the taped area, and determine the degree of muscle activity completion on the affected side of the patient when performing the movement tasks;

[0041] Dynamically adjust the task difficulty level based on the strength of the electromyographic signal during the action task, detect whether the muscle strength is uniform, and lower the response time threshold of the action task if it is uneven;

[0042] By measuring the patient's unit action potential before and after the movement task during the preset rehabilitation cycle, the progress of nerve regeneration is quantified and the proportion of patches for the recommended items is determined.

[0043] In combination with the first aspect, the patient-side execution feedback module further includes:

[0044] Configure an acupoint electrical stimulation parameter library to store the optimal stimulation frequency, waveform, and duration of the acupoints in the area to be taped, and link it with the spatiotemporal association database of the taping scheme;

[0045] Configure a collaborative control algorithm to calculate the superposition effect of acupuncture depth and the mechanical action of the tape in the same muscle group before and after acupuncture taping through finite element simulation to determine whether there is local microcirculatory overload;

[0046] Configure the IoT synchronization interface to detect whether the stretching force of the patch stretching function returns to the preset tension parameters when the acupuncture device starts the penetration operation.

[0047] In combination with the first aspect, the patient-side execution feedback module is also used to fuse ultrasonic elastic imaging data with real-time electromyographic signals in an augmented reality interface to generate a three-dimensional biomechanical action vector field, and optimize the spatial distribution density of patch anchor points through a gradient descent algorithm.

[0048] Other features and advantages of the present application will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0049] The technical solution of the present invention is described in detail below in conjunction with the first aspect through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present application and do not constitute a limitation of the present application.

[0051] In the attached figure:

[0052] Figure 1 This is a system execution process diagram of a remote kinesiology taping treatment guidance and quality control system in an embodiment of the present application;

[0053] Figure 2 This is a schematic diagram of the image acquisition process in the embodiment of the present application;

[0054] Figure 3 This is a schematic diagram of the process in which the AI ​​engine includes a multimodal fusion training module in an embodiment of the present application;

[0055] Figure 4 This is a process diagram of the blockchain quality control module in an embodiment of the present application;

[0056] Figure 5 This is a schematic diagram of the report generation process of the feedback module executed by the patient side in an embodiment of the present application;

[0057] Figure 6 This is a functional diagram of the program review interface for this application embodiment;

[0058] Figure 7 Schematic diagram of the recommendation item update process of the AI ​​analysis engine in an embodiment of the present application;

[0059] Figure 8 Schematic diagram of the acupoint stimulation linkage process of the feedback module in the embodiment of the present application;

[0060] Figure 9 This is a functional diagram of the patient-side execution feedback module in an embodiment of the present application. DETAILED DESCRIPTION

[0061] The preferred embodiments of the present application are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application and are not used to limit the present application.

[0062] See Figure 1 In order to solve the three core problems of traditional kinesio taping treatment, namely the lack of dynamic evaluation, inaccurate remote execution, and broken quality control chain, this application proposes a remote kinesio taping treatment guidance and quality control system with a three-layer architecture, including a front-end system 10, an AI core layer 20, and a quality control execution layer 30. The front-end system 10 is mainly configured in a smart terminal device to collect high-definition image data. The AI ​​core layer 20 is configured in a local or cloud-based AI server to perform feature extraction and determine the user's taping path. The quality control execution layer 30 is mainly used to determine the final taping plan, determine whether the taping plan meets actual needs, and then realize the processing of feedback information on the patient side.

[0063] In traditional solutions, a single static photo and an ordinary camera are mainly used to record the muscle status of the person being taped. In terms of effect, it is impossible to judge the dynamic contraction status of the muscles. The positioning of feature points mainly relies on manual labeling, which is also a clinical defect.

[0064] The image acquisition module 100 is used to obtain high-definition image data of the area to be taped through the smart terminal device and mark anatomical feature points; wherein the high-definition image data includes static images and dynamic muscle movement video streams;

[0065] Smart terminal devices include mobile phones, tablets and interactive devices. They use high-definition cameras to collect static images and dynamic video streams of the patient's area to be taped. In this process, computer vision algorithms, such as OpenCV or YOLO models, automatically identify anatomical features and automatically mark anatomical feature points, such as muscle starting points, insertion points, and fascia directions. In this application, the dynamic video stream uses the optical flow method to analyze the muscle movement trajectory, assist in determining the area of ​​abnormal muscle tension, and finally combine the anatomical structure with the taping direction to ensure that the feature point marking is consistent with the physiological direction of the muscle and fascia. The image acquisition module 100 of this application mainly rejects the difficulty of collecting dynamic muscle movement data, realizes dynamic and static comparison, and realizes the recognition and automatic labeling of muscle feature breakpoints.

[0066] In view of the fact that the image recognition of the rule engine in the traditional solution can only judge the surface skin condition but cannot quantify the muscle tension, the present application constructs an AI analysis engine 200, and constructs a muscle dynamics model of the area to be taped based on a convolutional neural network. It extracts features from the high-definition image data, identifies areas of abnormal muscle tension and the degree of nerve damage, and generates an initial taping path recommendation plan;

[0067] In this process, the present application constructs a muscle dynamics model based on a convolutional neural network (CNN), inputs static images and dynamic video data, extracts features such as muscle morphology, motion trajectory, fascia tension, etc., and then based on transfer learning, for example: ResNet or U-Net architecture, determines and identifies abnormal tension areas, such as: spasmodic or atrophied muscle areas, and classifies the degree of nerve damage by extracting features of high-definition impact data and combining clinical databases to classify, for example: mild, moderate, severe, and areas with abnormal muscle tension corresponding to different degrees of nerve damage, to determine the location where patching may be required, and generates a recommended patching path plan. The recommended plan includes patch shapes, I-type, Y-type, X-type, etc., as well as tension parameters and patching directions corresponding to different patching shapes, so as to accurately patch the area that needs to be patched while reducing the force that the patching area may bear exceeding the necessary limit.

[0068] In response to the problem that data integrity cannot be guaranteed in the existing technology, this application configures a remote review interface 300 to transmit the recommended plan generated by AI to the physician's interactive interface, obtain the target taping recommendation items after the physician modifies the plan based on the patient's medical history data, and confirm the final taping instructions through a digital signature; during actual implementation, the AI-generated plan is transmitted to the physician through an encrypted communication protocol. The physician adjusts the taping parameters based on the patient's medical history, such as fracture history and allergy history, and confirms the final plan through a digital signature to ensure that medical responsibility is traceable.

[0069] In traditional solutions, centralized SQL databases store operation logs, which pose a risk of administrator tampering. The blockchain quality control module 310 encrypts key operational data from the taping process and writes it to a distributed ledger, creating an unalterable traceability chain for the taping process. In practical implementation, this application preferably utilizes a Hyperledger Fabric or Ethereum private blockchain. Key data from the taping process, such as the taping path, tension value, and physician signature, is encrypted using the SHA-256 algorithm and written to a distributed ledger, generating a timestamped traceability chain. Patients, physicians, and regulatory agencies can query the on-chain data through hierarchical permissions, ensuring transparency and unalterability.

[0070] For the two-dimensional plane projection guidance, the actual application pressure distribution deviation is relatively large. The patient-side execution feedback module 320 obtains the taping path through augmented reality and superimposes it on the patient's real-time image screen to guide the patient or caregiver to complete the taping operation, and collects the actual taping parameters based on the pressure sensor patch, compares the deviation with the recommended solution, and triggers abnormal warnings or automatic calibration instructions. In actual implementation, the taping path is superimposed on the patient's real-time image screen through AR technology to guide the operation. Pressure sensors, such as flexible piezoelectric films, collect the tension and fit of the patch in real time, and then compare them with the recommended items. If the deviation exceeds the threshold, such as a five percent tension value, a voice warning is triggered or the patch position is automatically calibrated through the PID algorithm.

[0071] In terms of technical effectiveness, this application combines AI with anatomical features and dynamic models to overcome the limitations of traditional taping, which relies on empirical evidence. This can improve muscle tone regulation and pain relief. Remote taping can overcome geographical limitations, allowing grassroots patients to access expert taping solutions through the system and reduce the impact of uneven medical resource distribution. Blockchain technology ensures traceability of the taping process, meeting medical compliance requirements such as CFDA audits. AR guidance and sensor calibration reduce the risk of misuse and prevent adverse effects caused by random taping. Accumulated taping data can be fed back into the AI ​​model to continuously improve the accuracy of recommended solutions.

[0072] See Figure 2In order to solve the problem of acupoint collaborative positioning during image acquisition, the image acquisition module 100 includes a preprocessing submodule 110, a dynamic feature extraction submodule 120 and an anatomical landmark database 130.

[0073] In order to solve the problem in the traditional technical solution that a large number of feature points are lost in fitting and reflective scenes by simply processing the image with histogram equalization, the present application configures the pre-processing submodule 110 and uses the OpenCV library to perform illumination equalization on the original image in actual implementation, eliminates ambient light interference through histogram matching, and uses a non-rigid registration algorithm to align the frontal and lateral images of the patient's area to be taped. In this process, an adaptive illumination equalization algorithm is used to reduce the image contrast difference under different illumination conditions, and non-rigid registration of the frontal and lateral images is performed through thin plate spline transformation, and the registration error is also reduced synchronously, thereby eliminating the image quality fluctuation caused by ambient light interference, establishing a spatial consistency coordinate system for multi-view images, and performing pixel enhancement on different acupuncture points through the three-dimensional muscle motion vector field corresponding to the coordinate system, that is, clarifying the pixels of each acupuncture point.

[0074] In actual implementation, the cumulative distribution function of the original image and the reference image is calculated to establish a mapping relationship between pixel values ​​so that the brightness distribution of the two images tends to be consistent.

[0075] For example, using OpenCV's Histogram Matching technology, the histogram of the image of the area to be taped is adjusted to be consistent with the standard reference template to eliminate ambient light interference (such as strong backlight or shadows). A B-spline deformation model is used to align the frontal and lateral images to compensate for anatomical deformation caused by body position changes. The control point grid is used to optimize the matching error and preserve the muscle morphological characteristics.

[0076] In order to solve the problem that traditional color marking and two-dimensional motion capture cannot display muscle movement, the present application adopts a dynamic feature extraction submodule 120 to track the muscle movement trajectory of the patient when performing the action of the designated area to be taped based on the optical flow method, construct a three-dimensional muscle motion vector field, and quantitatively analyze the symmetry difference between the affected side and the healthy side;

[0077] In practical implementation, local motion vectors are estimated by calculating the spatiotemporal gradients of pixels between consecutive frames. A three-dimensional vector field is generated using a binocular camera or depth sensor. The Lucas-Kanade optical flow algorithm is then used to track the displacement vectors of key muscle points in dynamic videos, constructing a three-dimensional motion trajectory. This quantifies the difference in movement amplitude and velocity between the affected and unaffected muscles. For example, this involves calculating the offset of the trajectory of the deltoid muscle on the affected side during abduction in patients with rotator cuff injuries. If the optical flow algorithm used is a dense optical flow method, the constructed three-dimensional motion vector field can detect micro-movements. Establishing a gentle motion symmetry index can correlate surface skin movement with deep muscle movement, enabling symmetry detection.

[0078] In order to solve the problem that the fixed acupoint coordinate table in the traditional technology does not take into account the differences caused by individual anatomy, the present application stores the coordinate mapping relationship of each acupoint in the area to be taped through the anatomical landmark database 130, and establishes an acupoint-muscle group association model in combination with FACS to achieve the coordinated positioning of the acupoints guiding the anchor points of the patch. In the specific implementation, the synergistic rules of acupoints and muscles are defined by anatomical prior knowledge (such as "Meridian Acupoints"). For example, the Jianyu point corresponds to the activation of the anterior deltoid muscle, and the Quyuan point is associated with the tension regulation of the supraspinatus muscle. The coordinates of the anatomical landmark points of the area to be taped are stored, for example: the Jianyu point and the Quyuan point of the shoulder, and the extended logic of the taping area action coding system (FACS) is combined to establish a mapping relationship between the acupoints and the target muscle group, which can realize personalized acupoint movement. When taping obese patients, the error correction of the patch aiming point positioning can be realized, and the acupoint coordinates are automatically adjusted according to the patient's body shape.

[0079] In actual implementation, the dynamic feature extraction submodule 120 and the anatomical landmark database 130 automatically optimize the mapping relationship of the patella-related acupuncture points by calculating the reduction of the motion trajectory error after eliminating the motion blurred image.

[0080] The AI ​​analysis engine 200 includes a multimodal fusion training module for performing multi-source fusion, thereby solving the problem of parameter solidification of the taping scheme. The multimodal fusion training module includes a prediction submodule 210, a real-time optimization submodule 220 and an interpretable output unit 320, see Figure 3 :

[0081] Traditional visual algorithms for marking anatomical points do not combine electromyographic signals and are not based on the time-frequency characteristics of electromyographic signals. In clinical data, the prediction model constructed solely by imaging has an error rate of over 40% in determining nerve damage. Existing technologies only use models built through muscle movement, which has errors in spatiotemporal alignment. There are spatiotemporal alignment errors between the dynamic signals of the muscles and the actually displayed anatomical markers.

[0082] This application uses a prediction submodule 210 to determine the electromyographic (EMG) features of the taping area based on the marked anatomical feature points, thereby forming an efficacy prediction model. This process is summarized by developing a spatiotemporal synchronization algorithm to align the signals on the surface of the patient's taping area with the motion trajectory of the anatomical landmarks at the millisecond level. The efficacy prediction model, based on a multivariate feature extraction network, can achieve synchronous fusion of multidimensional features in the signal trait fusion, construct a dynamic authority mechanism, and automatically adjust the weight contribution of the factors affecting the fetus based on the type of muscle or nerve injury. In clinical practice, a transfer learning method is used to integrate the House-Brackmann grading data, FDI scale results, and electromyography (EMG) features from clinical studies based on a pre-trained ResNet-50 model to construct an end-to-end efficacy prediction model. In actual implementation, a pre-trained ResNet-50, such as a medical image analysis backbone network verified in a preset knowledge base, is used as a feature extractor. The underlying convolutional layers are frozen to retain common image features, and only the fully connected layers are fine-tuned to adapt to the kinesio taping task. One path processes image data (ResNet-50), and the other processes structured clinical data (MLP). The features are fused through the concatenate layer, and finally the efficacy prediction model outputs the efficacy prediction probability, eliminating the bias of the data.

[0083] In the present application, in order to address the defect of capturing the nonlinear characteristics of nerves, the real-time optimization submodule 220 of the present application performs a time series analysis on historical taping cases based on the efficacy prediction model and the LSTM network, that is: based on the LSTM network, the historical taping cases are subjected to a time series analysis, and the tape stretching parameters in the recommended scheme are dynamically adjusted to adapt to the progressive characteristics of nerve function recovery during the patient's taping cycle; the LSTM network is a bidirectional network architecture, and the input dimensions include the daily amplitude change rate, joint range of motion, enhancement and other multi-dimensional time series characteristics, and then the optimal stretching coefficient is automatically calculated based on the real-time detection value of the nerve conduction velocity (the detection value in the historical taping case), thereby determining the tape stretching parameters. The LSTM network's time series analysis of historical taping cases is also a negative sample reinforcement learning mechanism, which can limit the static parameters of the rabbit skin and summarize the nerve injury recovery process to achieve a positive correlation calculation between the taping cycle and the tape stretching parameters.

[0084] This application addresses the problem that the AI-generated solutions in traditional solutions may not be able to understand the biomechanical mechanism of action. This application is configured with an interpretable output unit 230 to generate muscle state characteristics in the muscle dynamics model based on the tape stretching parameters. During implementation, a visual heat map will be used to identify the expected mechanical effect intensity of the taping solution on a specific muscle group. The constructed muscle dynamics model can display the attenuation gradient of the shear force applied by the tape in the fascia layer, and for any anatomical point, the functional relationship between the signal change rate in the area and the stretching parameters can be displayed. This can achieve real-time identification of muscle state characteristics and tracing of different diagnosis and treatment plans in the mechanical action chain.

[0085] In practice, the AI-recommended taping regimen is linked to the mechanical strength of muscle groups using gradient-weighted class activation mapping. Specifically, the feature map of the last convolutional layer of ResNet-50 is extracted. For the target efficacy prediction category, such as spasticity relief, the gradient is calculated to generate a weight matrix. The weighted feature maps are then overlaid onto the patient image to create a heat map, identifying areas of high stress (e.g., red indicates concentrated tension on the trapezius muscle due to the tape).

[0086] Blockchain quality control module 310 can realize the tamper-proof traceability chain of the taping process, see Figure 4 ;

[0087] Hyperledger Fabric chaincode is the core of smart contracts, and its function is to handle business logic processing, data management, security control and other aspects of the blockchain network.

[0088] This application targets centralized image storage devices. Cross-hospital retrieval requires manual application. The integrity of the data relies on MD5 hash comparison, which cannot prevent malicious tampering by administrators and the high loss rate of clinical image data. The high-resolution image raw data, i.e., high-definition image data, is stored based on the IPFS protocol through the tagging supervision submodule 311. The file hash value is written into the Hyperledger Fabric chain code to ensure data integrity and cross-institutional retrieval efficiency.

[0089] In actual implementation, high-resolution imaging data (such as DICOM format) is stored in blocks on IPFS (InterPlanetary File System), generating a unique file hash (CID). IPFS addresses the single point of failure risk of traditional cloud storage through content addressing and decentralized node redundant storage. The file hash is written to the chain code (smart contract) of Hyperledger Fabric to ensure that the data cannot be tampered with. For example, the CID of a patient image is stored using the putState function, and subsequent access requires hash consistency verification. Medical institutions use IPFS nodes to achieve cross-regional data retrieval, quickly obtain original data, and record access logs on the blockchain.

[0090] In view of the fact that the system alarm threshold in traditional solutions is fixed and cannot adapt to the needs of different types of diseases, the smart contract-driven quality control rule library 312 defines automated compliance checks including patch replacement cycle threshold, tension deviation tolerance, and physician response time. Illegal operations trigger on-chain alarms and quality control score downgrades. The automated execution of quality control rules reduces the rate of operational violations.

[0091] Pre-set quality control rules, for example: if the patch is not changed after 72 hours, an alarm will be triggered; if the tension exceeds the recommended value by ±5%, it will be marked as "irregular operation", etc.

[0092] The use of AES encryption causes the failure of on-chain data analysis, and multi-institutional sharing requires third-party notarization. The privacy computing submodule 313 of this application uses homomorphic encryption technology to desensitize patients' sensitive information, and performs zero-knowledge proof verification when sharing on-chain data to ensure compliance with data protection regulations such as GDPR, and compliance with data protection regulations such as GDPR based on compliance values ​​to achieve preset target values. In actual implementation, homomorphic encryption (HE) uses the Paillier algorithm to encrypt patient sensitive information (such as name, medical history), supports direct calculation of ciphertext (such as calculating quality control scores), and does not require decrypting the original data. When sharing data, the verifier verifies the data compliance (such as "patient age > 18 years old") through the zk-SNARKs protocol, without exposing the specific age value, which complies with the "data minimization" principle of GDPR.

[0093] The patient-side execution feedback module 320 of this application includes smart patch hardware 321 and efficacy tracking submodule 323; it is used to upload the patient's patch data, see Figure 5 ;

[0094] This application targets external pressure sensing patches, which cannot capture dynamic pressure changes. Moreover, if the pressure sensing device is installed independently of the patch, there will be angle deviations during measurement, and lymphatic flow obstruction may occur.

[0095] Smart patch hardware 321 integrates a flexible pressure sensor array and a micro inertial measurement unit to monitor the pressure distribution, tensile deformation and angular offset data of the patch and skin contact surface in real time; this application uses piezoresistive or capacitive flexible sensors (such as Tekscan pressure mapping technology) embedded in the inner layer of the patch to monitor the pressure distribution of the patch and skin contact surface in real time (accuracy ±5mmHg). The sensor array captures local pressure differences through a grid layout (such as an 8×8 dot matrix) to identify areas where the patch edge is curled or not firmly adhered. An integrated 6-axis IMU (accelerometer + gyroscope) with a sampling frequency of ≥100Hz measures the tensile deformation (such as the strain rate along the muscle fiber direction) and angular offset (such as the angular deviation between the patch and the muscle direction) of the patch in real time. The data is transmitted to the patient-side APP via low-power Bluetooth.

[0096] The identification submodule 322 is based on an adaptive calibration algorithm. When it is detected that the deviation between the actual taping parameters and the recommended scheme exceeds the preset threshold, the correction guide arrow is dynamically rendered through the AR interface, and a vibration prompt is given to indicate an operation error. The calibration process is triggered when the sensor data exceeds the threshold by presetting the clinical safety threshold (such as pressure deviation ±10%, angle deviation>5°). The correction arrow is superimposed on the real-time image through ARKit or ARCore to indicate the direction the operator needs to adjust (such as "stretch 2cm to the left"). At this time, the arrow color code will be displayed (red = urgent calibration, yellow = slight deviation).

[0097] The efficacy tracking submodule 323 is used to trigger the patient's self-inspection process of the taping area on a daily basis, calculate the HB score improvement rate through video analysis, generate a taping progress report and synchronize it to the physician-side decision support system, and guide the patient to complete standardized movements on a daily basis, such as frowning, smiling, and puffing cheeks, and collect videos through the front camera of the mobile phone.

[0098] The remote audit interface 300 of the present application is also used to prevent the homogeneity and poor effect of diagnosis and treatment plans, see Figure 6 :

[0099] First, since centralized servers may store a large amount of medical data and may leak privacy, the interface of this application is configured with a learning framework 301, which is used to establish a federated learning framework across medical institutions. Each node hospital stores patient taping data locally, uploads model gradients to the central server through homomorphic encryption and secure aggregation protocols, and updates the global efficacy prediction model. Specifically, the horizontal federated learning model defined in the knowledge base is adopted. Each node hospital shares the same feature space, such as taping parameters and efficacy scores, and the patient samples are different. At the same time, the patient taping data, namely the patient's images, pressure sensor logs, HB scores and other data will be retained on the local server.

[0100] When implementing this application, traditional fixed noise parameters cannot adapt to different data sensitivities. Gaussian noise is added before gradient transmission to ensure that individual patient data cannot be reversely inferred, meeting HIPAA medical privacy protection standards; gradients are encrypted using the Paillier encryption algorithm, supporting direct aggregation of ciphertexts, such as additive homomorphism, to ensure that the central server cannot reversely deduce individual data; and Gaussian noise that meets differential privacy is added before gradient transmission, such as a noise scale of σ = 0.5, making it impossible for attackers to reversely infer the original data through the gradient, and the privacy leakage risk is controlled at ε = 1.2.

[0101] This application targets the regression linear model in the traditional solution, which cannot capture the factor interaction. The interface is configured with an XGBoost model 302, which is trained based on multi-center data to identify the key factor combinations that affect the efficacy and generate a priority list of combined taping solutions for different stages of the disease. Based on the multi-center data aggregated by federated learning, the XGBoost model is trained to analyze efficacy-related factors, such as the resulting taping force, disease stage, HB baseline score, and other efficacy-related factors.

[0102] The contribution of each factor is then quantified using the SHAP value, for example:

[0103] A counter 303 is also provided for generating task counting information according to the task thread of the signal sequence after being allocated by the scheduler; wherein the technical information of the synchronization task has a synchronization identifier.

[0104] In the actual implementation process, the asynchronous federated learning framework can allow node hospitals to upload gradients asynchronously, avoiding network delay bottlenecks.

[0105] The AI ​​analysis engine 200 of the present application also includes a reinforcement learning agent module 240 and a stagnation judgment module 250; it has the ability to monitor the taping area in real time, thereby judging the taping effect and updating the recommended items. Figure 7 :

[0106] Traditional taping programs are based on fixed rules, such as daily adjustments, which cannot identify changes in muscle status in real time, resulting in insufficient taping and ineffective taping intervention.

[0107] This application obtains real-time image data of the patient's taping area through a reinforcement learning agent module, determines the real-time input status of the taping area, and generates a first set of recommended items based on the displacement of the patch anchor point and the tension adjustment amplitude; because the input of the existing technology only contains static image features and ignores the spatiotemporal correlation of the anchor point displacement, this application uses real-time image data to continuously optimize the changing parameters of the patch anchor point position and tension adjustment. This process integrates the muscle texture features of the real-time image and the anchor point displacement vector.

[0108] In view of the decrease in detection sensitivity and the stagnation of abnormal muscle regions in the nerve injury taping in the prior art, a double-threshold monitoring mechanism is established in the present application:

[0109] The change rate of muscle tension standard deviation is determined by configuring a primary threshold value;

[0110] The nerve injury classification is determined by configuring a secondary threshold value;

[0111] When the double thresholds are met at the same time, the PPO algorithm is started to recalculate the parameter exploration strategy, and then the taping parameters in the first recommended item set are updated according to the calculation result to generate a second recommended item set, so that the parameters during taping can be quickly adjusted.

[0112] Referring to Figure 8 In the processing work of the feedback module 320 at the patient end, a virtual social environment is also built, and the taping recommendation items that need to be updated are determined according to the action tasks in the rehabilitation process.

[0113] The traditional scheme is a 2D plane training interface, which lacks social motivation and action specification feedback. The present application builds a virtual social environment, determines the action tasks of the taping area, and determines the completion degree of the muscle activity of the affected side of the patient when performing the action tasks. In actual implementation, a 3D rehabilitation community can be built in the VR helmet, and the patient avatar can complete tasks with other rehabilitators, capture the motion trajectory through Kinect V2, compare the actions, determine the muscle state of the patient, and determine the muscle activity state. This way, the patient also does not have psychological burnout.

[0114] In view of the muscle compensatory fatigue in the rehabilitation process caused by the fixed difficulty in the traditional scheme, the present application dynamically adjusts the task difficulty level according to the electromyographic signal intensity in the action task, detects whether the muscle strength is uniform, and reduces the response time threshold of the action task when it is not uniform. In implementation, the response time of the patient at different action difficulties is determined by calculating the bilateral muscle activation balance index to determine the recovery state of the patient.

[0115] The unit action potential of the patient before and after the action task before and after the preset rehabilitation period is quantified to determine the recommended item taping ratio. In determining the recovery state of the patient, the unit action point of the patient's muscle in a unit of time before and after the rehabilitation period can also be determined to decompose the movement unit, calculate the regeneration index of the nerve unit, and determine whether the taping coverage density needs to be increased, the stretching coefficient needs to be adjusted, or the taping anchor point distance needs to be reduced.

[0116] The patient end execution feedback module 320 of the present application can also judge the effect of the paste and stick of the patient in combination with the acupoint stimulation from the perspective of traditional Chinese medicine (mainly based on electronic physiotherapy acupoint stimulation), and then adjust the stretching force during the application of the medicine package. See Figure 9 :

[0117] In the traditional scheme, when acupoint stimulation is performed, the stretching of the acupoint patch is not considered, which leads to a mismatch between the clinical paste and stick and the acupoint strength, resulting in spasm. The present application configures an electric stimulation parameter library to store the optimal stimulation frequency, waveform and duration of the acupoint to be pasted and stuck, and is associated with the space-time correlation database of the paste and stick scheme; by establishing the mechanical mapping relationship between the acupoint coordinates and the patch anchor points, and automatically matching the electric stimulation frequency with the real-time tension of the patch, the interference of the mechanical field of the patch caused by the fixed parameter electric stimulation is relieved, and the rehabilitation effect is not good.

[0118] In view of the problem that the traditional empirical formula may not be able to determine the non-linear relationship between the patch tension and the depth of acupoint stimulation, the present application configures a cooperative control algorithm to calculate the superposition effect of the acupuncture depth and the mechanical action of the patch in the same muscle group area before and after acupuncture and sticking, and to determine whether there is local microcirculation overload; by constructing a three-dimensional grid model of the muscle group area, the vector superposition stress field of the acupuncture depth and the patch tension is calculated, and then when the local stress exceeds the preset value, an overload alarm is executed; thereby reducing the electric acupuncture intensity of acupuncture.

[0119] The present application also configures an Internet of Things synchronization interface for detecting whether the stretching force of the patch stretching function returns to the preset tension parameter when the acupuncture device starts the penetration operation. The present application ensures that the penetration operation and the patch tension adjustment are synchronized, and the patch tension is detected at the moment of penetration of the acupuncture device. If the deviation is greater than the preset value, a PID adjustment is triggered.

[0120] The patient end execution feedback module 320 is also used to fuse the ultrasonic elastography data and real-time electromyographic signals in the augmented reality interface to generate a three-dimensional biomechanical action vector field, and to optimize the spatial distribution density of the patch anchor points by a gradient descent algorithm.

[0121] The technical scheme fuses the ultrasonic elastography data and real-time electromyographic signals in the augmented reality interface to construct a three-dimensional biomechanical action vector field, and realizes the precise optimization of the distribution of the patch anchor points.

[0122] In terms of technical principles:

[0123] The real-time electromyographic signals are extracted by ultrasonic elastography to capture the elastic modulus of muscle tissue, and the timing characteristics of motor unit recruitment are extracted;

[0124] The two types of data are mapped to the same coordinate system by a spatial registration algorithm to generate a three-dimensional vector field containing the shear direction, tissue stiffness and muscle activation intensity.

[0125] According to the gradient descent algorithm based on the mechanical gradient of the vector field, the anchor point position is iteratively calculated to minimize the pressure distribution variance. The optimization process adopts an adaptive learning rate to achieve accurate recommendation of patch anchor point distribution.

[0126] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A remote kinesio taping treatment guidance and quality control system, the system is for non-disease treatment purposes, characterized in that: include: An image acquisition module is used to acquire high-definition image data of the treatment area through a smart terminal device and mark anatomical feature points; the high-definition image data includes static images and dynamic muscle movement video streams; The AI ​​analysis engine uses a convolutional neural network to build a muscle dynamics model of the taping area, extract features from high-definition image data, identify areas of abnormal muscle tension and grade the degree of nerve damage, and generate initial taping path recommendations; The remote review interface transmits the initial taping path recommendations generated by AI to the physician-side interactive interface, obtains the target taping recommendations after the physician-side has revised the plan based on the patient's medical history data, and confirms the final taping instructions through digital signatures; The blockchain quality control module is used to encrypt the key operation data in the final taping instructions and write them into the distributed ledger, generating an unalterable traceability chain for the taping process; The patient-side execution feedback module obtains the target taping path corresponding to the final taping instruction through augmented reality, superimposes it on the patient's real-time image, collects the actual taping parameters through the pressure sensor patch, and compares the deviation with the target taping recommendation, triggering abnormal warnings or automatic calibration instructions.

2. A remote kinesiology taping treatment guidance and quality control system as claimed in claim 1, characterized in that: The image acquisition module includes: The preprocessing submodule uses a preset OpenCV library to perform illumination equalization on the original image of the taping area, and aligns the original image after illumination equalization with the anteroposterior and lateral images of the taping area of ​​the patient using a non-rigid registration algorithm to generate a first image; The dynamic feature extraction submodule tracks the muscle motion trajectory of the taping area in the first image based on the optical flow method and constructs a three-dimensional muscle motion vector field; The anatomical landmark database is used to perform pixel enhancement on each acupuncture point in the taping area according to the three-dimensional muscle motion vector field to generate high-definition image data.

3. The remote kinesiology taping treatment guidance and quality control system according to claim 1, characterized in that: The AI ​​engine includes a multimodal fusion training module, including: Prediction submodule: used to determine the electromyographic characteristics of the taping area based on the marked anatomical feature points, and form a therapeutic effect prediction model; The real-time optimization submodule uses a time-series analysis of historical taping cases based on an efficacy prediction model and an LSTM network to determine the tape stretching parameters for the target taping recommendation. These parameters are positively correlated with the progressive nature of neurological recovery within the patient's taping cycle. The interpretable output unit determines the muscle state characteristics in the muscle dynamics model based on the tape stretch parameters.

4. The remote kinesiology taping treatment guidance and quality control system according to claim 1, characterized in that: The blockchain quality control module includes: The tagging supervision submodule stores high-definition image data based on the IPFS protocol, determines the file hash value, and writes the file hash value to the Hyperledger Fabric chain code rate; Smart contract-driven quality control rule base, which defines automated compliance checks for patch replacement cycle thresholds, tension deviation tolerance, and physician response time, and responds to violations; The privacy computing sub-module uses homomorphic encryption technology to desensitize patients' sensitive information, performs zero-knowledge proof verification when sharing data on the chain, and generates compliance values.

5. The remote kinesiology taping treatment guidance and quality control system according to claim 1, characterized in that: The patient-side execution feedback module includes: Smart patch hardware integrates a flexible pressure sensor array and a micro inertial measurement unit to monitor the pressure distribution, tensile deformation, and angular offset data of the patch-skin contact surface in real time; Identification submodule: Based on an adaptive calibration algorithm, when the deviation between the actual taping parameters and the recommended benchmark taping parameters exceeds a preset threshold, a correction guidance arrow is dynamically rendered through the AR interface and a vibration prompt message is generated; The efficacy tracking sub-module triggers the patient's movement self-inspection process in the taping area on a daily basis, calculates the HB score improvement rate through video analysis, generates a taping progress report and synchronizes it to the physician end.

6. The remote kinesiology taping treatment guidance and quality control system according to claim 1, characterized in that: The remote audit interface is also used to: Establish a federated learning framework across medical institutions. Each node hospital locally stores patient taping data, uploads model gradients to a central server via homomorphic encryption and secure aggregation protocols, and updates the global efficacy prediction model in real time. An XGBoost model was trained based on multi-center data to identify key factor combinations that affect efficacy and generate a priority list of combined taping regimens for different disease stages. And according to the task thread after the signal sequence is assigned by the scheduler, task counting information is generated; wherein, the technical information of the synchronization task has a synchronization identifier.

7. The remote kinesiology taping treatment guidance and quality control system according to claim 6, characterized in that: The AI ​​analysis engine also includes: A reinforcement learning agent module is used to obtain real-time image data of the taping area of ​​the patient, determine the real-time input state of the taping area, and generate a first set of recommended items based on the displacement of the taping anchor point and the tension adjustment amplitude; Stagnation judgment module: When the level of abnormal muscle tension area and nerve damage degree in real-time image data does not change within the preset supervision time, it automatically triggers the PPO algorithm to recalculate the taping parameters in the first recommendation item set and generate the second recommendation item set.

8. The remote kinesiology taping treatment guidance and quality control system according to claim 5, characterized in that: The patient-side execution feedback module is further used for: Construct a virtual social environment by determining the movement tasks of the taped area and determining the degree of completion of the side muscle activity when performing the movement tasks; Dynamically adjust the task difficulty level based on the strength of the electromyographic signal during the action task, detect whether the muscle strength is uniform, and lower the response time threshold of the action task if it is uneven; By measuring the patient's unit action potential before and after the movement task within the preset rehabilitation cycle, the progress of nerve regeneration is quantified and the proportion of patches for the recommended items is determined.

9. The remote kinesiology taping treatment guidance and quality control system according to claim 1, characterized in that: The patient-side execution feedback module also includes: Configure an acupoint electrical stimulation parameter library to store the optimal stimulation frequency, waveform, and duration of the acupoints in the area to be taped, and link it with the spatiotemporal association database of the taping scheme; Configure a collaborative control algorithm to calculate the superposition effect of acupuncture depth and the mechanical action of the tape in the same muscle group before and after acupuncture taping through finite element simulation to determine whether there is local microcirculatory overload; Configure the IoT synchronization interface to detect whether the stretching force of the patch stretching function returns to the preset tension parameters when the acupuncture device starts the penetration operation.

10. The remote kinesiology taping treatment guidance and quality control system according to claim 1, characterized in that: The patient-side execution feedback module is also used to fuse ultrasonic elastic imaging data with real-time electromyographic signals in an augmented reality interface to generate a three-dimensional biomechanical action vector field and optimize the spatial distribution density of patch anchor points through a gradient descent algorithm.

Images