A monitoring system for arthritis correction process based on motion monitoring

Abstract

The present invention relates to an arthritis correction process monitoring system based on motion monitoring, which integrates a monitoring host, a pressure sensor, an image acquisition module, a storage module and a user terminal. The system monitors the pressure distribution of both feet during walking in real time through the pressure sensor in the shoe, and combines the infrared and visible light cameras in the patient's activity area to collect dynamic images of key body parts marked by the wearer's light emitter. Using a deep learning algorithm, the system can accurately align infrared and visible light images, integrate pressure changes, joint movements and gait profile data, and realize intelligent analysis and prediction of the correction stage. The system not only provides personalized correction effect feedback, but also optimizes the treatment plan and improves the accuracy and efficiency of arthritis management. This technology greatly improves the patient's treatment experience through non-invasive and real-time monitoring, while providing doctors with powerful remote monitoring and analysis tools, promoting the modernization of arthritis treatment.

Description

Technical Field

[0001] The invention relates to the field of medical devices, and in particular to an arthritis correction process monitoring system based on motion monitoring. Background Art

[0002] As the aging population accelerates, the incidence of arthritis, as a common chronic disease, continues to rise worldwide, seriously affecting the quality of life of patients. Arthritis is mainly manifested by joint pain, stiffness and dysfunction, among which gait abnormality is one of the most direct clinical manifestations. In order to effectively monitor the progression of the disease and evaluate the treatment effect, the corrective treatment of arthritis pays more and more attention to personalization and dynamic monitoring.

[0003] Traditionally, the assessment of arthritis mainly relies on the clinical experience of doctors, the subjective description of patients, and regular X-ray or MRI examinations. These methods are not only highly subjective and have low detection frequency, but also difficult to reflect subtle changes in the treatment process in real time. Therefore, it is particularly important to develop a technology that can monitor the arthritis correction process in real time, objectively, and accurately.

[0004] In recent years, with the rapid development of the Internet of Things, intelligent sensing technology and computer vision, health management systems based on motion monitoring have gradually become a research hotspot. By integrating a variety of sensor devices, such as pressure sensors and image acquisition devices, motion monitoring systems can continuously and non-invasively collect a large amount of physiological and motion data during patients' daily activities. These data provide the possibility of accurately analyzing the range of joint motion, gait patterns and correction effects.

[0005] Pressure sensing technology plays an important role in biomechanical analysis, especially for evaluating gait characteristics and load distribution. The built-in pressure sensor in the sole can monitor the force on the patient's feet in real time when walking, revealing abnormal gait patterns, such as uneven plantar pressure distribution and changes in gait cycle, which are important indicators for arthritis diagnosis and efficacy evaluation.

[0006] At the same time, image acquisition and processing technology, especially the application of infrared light technology, provides a new way for non-contact three-dimensional motion capture. The combination of infrared illuminators and special cameras can penetrate clothing to capture the deep joint movement trajectory, and combined with the surface morphology captured by visible light cameras, it provides three-dimensional and comprehensive information for kinematic and dynamic analysis. Through deep learning analysis of these image data, the spatiotemporal characteristics of the patient's joint movement can be accurately extracted to further evaluate the effect of the correction device and the patient's rehabilitation progress. Summary of the invention

[0007] In order to solve the above problems, the present invention provides an arthritis correction process monitoring system based on motion monitoring, including a monitoring host, a pressure sensor, an image acquisition module, a storage module and a user terminal;

[0008] The monitoring host is connected to the pressure sensor, image acquisition module, storage module and user terminal to control the operation of the entire monitoring system;

[0009] The pressure sensor is installed on the sole of the shoe to detect the pressure on the patient's feet in real time and transmit it to the monitoring host in real time to monitor the patient's walking process;

[0010] The image acquisition module is set at the patient's activity site to collect video images of the patient when he / she is in action and send them to the monitoring host;

[0011] The storage module is used to store the collected data and analysis models for the monitoring host to analyze the patient's correction stage;

[0012] The user terminal sets a reminder interface for displaying the analysis results to doctors and patients.

[0013] The pressure sensor is set on the bottom plate of the shoe. Each shoe is set with two pressure sensors, and two shoes are set with a total of four pressure sensors; each shoe is set with a pressure detection chip, which is connected to the two pressure sensors on the shoe, collects the data of the pressure sensors and sends it to the monitoring host in real time;

[0014] A pressure sensor is set at the front and back of each shoe. The front pressure sensor is located below the base of the toes of the sole of the foot, and the back pressure sensor is located below the base of the sole of the foot. The pressure detection chip is set under the arch of the foot.

[0015] The image acquisition module includes a wearable light emitter, which is provided with a strap and an infrared light emitter; the strap is used to fix the infrared light emitter to the patient's feet, calves, knees, thighs and waist; the infrared light emitter is a point light source, and the wavelength of the emitted infrared light has a penetration rate of more than 50% for fabrics;

[0016] Wear 9 or more light emitters, at least on the feet, calves, knees, thighs and waist;

[0017] The image acquisition module is also provided with an infrared camera and a visible light camera. The infrared camera is used to collect the light emitted by the wearable light emitter in real time, and the visible light camera is used to collect the visible light image of the patient in real time.

[0018] The image acquisition module sends the collected infrared images and visible light images to the monitoring host in real time.

[0019] The monitoring host performs position registration of infrared images and visible light images:

[0020] Feature point extraction: In the infrared image and the corresponding visible light image, algorithms such as SIFT (Scale Invariant Feature Transform), ORB (Oriented FAST and Rotated BRIEF) or AKAZE (Accelerated-KAZE) are used to automatically extract stable feature points;

[0021] Feature matching: Match the feature points in the two images through KNN (k-nearest neighbor) search or BFMatcher (brute force matching) algorithm, use the RatioTest strategy to remove false matches, and obtain the corresponding relationship of feature points;

[0022] Using the matched feature point pairs, the infrared image and the visible light image are geometrically transformed and registered by using perspective transformation or homography matrix calculation method;

[0023] The monitoring host accurately calibrates the position of the wearable light emitter in the infrared image as follows:

[0024] Using image segmentation algorithm, the light spot area produced by the light emitter is identified and segmented from the infrared image;

[0025] The contour of each light spot is extracted, and the actual position of each wearable light emitter is calculated using the centroid method according to the shape, size and intensity distribution characteristics of the light spot, that is, the centroid of the light spot shape is determined and the centroid is used as the actual position of the wearable light emitter;

[0026] The monitoring host extracts the contour of the patient in the visible light image as follows:

[0027] First, the visible light image is preprocessed, including grayscale conversion and filtering to reduce noise interference in the image;

[0028] Apply edge detection algorithms to identify boundaries in images; edge detection algorithms include Canny edge detection, Sobel operator or Laplacian operator;

[0029] Based on the results of edge detection, a contour extraction algorithm is used to obtain closed contours in the image.

[0030] The centroid of the spot shape is calculated as follows:

[0031] Traverse each light spot, for each pixel (x i ,y i ) Its gray value I(x i ,y i) is regarded as the "mass" of the position; these "masses" and their corresponding coordinates need to be accumulated to calculate the center of mass; the center of mass coordinates of the spot (C x ,C y ) is calculated using the following weighted average formula:

[0032]

[0033]

[0034] Here, the summation is performed on all pixels i within the spot, x i and i are the horizontal and vertical coordinates of the pixel, I(x i ,y i ) is the intensity value of the pixel.

[0035] The monitoring host integrates an analysis module, which analyzes the patient's correction stage based on the data from the pressure sensor and image acquisition module, as follows:

[0036] Time Series Data Integration:

[0037] Directly collect the real-time pressure values ​​of the front and back sole pressure sensors to form two sets of one-dimensional sequence data that change over time, with two points on each foot for a total of four data sequences; ensure that the pressure data sequence is accurately aligned with the time axis of the position change of the worn light device in the infrared image and the change of the patient's contour in the visible light image, and use timestamps to ensure synchronization across data sources;

[0038] Feature extraction:

[0039] The following pressure features are extracted:

[0040] P peak : The maximum pressure value sequence of the front and back soles of the feet, reflecting the key load-bearing conditions during walking;

[0041] P ratio : Front and rear foot pressure ratio sequence to evaluate gait balance;

[0042] P rate : Pressure change rate, capturing gait dynamics;

[0043] Extract the following wearable light device position features:

[0044] G ankle ,G knee ,G hip : Represents the angle change sequence of ankle joint, knee joint and hip joint respectively, assessing the range of motion and flexibility of lower limb joints;

[0045] G amplitude : Joint motion amplitude sequence, reflecting the range of motion;

[0046] G smoothness : Movement smoothness index, obtained through analysis of joint angular velocity and acceleration;

[0047] The following patient contour features are extracted:

[0048] C gait_cycle : gait cycle sequence, obtained by similarity matching between contour sequences;

[0049] C step_width ,C step_length : Step width and step length sequence to evaluate gait characteristics;

[0050] C swing_range : Limb swing amplitude sequence, reflecting the movement amplitude of lower limbs;

[0051] Deep learning model construction:

[0052] A multimodal fusion network is used, where:

[0053] Pressure feature processing: Using bidirectional LSTM to process P peak ,P ratio ,P rate , capturing the time dependence of pressure changes;

[0054] Joint angle change processing: LSTM processing G ankle ,G knee ,G hip ,G amplitude ,G smoothness , extract time series features;

[0055] Contour feature processing: Lightweight CNN processes contour image sequences and extracts C gait_cycle ,C step_width ,C step_length ,C swing_range The spatial and temporal characteristics of

[0056] Feature fusion and prediction: The outputs of BLSTM, LSTM, and CNN are fused through the attention mechanism to generate a comprehensive feature vector, which is input into the fully connected layer for classification in the correction stage.

[0057] Model training and validation are as follows:

[0058] Data segmentation: 80% of the data is used as the training set, 10% as the validation set, and the remaining 10% as the test set;

[0059] Loss function: The cross entropy loss function is combined with the customized correction stage classification loss and trend prediction error loss.

[0060] The beneficial effects of the present invention are:

[0061] This system integrates multi-source data acquisition technologies such as pressure sensors and image acquisition modules to achieve all-round, high-precision monitoring of the correction process of arthritis patients. The real-time capture of the pressure changes of the front and back soles, joint movement trajectories, and gait morphology enables doctors to accurately evaluate the correction effect, adjust the treatment plan in time, and improve the pertinence and effectiveness of the treatment.

[0062] Through the analysis of fused data by deep learning models, the system can identify each patient's unique correction pattern and progression rate, provide data support for doctors, and help formulate more personalized treatment plans. This promotes the scientificity and accuracy of treatment plans and improves the success rate of patient recovery.

[0063] The user terminal settings can not only display the monitoring and analysis results in real time, but also provide early warning prompts according to the changes in the patient's correction stage. This helps patients to manage themselves, respond to doctor's orders in a timely manner, and reduce unnecessary risks of disease deterioration or recurrence. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0065] Attached Figure 1 It is a schematic diagram of the structure of the present invention;

[0066] Attached Figure 2 It is a schematic diagram of the arrangement of the pressure sensor of the present invention;

[0067] Attached Figure 3 This is a schematic diagram of the patient's contour structure and joint lines captured by the image acquisition module of the present invention.

[0068] Among them: 1 pressure sensor, 2 pressure detection chip. DETAILED DESCRIPTION

[0069] Embodiment 1:

[0070] See also Figure 1-Figure 3 , the present invention provides an arthritis correction process monitoring system based on motion monitoring, including a monitoring host, a pressure sensor, an image acquisition module, a storage module and a user terminal;

[0071] The monitoring host is connected to the pressure sensor, image acquisition module, storage module and user terminal to control the operation of the entire monitoring system;

[0072] The pressure sensor is installed on the sole of the shoe to detect the pressure on the patient's feet in real time and transmit it to the monitoring host in real time to monitor the patient's walking process;

[0073] The image acquisition module is set at the patient's activity site to collect video images of the patient when he / she is in action and send them to the monitoring host;

[0074] The storage module is used to store the collected data and analysis models for the monitoring host to analyze the patient's correction stage;

[0075] The user terminal sets a reminder interface for displaying the analysis results to doctors and patients.

[0076] The pressure sensor is set on the bottom plate of the shoe. Each shoe is set with two pressure sensors, and two shoes are set with a total of four pressure sensors; each shoe is set with a pressure detection chip, which is connected to the two pressure sensors on the shoe, collects the data of the pressure sensors and sends it to the monitoring host in real time;

[0077] A pressure sensor is set at the front and back of each shoe. The front pressure sensor is located below the base of the toes of the sole of the foot, and the back pressure sensor is located below the base of the sole of the foot. The pressure detection chip is set under the arch of the foot.

[0078] The design and implementation of the pressure sensor module demonstrates significant superiority. It integrates the sole of the shoe to achieve real-time monitoring of the pressure distribution of the feet of arthritis patients when walking or standing. The pressure sensor module provides a quantitative assessment tool by continuously and dynamically recording the pressure changes in the front and back soles of the feet.

[0079] The image acquisition module includes a wearable light emitter, which is provided with a strap and an infrared light emitter; the strap is used to fix the infrared light emitter to the patient's feet, calves, knees, thighs and waist; the infrared light emitter is a point light source, and the wavelength of the emitted infrared light has a penetration rate of more than 50% for fabrics;

[0080] Wear 9 or more light emitters, at least on the feet, calves, knees, thighs and waist;

[0081] The image acquisition module is also provided with an infrared camera and a visible light camera. The infrared camera is used to collect the light emitted by the wearable light emitter in real time, and the visible light camera is used to collect the visible light image of the patient in real time.

[0082] The image acquisition module sends the collected infrared images and visible light images to the monitoring host in real time.

[0083] The monitoring host performs position registration of infrared images and visible light images:

[0084] Feature point extraction: In the infrared image and the corresponding visible light image, algorithms such as SIFT (Scale Invariant Feature Transform), ORB (Oriented FAST and Rotated BRIEF) or AKAZE (Accelerated-KAZE) are used to automatically extract stable feature points;

[0085] Feature matching: Match the feature points in the two images through KNN (k-nearest neighbor) search or BFMatcher (brute force matching) algorithm, use the RatioTest strategy to remove false matches, and obtain the corresponding relationship of feature points;

[0086] Using the matched feature point pairs, the infrared image and the visible light image are geometrically transformed and registered by using perspective transformation or homography matrix calculation method;

[0087] The monitoring host accurately calibrates the position of the wearable light emitter in the infrared image as follows:

[0088] Using image segmentation algorithm, the light spot area produced by the light emitter is identified and segmented from the infrared image;

[0089] The contour of each light spot is extracted, and the actual position of each wearable light emitter is calculated using the centroid method according to the shape, size and intensity distribution characteristics of the light spot, that is, the centroid of the light spot shape is determined and the centroid is used as the actual position of the wearable light emitter;

[0090] This design overcomes the problems of clothing occlusion and light limitation in traditional visual monitoring, significantly enhancing the accuracy and practicality of monitoring. Infrared light can effectively penetrate clothing and illuminate the light emitters bound to key body parts, clearly recording the trajectory of joint movement even in complex environments. Using advanced image processing algorithms, the system can accurately extract and analyze these light spot data, converting them into key indicators such as joint angles and range of motion, providing intuitive biomechanical parameters.

[0091] The monitoring host extracts the contour of the patient in the visible light image as follows:

[0092] First, the visible light image is preprocessed, including grayscale conversion and filtering to reduce noise interference in the image;

[0093] Apply edge detection algorithms to identify boundaries in images; edge detection algorithms include Canny edge detection, Sobel operator or Laplacian operator;

[0094] Based on the results of edge detection, a contour extraction algorithm is used to obtain closed contours in the image.

[0095] The centroid of the spot shape is calculated as follows:

[0096] Traverse each light spot, for each pixel (x i ,y i ) Its gray value I(x i ,y i ) is regarded as the "mass" of the position; these "masses" and their corresponding coordinates need to be accumulated to calculate the center of mass; the center of mass coordinates of the spot (C x ,C y ) is calculated using the following weighted average formula:

[0097]

[0098]

[0099] Here, the summation is performed on all pixels i within the spot, x i and i are the horizontal and vertical coordinates of the pixel, I(x i ,y i ) is the intensity value of the pixel.

[0100] The monitoring host integrates an analysis module, which analyzes the patient's correction stage based on the data from the pressure sensor and image acquisition module, as follows:

[0101] Time Series Data Integration:

[0102] Directly collect the real-time pressure values ​​of the front and back sole pressure sensors to form two sets of one-dimensional sequence data that change over time, with two points on each foot for a total of four data sequences; ensure that the pressure data sequence is accurately aligned with the time axis of the position change of the worn light device in the infrared image and the change of the patient's contour in the visible light image, and use timestamps to ensure synchronization across data sources;

[0103] Feature extraction:

[0104] The following pressure features are extracted:

[0105] P peak : The maximum pressure value sequence of the front and back soles of the feet reflects the key load-bearing conditions during walking;

[0106] P ratio : Front and rear foot pressure ratio sequence to evaluate gait balance;

[0107] P rate : Pressure change rate, capturing gait dynamics;

[0108] Extract the following wearable light device position features:

[0109] G ankle ,G knee ,Ghip : Represents the angle change sequence of ankle joint, knee joint and hip joint respectively, assessing the range of motion and flexibility of lower limb joints;

[0110] G amplitude : Joint motion amplitude sequence, reflecting the range of motion;

[0111] G smoothness : Movement smoothness index, obtained through analysis of joint angular velocity and acceleration;

[0112] The following patient contour features are extracted:

[0113] C gait_cycle : gait cycle sequence, obtained by similarity matching between contour sequences;

[0114] C step_width ,C step_length : Step width and step length sequence to evaluate gait characteristics;

[0115] C swing_range : Limb swing amplitude sequence, reflecting the movement amplitude of lower limbs;

[0116] Deep learning model construction:

[0117] A multimodal fusion network is used, where:

[0118] Pressure feature processing: Using bidirectional LSTM to process P peak ,P ratio ,P rate , capturing the time dependence of pressure changes;

[0119] Joint angle change processing: LSTM processing G ankle ,G knee ,G hip ,G amplitude ,G smoothness , extract time series features;

[0120] Contour feature processing: Lightweight CNN processes contour image sequences and extracts C gait_cycle ,C step_width ,C step_length ,C swing_range The spatial and temporal characteristics of

[0121] Feature fusion and prediction: The outputs of BLSTM, LSTM, and CNN are fused through the attention mechanism to generate a comprehensive feature vector, which is input into the fully connected layer for classification in the correction stage.

[0122] Model training and validation are as follows:

[0123] Data segmentation: 80% of the data is used as the training set, 10% as the validation set, and the remaining 10% as the test set;

[0124] Loss function: The cross entropy loss function is combined with the customized correction stage classification loss and trend prediction error loss.

[0125] The correction stages referred to in the present invention are divided into: initial adaptation stage, functional recovery stage, functional strengthening and adaptive training stage, maintenance and relapse prevention stage;

[0126] Collect various data of patients in different correction stages, including pressure sensor data and various data collected by the image acquisition module; analyze, integrate and process the data according to the various functions of the aforementioned analysis module;

[0127] At the same time, a deep learning model is constructed according to the aforementioned analysis model, and the various collected data are used as training data for model training, and the output is the correction stage.

[0128] By integrating pressure sensing, infrared image processing and deep learning analysis, it has pioneered innovation in the monitoring of arthritis correction process. It not only realizes real-time and accurate monitoring of patients' gait and joint movement, but also improves the comfort and convenience of monitoring through non-invasive methods. The system comprehensively analyzes pressure distribution, joint dynamics and contour changes, provides doctors and patients with instant feedback and personalized treatment adjustment basis, and accelerates the recovery process. In addition, this technology optimizes the utilization of medical resources through intelligent data processing and remote monitoring capabilities, promotes the scientific and efficient management of arthritis, and makes important contributions to improving patients' quality of life and promoting medical progress.

[0129] Thus far, the description of the above-described embodiments is provided for the purpose of illustration and description. It is not intended to be exhaustive or to limit the present disclosure. The individual elements or features of a specific embodiment are generally not subject to the limitations of the specific embodiment, but when applicable, they can be interchanged and used for selected embodiments even if not specifically shown or described. In many respects, the same element or feature also can be changed. This variation is not considered to be a departure from the present disclosure, and all such modifications are intended to be included in the scope of the present disclosure.

[0130] Example embodiments are provided so that the present disclosure will be thorough and the scope will be fully conveyed to those skilled in the art. In order to thoroughly understand the embodiments of the present disclosure, numerous details are set forth, such as examples of specific parts, devices, and methods. Obviously, for those skilled in the art, it is not necessary to use specific details, the example embodiments can be implemented in many different forms, and neither should be construed as limiting the scope of the present disclosure. In certain example embodiments, well-known processes, well-known device structures, and well-known technologies are not described in detail.

[0131] Here, specialized vocabulary is used only for the purpose of describing specific example embodiments, and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "a" and "the" used herein may be intended to include plural forms as well. The terms "including" and "having" are inclusive, and therefore specify the presence of the claimed features, wholes, steps, operations, elements and / or components, but do not exclude the presence or additional presence of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof. Unless the order of execution is clearly indicated, the method steps, processing and operations described herein are not interpreted as necessarily needing to be performed in the specific order discussed and shown. It should also be understood that additional or optional steps may be adopted.

Claims

1. An arthritis correction process monitoring system based on motion monitoring, characterized in that: It includes a monitoring host, a pressure sensor, an image acquisition module, a storage module and a user terminal; The monitoring host is connected to the pressure sensor, image acquisition module, storage module and user terminal to control the operation of the entire monitoring system; The pressure sensor is installed on the sole of the shoe to detect the pressure on the patient's feet in real time and transmit it to the monitoring host in real time to monitor the patient's walking process; The image acquisition module is set at the patient's activity site to collect video images of the patient when he / she is in action and send them to the monitoring host; The storage module is used to store the collected data and analysis models for the monitoring host to analyze the patient's correction stage; The user terminal sets a reminder interface for displaying analysis results to doctors and patients; The pressure sensors are set on the bottom plate of the shoes. Each shoe is set with two pressure sensors, and two shoes are set with a total of four pressure sensors; Each shoe is equipped with a pressure detection chip, which is connected to two pressure sensors on the shoe to collect data from the pressure sensors and send it to the monitoring host in real time; Each shoe is provided with a pressure sensor at the front and back, the front pressure sensor is located below the base of the toe bones of the sole, and the back pressure sensor is located below the base of the sole; the pressure detection chip is located below the arch of the foot; The image acquisition module includes a wearable light emitter, which is provided with a strap and an infrared light emitter; the strap is used to fix the infrared light emitter to the patient's feet, calves, knees, thighs and waist; the infrared light emitter is a point light source, and the wavelength of the emitted infrared light has a penetration rate of more than 50% for fabrics; Wear 9 or more light emitters, at least on the feet, calves, knees, thighs and waist; The image acquisition module is also provided with an infrared camera and a visible light camera. The infrared camera is used to collect the light emitted by the wearable light emitter in real time, and the visible light camera is used to collect the visible light image of the patient in real time. The image acquisition module sends the acquired infrared images and visible light images to the monitoring host in real time; The monitoring host performs position registration of infrared images and visible light images: Feature point extraction: In the infrared image and the corresponding visible light image, scale-invariant feature transformation, ORB or AKAZE algorithm are used to automatically extract stable feature points; Feature matching: Match the feature points in the two images through KNN search or BFMatcher algorithm, use RatioTest strategy to remove false matches, and obtain the corresponding relationship of feature points; Using the matched feature point pairs, the infrared image and the visible light image are geometrically transformed and registered by using perspective transformation or homography matrix calculation method; The monitoring host accurately calibrates the position of the wearable light emitter in the infrared image as follows: Using image segmentation algorithm, the light spot area produced by the light emitter is identified and segmented from the infrared image; The contour of each light spot is extracted, and the actual position of each wearable light emitter is calculated using the centroid method according to the shape, size and intensity distribution characteristics of the light spot, that is, the centroid of the light spot shape is determined and the centroid is used as the actual position of the wearable light emitter; The monitoring host extracts the contour of the patient in the visible light image as follows: First, the visible light image is preprocessed, including grayscale conversion and filtering to reduce noise interference in the image; Apply edge detection algorithms to identify boundaries in images; edge detection algorithms include Canny edge detection, Sobel operator or Laplacian operator; Based on the results of edge detection, a contour extraction algorithm is used to obtain closed contours in the image.

2. The arthritis correction process monitoring system based on motion monitoring according to claim 1 is characterized in that: The centroid of the spot shape is calculated as follows: Traverse each light spot, for each pixel (x i ,y i ) Its gray value I(x i ,y i ) is regarded as the "mass" of the position; these "masses" and their corresponding coordinates need to be accumulated to calculate the center of mass; the center of mass coordinates of the spot (C x ,C y ) is calculated using the following weighted average formula: Here, the summation is performed on all pixels i within the spot, x i and i are the horizontal and vertical coordinates of the pixel, I(x i ,y i ) is the intensity value of the pixel.

3. The arthritis correction process monitoring system based on motion monitoring according to claim 2 is characterized in that: The monitoring host integrates an analysis module, which analyzes the patient's correction stage based on the data from the pressure sensor and image acquisition module, as follows: Time Series Data Integration: Directly collect the real-time pressure values ​​of the front and back sole pressure sensors to form two sets of one-dimensional sequence data that change over time, with two points on each foot for a total of four data sequences; ensure that the pressure data sequence is accurately aligned with the time axis of the position change of the worn light device in the infrared image and the change of the patient's contour in the visible light image, and use timestamps to ensure synchronization across data sources; Feature extraction: The following pressure features are extracted: P peak : The maximum pressure value sequence of the front and back soles of the feet, reflecting the key load-bearing conditions during walking; P ratio : Front and rear foot pressure ratio sequence to evaluate gait balance; P rate : Pressure change rate, capturing gait dynamics; Extract the following wearable light device position features: G ankle ,G knee ,G hip : Represents the angle change sequence of ankle joint, knee joint and hip joint respectively, assessing the range of motion and flexibility of lower limb joints; G amplitude : Joint motion amplitude sequence, reflecting the range of motion; G smoothness : Movement smoothness index, obtained by analyzing joint angular velocity and acceleration; extracting the following patient contour features: C gait_cycle : gait cycle sequence, obtained by similarity matching between contour sequences; C step_width ,C step_length : Step width and step length sequence to evaluate gait characteristics; C swing_range : Limb swing amplitude sequence, reflecting the movement amplitude of lower limbs; Deep learning model construction: Using a multimodal fusion network, in: Pressure feature processing: Using bidirectional LSTM to process P peak ,P ratio ,P rate , capturing the time dependence of pressure changes; Joint angle change processing: LSTM processing G ankle ,G knee ,G hip ,G amplitude ,G smoothness , extract time series features; Contour feature processing: Lightweight CNN processes contour image sequences and extracts C gait_cycle ,C step_width ,C step_length ,C swing_range The spatial and temporal characteristics of Feature fusion and prediction: The outputs of BLSTM, LSTM, and CNN are fused through the attention mechanism to generate a comprehensive feature vector, which is input into the fully connected layer for classification in the correction stage.

4. The arthritis correction process monitoring system based on motion monitoring according to claim 3 is characterized in that: Model training and validation are as follows: Data segmentation: 80% of the data is used as the training set, 10% as the validation set, and the remaining 10% as the test set; Loss function: The cross entropy loss function is combined with the customized correction stage classification loss and trend prediction error loss.

Images