Wearable ankle joint rehabilitation monitoring system

Abstract

The invention discloses a wearable ankle joint rehabilitation monitoring system, which comprises a signal acquisition module, a signal conditioning and conversion module, a data processing and scoring module and a visual output module, the signal acquisition module is used for acquiring micro-strain signals on the inner side and the outer side during ankle joint movement in real time; the signal conditioning and conversion module is connected with the signal acquisition module and is used for carrying out amplification, filtering and analog-to-digital conversion on the micro strain signal to obtain an internal and external ankle strain signal; the data processing and scoring module is connected with the signal conditioning and conversion module and is used for performing two-stage filtering denoising, normalization, feature extraction and fusion machine learning model scoring on the internal and external ankle strain signals to obtain ankle joint function scores; the visual output module is connected with the data processing and scoring module and used for displaying ankle joint function scores and generating strain curves of the inner and outer ankles based on the strain signals of the inner and outer ankles.

Description

Technical Field

[0001] This invention belongs to the field of medical device and rehabilitation engineering technology, specifically relating to a wearable ankle joint rehabilitation monitoring system. Background Technology

[0002] The ankle joint is a crucial weight-bearing and movement joint in the human body, and rehabilitation after injuries such as sprains, fractures, and ligament tears is of paramount importance. As the starting point of the body's kinetic chain, it plays a key role in maintaining postural control, gait transitions, and lower limb motor function. Ankle stability is not only fundamental for daily activities such as walking, running, and jumping, but also plays a vital role in health management and rehabilitation for various populations. Currently, the assessment of ankle rehabilitation effectiveness largely relies on: 1. Subjective assessment: Physician manual examination, patient-reported pain level (VAS score), and functional questionnaires (such as AOFAS score). This method is highly dependent on physician experience, lacks objective quantitative data, and cannot reflect the dynamic situation in rehabilitation training in real time.

[0003] 2. Imaging assessment: X-rays, MRI, etc. can observe the static structure of bones and soft tissues, but cannot assess the functional status of joints during movement, and have problems such as radiation, high cost, and infrequent use.

[0004] 3. In recent years, several wearable sensor-based technologies have been proposed, such as pressure insoles and inertial measurement units (IMUs). Pressure insoles primarily monitor plantar pressure distribution and are unable to specifically reflect tension changes in the medial and lateral ligament groups of the ankle joint; IMU goniometers, on the other hand, cannot directly measure micro-strain within soft tissues. More importantly, existing sensor solutions are prone to artifacts during ankle torsion (pronation / supination), leading to a significant decrease in measurement accuracy and failing to meet the precision requirements of rehabilitation monitoring.

[0005] Full-bridge strain gauges are characterized by high measurement accuracy and strong anti-interference capabilities, and are commonly used to measure minute deformations. However, by innovatively applying them to specific points on the inner and outer ankle joints of the human body, the difference between the two sensors is calculated to dynamically and objectively quantify joint stability and range of motion, and further construct an intuitive scoring system.

[0006] Therefore, there is an urgent need for a system and method that can objectively, quantitatively, in real time, and visually monitor and evaluate ankle rehabilitation. Summary of the Invention

[0007] This invention aims to address the shortcomings of existing technologies and provides the following solutions: A wearable ankle joint rehabilitation monitoring system includes: a signal acquisition module, a signal conditioning and conversion module, a data processing and scoring module, and a visualization output module; The signal acquisition module is used to acquire micro-strain signals on both the inner and outer sides of the ankle joint in real time during movement; The signal conditioning and conversion module is connected to the signal acquisition module and is used to amplify, filter and convert the micro-strain signal to analog-to-digital to obtain the inner and outer ankle strain signals; The data processing and scoring module is connected to the signal conditioning and conversion module, and is used to perform dual-stage filtering and denoising, normalization, feature extraction and machine learning model fusion scoring on the medial and lateral ankle strain signals to obtain an ankle joint function score. The visualization output module is connected to the data processing and scoring module and is used to display the ankle joint function score and generate strain curves of the medial and lateral ankles based on the medial and lateral ankle strain signals.

[0008] Preferably, the signal acquisition module includes: Two manganese steel sheets with a biocompatible insulating layer on the surface and two full-bridge strain sensors; The two full-bridge strain sensors are respectively attached to the two manganese steel sheets to form an inner ankle sensing unit and an outer ankle sensing unit.

[0009] Preferably, the signal conditioning and conversion module includes: an instrumentation amplifier, a filter circuit, an analog-to-digital converter, and a Bluetooth unit; The instrumentation amplifier is used to initially amplify the millivolt-level analog voltage signal output by the full-bridge strain sensor, and the common-mode rejection ratio of the instrumentation amplifier is not less than 120dB. The filtering circuit consists of a second-order Sallen-Key low-pass filter with a cutoff frequency of 25Hz and a double-T notch filter with a center frequency of 50Hz, used to filter the initially amplified signal. The analog-to-digital converter uses a 16-bit resolution and a sampling rate of 200Hz to convert the filtered analog voltage signal into the digital inner and outer ankle strain signals. The Bluetooth unit is used to send the medial and lateral ankle strain signals to the data processing and scoring module.

[0010] Preferably, the data processing and scoring module includes: a two-stage filtering unit, a normalization unit, a feature extraction unit, and a model scoring unit; The dual-stage filtering unit includes a translational filtering subunit and an exponential smoothing filtering subunit, which are used to perform dual-stage filtering on the medial and lateral ankle strain signals to obtain filtered digital signals. The normalization unit is used to normalize the filtered digital signal to obtain a normalized digital signal; The feature extraction unit is used to extract features from the normalized digital signal to obtain the extracted medial and lateral malleolar features: mean medial malleolar strain, mean lateral malleolar strain, maximum medial malleolar strain, minimum medial malleolar strain, maximum lateral malleolar strain, minimum lateral malleolar strain, medial malleolar strain amplitude, lateral malleolar strain amplitude, difference between mean medial and lateral malleolar strain and difference between medial and lateral malleolar strain amplitude; The model scoring unit obtains a first predictive sub-score and a second predictive sub-score for the stability, symmetry and mobility of the ankle joint based on the medial and lateral malleolar features. The first predictive sub-score and the second predictive sub-score are then fused using Bayesian weighted fusion to obtain a fused three-dimensional sub-score vector of stability, symmetry and mobility sub-scores. Finally, the fused three-dimensional sub-score vector is mapped to the ankle joint function score in the range of 0 to 100.

[0011] Preferably, in the dual-stage filtering unit, the workflow of the translational filtering subunit includes: in, This indicates the signal after translation and filtering. N Indicates the length of the sliding window. This represents current and historical sampling data; The workflow of the exponential smoothing filter subunit includes: in, This represents the output signal after exponential smoothing at the current moment. This represents the output signal after exponential smoothing filtering at the previous time step, where α represents the smoothing factor.

[0012] Preferably, the model scoring unit includes: a first prediction subunit, a second prediction subunit, a weighted fusion subunit, and a mapping subunit; Based on the medial and lateral malleolar features, the first prediction subunit uses the LightGBM submodel of statistical feature vectors to obtain the first prediction sub-scores of ankle joint stability, symmetry and range of motion. The second prediction subunit, based on the medial and lateral malleolar features, uses a deep temporal network sub-model combining a 1D-CNN with a bidirectional long short-term memory network of the original temporal data to obtain the second prediction sub-scores for ankle joint stability, symmetry, and range of motion. The weighted fusion subunit is used to perform Bayesian weighted fusion of the first predicted sub-score and the second predicted sub-score to obtain a fused three-dimensional sub-score vector of stability, symmetry and activity sub-scores; The mapping subunit includes a scoring synthesis network for mapping the fused three-dimensional sub-score vector to the ankle joint function score in the range of 0 to 100.

[0013] Preferably, in the weighted fusion subunit, the formula for calculating the sub-score of each dimension is as follows: in, and These represent the predicted value and uncertainty estimate output by the LightGBM sub-model, respectively. and represents the predicted value and uncertainty estimate output by the deep temporal sub-model, respectively, and S represents the fused three-dimensional sub-fraction vector obtained after fusion.

[0014] Preferably, the visualization output module includes: a patient terminal and a doctor terminal; The patient terminal consists of a real-time display unit, a data review unit, and a rehabilitation trend analysis unit shared by both the patient and doctor terminals. The doctor's end consists of a patient list unit, a patient monitoring record unit, a patient monitoring record details unit shared by both the patient's end and the doctor's end, and a doctor's end recovery trend analysis unit.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention utilizes symmetrically arranged full-bridge strain sensors on both the medial and lateral ankles to simultaneously acquire medial and lateral ankle signals, ensuring comprehensiveness and accuracy of monitoring results. Based on feature extraction, it accurately reflects the stability, symmetry, and range of motion of the ankle joint during rehabilitation. By integrating statistical methods (based on coefficient of variation, amplitude comparison, etc.) with machine learning / deep learning methods (LightGBM, 1D-CNN combined with Bi-LSTM) and introducing a Bayesian weighted fusion layer, it ensures the interpretability of model results and improves the robustness and accuracy of score prediction. By setting an individual baseline reference range Rnorm, it can provide personalized rehabilitation assessment results for different patients, avoiding evaluation distortion caused by population differences. Monitoring results can be displayed in real time in the software interface, and historical data review, trend analysis, and data export are supported, providing scientific evidence for doctors and intuitive rehabilitation feedback for patients. The monitoring and assessment method of this invention is applicable to clinical rehabilitation, home rehabilitation, and sports protection scenarios, and has broad promotional value and application prospects. Attached Figure Description

[0016] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Picture 1This is a schematic diagram of the system structure according to an embodiment of the present invention; Picture 2 This is a flowchart of the ankle joint rehabilitation monitoring system of the present invention; Picture 3 This is a schematic diagram of the time-domain plot of the average strain of the medial and lateral malleoli in this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] Example In this embodiment, as Picture 1 As shown, a wearable ankle joint rehabilitation monitoring system includes: a signal acquisition module, a signal conditioning and conversion module, a data processing and scoring module, and a visualization output module.

[0021] The signal acquisition module is used to acquire micro-strain signals from both the inner and outer sides of the ankle joint in real time during movement.

[0022] The signal acquisition module includes two manganese steel sheets with biocompatible insulating layers on their surfaces and two full-bridge strain sensors. The two full-bridge strain sensors are respectively attached to the two manganese steel sheets to form an medial malleolus sensing unit and a lateral malleolus sensing unit.

[0023] In this embodiment, the full-bridge strain sensor uses a YX-350 constantan foil strain sensor with a resistance of 350Ω, a range of ±20000Ω, and a sensitivity coefficient of 2.11±1%. The manganese steel substrate is made of 65Mn spring steel, precision wire-cut into strips with a thickness of 0.3mm, a width of 15mm, and a length of 130mm. Subsequently, a polyimide insulating coating is applied to its surface to form a biocompatible insulating layer. Finally, epoxy-based conductive silver paste is used to firmly attach the strain gauge to the manganese steel substrate, forming an integrated medial and lateral malleolus sensing unit. The two sensor units are respectively embedded in the medial and lateral malleolus slots of a customized medical-grade elastic ankle brace, symmetrically arranged with the Achilles tendon as the anatomical center. The sensors are fixed by special slots inside the ankle brace to ensure they do not shift during movement.

[0024] The signal conditioning and conversion module is connected to the signal acquisition module and is used to amplify, filter and convert the micro-strain signal into analog and digital signals to obtain the inner and outer ankle strain signals.

[0025] The signal conditioning and conversion module includes an instrumentation amplifier, a filtering circuit, an analog-to-digital converter (ADC), and a Bluetooth unit. The instrumentation amplifier is used to initially amplify the millivolt-level analog voltage signal output from the full-bridge strain sensor, with a common-mode rejection ratio (CMRR) of no less than 120 dB. The filtering circuit consists of a second-order Sallen-Key low-pass filter with a cutoff frequency of 25 Hz and a dual-T notch filter with a center frequency of 50 Hz, used to filter the initially amplified signal. The ADC uses 16-bit resolution and a sampling rate of 200 Hz to convert the filtered analog voltage signal into digital medial and lateral malleolar strain signals. The Bluetooth unit is used to transmit the medial and lateral malleolar strain signals to the data processing and scoring module.

[0026] In this embodiment, the module is integrated onto a 30mm x 20mm PCB board and encapsulated in a 3D-printed ABS shell, and connected to the signal acquisition module.

[0027] The instrumentation amplifier uses the INA333 instrumentation amplifier. Its common-mode rejection ratio (CMRR) is typically 120dB. The gain is fixed at 32x by configuring a 1% precision surface-mount resistor to accommodate the typical amplitude range of ankle strain signals. The filtering circuit includes a second-order Sallen-Key low-pass filter with a cutoff frequency of 25Hz, using an AD8606 op-amp to filter out electromyographic signals above 20Hz; and a dual-T active notch filter with a center frequency of 50Hz. The analog-to-digital converter (ADC) uses an ADS1115 16-bit ADC. Its sampling rate is configured to 200Hz via I2C, fully satisfying the Nyquist sampling theorem. The Bluetooth unit uses an nRF52832 low-power Bluetooth SoC. It is responsible for packaging ADC data into custom protocol data frames and transmitting them to the smart terminal. The communication distance is greater than 10 meters, and the packet loss rate is less than 1%.

[0028] The data processing and scoring module is connected to the signal conditioning and conversion module. It is used to perform two-stage filtering, normalization, feature extraction, and fusion machine learning model scoring on the medial and lateral ankle strain signals to obtain an ankle joint function score.

[0029] The data processing and scoring module includes: a two-stage filtering unit, a normalization unit, a feature extraction unit, and a model scoring unit.

[0030] The dual-stage filtering unit includes a translational filtering subunit and an exponential smoothing filtering subunit, which are used to perform dual-stage filtering on the inner and outer ankle strain signals to obtain the filtered digital signal.

[0031] In this embodiment, a cascaded dual-filter architecture is adopted to effectively reduce noise in the original strain signal and extract physiological signals reflecting the gait cycle. A moving average filter is embedded in the ankle joint rehabilitation monitoring system, and an exponential smoothing filter is added on top of the moving average filter. This performs sequence processing on the real-time medial and lateral ankle curves displayed on the monitoring interface, ultimately obtaining a smooth, well-periodic, sinusoidal waveform. Specifically, the workflow of the Moving Average (MA) filter subunit in the dual-stage filtering unit includes: in, This indicates the signal after translation and filtering. N Indicates the length of the sliding window. This represents current and historical sampling point data; in this embodiment, the window length N is not a fixed value, but rather depends on the sampling rate of the monitoring system. Dynamic association, its calculation formula is: This ensures the filtering window consistently corresponds to a 50ms physiological time interval, guaranteeing that the filtering process matches the physiological timing of gait fine-tuning. The workflow of the Exponential Smoothing (ES) filtering subunit includes: deep smoothing the output of the first stage to generate the final medial and lateral malleolar strain waveforms for display and analysis, whose mathematical expression is: in, This represents the output signal after exponential smoothing at the current moment. This represents the output signal after exponential smoothing filtering at the previous moment, where α represents the smoothing factor. In this embodiment, α is set to 0.15. This value strikes the best balance between the speed of signal tracking and the thoroughness of smoothing, making it suitable for the smoothing requirements of human gait signals. It can effectively suppress physiological tremors and equipment noise while preserving the true signal trend.

[0032] The normalization unit is used to normalize the filtered digital signal, normalize the amplitude of the medial and lateral ankle strain signals to the [-1,1] interval, eliminate the amplitude difference caused by the different tightness of each wearing, and allow the data from different times to be compared to obtain the normalized digital signal.

[0033] The feature extraction unit is used to extract features from the normalized digital signal, specifically from the medial malleolus time series signal. The timing of the lateral malleolus is Extracting the medial and lateral malleolar features, the medial malleolar strain mean was obtained. Mean strain of lateral malleolus Maximum strain of the medial malleolus Minimum strain of medial malleolus , maximum strain of the lateral ankle Minimum strain of the lateral malleolus Medial malleolus strain amplitude lateral ankle strain amplitude Difference between mean strain values ​​of medial and lateral malleoli and the difference in strain amplitude between the medial and lateral malleoli .

[0034] The model scoring unit obtains the first and second predicted sub-scores of ankle joint stability, symmetry, and mobility based on the features of the medial and lateral malleoli. The first and second predicted sub-scores are then fused using Bayesian weighting to obtain a fused three-dimensional sub-score vector of stability, symmetry, and mobility. Finally, the fused three-dimensional sub-score vector is mapped to an ankle joint function score in the range of 0 to 100.

[0035] The model scoring unit includes: a first prediction subunit, a second prediction subunit, a weighted fusion subunit, and a mapping subunit. The first prediction subunit, based on medial and lateral malleolar features, uses a LightGBM submodel with statistical feature vectors to obtain first predicted sub-scores for ankle joint stability, symmetry, and range of motion. The second prediction subunit, also based on medial and lateral malleolar features, uses a deep temporal network submodel combining a 1D-CNN with a bidirectional long short-term memory network based on the original temporal data to obtain second predicted sub-scores for ankle joint stability, symmetry, and range of motion. The weighted fusion subunit performs Bayesian weighted fusion of the first and second predicted sub-scores to obtain a fused three-dimensional sub-score vector for stability, symmetry, and range of motion. The mapping subunit contains a scoring synthesis network that maps the fused three-dimensional sub-score vector to an ankle joint function score within the range of 0 to 100.

[0036] Specifically, the stability score assesses the ankle joint's ability to resist interference and maintain stability during movement. This invention uses an exponential decay function based on the coefficient of variation (CV) for quantification. Specifically, it collects and calculates the standard deviation and coefficient of variation for two channels (medial and lateral malleoli): in, Represents the medial malleolus strain signal sequence The standard deviation of a sequence is given by std, which represents the standard mathematical function for calculating the standard deviation of a sequence. The coefficient of variation represents the strain signal of the medial malleolus. ε This represents a very small positive protection factor. Represents the strain signal sequence of the lateral malleolus standard deviation The coefficient of variation of the lateral malleolus strain signal is represented; further, the stability sub-scores of the medial and lateral malleoli are defined as exponentially decaying forms respectively: in, This represents the stability sub-score calculated based on the medial malleolus signal, where exp represents an exponential function with the natural constant e as the base. This represents the stability sub-score calculated based on the lateral malleolus signal. and Both represent configurable adjustment parameters greater than 0, used to control the sensitivity of CV to the score; subsequently, the two stability sub-scores are geometrically merged to synthesize a total stability score: The more stable the gait signal CV The smaller the value, the higher the stability score. The higher.

[0037] Symmetry score: quantifies the balance of forces on both sides, taking the mean strain of the medial malleolus. Mean strain of lateral malleolus Then the mathematical expression for symmetry scoring is defined as: when and The closer they get, The closer to 100, the lower the score; if there is extreme asymmetry between the two sides, the lower the score. To avoid the influence of a zero denominator or outliers, add... ε Protect.

[0038] Range of motion score: Evaluates the range of motion of the ankle joint during rehabilitation training. Range of motion is calculated using strain amplitudes at both the medial and lateral malleoli. , The composite amplitude is obtained by using the geometric mean and multiplying by a symmetry penalty factor: This leads to an activity score: Reference amplitude In the initial calibration phase, the subjects were guided to complete several standard movement sequences (such as standing, walking, etc.), and a total of M sessions were collected (M≥3). For each session, the robust amplitudes of the medial and lateral malleoli were calculated. Where j = 1, ..., M. Take the median of the multiple results: Geometric composition: .

[0039] Furthermore, multiple models are integrated for ankle joint scoring, combining the LightGBM sub-model with a 1D-CNN+Bi-LSTM deep temporal network, and generating sub-scores and comprehensive scores through a Bayesian weighted fusion layer, thereby improving prediction accuracy and robustness while ensuring interpretability.

[0040] Specifically, the first prediction sub-unit constructs a LightGBM sub-model, which is an efficient gradient boosting decision tree model. The input is the aforementioned statistical feature vector: in, , LightGBM is based on Output three predicted sub-scores and uncertainty estimation In parallel, the second prediction sub-unit constructs a 1D-CNN+Bi-LSTM deep temporal network sub-model. This network is used to process temporal signals: specifically, it first uses a one-dimensional convolutional layer (1D-CNN) to extract high-dimensional features at local time scales; then, it uses a bidirectional long short-term memory network (Bi-LSTM) to capture the long-term temporal dependencies of the entire sequence from both forward and backward directions. The network input is the original two-channel temporal segment. Local temporal features are extracted using 1D-CNN and bidirectional temporal dependencies are captured by Bi-LSTM, finally outputting three corresponding temporal driving sub-scores. Meanwhile, the prediction uncertainty is estimated to obtain .

[0041] Furthermore, Bayesian weighted fusion is performed on the sub-scores for each dimension (stability, symmetry, activity). The fused sub-scores are obtained by fusion using Bayesian minimum variance fusion. The formula for calculating the sub-score for each dimension in the weighted fusion sub-unit is as follows: in, and These represent the predicted value and uncertainty estimate output by the LightGBM sub-model, respectively. and Let represent the predicted value and uncertainty estimate output by the deep temporal sub-model, respectively, and S represent the fused 3D sub-fraction vector. The 3D results are then analyzed. The mapping sub-unit further constructs the input vector from the fused 3D sub-fractions. The input is given to a fully connected neural network. The system uses a nonlinear mapping to output an ankle joint function score after rehabilitation training, monitored by the system, and is in the rehabilitation phase. .

[0042] The visualization output module is connected to the data processing and scoring module to display ankle joint function scores and generate strain curves for the medial and lateral malleoli based on the medial and lateral malleoli strain signals.

[0043] The visualization output module includes a patient end and a doctor end. The patient end consists of a real-time display unit, a data review unit, and a rehabilitation trend analysis unit shared by both the patient and doctor ends. The doctor end consists of a patient list unit, a patient monitoring record unit, a patient monitoring record details unit shared by both the patient and doctor ends, and a rehabilitation trend analysis unit shared by the doctor end.

[0044] Based on clinical rehabilitation medicine, the ankle rehabilitation process is divided into three stages: (1) Acute phase: This phase corresponds to the early stage of rehabilitation, characterized by severely limited range of motion, significant gait asymmetry, and poor stability. (2) Recovery phase: This phase corresponds to the gradual recovery of function, with improved range of motion and gait symmetry, but insufficient dynamic stability. (3) Functional recovery phase: This phase corresponds to the basic recovery of function, approaching normal exercise levels, with a large range of motion and good dynamic stability and strength symmetry.

[0045] Furthermore, ankle joint function scores will be... Defined as the Recovery Index (RI), i.e. The RI threshold range is set for each stage. The rehabilitation stage diagnostic unit calculates the RI value in real time and compares it with the preset threshold to automatically determine the current rehabilitation stage. Specifically: It was determined to be in the acute phase; It was determined to be in the recovery period; It is determined to be in the functional recovery period.

[0046] In this embodiment, during ankle joint rehabilitation monitoring, from the start to the end of monitoring, the left side of the real-time display unit monitors and displays the strain change waveforms of the medial and lateral malleoli after dual-stage filtering; the right side continuously updates the maximum / minimum and mean values ​​of the medial and lateral malleoli strain in real time; the background calculates the ankle joint stability score, symmetry score, range of motion score, and comprehensive score based on the collected values ​​of the medial and lateral malleoli. RI "Value, and based on the overall score" RIThe value determines the patient's rehabilitation stage: acute phase / recovery phase / functional recovery phase. When monitoring ceases, all data is automatically saved to the doctor's patient monitoring record details module, the patient's data review module, and the rehabilitation trend analysis module.

[0047] Data Review Unit: Patients can click on any monitoring record to enter the review interface, which displays the time-domain waveforms of medial and lateral malleolar strain, statistical charts of the maximum, minimum, and average values ​​of medial and lateral malleolar strain, a time-domain graph of the mean, ankle stability score, symmetry score, range of motion score, and comprehensive score for that monitoring session. RI The data includes values ​​and corresponding recovery stages: acute phase / recovery phase / functional recovery phase, and supports data export as PNG images and Excel files.

[0048] Furthermore, the patient monitoring and recording unit on the doctor's end displays a list of all patients. Clicking on any patient leads to the patient monitoring and recording module, which displays all monitoring records for that patient. Each record shows the monitoring date (year-month-day hour:minute:second) and the monitoring duration. Clicking on any monitoring record leads to the patient monitoring record details module, where users can view the strain time-domain waveforms of the medial and lateral malleoli for that monitoring record, statistical charts of the maximum, minimum, and average strain values ​​of the medial and lateral malleoli, a mean time-domain graph, ankle joint stability score, symmetry score, range of motion score, and comprehensive score. RI Values ​​and corresponding rehabilitation stages: acute phase / recovery phase / functional recovery phase.

[0049] Specifically, such as Picture 3 The mean time-domain plot shown: The mean medial malleolus strain during the monitoring period is indicated by two horizontal dashed lines. and mean strain of the lateral malleolus The vertical distance between the two mean lines It visually reflects the symmetry and balance of forces on both sides of the ankle joint. When A smaller value indicates that the strains of the medial and lateral malleoli are similar, indicating good ankle joint stability and a better rehabilitation effect; when... A large value indicates a significant difference in force distribution between the two sides, poor recovery status, and the need for further intervention.

[0050] Rehabilitation trend analysis unit: Further supports rehabilitation trend analysis on both ends. The interface plots the trend curve of the patient's ankle joint rehabilitation status over time, including stability score, symmetry score, range of motion score, comprehensive score, and the curve trajectory of the changes in multiple monitoring results of the corresponding rehabilitation stages, to assist doctors in evaluating the stage rehabilitation effect (week 1, week 2, week 3 of rehabilitation, etc.). Picture 2 This is a flowchart illustrating the workflow of the ankle joint rehabilitation monitoring system of the present invention.

[0051] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A wearable ankle joint rehabilitation monitoring system, characterized in that, include: Signal acquisition module, signal conditioning and conversion module, data processing and scoring module, and visualization output module; The signal acquisition module is used to acquire micro-strain signals on both the inner and outer sides of the ankle joint in real time during movement; The signal conditioning and conversion module is connected to the signal acquisition module and is used to amplify, filter and convert the micro-strain signal to analog-to-digital to obtain the inner and outer ankle strain signals; The data processing and scoring module is connected to the signal conditioning and conversion module, and is used to perform two-stage filtering and denoising, normalization, feature extraction and machine learning model fusion scoring on the medial and lateral ankle strain signals to obtain an ankle joint function score. The visualization output module is connected to the data processing and scoring module and is used to display the ankle joint function score and generate strain curves of the medial and lateral ankles based on the medial and lateral ankle strain signals.

2. The wearable ankle joint rehabilitation monitoring system according to claim 1, characterized in that, The signal acquisition module includes: Two manganese steel sheets with a biocompatible insulating layer on the surface and two full-bridge strain sensors; The two full-bridge strain sensors are respectively attached to the two manganese steel sheets to form an inner ankle sensing unit and an outer ankle sensing unit.

3. The wearable ankle joint rehabilitation monitoring system according to claim 2, characterized in that, The signal conditioning and conversion module includes: an instrumentation amplifier, a filter circuit, an analog-to-digital converter, and a Bluetooth unit; The instrumentation amplifier is used to initially amplify the millivolt-level analog voltage signal output by the full-bridge strain sensor, and the common-mode rejection ratio of the instrumentation amplifier is not less than 120dB. The filtering circuit consists of a second-order Sallen-Key low-pass filter with a cutoff frequency of 25Hz and a double-T notch filter with a center frequency of 50Hz, used to filter the initially amplified signal. The analog-to-digital converter uses a 16-bit resolution and a sampling rate of 200Hz to convert the filtered analog voltage signal into the digital inner and outer ankle strain signals. The Bluetooth unit is used to send the medial and lateral ankle strain signals to the data processing and scoring module.

4. The wearable ankle joint rehabilitation monitoring system according to claim 1, characterized in that, The data processing and scoring module includes: a two-stage filtering unit, a normalization unit, a feature extraction unit, and a model scoring unit; The dual-stage filtering unit includes a translational filtering subunit and an exponential smoothing filtering subunit, which are used to perform dual-stage filtering on the medial and lateral ankle strain signals to obtain filtered digital signals. The normalization unit is used to normalize the filtered digital signal to obtain a normalized digital signal; The feature extraction unit is used to extract features from the normalized digital signal to obtain the extracted medial and lateral malleolar features: mean medial malleolar strain, mean lateral malleolar strain, maximum medial malleolar strain, minimum medial malleolar strain, maximum lateral malleolar strain, minimum lateral malleolar strain, medial malleolar strain amplitude, lateral malleolar strain amplitude, difference between mean medial and lateral malleolar strain and difference between medial and lateral malleolar strain amplitude; The model scoring unit obtains a first predictive sub-score and a second predictive sub-score for the stability, symmetry and mobility of the ankle joint based on the medial and lateral malleolar features. The first predictive sub-score and the second predictive sub-score are then fused using Bayesian weighted fusion to obtain a fused three-dimensional sub-score vector of stability, symmetry and mobility sub-scores. Finally, the fused three-dimensional sub-score vector is mapped to the ankle joint function score in the range of 0 to 100.

5. The wearable ankle joint rehabilitation monitoring system according to claim 4, characterized in that, The workflow of the translational filtering subunit in the dual-stage filtering unit includes: in, This indicates the signal after translation and filtering. N Indicates the length of the sliding window. This represents current and historical sampling data; The workflow of the exponential smoothing filter subunit includes: in, This represents the output signal after exponential smoothing at the current moment. This represents the output signal after exponential smoothing filtering at the previous time step, where α represents the smoothing factor.

6. The wearable ankle joint rehabilitation monitoring system according to claim 4, characterized in that, The model scoring unit includes: a first prediction subunit, a second prediction subunit, a weighted fusion subunit, and a mapping subunit; Based on the medial and lateral malleolar features, the first prediction subunit uses the LightGBM submodel of statistical feature vectors to obtain the first prediction sub-scores of ankle joint stability, symmetry and range of motion. The second prediction subunit, based on the medial and lateral malleolar features, uses a deep temporal network sub-model combining a 1D-CNN with a bidirectional long short-term memory network of the original temporal data to obtain the second prediction sub-scores for ankle joint stability, symmetry, and range of motion. The weighted fusion subunit is used to perform Bayesian weighted fusion of the first predicted sub-score and the second predicted sub-score to obtain a fused three-dimensional sub-score vector of stability, symmetry and activity sub-scores; The mapping subunit includes a scoring synthesis network for mapping the fused three-dimensional sub-score vector to the ankle joint function score in the range of 0 to 100.

7. The wearable ankle joint rehabilitation monitoring system according to claim 6, characterized in that, In the weighted fusion subunit, the formula for calculating the sub-score of each dimension is as follows: in, and These represent the predicted value and uncertainty estimate output by the LightGBM sub-model, respectively. and represents the predicted value and uncertainty estimate output by the deep temporal sub-model, respectively, and S represents the fused three-dimensional sub-fraction vector obtained after fusion.

8. The wearable ankle joint rehabilitation monitoring system according to claim 1, characterized in that, The visualization output module includes: a patient terminal and a doctor terminal; The patient terminal consists of a real-time display unit, a data review unit, and a rehabilitation trend analysis unit shared by both the patient and doctor terminals. The doctor's end consists of a patient list unit, a patient monitoring record unit, a patient monitoring record details unit shared by both the patient's end and the doctor's end, and a doctor's end recovery trend analysis unit.

Images