Intelligent measuring device and method for post-stroke shoulder-hand syndrome comprehensive rehabilitation evaluation

Abstract

The application provides an intelligent measuring device and method for comprehensive rehabilitation evaluation of post-stroke shoulder-hand syndrome, wherein the device comprises: a motion data acquisition module for acquiring target patient motion data; a standard action guiding module for guiding the target patient to sequentially perform a standard evaluation action sequence; a compensation separation module for calculating a shoulder joint true range of motion according to first multi-axis motion data and second multi-axis motion data collected in each evaluation action; a multi-dimensional data acquisition module for acquiring pain evaluation data and swelling evaluation data of the target patient; and a comprehensive evaluation module for inputting the shoulder joint true range of motion, the pain evaluation data and the swelling evaluation data into a multi-dimensional comprehensive evaluation model to obtain a comprehensive rehabilitation evaluation score. The intelligent measuring device and method for comprehensive rehabilitation evaluation of post-stroke shoulder-hand syndrome can measure the data of three dimensions of range of motion, pain and swelling to perform multi-dimensional quantitative evaluation of shoulder-hand syndrome, and the evaluation is more accurate.

Description

Technical Field

[0001] This invention relates to the field of intelligent assessment technology in rehabilitation medicine, and in particular to an intelligent measurement device and method for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome. Background Technology

[0002] Stroke is one of the leading causes of disability in my country. Currently, approximately 12% to 70% of stroke patients develop shoulder-hand syndrome during the course of their illness. The core symptoms include shoulder pain, hand swelling, and limited shoulder range of motion on the affected side. Early intervention and continuous rehabilitation exercises are crucial to preventing shoulder-hand syndrome from progressing to serious complications such as joint stiffness and muscle atrophy. In this process, shoulder range of motion is a core quantitative indicator for evaluating rehabilitation effectiveness. The accuracy, convenience, and continuity of its measurement directly affect the dynamic adjustment of rehabilitation programs and the objective assessment of treatment outcomes.

[0003] Currently, the measurement of shoulder joint range of motion in clinical practice mainly relies on traditional goniometers, which are manually operated by rehabilitation therapists. This method has the following prominent drawbacks: Measurements are highly subjective and have poor repeatability. Traditional protractor measurements rely heavily on the operator's experience and technique, and the measurement results vary significantly between different evaluators (inter-evaluator reliability) and between different time points for the same evaluator (intra-evaluator reliability).

[0004] The measurement scenarios are limited, making continuous home monitoring impossible. Goniometer measurements must be performed by professionals in medical institutions, and patients cannot obtain objective range of motion data on their own during rehabilitation exercises outside the hospital. Furthermore, rehabilitation for shoulder-hand syndrome is a process that lasts for weeks to months, and data from outpatient follow-up visits spaced several days or even weeks apart is insufficient to capture subtle changes and turning points in the rehabilitation process.

[0005] The assessment is limited to a single dimension and lacks comprehensive evaluation capabilities. Rehabilitation assessment for shoulder-hand syndrome involves not only shoulder joint range of motion but also pain intensity, hand swelling, and upper limb function. Traditional goniometers can only measure the single physical quantity of angle, and pain scores (such as VAS scores) and swelling assessments still need to be performed independently. The data are recorded in a scattered manner, making it impossible to achieve correlation analysis and comprehensive evaluation of multidimensional indicators.

[0006] Data cannot be digitally archived or transmitted remotely. Protractor measurement results must be manually recorded on paper rehabilitation assessment forms, which is not conducive to long-term data management and statistical analysis, nor can it achieve data sharing between patients and remote clinicians, thus hindering the implementation of remote rehabilitation guidance.

[0007] In recent years, several intelligent measurement tools for shoulder joint range of motion have emerged both domestically and internationally, including camera-based motion capture systems (such as iBalance and MARS), wearable inertial measurement units (such as CuffLink and Romiumeter), and mobile applications (such as Mobile Constant). However, existing intelligent tools still suffer from issues such as fixed usage scenarios, the ease with which compensatory components can be incorporated into measurement results, and the difficulty in ensuring standardized usage.

[0008] In view of this, there is an urgent need for intelligent measurement devices and methods for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome, in order to at least address the above-mentioned shortcomings. Summary of the Invention

[0009] One objective of this invention is to provide an intelligent measurement device and method for comprehensive rehabilitation assessment of shoulder-hand syndrome after stroke. By collecting motion data from the target patient and combining posture calculation and motion decomposition algorithms, it achieves objective and digital measurement of shoulder joint range of motion, eliminating the reliance on operator experience associated with traditional goniometers and significantly improving measurement accuracy and consistency. This effectively solves the problems of large inter-measurer errors and poor repeatability in manual assessment. The invention quantifies compensatory motion components during patient assessment and outputs the true range of motion of the shoulder joint after removing compensation, addressing the clinical challenge of inflated range of motion measurements in stroke patients due to compensatory movements such as trunk forward tilting and lateral flexion. This provides more reliable data for treatment planning. By integrating data from the three dimensions of range of motion, pain, and swelling into a comprehensive evaluation model, it achieves multi-dimensional quantitative assessment of shoulder-hand syndrome, overcoming the limitations of isolated analysis of indicators in traditional assessments.

[0010] The intelligent measurement device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome provided in this embodiment of the invention includes: The motion data acquisition module is used to collect real-time multi-axis motion data of the upper arm and the second multi-axis motion data of the trunk of the target patient; the target patient is a patient with post-stroke shoulder-hand syndrome. A standard movement guidance module is used to guide the target patient to perform a preset standard assessment movement sequence sequentially via a smart terminal application; the standard assessment movement sequence covers at least two degrees of freedom of movement of the shoulder joint; The compensation separation module is used to perform posture calculation and motion decomposition processing on the first and second multi-axis motion data collected in each evaluation action, and to calculate the true range of motion of the shoulder joint. The multidimensional data acquisition module is used to collect pain assessment data and swelling assessment data of the target patient through an application on a smart terminal during and / or after the execution of the standard assessment action sequence. The comprehensive evaluation module is used to input the actual range of motion of the shoulder joint, pain assessment data, and swelling assessment data into a preset multi-dimensional comprehensive evaluation model for fusion calculation to obtain a comprehensive rehabilitation assessment score.

[0011] Preferably, the compensation separation module performs posture calculation and motion decomposition processing on the first and second multi-axis motion data collected in each evaluation action, and calculates the true range of motion of the shoulder joint, including: The attitude calculation is performed independently on the first multi-axis motion data and the second multi-axis motion data to obtain the upper arm attitude quaternion and the torso attitude quaternion. Calculate the relative quaternion of the upper arm to the torso by multiplying the torso posture quaternion inversely by the upper arm posture quaternion. The relative quaternion is decomposed into angles according to the preset shoulder joint rotation decomposition sequence, and converted into flexion-extension angle, abduction-adduction angle and rotation angle; Based on the type of assessment action being performed, the peak values ​​of the corresponding angular components of the main motion plane are extracted from the flexion-extension angle, abduction-adduction angle, and rotation angle, and used as the true range of motion of the shoulder joint for the corresponding assessment action.

[0012] Preferably, the compensation separation module performs posture calculation and motion decomposition processing on the first and second multi-axis motion data collected in each evaluation action. After calculating the true range of motion of the shoulder joint, it further includes a step of multi-plane classification detection and degree quantification of the compensatory motion; the multi-plane classification detection and degree quantification include: During the execution of each evaluation action, the attitude calculation is performed on the second multi-axis motion data on a sampling cycle to obtain the trunk attitude quaternion at each sampling time, forming a time sequence of trunk attitude quaternions within the execution time period of each evaluation action. Using the trunk posture quaternion at the start of the evaluation action as the reference posture quaternion, the posture change of the trunk posture quaternion at each sampling time relative to the reference posture quaternion is calculated to obtain the time series data of trunk motion change during the execution of each evaluation action. The time-series data of trunk motion changes were decomposed into three anatomical planes of the human trunk to obtain time-series data of sagittal angle changes, coronal angle changes, and horizontal angle changes. Among them, sagittal angle changes correspond to trunk forward or backward tilting movements, coronal angle changes correspond to trunk lateral tilting movements towards the affected or healthy side, and horizontal angle changes correspond to trunk rotation movements. The peak change of the sagittal angle change time series data, coronal angle change time series data and horizontal angle change time series data during the evaluation action execution process are extracted respectively. The peak change of each anatomical plane is compared with the preset compensation judgment threshold corresponding to the corresponding anatomical plane. The movement type corresponding to the anatomical plane where the peak change exceeds the corresponding preset compensation judgment threshold is determined as having a compensatory movement, forming a compensatory movement type identification result for the evaluated action; wherein, the preset compensation judgment threshold is set according to the evaluated action type. For each anatomical plane where compensatory movement is determined to exist, the peak value change of the corresponding anatomical plane is used as the compensation quantification value of the corresponding compensatory movement type of the corresponding anatomical plane. The results of the compensation movement type identification and the compensation quantification value are correlated with the actual range of motion of the shoulder joint in the evaluated movement, and the compensation evaluation result of the evaluated movement is output.

[0013] Preferably, the attitude calculation uses the extended Kalman filter algorithm, which performs short-time attitude estimation using angular velocity data from multi-axis motion data, and uses acceleration data and geomagnetic data from multi-axis motion data as absolute references to correct integral drift. During the attitude calculation process, the vector magnitude of the acceleration data is calculated in real time. When the vector magnitude deviates from the gravitational acceleration value by more than a preset dynamic threshold, the weight of the acceleration data in the fusion calculation is reduced.

[0014] Preferably, before performing attitude calculation, the sensor coordinate system and the human skeletal segment anatomical coordinate system are functionally aligned and calibrated.

[0015] Preferably, functional alignment calibration is performed between the sensor coordinate system and the human skeletal segment anatomical coordinate system, including: Before the standard assessment sequence is executed, the target patient is guided to maintain a preset static calibration posture. Upper arm acceleration data and trunk acceleration data are collected in the static calibration posture. The axis corresponding to the gravity direction of the upper arm sensor coordinate system and the trunk sensor coordinate system is determined according to the direction of the acceleration vector in each sensor coordinate system. The first alignment relationship between the upper arm sensor coordinate system and the vertical direction and the third alignment relationship between the trunk sensor coordinate system and the vertical direction are established. Guide the target patient to perform a preset single-degree-of-freedom simplified calibration movement with the affected upper arm, and collect upper arm angular velocity data during the simplified calibration movement; The variation amplitude of the component sequence of upper arm angular velocity data on each axis of the sensor coordinate system is calculated. The axis with the largest variation amplitude is determined as the direction of the main motion axis in the upper arm sensor coordinate system. A second alignment relationship is established between the direction of the main motion axis and the preset anatomical axis of the upper arm bone segment. The simplified calibration action is a small-amplitude reciprocating motion for a single joint. For the upper arm sensor coordinate system, based on the vertical direction axis determined by the first alignment relationship and the direction of the main motion axis determined by the second alignment relationship, the third axis direction is solved by jointly solving the orthogonal constraint conditions between the vertical direction axis and the direction of the main motion axis to obtain the upper arm alignment transformation parameters from the upper arm sensor coordinate system to the upper arm bone segment anatomical coordinate system. For the torso sensor coordinate system, based on the vertical axis determined by the third alignment relationship and the horizontal plane direction constraint provided by the known standard body position orientation of the torso under static calibration posture, the torso alignment transformation parameters from the torso sensor coordinate system to the anatomical coordinate system of the torso skeletal segment are solved. In the subsequent posture calculation process for each evaluation action, the coordinate transformation of the real-time acquired first multi-axis motion data and second multi-axis motion data is first performed on the upper arm alignment transformation parameters and the torso alignment transformation parameters, respectively, and then the posture calculation is performed.

[0016] Preferably, the comprehensive evaluation module inputs the actual range of motion of the shoulder joint, pain assessment data, and swelling assessment data into a preset multi-dimensional comprehensive evaluation model for fusion calculation to obtain a comprehensive rehabilitation assessment score, including: The normalized range of motion of the shoulder joint for each assessment movement is obtained by dividing the actual range of motion of the shoulder joint by the upper limit of the normal reference range of each assessment movement. The normalized range of motion values ​​of each assessment movement are then weighted according to the preset clinical importance weights of each assessment movement to obtain the range of motion dimension score. Among them, the weights of shoulder flexion and abduction movements are higher than those of adduction and extension movements. Pain assessment data includes rest pain scores and movement pain scores collected using the visual analog scale (VAS). Movement pain scores are collected immediately after each assessment action. The larger of the rest pain score and movement pain score is taken, and the larger value is subtracted from the maximum value of the preset pain score range. Then, the result is divided by the maximum value of the pain score range to obtain the pain dimension score. The swelling dimension score is obtained by subtracting the swelling assessment data level from the preset maximum swelling assessment level and then dividing by the maximum swelling assessment level. The scores for activity level, pain, and swelling are weighted linearly according to preset dimensional weights to obtain the comprehensive rehabilitation assessment score.

[0017] The intelligent measurement method for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome provided in this invention includes: Real-time acquisition of first-axis multi-axis motion data of the upper arm and second-axis multi-axis motion data of the trunk of the target patient; the target patient is a patient with post-stroke shoulder-hand syndrome; The target patient is guided to perform a preset standard assessment sequence of actions through a smart terminal application; the standard assessment sequence of actions covers at least two degrees of freedom of movement of the shoulder joint. The first and second multi-axis motion data collected in each evaluation action are processed for posture calculation and motion decomposition to calculate the true range of motion of the shoulder joint. During and / or after the execution of the standard assessment sequence, pain assessment data and swelling assessment data of the target patient are collected via a smart terminal application. The actual range of motion of the shoulder joint, pain assessment data, and swelling assessment data are input into a preset multi-dimensional comprehensive evaluation model for fusion calculation to obtain a comprehensive rehabilitation assessment score.

[0018] The beneficial effects of this invention are as follows: This invention achieves objective and digital measurement of shoulder joint range of motion by collecting motion data from target patients and combining posture calculation and motion decomposition algorithms. It eliminates the reliance on operator experience associated with traditional goniometers, significantly improving measurement accuracy and consistency, and effectively solving the problems of large inter-measurer errors and poor repeatability in manual assessment. It quantifies compensatory motion components during patient assessment and outputs the true range of motion of the shoulder joint after removing compensation, addressing the clinical challenge of inflated range of motion measurements in stroke patients due to compensatory movements such as trunk forward tilting and lateral flexion, providing more reliable data for treatment planning. By integrating data from the three dimensions of range of motion, pain, and swelling into a comprehensive evaluation model, it achieves multi-dimensional quantitative assessment of shoulder-hand syndrome, overcoming the limitations of isolated analysis of indicators in traditional assessments.

[0019] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in this application.

[0020] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of an intelligent measurement device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome in an embodiment of the present invention; Figure 2 This is a schematic diagram of an intelligent measurement method for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome in an embodiment of the present invention. Detailed Implementation

[0022] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0023] This invention provides an intelligent measurement device and method for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome. The device is implemented based on a collaborative measurement system that includes wearable inertial sensors and a smart terminal application.

[0024] Before delving into the detailed descriptions of each module, we will first provide an explanation of the core technical concepts appearing in this manual, aimed at those skilled in the art (the intersection of rehabilitation medicine and biomedical engineering): A quaternion is a number that uses four numbers (one real part and three imaginary parts, usually denoted as quaternion). Satisfying normalization constraints Quaternions are mathematical tools used to express the rotational attitude of an object in three-dimensional space. Compared to the more intuitive Euler angles (such as pitch, yaw, and roll) used in clinical practice, quaternions do not suffer from the "gimbal lock" problem (a mathematical singularity where a rotational degree of freedom is lost at certain angles, leading to computational failure) when performing continuous rotation calculations, and they offer superior numerical stability during computation. In this invention, the motion data measured by each sensor is processed by attitude calculation to output a quaternion, which fully describes the three-dimensional rotational attitude of the sensor (and the attached human skeletal segment) relative to the ground reference coordinate system.

[0025] Attitude estimation refers to the process of estimating the rotational attitude of an object in three-dimensional space in real time using raw measurement data from inertial sensors (accelerometers, gyroscopes, and magnetometers) through mathematical filtering and fusion algorithms. Essentially, it is a multi-source information fusion problem: angular velocity data from gyroscopes can be integrated to obtain short-term attitude changes with high dynamic response, but long-term integration inevitably produces cumulative drift errors; accelerometer outputs primarily reflect the direction of gravity when stationary or moving slowly, serving as an absolute reference for pitch and roll, but during acceleration, motion acceleration interference is superimposed; magnetometers measure the direction of the Earth's magnetic field, providing a horizontal orientation reference. The core task of attitude estimation algorithms is to optimally integrate these three complementary information sources, maintaining high dynamic tracking capabilities while eliminating long-term drift.

[0026] The sensor coordinate system and the anatomical coordinate system are two distinct three-dimensional spatial reference frames. The sensor coordinate system consists of three orthogonal sensing axes (usually labeled X, Y, and Z axes) defined by the sensor chip itself, determined by the sensor's physical manufacturing process. The anatomical coordinate system consists of three orthogonal axes defined for skeletal segments according to human anatomy (e.g., the long axis, medial / lateral direction, and anterior / posterior direction of the upper arm). Because the sensor is worn on the body surface via straps, its coordinate axes are often not aligned with the anatomical axes of the skeletal segments, resulting in a rotational bias. "Functional alignment calibration" is the process of calculating and eliminating this rotational bias by having the patient perform specific simple postures or movements, utilizing these known motion characteristics.

[0027] This invention provides an intelligent measurement device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome, such as... Figure 1 As shown, it includes: The motion data acquisition module 1 is used to acquire in real time the first multi-axis motion data of the upper arm and the second multi-axis motion data of the trunk of the target patient; the target patient is a patient with post-stroke shoulder-hand syndrome.

[0028] The motion data acquisition module is the physical sensing layer of the entire system, and its core hardware consists of two miniaturized wireless inertial measurement unit sensor nodes.

[0029] Each sensor node integrates a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer, forming a nine-axis inertial measurement unit. The accelerometer measures the linear acceleration (including the gravitational component) experienced by the sensor along the three orthogonal axes, measured in m / s² or g (gravitational acceleration, 1g ≈ 9.81 m / s²). The gyroscope measures the angular velocity of the sensor's rotation around the three orthogonal axes, measured in ° / s (degrees per second). The magnetometer measures the intensity of the Earth's magnetic field components along the three orthogonal axes, measured in μT (microtesla). The data from these three sensors are combined to constitute the "multi-axis motion data" described in this invention.

[0030] In this preferred embodiment, the accelerometer is a MEMS accelerometer with a range of ±16g and a resolution of 16 bits to ensure that the range does not saturate when the patient performs rapid movements; the gyroscope is a MEMS angular rate sensor with a range of ±2000° / s and a resolution of 16 bits, whose zero-bias stability is better than 10° / h to ensure the accuracy of attitude integration; the magnetometer is a geomagnetic sensor with a range of ±4900μT and a resolution of 16 bits to provide heading reference. The synchronous sampling frequency of the three sensors is uniformly set to 100Hz (i.e., 100 samples per second). This frequency can meet the Nyquist sampling theorem requirements for shoulder joint rehabilitation assessment movements (typical movement frequency 0.1Hz to 2Hz) (the sampling frequency must be greater than twice the highest frequency of the signal), while also taking into account the bandwidth limitations of low-power Bluetooth transmission.

[0031] The first sensor node is fixed to the outer side of the affected upper arm of the target patient using a medical elastic bandage, specifically about 3 to 5 centimeters distal to the deltoid tuberosity. This location has relatively flat muscles and minimizes skin slippage during movement. The bandage tightness is such that the fingers cannot slide under the bandage but do not compress the skin, ensuring that the sensor does not undergo displacement or rotation relative to the bone during movement. The data collected by the first sensor node is defined as the first multi-axis motion data, including time-series data from nine channels, encompassing three-axis acceleration, three-axis angular velocity, and three-axis geomagnetic data of the upper arm.

[0032] The second sensor node is fixed in the same manner directly above the manubrium of the sternum of the target patient, or to the spinous process region of the third to fifth thoracic vertebrae (T3-T5). This location, situated above the rigid skeletal segments of the trunk, accurately reflects the posture changes of the entire trunk and is less affected by skin displacement caused by respiratory movements. The data collected by the second sensor node is defined as second multi-axis motion data, which also includes time-series data from nine channels, including triaxial acceleration, triaxial angular velocity, and triaxial geomagnetic data of the trunk.

[0033] Data from the two sensor nodes is synchronously transmitted to the smart terminal via the BLE 5.0 protocol at 20-millisecond intervals (corresponding to an effective transmission frame rate of 50Hz, with two sampling points packed into each frame). To ensure the time synchronization of the dual sensor data, the system performs clock alignment via Bluetooth connection event before each evaluation begins. Afterward, the two nodes independently time based on their respective crystal oscillators, and the smart terminal receiver performs realignment and linear interpolation based on the timestamps embedded in the data frames.

[0034] The target patients are those with post-stroke shoulder-hand syndrome. This device is primarily intended for hemiplegic patients who experience clinical symptoms of shoulder-hand syndrome after stroke, such as shoulder pain, limited range of motion, and / or hand swelling.

[0035] The standard action guidance module 2 is used to guide the target patient to perform a preset standard assessment action sequence in sequence through an application on a smart terminal; the standard assessment action sequence covers at least two degrees of freedom of movement of the shoulder joint.

[0036] The standard assessment sequence covers at least two degrees of freedom of the shoulder joint. In this preferred embodiment, the standard assessment sequence includes four standard assessment movements: shoulder flexion (raising the upper arm to its maximum angle from the side of the body towards the front and overhead in the sagittal plane), shoulder extension (swinging the upper arm from the side of the body towards the rear in the sagittal plane to its maximum angle), shoulder abduction (raising the upper arm to its maximum angle from the side of the body towards the outside in the coronal plane), and shoulder adduction (swinging the upper arm from the side of the body towards the opposite side across the midline of the body in the coronal plane to its maximum angle). These four movements correspond to the two main degrees of freedom of flexion and extension, and abduction and adduction, respectively, covering the movement directions of greatest concern in the clinical assessment of shoulder-hand syndrome. Depending on clinical needs, this sequence can be further expanded into a six-movement sequence including internal and external rotation of the shoulder joint to cover the third degree of rotation.

[0037] The detailed process of the standard action guidance module guiding the target patient to perform a preset sequence of standard assessment actions through a smart terminal application is as follows: First, the system enters the motion demonstration phase. It displays a standard motion demonstration animation of the currently assessed movement on the smart terminal's screen. This animation is generated from a standard 3D human model recorded by a motion capture system, showcasing the standard shoulder joint movement trajectory from a third-person perspective. The animation playback speed is adjusted to allow the patient to clearly observe the starting posture, direction of movement, and ending posture. Simultaneously, the system plays voice commands. These commands include instructions on the direction of movement (e.g., "Please slowly raise the affected arm from in front of your body, as high as possible") and prompts to avoid compensatory movements (e.g., "Please keep your body straight; do not lean forward or bend to the side to help raise your arm").

[0038] Following this, the system proceeds to the initial posture confirmation phase. After the demonstration, the system performs posture estimation on the real-time multi-axis motion data. Specifically, the system calculates the tilt angle of the upper arm relative to the direction of gravity based on the acceleration data from the upper arm sensor, determining whether the target patient's current posture is within the preset initial posture window. Specifically, when the upper arm hangs naturally, the output vector of the accelerometer should point downwards along the long axis of the upper arm (i.e., basically consistent with the direction of gravity), at which point the upper arm tilt angle is close to 0°. The system calculates the angle between the acceleration vector and the vertical direction, determining whether it is within the allowable window of ±10°. For example, for shoulder flexion movements, the initial posture requires the upper arm to hang naturally at the side. If the patient's posture is not ready (e.g., the upper arm is slightly raised due to pain or tension), the screen will display the text "Please lower your arm naturally" and a downward arrow. After confirmation, the system issues a start signal, the screen displays the text "Start Movement," and simultaneously plays a short prompt sound, marking the official start of data recording for the current assessment movement.

[0039] During the execution of the action, the system enters a real-time feedback phase. The application displays the current angle value calculated from the first and second multi-axis motion data on the smart terminal in real time. This angle value is the actual range of motion of the shoulder joint after compensatory separation (the calculation method will be detailed in the compensatory separation module). The current angle is displayed on the screen (e.g., "45.2°").

[0040] Next, the system enters the automatic termination determination phase. The system continuously monitors the angular velocity data extracted from the first multi-axis motion data. Angular velocity data reflects the instantaneous rotational rate of the upper arm around each axis; when the upper arm stops moving, the angular velocity value approaches zero. The system calculates the three-axis vector magnitude of the angular velocity data in real time. This modulus value reflects the overall rotational rate of the upper arm in any direction. When this modulus value remains below a preset static threshold for a preset time window, the system determines that the current assessment action has been completed. In this preferred embodiment, the preset static threshold is set to 5° / s, and the preset time window is set to 1.5 seconds. When the patient raises their upper arm to their subjectively perceived maximum position, the limb will naturally stop moving and remain still. The 5° / s threshold effectively excludes slight muscle tremors and postural maintenance adjustments at the terminal position, while the 1.5-second duration requirement prevents short pauses during the movement from being mistakenly interpreted as termination. After termination is determined, the system locks the peak value of the angle curve before the start time of the time window as the maximum joint range of motion for the current action, and simultaneously plays a voice feedback message saying "Action completed".

[0041] Finally, once all the assessment movements in the standard assessment sequence (in this example, flexion, extension, abduction, and adduction) have been performed in sequence, the system completes the guidance and enters the data processing and result display stage.

[0042] The compensation separation module 3 is used to perform posture calculation and motion decomposition processing on the first and second multi-axis motion data collected in each evaluation action, and to calculate the true range of motion of the shoulder joint.

[0043] Before performing attitude calculations, a functional alignment calibration is required between the sensor coordinate system and the anatomical coordinate system of the human skeletal segment. As explained in the preceding conceptual explanation, the sensor coordinate system consists of the three sensitive axes of the sensor chip itself, while the anatomical coordinate system consists of the three directional axes defined anatomically by the human skeletal segment. Since the sensor is worn on the human body via straps, its orientation may differ each time it is worn, causing the rotational deviation between the sensor coordinate system and the anatomical coordinate system to vary with each wear. Without calibration, the joint angles subsequently calculated based on the sensor's attitude will not accurately correspond to meaningful anatomical motion planes (for example, the result intended for flexion-extension angles may contain components of abduction and adduction directions), leading to systematic errors in the measurement results.

[0044] Traditional methods for precisely placing sensors based on bony landmarks require a high level of anatomical knowledge from the operator and are difficult to locate on the body surface of hemiplegic patients. This invention employs a functional alignment calibration method, which requires only simple posture maintenance and minor movements by the patient to complete the calibration. It is convenient to operate and highly error-tolerant.

[0045] The specific steps for functional alignment calibration between the sensor coordinate system and the human skeletal segment anatomical coordinate system are as follows: Before the standard assessment sequence is executed, the application guides the target patient to maintain a preset static calibration posture via voice and on-screen text. In this preferred embodiment, the preset static calibration posture requires the patient to sit on a hard chair without armrests (or stand), with the torso upright, arms hanging naturally at the sides, and palms facing the sides.

[0046] The system instructs the patient to maintain this posture for approximately 3 seconds. During this period, the system collects upper arm acceleration data and trunk acceleration data under the static calibration posture, and calculates the time average of the acceleration data from each sensor within a 3-second time window to obtain the upper arm static acceleration vector and trunk static acceleration vector. The purpose of using time averages is to reduce the influence of random noise from the sensors and improve the accuracy of orientation estimation.

[0047] Since the human body does not generate motion acceleration under static conditions, the accelerometer output only reflects the projection direction of the gravity vector in the sensor coordinate system. Therefore, based on the direction of the acceleration vector in each sensor coordinate system, the system determines the axis corresponding to the gravity direction in the upper arm sensor coordinate system and the torso sensor coordinate system, respectively.

[0048] For the upper arm sensor, the static acceleration vector is normalized to obtain the unit vector of gravity. Under the preset static calibration posture (upper arm hanging naturally), the direction of gravity is basically consistent with the long axis direction of the upper arm skeletal segment (from shoulder to elbow), therefore... A correspondence was established between a certain direction in the upper arm sensor coordinate system and the vertical direction (major axis direction) in the upper arm anatomical coordinate system, i.e., the first alignment relationship.

[0049] For the torso sensor, the static acceleration vector is also normalized to obtain the unit vector of gravity. r. In the preset static calibration posture (torso upright), the direction of gravity is consistent with the longitudinal axis of the torso skeletal segment (pointing from head to toe), therefore A correspondence was established between a certain direction in the torso sensor coordinate system and the vertical direction (longitudinal axis) in the torso anatomical coordinate system, namely the third alignment relationship.

[0050] The first and third alignment relationships each define an axis direction (vertical axis), but the complete determination of a three-dimensional coordinate system requires at least two non-parallel axis directions. Therefore, a second direction needs to be determined through subsequent steps.

[0051] Subsequently, the system guides the target patient to perform a preset single-degree-of-freedom simplified calibration movement with the affected upper arm. In this preferred embodiment, the simplified calibration movement is a small-amplitude reciprocating motion in the shoulder joint flexion direction, that is, the upper arm swings back and forth with an amplitude of about 20° to 30° in the sagittal plane, repeated 3 to 5 times. This movement is required to be a small-amplitude reciprocating motion targeting a single joint (shoulder joint) and in a single plane of motion (sagittal plane), so that the rotational angular velocity of the movement is mainly concentrated in a predictable rotational axis direction (i.e., the medial and lateral axes perpendicular to the sagittal plane).

[0052] For patients with extremely low muscle strength (muscle strength ≤ grade 2) in the affected upper limb who are unable to actively complete small-amplitude reciprocating movements, the following alternative can be used: an assistant assessor assists the patient in completing the same amplitude reciprocating movement in the sagittal plane by passively assisting the movement. The angular velocity data collected by the sensor is equivalent to that under active movement conditions, and the operation logic of the calibration algorithm is not affected.

[0053] The system collects upper arm angular velocity data during the simplified calibration process, lasting approximately 5 seconds and comprising 3 to 5 complete reciprocating cycles.

[0054] The system calculates the variation amplitude of the upper arm angular velocity data component sequences on each axis of the sensor coordinate system (i.e., the X-axis angular velocity time sequence, the Y-axis angular velocity time sequence, and the Z-axis angular velocity time sequence). In this preferred embodiment, the variation amplitude is measured by the peak-to-peak value (i.e., the difference between the maximum and minimum values) of each component sequence during the motion period; standard deviation can also be used as an alternative measure. Since the simplified calibration action is a motion around a single rotation axis, the gyroscope component in that rotation axis direction will exhibit a significant sinusoidal waveform change (angular velocity alternating between positive and negative), while the variation amplitude of the components on the other two orthogonal axes is extremely small (only due to sensor noise and minor cross-coupling).

[0055] The system determines the axis with the largest change as the direction of the principal axis of motion in the upper arm sensor coordinate system. This establishes a second alignment between the main axis of motion and the pre-defined anatomical axis of the upper arm skeletal segment. Specifically, if the simplified calibration movement is pre-defined as sagittal forward flexion of the shoulder joint, then the axis of rotation should be the medial and lateral axes in the anatomical coordinate system of the upper arm skeletal segment. That is, the inner and outer axes corresponding to the upper arm anatomical coordinate system are expressed in the sensor coordinate system.

[0056] It should be noted that the second alignment relationship only needs to be established for the upper arm sensor; the calibration of the torso sensor does not require an additional dynamic calibration step.

[0057] For the upper arm sensor coordinate system, the system has obtained two physically meaningful directions: the vertical axis determined by the first alignment relationship ( (corresponding to the major axis direction of the anatomical coordinate system), and the direction of the principal axis of motion determined by the second alignment relationship ( (corresponding to the inner and outer axes of the anatomical coordinate system). Since the sensor coordinate system is a three-dimensional space, three mutually perpendicular axes are needed to fully determine its rotational relationship with the anatomical coordinate system.

[0058] However, due to the limited precision with which patients maintain their posture and perform actions in actual practice, and Typically, the orientation is not perfectly perpendicular; therefore, the system uses orthogonal constraints between the vertical axis and the principal axes of motion to jointly solve for the third axis direction. The specific orthogonalization process employs the Gram-Schmidt orthogonalization method, with the following steps: First, using… As the first priority axis (The major axis direction has the highest priority because the measurement accuracy in the direction of gravity is the most reliable); then... Projected onto perpendicular to On the plane, remove its along The directional component, after normalizing the remaining components, is used as the second axis. (Inner and outer axes); finally, the vector cross product operation is performed ( Solving for the direction of the third axis (Anterior and posterior axes) This yields the upper arm alignment transformation parameters from the upper arm sensor coordinate system to the upper arm skeletal segment anatomical coordinate system. Mathematically, the upper arm alignment transformation parameters are expressed as a 3×3 rotation matrix (its three columns are...). , , ), or equivalently expressed as a unit quaternion.

[0059] For the torso sensor coordinate system, the system has determined the vertical axis through a third alignment relationship. (corresponding to the vertical axis direction of the torso anatomical coordinate system). It is also necessary to determine a direction within the horizontal plane to fully constrain the three-dimensional rotational relationships.

[0060] Since the trunk is the central structure of the human body, sensors on it are not as convenient as those on the limbs for performing single-joint, single-degree-of-freedom calibration movements (trunk movement involves coupled movements of multiple segments of the spine, making it difficult to ensure that the movement occurs only on a single axis of rotation). Therefore, this invention employs a different strategy for trunk sensors: utilizing the known standard posture of the trunk under static calibration to provide a horizontal plane constraint. Specifically, in the aforementioned steps, the patient is asked to maintain a standard sitting posture and face a known direction (e.g., facing the screen of a smart device, facing a wall directly in front, or verbally confirmed by the assessor that the patient is facing forward). Under this known standard posture, combined with the magnetic north direction information in the horizontal plane provided by the magnetometer readings under static calibration, the system can determine the projection direction of the anterior-posterior axis (sagittal plane normal) of the trunk anatomical coordinate system onto the sensor coordinate system. Alternatively, in a simplified scheme that does not use a magnetometer reference, the system directly utilizes the gravity vector. The reference orientation in the horizontal plane is obtained by cross product relationship between the sensor coordinate system and a predetermined axis (e.g., the orientation axis marked on the sensor housing, which is required to be roughly facing the patient when worn).

[0061] Based on the vertical axis and horizontal plane constraints, the three orthogonal axis directions are also solved through the Gram-Schmidt orthogonalization process to obtain the trunk alignment transformation parameters from the trunk sensor coordinate system to the trunk skeletal segment anatomical coordinate system.

[0062] In the subsequent attitude calculation process for each evaluation action, the system first performs coordinate transformation on the real-time acquired first multi-axis motion data and second multi-axis motion data in each sampling period using the upper arm alignment transformation parameter and the torso alignment transformation parameter respectively (that is, multiply the original acceleration vector, angular velocity vector and magnetic force vector in the sensor coordinate system by the alignment rotation matrix on the left and transform them to the anatomical coordinate system of the corresponding bone segment), and then performs the attitude calculation described later.

[0063] Attitude calculation employs the Extended Kalman Filter (EKF) algorithm. In this invention, the state variables in EKF are attitude quaternions. and gyroscope zero bias vector There are a total of 7 states. The EKF operation consists of two alternating steps: prediction and update. In the prediction step, the system performs short-time attitude estimation using angular velocity data from multi-axis motion data. Specifically, it uses angular velocity measured by a gyroscope. Subtract the estimated zero bias The corrected angular velocity was then obtained. Using the kinematic differential equations of quaternions Numerical integration (using the first-order Runge-Kutta method with a step size of 10 milliseconds sampling interval) yields the prior attitude quaternion estimate. Simultaneously propagating the state error covariance moment ,in Let Jacobian matrix be the system state transition matrix. Let be the process noise covariance matrix. Gyroscope integration has extremely high accuracy within a single sampling period (10 milliseconds), but due to the presence of zero-bias drift and random walk noise, long-term integration will produce a non-negligible cumulative drift error.

[0064] In the update process, the system uses acceleration and geomagnetic data from multi-axis motion data as absolute references to correct integral drift. When the human body is at rest or near rest, the accelerometer output primarily reflects the direction of the gravity vector. By comparing the theoretical gravity direction prediction under a prior attitude with the actual accelerometer measurements, corrections for pitch and roll directions can be obtained. The magnetometer outputs the direction of the Earth's magnetic field vector; combined with the known gravity direction, it can provide corrections for yaw direction. EKF utilizes the Kalman gain matrix... The measurement residuals (the difference between the observed values ​​and the prior predictions) are weighted and added to the prior estimate to obtain the posterior attitude estimate. And update the covariance matrix, where Let Jacobian matrix be the equation of measurement. This is the measurement noise covariance matrix.

[0065] However, during shoulder joint motion assessment, the patient's upper arm is in an accelerated motion state, and the accelerometer output includes not only the gravity component but also the motion acceleration component. If the accelerometer output is still used as the gravity reference at this time, it will introduce significant errors. Therefore, during attitude calculation, the system calculates the vector magnitude of the acceleration data in real time. When the vector magnitude deviates from the gravitational acceleration value by more than a preset dynamic threshold, the weight of acceleration data in the fusion calculation is reduced.

[0066] In the specific calculation, the system dynamically adjusts the accelerometer noise covariance matrix in the EKF based on the deviation between the vector magnitude and the gravitational acceleration value. The diagonal element value corresponding to the accelerometer. In this preferred embodiment, the preset dynamic threshold is set to 0.3g. When the deviation does not exceed 0.3g, the accelerometer measurement noise variance maintains its calibrated reference value (e.g., 0.01). The accelerometer correction functions normally; however, when the deviation exceeds 0.3g, the accelerometer measurement noise variance is amplified to 100 to 1000 times the reference value (e.g., increased to 1 to 10). This causes the accelerometer's contribution to the Kalman gain to approach zero. During this period, the system mainly relies on the gyroscope's integral calculation to maintain attitude tracking, avoiding incorrect corrections that mistakenly interpret motion acceleration as a shift in the direction of gravity. Since the duration of shoulder joint assessment movements is typically on the order of 3 to 10 seconds, and the movement speed is relatively slow, the cumulative time for the accelerometer weight to be reduced during the movement is short, and the cumulative amount of gyroscope integral drift is within an acceptable range (tested to show that the drift within 10 seconds typically does not exceed 0.5°). When the patient reaches the terminal position and remains stationary, the vector magnitude returns to around 1g, the accelerometer weight immediately recovers, and the drift is quickly corrected.

[0067] The compensatory separation module performs posture calculation and motion decomposition processing on the first and second multi-axis motion data collected in each evaluation action. The detailed process of calculating the true range of motion of the shoulder joint is as follows: First, the system independently performs attitude calculations on the first and second multi-axis motion data (as previously mentioned) to obtain the upper arm attitude quaternion. ) and trunk posture quaternions ( These two quaternions describe the absolute attitude of the upper arm and torso in the global reference frame (a geodetic coordinate system based on the direction of gravity and magnetic north), and output a pair of quaternion values ​​at each sampling time.

[0068] Next, the system takes the inverse of the torso posture quaternion (equivalent to the conjugate quaternion in the case of a unit quaternion). Multiply by the upper arm posture quaternion to calculate the relative quaternion of the upper arm relative to the torso: ; in, This represents the quaternion multiplication operation.

[0069] To facilitate understanding, a visual example is used here: Suppose a patient is undergoing a shoulder flexion assessment, and their shoulder joint actually only elevates by 60°, but simultaneously their trunk tilts forward by 15°. If only a single sensor on the upper arm is used for measurement, because the sensor simultaneously senses the combined effect of the shoulder joint movement (60°) and the trunk tilt (15°), the output absolute posture change of the upper arm will be close to 75°, resulting in an inflated measurement value. However, in the dual-sensor solution of this invention, the absolute posture quaternion of the upper arm sensor... It does indeed contain comprehensive change information of 75°, but at the same time, the absolute attitude quaternion of the torso sensor... A precise 15° forward tilt of the torso was recorded. Through the aforementioned relative quaternion operations, the 15° torso motion component was automatically removed from the absolute motion of the upper arm, resulting in the final... It only reflects 60° of pure shoulder joint movement, thus decoupling shoulder joint movement from trunk compensatory movement.

[0070] Then, the system decomposes the relative quaternions into angles according to a preset shoulder joint rotation decomposition sequence, transforming them into anatomically significant flexion-extension angles, abduction-adduction angles, and rotation angles.

[0071] A predefined shoulder joint rotation decomposition sequence refers to the rule of decomposing the three-dimensional rotation expressed by relative quaternions into three sequentially superimposed single-axis rotation angles according to a specific axis order and rotation sequence. In sports biomechanics, Euler angle decomposition is commonly used to achieve this. Different axis order choices will lead to different angle decomposition results, and there is also a "gimbal lock" problem—when the second rotation angle is close to ±90°, the first and third rotation angles will produce coupling singularities, resulting in abrupt changes in angle values ​​or inability to uniquely determine them.

[0072] In selecting the rotational decomposition sequence, this invention chooses a sequence that ensures the gimbal lock position is far from the maximum range of motion of the shoulder joint in each assessed movement. Specifically, in this preferred embodiment, for flexion / extension and abduction / adduction movements, the system uses a ZXY sequence (i.e., the first axis rotation is around the longitudinal axis Z of the trunk coordinate system, the second axis rotation is around the rotated medial and lateral axes X, and the third axis rotation is around the upper arm long axis Y after two rotations). The gimbal lock position of this sequence occurs when the X-axis (medial and lateral axes) rotation angle is close to ±90°, while the range of motion in each direction of patients with shoulder-hand syndrome is usually much less than 90°, thus avoiding the occurrence of singular regions. (From quaternions) The conversion to the three Euler angles of the ZXY sequence is accomplished using the standard quaternion-Euler angle analytical formula, which is a well-known technique in the field of sports biomechanics. Its derivation process will not be elaborated here.

[0073] Finally, based on the type of assessment movement being performed, the system extracts the peak values ​​of the angular components corresponding to the principal motion plane from the flexion-extension, abduction-adduction, and rotation angles, using these as the true range of motion of the shoulder joint for the corresponding assessment movement. The "principal motion plane" refers to the anatomical plane in which the primary movement was expected to occur during the design of the current assessment movement. Specifically: When a patient performs a shoulder flexion movement, the main movement occurs in the sagittal plane. The system extracts the maximum value of the flexion-extension angle component during the movement execution time (the positive value represents flexion, i.e., raising the arm forward and upward), as the true range of motion of the shoulder joint in the flexion direction (e.g., outputting "flexion 85°"). When the patient performs a shoulder extension movement, the main movement also occurs in the sagittal plane. The system extracts the extreme values ​​of the flexion and extension angle components in the negative direction (the absolute value represents the extension angle, i.e., swinging the arm backward) as the true range of motion of the shoulder joint in the extension direction (e.g., outputting "extension 30°"). When a patient performs a shoulder abduction movement, the main movement occurs in the coronal plane. The system extracts the maximum value of the abduction and adduction angle components during the movement execution time (the positive value represents abduction, i.e., raising the arm to the side), as the true range of motion of the shoulder joint in the abduction direction (e.g., outputting "abduction 90°"). When a patient performs a shoulder adduction movement, the main movement occurs in the coronal plane. The system extracts the extreme values ​​of the abduction and adduction angle components in the negative direction (the absolute value represents the adduction angle, i.e., swinging the arm to the opposite side across the body midline) as the true range of motion of the shoulder joint in the adduction direction (e.g., outputting "adduction 20°").

[0074] Through this complete posture calculation and motion decomposition processing chain, the system ultimately outputs a more accurate true range of motion of the shoulder joint.

[0075] In addition to separating compensation and calculating the actual range of motion, the compensation separation module also includes steps for multi-plane classification detection and degree quantification of compensatory movements, which are used to accurately describe the characteristics of compensation (which direction the trunk compensates and how much it compensates), providing a basis for developing precise rehabilitation training programs.

[0076] First, during the execution of each evaluation action, attitude calculation is performed on the second multi-axis motion data (torso sensor data) on a sampling cycle basis (every 10 milliseconds in this embodiment) to obtain the torso attitude quaternion at each sampling time. ( ,in The start signal is emitted at this time. (To determine the termination time), a quaternion time sequence of trunk posture is formed within the execution time period of each evaluation action.

[0077] Then, to assess the start time of the action. torso posture quaternion Using the baseline attitude quaternion as the reference attitude quaternion, the attitude change of the torso attitude quaternion at each sampling time relative to the baseline attitude quaternion is calculated: ; Time-series data of trunk motion changes during the execution of each evaluation action were obtained. The design with a zero-point reference time means that: unit quaternion (Indicates zero rotation), subsequent time steps It reflects the amount of trunk offset relative to the starting position during the execution of the movement. This design ensures that even if the patient has a slight postural deviation in the initial position (such as habitual mild scoliosis), as long as the deviation exists before the movement begins, it will not be mistakenly counted as compensatory movement during the execution of the movement.

[0078] Next, the system decomposes the time-series data of trunk motion changes into angles according to the three anatomical planes of the human trunk (reducing...). (After being converted into three angular components according to the Euler angle decomposition rule), the time series data of sagittal angle changes were obtained respectively. Time series data of coronal angle changes Time series data of horizontal plane angle change Changes in the sagittal plane angle correspond to forward or backward tilting movements of the trunk (e.g., a patient bends forward to increase the apparent angle of flexion), changes in the coronal plane angle correspond to lateral tilting movements of the trunk towards the affected or unaffected side (e.g., a patient tilts their body towards the unaffected side to increase the apparent height of their raised arm during abduction), and changes in the horizontal plane angle correspond to trunk rotation movements (e.g., a patient rotates their trunk towards the affected side during forward flexion to make their arm appear to be raised higher).

[0079] The system extracts the peak change (i.e., the maximum absolute value of each time series data) of the angle changes in the three anatomical planes during the evaluation action execution process: , , The absolute value is taken because compensatory movements can occur in either direction (e.g., leaning forward or backward both count as sagittal plane compensation), while the peak value is taken to capture the moment when compensation is most severe during the entire movement.

[0080] The system compares the peak value changes at each anatomical plane with the corresponding preset compensation thresholds. These preset compensation thresholds are set according to the type of action being assessed and can be adjusted via system configuration parameters based on different clinical scenarios or patient groups. For example, flexion and extension movements most easily induce compensation in the sagittal plane (forward tilting or backward leaning), therefore the sagittal threshold is set relatively strictly (5°); abduction and adduction movements most easily induce compensation in the coronal plane (lateral tilting), therefore the coronal threshold is also set relatively strictly (5°). For compensation directions that are less likely to be induced by the action, the threshold is appropriately relaxed to 8° to avoid misjudging normal, minor accompanying movements as compensation. Horizontal rotational compensation may occur in all movements and is of high clinical concern, therefore it is uniformly set at 5°.

[0081] The system identifies the movement type corresponding to the anatomical plane where the peak change exceeds the corresponding preset compensation threshold as having compensatory movement, thus forming a compensation movement type identification result for the evaluated movement. For example, during the execution of a forward flexion movement, if the peak change in the sagittal plane is 12° (exceeding the 5° threshold), the peak change in the coronal plane is 3° (not exceeding the 8° threshold), and the peak change in the horizontal plane is 7° (exceeding the 5° threshold), then the compensation movement type identification result is "the forward flexion movement has sagittal forward tilt compensation and horizontal rotation compensation".

[0082] For each anatomical plane where compensatory movement is identified, the system uses the peak change in the corresponding anatomical plane as the quantified compensation value for the corresponding type of compensatory movement. Continuing the previous example, the quantified compensation value for sagittal anterior tilt compensation is 12°, and the quantified compensation value for horizontal plane rotation compensation is 7°. The quantified compensation value is expressed in degrees, indicating how many degrees the trunk has deviated in that direction, which is convenient for therapists to understand and record.

[0083] Finally, the system correlates the compensation movement type identification result and the compensation quantification value with the actual range of motion of the shoulder joint in the evaluated movement, and outputs the compensation evaluation result of the evaluated movement. The output format is as follows: "Shoulder flexion - actual range of motion: 85°; compensation detection: sagittal forward tilt compensation 12°, horizontal plane rotation compensation 7°".

[0084] In addition to the angle values ​​of joint range of motion and compensation analysis, the compensation separation module also performs multi-dimensional quantitative assessment of motion quality. The motion quality assessment focuses not on "how far the movement was" (range of motion angle), but on "how well the movement was performed" (the fineness of motor control). Its assessment indicators can more sensitively reflect subtle changes in neuromotor control function.

[0085] First, the system performs spectral analysis on the angular velocity data extracted from the first multi-axis motion data, calculating the spectral arc length of the angular velocity signal during the execution of each evaluation action, as an indicator of motion smoothness. The calculation process for the spectral arc length is as follows: A Fast Fourier Transform (FFT) is performed on the time series of the angular velocity vector magnitude during the evaluation action execution period to obtain the frequency domain amplitude spectrum; the amplitude spectrum is normalized (based on the maximum amplitude); within the effective frequency range of the frequency axis (set to 0 to 4 Hz in this embodiment, covering the frequency range of voluntary shoulder joint movement), the arc length of the normalized amplitude spectrum curve is calculated (i.e., the summation of the curve segment lengths between adjacent points in the frequency dimension). The arc length value is always negative (negative according to the standard definition of spectral arc length). Because the spectral energy of smooth motion is concentrated in the low-frequency region, the amplitude spectrum curve is concise and short; while unsmooth motion (full of jitter, pauses, and restarts) exhibits energy diffusion in the high-frequency region, causing the amplitude spectrum curve to bend and lengthen, and the absolute value of the arc length to increase. Therefore, the smaller the absolute value of the arc length (closer to zero), the smoother the motion.

[0086] Secondly, the system performs time-domain envelope extraction on the angular velocity data extracted from the first multi-axis motion data to obtain the motion velocity curves for each evaluated action. Specifically, this involves calculating the sequence of angular velocity three-axis vector magnitudes. The envelope of the motion velocity curve is extracted using a low-pass Butterworth filter with a cutoff frequency of 2Hz, resulting in a smooth curve reflecting the change in motion velocity over time. The system performs peak detection and symmetry analysis on the motion velocity curve.

[0087] Peak detection is used to obtain the number of peak values. A normal single shoulder lift (from the rest position to the terminal position) should exhibit a single, smooth velocity peak—accelerating first and then decelerating, with a velocity curve resembling a bell shape. If multiple velocity peaks are detected (i.e., the velocity rises and falls repeatedly), it indicates that there is a pause followed by re-acceleration during the movement. The more peaks there are, the worse the motion control ability and the more fragmented the motion strategy. The system uses a peak detection algorithm with minimum peak spacing and minimum peak height constraints (minimum peak spacing is set to 0.3 seconds, and minimum peak height is set to 15% of the maximum peak value) to eliminate false peaks caused by noise.

[0088] Symmetry analysis calculates the ratio of the time of the rising edge (acceleration phase, the time from the start of motion to the maximum velocity) to the time of the falling edge (deceleration phase, the time from the maximum velocity to the end of motion) of the velocity curve, yielding the symmetry index. The formula is: Symmetry index = Acceleration phase time / Deceleration phase time. The symmetry index in healthy individuals is typically between 0.8 and 1.2 (close to 1.0 indicates symmetrical acceleration and deceleration). Stroke patients, due to insufficient muscle strength or coordination impairment, often exhibit a pattern of rapidly reaching peak velocity but slowly decelerating (symmetry index less than 0.8), or a pattern of slowly accelerating to peak velocity and then suddenly stopping (symmetry index greater than 1.2).

[0089] Next, the system performs variability analysis on the motion trajectories generated by repeated execution of the same assessment action, calculating the trajectory variation coefficient as an indicator of motion repeatability. In this preferred embodiment, each assessment action requires the patient to repeat it three times. The system normalizes the shoulder joint's actual range of motion angle-time curves from the three executions (by linear interpolation, the time axis of each execution is uniformly mapped to a normalized time axis from 0% to 100%, resulting in 101 evenly distributed time points), and then calculates the standard deviation of the angle values ​​of the three curves at each normalized time point. Finally, the average of the standard deviations at all time points is taken as the trajectory coefficient of variation: Trajectory coefficient of variation = The unit is degrees. The smaller the repeatability index, the more stable the patient's control over the movement pattern and the more likely they are to repeat the same movement trajectory; a larger value suggests instability in the motor control strategy.

[0090] Finally, the system uses the motion smoothness index (SPARC value), peak count, symmetry index, and motion repeatability index (trajectory variation coefficient) as inputs for motion quality assessment dimensions, which are then fed into a multi-dimensional comprehensive evaluation model. In this model, the motion quality assessment dimension can participate in the weighted calculation as an independent fourth assessment dimension (in which case the SPARC value and other indicators need to be normalized and assigned dimension weights), or it can serve as a correction factor for the activity dimension score (for example, when motion smoothness is extremely poor, the activity dimension score is multiplied by a correction coefficient less than 1). The specific fusion method can be configured by clinical users according to their actual needs, and the motion quality assessment indicators are also displayed as independent supplementary information in the evaluation report.

[0091] The multidimensional data acquisition module 4 is used to acquire pain assessment data and swelling assessment data of the target patient through an application on a smart terminal during and / or after the execution of the standard assessment action sequence.

[0092] Pain assessment data included rest pain scores and movement pain scores collected using the Visual Analogue Scale (VAS). The VAS is the most widely used subjective pain quantification tool in clinical practice. It is represented by a horizontal line 10 cm long, with "no pain" (0 points) marked on the left and "severe pain" (10 points) marked on the right. Patients express the intensity of pain by marking a position on the line.

[0093] In terms of the application's interface, the system presents the VAS scale as a horizontal slider on the screen. Patients or assistant assessors select the pain level by sliding their fingers on the slider, and the corresponding value (accurate to 0.1 points) is displayed above the slider in real time.

[0094] Resting pain score is collected once before the assessment begins, while the patient remains at rest. The application displays the prompt: "Please rate the level of shoulder pain you currently experience while at rest."

[0095] Exercise pain scores are collected immediately after each assessment movement. That is, as soon as each assessment movement is determined to be completed and before the next assessment movement begins, the application immediately displays the VAS scoring interface, prompting: "Please rate the level of shoulder pain you experienced while performing this movement." This real-time data collection design ensures that the exercise pain score reflects the acute pain sensation caused by the most recent movement, avoiding memory decay bias caused by delayed assessment. The system records the exercise pain scores for all assessment movements and takes the highest value as the representative exercise pain score for that assessment.

[0096] During the assessment of the movement, if the patient stops the movement prematurely due to pain or other reasons, the automatic termination judgment mechanism can also identify the premature stop and record the current angle as the range of motion value of this movement. At the same time, the system collects the VAS score of movement pain for each movement in real time. If the score reaches the preset pain warning threshold (which can be configured by the assessor according to the clinical situation), the system can issue a prompt message, and the assessor can decide whether to stop the subsequent assessment movements to avoid the risk of secondary injury during the assessment process.

[0097] After the standard assessment sequence is completed, the multidimensional data acquisition module collects swelling assessment data through the infrared structured light scanning submodule integrated into the smart terminal or the independent scanning handle that comes with the device. The infrared structured light scanning submodule consists of a near-infrared speckle pattern projector and a near-infrared image sensor, with a depth resolution better than 1.5mm at a working distance of 25cm to 35cm. During the assessment, the application guides the patient to place the affected forearm and hand in a standard position (forearm in neutral rotation, fingers naturally extended, forearm placed on a standardized support pad), and the assessor holds the scanning handle along the guide. The system guides the affected limb to perform multi-angle scanning, covering a range from 10cm proximal to the wrist crease to 1cm distal to the metacarpophalangeal joint. The application displays a heatmap of point cloud coverage in real time and prompts that the data acquisition is complete once the coverage reaches a preset threshold. The same process is then used to scan the healthy limb to obtain an individualized self-control baseline. The system performs statistical outlier filtering and voxel downsampling preprocessing on the original depth images from each angle. The multi-angle local point cloud is registered and fused into a unified coordinate system using a point-to-surface iterative nearest point algorithm. Finally, a Poisson surface reconstruction algorithm is used to generate closed triangular meshes for the affected and healthy sides.

[0098] After the mesh is generated, the system sequentially performs the following steps to determine the standardized cut-off section: Principal component analysis is performed on the triangular mesh point cloud, and the eigenvector corresponding to the largest eigenvalue is taken as the limb's longitudinal axis Z-axis (positive direction towards the proximal end). A local coordinate system for the limb is established, and the coordinates of all vertices are transformed. Along the Z-axis, the mesh is cut off at equal intervals using tangent planes with a step size of 1 mm. The perimeter of each cross-section is calculated to obtain the cross-sectional circumference profile curve C(z). Within the far 40% segment of the scanning range, C(z) is smoothed, and the minimum point is searched. The coordinates corresponding to the most significant minimum point, where the average circumference decreases by no less than 5% within the near 20 mm and far 15 mm ranges, are marked as... The system highlights the automatically located wrist crease section in the 3D view. The evaluator must click "Confirm" to accept the location, or drag the marker line to the correct position and then confirm. If automatic positioning fails (no minimum point meeting the conditions is found within the target segment), the evaluator must manually click the specified wrist crease section in the 3D view. The system does not allow automatic volume calculation without manual confirmation. Based on, extract to For the mm segment, the open sections of the meshes on the affected and healthy sides are trimmed and closed to form closed local meshes for volume integration; the volume of the affected side is calculated using the divergence theorem method on the above closed local meshes. volume of the healthy side ; based on volume swelling rate As the primary quantitative indicator of swelling, the difference in circumference between the affected and healthy sides at 5cm proximal to the wrist crease and at the midpoint of the metacarpal bones in the circumference distribution curves of each cross section was simultaneously extracted as an auxiliary indicator; ultimately, the... The swelling dimension score is linearly mapped to the range of 0 to 1 according to the preset clinical grading threshold (e.g., SVI≤5% is no obvious swelling, 5% to 15% is mild, 15% to 30% is moderate, and greater than 30% is severe; the threshold can be configured according to the clinical scenario), and then input into the multi-dimensional comprehensive evaluation model for fusion calculation.

[0099] The comprehensive evaluation module 5 is used to input the actual range of motion of the shoulder joint, pain assessment data, and swelling assessment data into a preset multi-dimensional comprehensive evaluation model for fusion calculation to obtain a comprehensive rehabilitation assessment score.

[0100] The system normalizes the actual range of motion of the shoulder joint for each assessment movement. Specifically, the actual measured angle of each assessment movement is divided by the pre-stored upper limit of the normal reference range for that movement, resulting in a normalized range of motion value between 0 and 1. If the actual measured angle of a movement is exactly equal to the upper limit of the reference range, the normalized range of motion value for that movement is the full value of 1; if the actual angle is 0°, the normalized value is 0; in rare cases where the actual measured value exceeds the upper limit of the reference range, the normalized value is truncated to 1, and no single movement is allowed to score more than the full score, to ensure the stability of subsequent weighted calculations.

[0101] In this preferred embodiment, the upper limit of the normal reference range for each assessment movement is determined based on statistical data of healthy individuals in the same age group, referencing the normal range of motion standards for joint mobility published by the American Academy of Orthopaedic Surgeons and the American Medical Association. Specifically, the values ​​are: shoulder flexion 180°, extension 60°, abduction 180°, and adduction 45°.

[0102] After normalizing each movement, the system calculates a weighted average based on the preset clinical importance weights for each assessment movement to obtain a range of motion score. The weighting is based on the importance of each movement direction in the reconstruction of daily living functions in patients with post-stroke shoulder-hand syndrome. Shoulder flexion and abduction are core movements for basic daily activities such as picking up objects, combing hair, and dressing; restriction in these movements severely impacts independence in daily life, therefore they are assigned higher weights. While adduction and extension are also basic shoulder movement directions, their frequency of use and compensatory substitution in daily activities are relatively high, thus they are assigned lower weights. In this preferred embodiment, flexion and abduction each account for 35%, and extension and adduction each account for 15%, with the sum of the weights for all four directions being 100%. The closer the range of motion score is to 1, the closer the patient's overall shoulder joint function is to a normal level.

[0103] Pain assessment data included two sub-data sets: rest pain score and movement pain score. Both were collected using the visual analog scale, with a scale range of 0 (no pain) to 10 (severe pain).

[0104] Before merging the scores, the system first selects the larger of the resting pain score and the movement pain score, using this larger value as the aggregated value representing the severity of pain in this assessment. The reason for using the larger value instead of the average is that for patients in the acute phase of shoulder-hand syndrome, their resting pain is often quite severe, and the resting pain score may be higher than the movement pain score; while for patients in the subacute or chronic phase, pain is mainly induced by movement, and the movement pain score is usually higher than the resting pain score. Using the larger of the two values ​​ensures that the comprehensive evaluation always reflects the most severe pain state actually experienced by the patient, thus avoiding an underestimation of the severity of the condition.

[0105] After determining the total pain value, the system uses reverse scoring logic to convert it into a pain dimension score: the total pain value is subtracted from the maximum scale score of 10, and then the difference is divided by the maximum scale score of 10, resulting in a pain dimension score ranging from 0 to 1. The purpose of reverse scoring is to ensure that the pain dimension score and the activity dimension score are aligned in direction, meaning that a higher score indicates a better state: a maximum pain dimension score of 1 is obtained when there is no pain, and a score of 0 is obtained when the pain reaches the maximum scale value.

[0106] Swelling assessment data is presented as a four-level ordinal scale, corresponding to level 0 (no swelling), level 1 (mild swelling), level 2 (moderate swelling), and level 3 (severe swelling). The system uses the same inverse scoring logic as the pain dimension to convert the swelling level value into a dimension score: subtracting the current swelling level value from the highest scale value of 3, and then dividing the difference by the highest scale value of 3, yields a swelling dimension score ranging from 0 to 1. No swelling is scored as a perfect 1, severe swelling is scored as 0, and mild and moderate swelling correspond to scores of approximately 0.67 and 0.33, respectively.

[0107] After calculating the independent scores for the three dimensions mentioned above, the system performs a weighted linear combination of the activity level, pain, and swelling scores according to preset dimension weights to obtain a comprehensive rehabilitation assessment score. The sum of the weights for the three dimensions is 100%, with the activity level dimension having the highest weight at 50%, the pain dimension having a medium weight at 30%, and the swelling dimension having the lowest weight at 20%.

[0108] Limited range of motion is the most fundamental functional impairment of shoulder-hand syndrome and the most direct objective basis for developing and adjusting rehabilitation programs. Therefore, it is given the highest weight to ensure that the overall score is most sensitive to the state of motor function. Pain not only directly affects the patient's quality of life but also indirectly hinders the improvement of range of motion through a vicious cycle of "pain-avoidance of movement-worsening of joint contractures." It is the second core dimension that needs to be addressed during rehabilitation and is therefore given the second highest weight. Although swelling is an important characteristic of shoulder-hand syndrome, in most cases, the improvement of swelling often precedes and is relatively independent of the recovery of range of motion. Moreover, its severity has a relatively weak direct correlation with functional impairment, so it is given the lowest weight.

[0109] The comprehensive rehabilitation assessment score ranges from 0 to 1. In practice, it can be proportionally mapped to a percentage system of 0 to 100 for easier clinical communication and patient understanding. A higher score indicates a better overall rehabilitation status across the three dimensions of joint range of motion, pain control, and swelling.

[0110] Furthermore, when available dimensions for assessing motor quality, the system can incorporate motor smoothness, peak count, symmetry index, and motor repeatability as extended dimensions into the comprehensive evaluation model. In this case, the system first normalizes each of the four motor quality sub-indicators to the 0-1 range, and then takes a weighted average based on the clinical weights of each sub-indicator in reflecting motor control ability to obtain a motor quality dimension score. The motor quality dimension score is then introduced into the weighted fusion stage with a certain weight, while the weights of the other three dimensions are proportionally reduced to maintain the constraint that the sum of all dimension weights is always 100%.

[0111] In another embodiment, the standard assessment sequence of motion is expanded to include a six-motion sequence that includes internal and external rotation of the shoulder joint, covering three degrees of freedom of movement; correspondingly, the weighting of the range of motion dimension score is adjusted to flexion 30%, abduction 30%, extension 10%, adduction 10%, internal rotation 10%, and external rotation 10%.

[0112] In another embodiment, in a scenario configuration suitable for acute patients with severe hand swelling and pain as their main complaint, the dimensional weights are adjusted to 40% for pain, 30% for swelling, and 30% for mobility, so that the comprehensive assessment score is more sensitive to changes in acute symptoms.

[0113] This invention provides an intelligent measurement method for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome, such as... Figure 2 As shown, it includes: Step 1: Real-time acquisition of first multi-axis motion data of the upper arm and second multi-axis motion data of the trunk of the target patient; the target patient is a patient with post-stroke shoulder-hand syndrome; Step 2: Guide the target patient to perform a preset standard assessment sequence of movements sequentially via a smart terminal application; the standard assessment sequence of movements covers at least two degrees of freedom of movement of the shoulder joint; Step 3: Perform posture calculation and motion decomposition processing on the first and second multi-axis motion data collected in each evaluation action, and calculate the true range of motion of the shoulder joint; Step 4: During and / or after the execution of the standard assessment sequence, collect pain assessment data and swelling assessment data of the target patient through the application on the smart terminal; Step 5: Input the actual range of motion of the shoulder joint, pain assessment data, and swelling assessment data into the preset multi-dimensional comprehensive evaluation model for fusion calculation to obtain a comprehensive rehabilitation assessment score.

[0114] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An intelligent measuring device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome, characterized in that, include: The motion data acquisition module is used to collect real-time multi-axis motion data of the upper arm and the second multi-axis motion data of the trunk of the target patient; the target patient is a patient with post-stroke shoulder-hand syndrome. A standard movement guidance module is used to guide the target patient to perform a preset standard assessment movement sequence sequentially via a smart terminal application; the standard assessment movement sequence covers at least two degrees of freedom of movement of the shoulder joint; The compensation separation module is used to perform posture calculation and motion decomposition processing on the first and second multi-axis motion data collected in each evaluation action, and to calculate the true range of motion of the shoulder joint. The multidimensional data acquisition module is used to collect pain assessment data and swelling assessment data of the target patient through an application on a smart terminal during and / or after the execution of the standard assessment action sequence. The comprehensive evaluation module is used to input the actual range of motion of the shoulder joint, pain assessment data, and swelling assessment data into a preset multi-dimensional comprehensive evaluation model for fusion calculation to obtain a comprehensive rehabilitation assessment score.

2. The intelligent measuring device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome as described in claim 1, characterized in that, The standard action guidance module guides the target patient to perform a preset sequence of standard assessment actions sequentially via a smart terminal application, including: The standard motion demonstration animation of the current evaluation action is displayed on the smart terminal, and voice commands are played simultaneously. The voice commands include instructions on the direction of movement and prompts to avoid compensatory movements. The attitude is estimated based on the real-time first multi-axis motion data. It is determined whether the current attitude of the target patient is within the preset starting attitude window. After confirmation, a start signal is issued. During the execution of the action, the current angle value calculated from the first multi-axis motion data and the second multi-axis motion data is displayed in real time on the smart terminal; When the angular velocity data extracted from the first multi-axis motion data remains below the preset static threshold for a preset time window, the current evaluation action is determined to be completed. Once all assessment actions in the standard assessment action sequence have been performed, the guidance is complete.

3. The intelligent measuring device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome as described in claim 1, characterized in that, The compensation separation module performs posture calculation and motion decomposition processing on the first and second multi-axis motion data collected in each evaluation action, and calculates the true range of motion of the shoulder joint, including: The attitude calculation is performed independently on the first multi-axis motion data and the second multi-axis motion data to obtain the upper arm attitude quaternion and the torso attitude quaternion. Calculate the relative quaternion of the upper arm to the torso by multiplying the torso posture quaternion inversely by the upper arm posture quaternion. The relative quaternion is decomposed into angles according to the preset shoulder joint rotation decomposition sequence, and converted into flexion-extension angle, abduction-adduction angle and rotation angle; Based on the type of assessment action being performed, the peak values ​​of the corresponding angular components of the main motion plane are extracted from the flexion-extension angle, abduction-adduction angle, and rotation angle, and used as the true range of motion of the shoulder joint for the corresponding assessment action.

4. The intelligent measuring device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome as described in claim 3, characterized in that, The compensation separation module performs posture calculation and motion decomposition processing on the first and second multi-axis motion data collected in each assessment action. After calculating the true range of motion of the shoulder joint, it also includes the steps of multi-plane classification detection and degree quantification of the compensation motion. The multi-plane classification detection and degree quantization include: During the execution of each evaluation action, the attitude calculation is performed on the second multi-axis motion data on a sampling cycle to obtain the trunk attitude quaternion at each sampling time, forming a time sequence of trunk attitude quaternions within the execution time period of each evaluation action. Using the trunk posture quaternion at the start of the evaluation action as the reference posture quaternion, the posture change of the trunk posture quaternion at each sampling time relative to the reference posture quaternion is calculated to obtain the time series data of trunk motion change during the execution of each evaluation action. The time-series data of trunk motion changes were decomposed into three anatomical planes of the human trunk to obtain time-series data of sagittal angle changes, coronal angle changes, and horizontal angle changes. Among them, sagittal angle changes correspond to trunk forward or backward tilting movements, coronal angle changes correspond to trunk lateral tilting movements towards the affected or healthy side, and horizontal angle changes correspond to trunk rotation movements. The peak change of the sagittal angle change time series data, coronal angle change time series data and horizontal angle change time series data during the evaluation action execution process are extracted respectively. The peak change of each anatomical plane is compared with the preset compensation judgment threshold corresponding to the corresponding anatomical plane. The movement type corresponding to the anatomical plane where the peak change exceeds the corresponding preset compensation judgment threshold is determined as having a compensatory movement, forming a compensatory movement type identification result for the evaluated action; wherein, the preset compensation judgment threshold is set according to the evaluated action type. For each anatomical plane where compensatory movement is determined to exist, the peak value change of the corresponding anatomical plane is used as the compensation quantification value of the corresponding compensatory movement type of the corresponding anatomical plane. The results of the compensation movement type identification and the compensation quantification value are correlated with the actual range of motion of the shoulder joint in the evaluated movement, and the compensation evaluation result of the evaluated movement is output.

5. The intelligent measuring device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome as described in claim 3, characterized in that, Attitude calculation uses the extended Kalman filter algorithm, which uses angular velocity data from multi-axis motion data to perform short-time attitude estimation, and uses acceleration data and geomagnetic data from multi-axis motion data as absolute references to correct integral drift. During attitude calculation, the vector magnitude of acceleration data is calculated in real time. When the vector magnitude deviates from the gravitational acceleration value by more than a preset dynamic threshold, the weight of acceleration data in the fusion calculation is reduced.

6. The intelligent measuring device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome as described in claim 3, characterized in that, Before performing attitude calculation, the sensor coordinate system and the human skeletal segment anatomical coordinate system are functionally aligned and calibrated.

7. The intelligent measurement device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome as described in claim 6, comprising functional alignment calibration of the sensor coordinate system and the human skeletal segment anatomical coordinate system, including: Before the standard assessment sequence is executed, the target patient is guided to maintain a preset static calibration posture. Upper arm acceleration data and trunk acceleration data are collected in the static calibration posture. The axis corresponding to the gravity direction of the upper arm sensor coordinate system and the trunk sensor coordinate system is determined according to the direction of the acceleration vector in each sensor coordinate system. The first alignment relationship between the upper arm sensor coordinate system and the vertical direction and the third alignment relationship between the trunk sensor coordinate system and the vertical direction are established. Guide the target patient to perform a preset single-degree-of-freedom simplified calibration movement with the affected upper arm, and collect upper arm angular velocity data during the simplified calibration movement; The variation amplitude of the component sequence of upper arm angular velocity data on each axis of the sensor coordinate system is calculated. The axis with the largest variation amplitude is determined as the direction of the main motion axis in the upper arm sensor coordinate system. A second alignment relationship is established between the direction of the main motion axis and the preset anatomical axis of the upper arm bone segment. The simplified calibration action is a small-amplitude reciprocating motion for a single joint. For the upper arm sensor coordinate system, based on the vertical direction axis determined by the first alignment relationship and the direction of the main motion axis determined by the second alignment relationship, the third axis direction is solved by jointly solving the orthogonal constraint conditions between the vertical direction axis and the direction of the main motion axis to obtain the upper arm alignment transformation parameters from the upper arm sensor coordinate system to the upper arm bone segment anatomical coordinate system. For the torso sensor coordinate system, based on the vertical axis determined by the third alignment relationship and the horizontal plane direction constraint provided by the known standard body position orientation of the torso under static calibration posture, the torso alignment transformation parameters from the torso sensor coordinate system to the anatomical coordinate system of the torso skeletal segment are solved. In the subsequent posture calculation process for each evaluation action, the coordinate transformation of the real-time acquired first multi-axis motion data and second multi-axis motion data is first performed on the upper arm alignment transformation parameters and the torso alignment transformation parameters, respectively, and then the posture calculation is performed.

8. The intelligent measuring device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome as described in claim 5, characterized in that, The compensation separation module also performs the following operations: Spectral analysis is performed on the angular velocity data extracted from the first multi-axis motion data to calculate the spectral arc length of the angular velocity signal during the execution of each evaluation action, which serves as an indicator of motion smoothness. The angular velocity data extracted from the first multi-axis motion data is subjected to time-domain envelope extraction to obtain the motion velocity curves of each evaluated action. Peak detection and symmetry analysis are performed on the motion velocity curves to obtain the number of peaks and the symmetry index. A variability analysis was performed on the motion trajectories generated by repeated execution of the same evaluation action, and the trajectory variation coefficient was calculated as an indicator of motion repeatability. The motion smoothness index, peak count, symmetry index, and motion repeatability index are used as inputs for motion quality assessment dimensions and subsequently fed into a multi-dimensional comprehensive evaluation model.

9. The intelligent measuring device for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome as described in claim 1, characterized in that, The comprehensive evaluation module inputs the actual range of motion of the shoulder joint, pain assessment data, and swelling assessment data into a preset multi-dimensional comprehensive evaluation model for fusion calculation, and obtains a comprehensive rehabilitation assessment score, including: The normalized range of motion of the shoulder joint for each assessment movement is obtained by dividing the actual range of motion of the shoulder joint by the upper limit of the normal reference range of each assessment movement. The normalized range of motion values ​​of each assessment movement are then weighted according to the preset clinical importance weights of each assessment movement to obtain the range of motion dimension score. Among them, the weights of shoulder flexion and abduction movements are higher than those of adduction and extension movements. Pain assessment data includes rest pain scores and movement pain scores collected using the visual analog scale (VAS). Movement pain scores are collected immediately after each assessment action. The larger of the rest pain score and movement pain score is taken, and the larger value is subtracted from the maximum value of the preset pain score range. Then, the result is divided by the maximum value of the pain score range to obtain the pain dimension score. The swelling dimension score is obtained by subtracting the swelling assessment data level from the preset maximum swelling assessment level and then dividing by the maximum swelling assessment level. The scores for activity level, pain, and swelling are weighted linearly according to preset dimensional weights to obtain the comprehensive rehabilitation assessment score.

10. An intelligent measurement method for comprehensive rehabilitation assessment of post-stroke shoulder-hand syndrome, characterized in that, include: Real-time acquisition of first-axis multi-axis motion data of the upper arm and second-axis multi-axis motion data of the trunk of the target patient; the target patient is a patient with post-stroke shoulder-hand syndrome; The target patient is guided to perform a preset standard assessment sequence of actions through a smart terminal application; the standard assessment sequence of actions covers at least two degrees of freedom of movement of the shoulder joint. The first and second multi-axis motion data collected in each evaluation action are processed for posture calculation and motion decomposition to calculate the true range of motion of the shoulder joint. During and / or after the execution of the standard assessment sequence, pain assessment data and swelling assessment data of the target patient are collected via a smart terminal application. The actual range of motion of the shoulder joint, pain assessment data, and swelling assessment data are input into a preset multi-dimensional comprehensive evaluation model for fusion calculation to obtain a comprehensive rehabilitation assessment score.

Images