A closed-loop force feedback based control system for step-by-step quantitative spinal traction physiotherapy
By using a step-by-step quantitative traction physiotherapy control system for the spine based on closed-loop force feedback, combined with a self-learning fuzzy logic controller and a step-by-step traction strategy, the individualization and safety issues of existing spinal traction systems are solved, achieving refined and adaptive traction force adjustment and improving treatment effectiveness.
Patent Information
- Application Number
- CN202610476268.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-13
- Publication Date
- 2026-05-29
AI Technical Summary
Existing spinal traction systems lack individualized control and cannot perceive physiological status in real time, resulting in inaccurate treatment parameters, safety hazards, and a lack of self-learning and adaptive capabilities, making it difficult to adapt to the differentiated needs of different disease types and disease stages.
A step-by-step quantitative traction physiotherapy control system for the spine is adopted based on closed-loop force feedback. It combines a data acquisition module, a traction bed drive module, and a self-learning fuzzy logic controller. Adaptive adjustment is achieved through a fuzzy logic rule base and neural network. The step-by-step traction strategy includes preloading, creep, and maintenance phases, and the traction force is dynamically adjusted.
It improves the precision and safety of treatment, can respond to changes in electromyographic signals and pain feedback in real time, avoids soft tissue damage, adapts to individual differences, and enhances the individualization and safety of treatment.
Smart Images

Figure CN122097050A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of physical therapy equipment, and more particularly to a spinal step-by-step quantitative traction physical therapy control system based on closed-loop force feedback. Background Technology
[0002] Spinal diseases are common and frequently occurring in clinical practice, seriously affecting the quality of life of those requiring traction. Traction therapy, as an important conservative treatment for spinal diseases, applies axial traction force to increase intervertebral space, reduce nerve root compression, and relieve muscle spasms, and has been widely used in clinical practice.
[0003] Currently, existing traction systems generally suffer from the following technical defects: First, treatment parameters rely entirely on the operator's experience for setting, resulting in imprecise traction force control and an inability to adapt to individual differences in the patients. Second, they lack the ability to perceive the physiological state of the patients in real time, failing to respond promptly when the patients exhibit signs of muscle spasm, posing a safety hazard. Third, most systems employ simple PID control algorithms, with the traction force output linearly related to preset parameters, making it difficult to cope with nonlinear dynamic changes in muscle tension and spasm attacks during treatment. Fourth, the traction strategy is singular and fixed, lacking self-learning and adaptive capabilities, making it difficult to adapt to the differentiated needs of patients with different disease types and stages of disease.
[0004] To address the aforementioned issues, there is an urgent need to develop a closed-loop control system capable of real-time sensing of the physiological state of the subject to be traction, possessing self-learning capabilities, and adaptively adjusting the traction force, in order to improve the safety, effectiveness, and individualization of spinal traction therapy. Summary of the Invention
[0005] To address the aforementioned technical problems, this invention provides a spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback.
[0006] In a first aspect, the present invention provides a spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback, the system comprising: The data acquisition module is used to acquire the real-time traction force value output by the traction bed drive module, as well as the electromyographic signal data and displacement change rate of the object to be tractioned; The traction bed drive module is used to receive traction force control commands output by the self-learning fuzzy logic controller, and to perform traction actions on the traction bed body according to the traction force control commands. The self-learning fuzzy logic controller is electrically connected to the data acquisition module and the traction bed drive module, respectively, and is used to generate the traction force control command based on the real-time traction force value, the electromyographic signal data, the displacement change rate and the pain level signal fed back by the object to be traction, using a preset step-by-step traction strategy. The self-learning fuzzy logic controller includes a fuzzy logic rule base and an adaptive fuzzy adjustment engine. The fuzzy logic rule base stores first fuzzy logic rules constructed based on expert experience and historical clinical data. The adaptive fuzzy adjustment engine is configured to use a fuzzy parameter adjustment model constructed based on neural networks and fuzzy inference to dynamically generate the traction control command by updating the shape parameters of the membership function of the first fuzzy logic rule.
[0007] Preferably, the step-by-step traction strategy includes: During the preloading phase, an adaptive percentage of the weight of the object to be traction is used as the initial target traction force, which is maintained for a first predetermined time. During the creep stage, the traction force is increased linearly at an adaptive rate. When the displacement change rate reaches the adaptive displacement change rate threshold, the linear increase stops, and the traction force at the current moment is used as the maintenance traction force. During the maintenance phase, the maintenance traction force is dynamically adjusted by the self-learning fuzzy logic controller based on the real-time traction force value, the electromyographic signal data, and the pain level signal, and is maintained for a second preset time. The adaptive percentage, the adaptive rate, and the adaptive displacement change rate threshold are all determined based on the physiological characteristic parameters of the object to be tractioned and the tolerance level of the traction force in the previous treatment cycle.
[0008] Preferably, the self-learning fuzzy logic controller further includes: The tolerance assessment module is used to acquire the electromyographic signal data of the subject to be traction during the previous physical therapy cycle, and to extract the body tolerance characterization features from the electromyographic signal data using a pre-trained signal recognition network; based on the body tolerance characterization features, the pain level signal, and the traction force convergence value of the maintenance phase, the tolerance traction force level is generated based on the third fuzzy logic rule stored in the fuzzy logic rule base. The body tolerance characterization features include at least: initial tolerance threshold, peak electromyographic signal, and spasm duration.
[0009] Preferably, the fuzzy logic rule base also stores a second fuzzy logic rule, which is used to perform fuzzy reasoning on the adaptive percentage of the first physiotherapy cycle based on the physiological characteristic parameters of the object to be tractioned, to obtain the first adaptive percentage; the physiological characteristic parameters include at least: age, gender, body mass index, disease type, disease stage and initial electromyography baseline value.
[0010] Preferably, the fuzzy logic rule base also stores a fourth fuzzy logic rule, which is used to perform fuzzy reasoning on the adaptive rate of the first physiotherapy cycle based on the physiological characteristic parameters to obtain the first adaptive rate.
[0011] Preferably, the fuzzy logic rule base also stores a fifth fuzzy logic rule, which is used to perform fuzzy reasoning on the adaptive displacement change rate threshold of the first physiotherapy cycle based on the physiological characteristic parameters, so as to obtain the first adaptive displacement change rate threshold.
[0012] Preferably, the self-learning fuzzy logic controller further includes: The personalized parameter adjustment module is used to dynamically adjust the adaptive percentage in subsequent treatment cycles using a first exponential smoothing model with limited amplitude, based on the tolerance traction level and the initial adaptive percentage of the previous treatment cycle, to obtain the adaptive percentage of the subsequent treatment cycle; to dynamically adjust the adaptive rate in subsequent treatment cycles using a second exponential smoothing model with limited amplitude, based on the tolerance traction level and the initial adaptive rate of the previous treatment cycle, to obtain the adaptive rate of the subsequent treatment cycle; and to dynamically adjust the adaptive displacement change rate threshold in subsequent treatment cycles using a third exponential smoothing model with limited amplitude, based on the tolerance traction level and the initial adaptive displacement change rate threshold of the previous treatment cycle, to obtain the adaptive displacement change rate threshold of the subsequent treatment cycle.
[0013] Preferably, the fuzzy parameter adjustment model includes: an input layer, a fuzzification layer, a rule layer, a normalization layer, a conclusion layer, and an output layer; The input layer is used to receive the real-time traction force value, the electromyographic signal data, and the pain level signal, and to extract the electromyographic signal deviation of the electromyographic signal data relative to the resting baseline and the rate of change of the electromyographic signal amplitude of the electromyographic signal data. The blurring layer is used to blur the real-time traction force value, the electromyographic signal deviation, the real-time electromyographic signal amplitude change rate, and the pain level signal using a Gaussian membership function; The rule layer is used to calculate the trigger strength of each rule in the first fuzzy logic rule using the multiplicative T-norm, and to correct the trigger strength based on the rule weight set for each rule in the first fuzzy logic rule to obtain an effective trigger strength; The normalization layer is used to calculate the normalized value of the effective trigger strength of each rule in the first fuzzy logic rule; The conclusion layer is used to calculate the output conclusion of each rule in the first fuzzy logic rule using a linear function; The output layer is used to obtain the traction force adjustment amount based on the normalized value of the effective trigger intensity and the output conclusion.
[0014] Preferably, the system further includes: The human-computer interaction interface is used to receive the pain level signal and use the pain level signal as a constraint condition. When the pain level signal exceeds a preset pain threshold, the traction force is forcibly reduced or the traction is suspended.
[0015] The present invention has the following technical effects: This invention provides a step-by-step quantitative traction physiotherapy control system for the spine based on closed-loop force feedback. Compared with the prior art, the embodiments of this invention have the following advantages: The spinal step-by-step quantitative traction physiotherapy control system provided in this application adopts a step-by-step traction strategy, which not only conforms to the viscoelastic mechanical properties of spinal tissue under traction, but also makes the control target more specific, improving the precision of treatment. In the creep phase, the rate of displacement change is used as the criterion for stopping linear increases, rather than simply relying on time or traction force thresholds. The rate of displacement change directly reflects the real-time response of the spinal structure under traction. When the rate of change reaches a preset adaptive threshold, the increase stops, effectively avoiding soft tissue damage, muscle spasms, or nerve root stimulation caused by excessive traction speed or amplitude. Simultaneously, in the maintenance phase, the traction force is dynamically adjusted through a self-learning fuzzy logic controller, which can respond in real-time to changes in the electromyographic signals and subjective pain level feedback of the subject being tractioned. The traction force is promptly reduced when pre-spasm symptoms appear, significantly improving the safety of the treatment process. The fuzzy parameter adjustment model is set as a six-layer structure consisting of an input layer, a fuzzification layer, a rule layer, a normalization layer, a conclusion layer, and an output layer, fully parameterizing the fuzzy logic reasoning process into a differentiable neural network. This enables the entire fuzzy parameter adjustment model to be trained end-to-end via backpropagation, achieving adaptive optimization of the fuzzy rule base. The physical meaning of each layer of the neural network is clear, giving the system's decision-making process a transparent logical chain. Clinicians can understand the contribution of each rule, facilitating clinical validation and intervention. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of a step-by-step quantitative traction physiotherapy control system for the spine based on closed-loop force feedback according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the self-learning fuzzy logic controller according to an embodiment of the present invention; Figure Labels 1-Data acquisition module, 2-Traction bed drive module, 3-Self-learning fuzzy logic controller, 31-Fuzzy logic rule base, 32-Adaptive fuzzy adjustment engine, 33-Personalized parameter adjustment module, 34-Tolerance assessment module. Detailed Implementation
[0017] The embodiments of the present invention are described in detail below with reference to the accompanying drawings. The embodiments are provided for illustrative purposes only and should not be construed as limiting the scope of the invention. The accompanying drawings are for reference and illustration only and do not constitute a limitation on the scope of protection of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of this invention.
[0018] In the description of this invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0019] This invention discloses a step-by-step quantitative traction physiotherapy control system for the spine based on closed-loop force feedback, referring to... Figure 1 The system includes: a data acquisition module, a traction bed drive module 2, and a self-learning fuzzy logic controller 3.
[0020] The spinal step-by-step quantitative traction physiotherapy control system disclosed in this application is applied to a traction bed including a fixed bed, a movable bed board, and a traction end connection part. The fixed bed serves as the main support frame and is fixedly installed on the ground or base. The movable bed board is slidably set on the fixed bed and is used to support the object to be tractioned. The traction end connection part is set at one end of the movable bed board and is used to mechanically connect with the high-precision tension sensor of the data acquisition module 1 and the traction bed drive module 2 to realize the transmission of traction force.
[0021] The data acquisition module 1 includes a high-precision tension sensor, a wireless surface electromyography (EMG) sensor, and a displacement sensor. The high-precision tension sensor is installed at the traction end connection of the traction bed, accurately acquiring the real-time traction force value acting on the spine of the object to be traction. The high-precision tension sensor of this application includes an elastomer, strain gauges, a signal conditioning circuit, and structural connectors. The elastomer, as a force-sensitive element, is made of high-strength alloy steel or aluminum alloy and undergoes elastic deformation under traction force. Strain gauges are attached to the stress concentration area of the elastomer, typically forming a Wheatstone bridge, converting the small deformation of the elastomer into a change in resistance. The signal conditioning circuit is electrically connected to the strain gauges and includes an amplification circuit, a filtering circuit, and an analog-to-digital conversion circuit, used to amplify the weak resistance change signal step by step, filter out high-frequency noise, and convert it into a digital signal. Structural connectors are located at both ends of the elastomer. One end is mechanically fixed to the movable bed plate of the traction bed via threaded fasteners, and the other end is detachably connected to the traction rope or transmission mechanism of the traction bed drive module 2 via a pin or hook, ensuring that all traction force flows through the elastomer.
[0022] The high-precision tension sensor collects real-time traction force values at a sampling frequency of not less than 100Hz and feeds them back to the self-learning fuzzy logic controller 3. In a preferred embodiment of this application, the high-precision tension sensor adopts an S-type tension and compression sensor, which has a compact structure and is easy to connect in series in the traction transmission chain through threaded connectors. It also has high resistance to eccentric loads and is suitable for physiotherapy scenarios where the traction force direction is stable and there is no obvious lateral force component.
[0023] In an alternative embodiment of this application, if the object to be traction is obese or severely stiff and requires greater traction force, a column-type tension sensor is used. The column-type tension sensor is usually mounted on a base and fixed between the traction bed and the drive mechanism by a flange or thread. It has a larger range and a more robust structure, making it suitable for high-load traction scenarios.
[0024] In a preferred embodiment of this application, the wireless surface electromyography (EMG) sensor includes: Ag / AgCl surface electrodes, a preamplifier, a main amplification and filtering circuit, an analog-to-digital converter, a microcontroller, and a Bluetooth communication module. The Ag / AgCl surface electrodes are configured in a bipolar differential pattern and are attached to the corresponding body surface locations of the multifidus and erector spinae muscles of the lumbar region of the subject to be traction, for non-invasive acquisition of muscle electrical activity signals. The preamplifier is positioned close to the electrodes to minimize noise coupling in the signal transmission path. The acquired EMG signals are amplified, filtered, and digitized, and then transmitted in real-time to the self-learning fuzzy logic controller 3 via the Bluetooth communication module.
[0025] In a preferred embodiment of this application, a displacement sensor measures the displacement of the movable bed board relative to the fixed bed frame, and the displacement is corrected by combining the traction angle parameter. Based on the corrected displacement, the displacement change rate between the movable bed board and the fixed bed frame is obtained. The traction angle parameter is obtained in real time by a preset tilt angle sensor or an angle sensor. To achieve accurate displacement measurement, a wire-type displacement sensor is selected. The sensor body is fixedly installed on the side of the fixed bed frame by a bracket, located outside the travel range of the movable bed board. The free end of the wire is fixed to the same side of the movable bed board, and the fixing point is located at the center of the travel direction of the movable bed board to reduce measurement errors caused by bed board tilting or deflection. When the servo motor drives the movable bed board to move, the wire extends or retracts with the displacement of the movable bed board. The rotary encoder inside the sensor outputs a pulse signal proportional to the displacement, which is calculated by the controller to obtain the displacement change rate.
[0026] In a preferred embodiment of this application, the traction bed drive module 2 is connected to the movable bed plate of the traction bed. As the system's power execution unit, the traction bed drive module 2 precisely executes traction actions on the traction bed and dynamically adjusts the traction force according to the traction force control commands output by the self-learning fuzzy logic controller 3. In a preferred embodiment of this application, the traction bed drive module 2 includes: a servo motor, a precision reducer, a transmission conversion structure, and an encoder. The servo motor, as the power source, is a permanent magnet synchronous AC servo motor or a DC brushless servo motor, possessing a high torque-to-inertia ratio and fast start-stop response characteristics. The input shaft of the precision reducer is mechanically connected to the output shaft of the servo motor to reduce the speed, increase the output torque, and match the output characteristics to the traction load. The transmission conversion structure is connected to the output end of the precision reducer and the movable bed plate of the traction bed, respectively, to convert the rotational motion of the servo motor into linear reciprocating motion, thereby applying or releasing traction force to the spine of the object to be tractioned. The encoder is integrated into the tail of the servo motor to monitor the position, speed, and direction of the motor rotor in real time, forming a semi-closed-loop or fully closed-loop control of the speed and position loops.
[0027] In a preferred embodiment of this application, the control input terminal of the traction bed drive module 2 is electrically connected to the control output terminal of the self-learning fuzzy logic controller 3. The position / velocity signal output by its internal encoder is fed back to the servo driver, forming an internal velocity loop and position loop; simultaneously, the real-time traction force value signal collected by the high-precision tension sensor is fed back to the self-learning fuzzy logic controller 3, forming an external force loop. This dual closed-loop control architecture, combining internal and external loops, ensures rapid response in the inner loop and accurate output in the outer loop, together achieving high-precision, low-overshoot, and fast-response adaptive control of the spinal traction force.
[0028] The self-learning fuzzy logic controller 3 is the core operational control module of the spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback in this application. It is electrically connected to the high-precision tension sensor, the wireless surface electromyography sensor, the displacement sensor, and the traction bed drive module 2. In a preferred embodiment of this application, the self-learning fuzzy logic controller 3 is configured to: receive in real time the traction force value fed back by the high-precision tension sensor, the electromyography signal data collected by the wireless surface electromyography sensor, the intervertebral disc height data collected by the displacement sensor, and the pain level signal, which is fed back in real time by the subject to be traction via a handheld terminal; perform fuzzy reasoning and decision-making based on the built-in step-by-step traction strategy, generate corresponding traction force control commands, and output the traction force control commands to the traction bed drive module 2 to precisely control the loading, maintenance, and release of traction force, thereby achieving adaptive traction physiotherapy of the spine of the subject to be traction.
[0029] To achieve adaptive traction therapy for different pathological stages and individual tolerance differences, the stepwise traction strategy of this application includes the following three stages: In the preloading phase, an adaptive percentage of the body weight of the object to be pulled is used as the initial target traction force, which is maintained for a first predetermined time. The purpose of the preloading phase is to overcome the initial viscous resistance of the muscles and establish a stable starting state for subsequent creep stretching. Specifically, the self-learning fuzzy logic controller 3 operates in force control mode, and the control objective is to ensure that the actual traction force quickly and smoothly reaches the initial target traction force and maintains it for the first preset time. The initial target traction force in this application is calculated based on the body weight of the object to be pulled, and the calculation formula is as follows: in, Indicates the initial target traction force. Indicates an adaptive percentage. This indicates the weight of the object to be pulled.
[0030] During the creep phase, once the preloading phase is complete and the object to be traction is stable, the self-learning fuzzy logic controller 3 switches to a speed-force hybrid control mode. During the creep phase, the traction force is increased linearly at an adaptive rate, while receiving displacement change rate signals from the displacement sensor in real time. The self-learning fuzzy logic controller 3 compares this real-time displacement change rate with an adaptive displacement change rate threshold. When it determines that the displacement change rate equals the adaptive displacement change rate threshold, it means that the ligament tissue has entered the creep plateau phase. The self-learning fuzzy logic controller 3 immediately stops the linear increase and locks the current traction force value as the maintenance traction force.
[0031] During the maintenance phase, the self-learning fuzzy logic controller 3 uses the traction force as the baseline traction force, introduces electromyographic signal data and the active input signal of the object to be traction as feedback, and dynamically adjusts the traction force through the self-learning fuzzy logic controller 3 to maintain the traction force for a second preset time.
[0032] This application employs a three-stage, step-by-step traction strategy comprising a preloading stage, a creep stage, and a maintenance stage. This decomposes the traction process into three clearly defined stages: initial adaptation, linear loading, and dynamic maintenance. The preloading stage uses an adaptive percentage of the subject's body weight as the initial target traction force, avoiding the drawbacks of fixed initial traction forces in traditional methods. The creep stage linearly increases the traction force at an adaptive rate, using the displacement change rate as the criterion for stopping the linear increase, achieving precise capture of the spinal biomechanical response. The maintenance stage dynamically adjusts the traction force based on real-time feedback, forming a complete closed-loop control. This three-stage structure not only conforms to the viscoelastic mechanical properties of spinal tissue under traction but also makes the control target more clearly defined, improving the precision of treatment. The creep stage uses the displacement change rate as the criterion for stopping the linear increase, rather than simply relying on time or a traction force threshold. The displacement change rate directly reflects the real-time response of the spinal structure under traction; when the change rate reaches a preset adaptive threshold, the increase stops, effectively avoiding soft tissue damage, muscle spasms, or nerve root stimulation caused by excessive traction speed or amplitude. Meanwhile, during the maintenance phase, the traction force is dynamically adjusted by the self-learning fuzzy logic controller 3, which can respond in real time to changes in the electromyographic signals and subjective pain level feedback of the subject to be traction. When signs of spasm appear, the traction force is reduced in time, which significantly improves the safety of the treatment process.
[0033] In the preferred embodiment of this application, the adaptive percentage in the preloading stage, the adaptive rate in the creep stage, and the adaptive displacement change rate threshold are not fixed constants, but are iteratively optimized based on adaptive fuzzy inference for different objects to be tractioned and the tolerance traction force level of the previous treatment cycle. This allows the traction parameters to continuously converge to the individual optimal value as the number of treatments increases, overcoming the technical defects of traditional traction systems with fixed parameters that cannot adapt to individual differences in the objects to be tractioned.
[0034] In the preferred embodiment of this application, such as Figure 2The schematic diagram shown illustrates the structure of the self-learning fuzzy logic controller 3. The self-learning fuzzy logic controller 3 includes a fuzzy logic rule base 31 and an adaptive fuzzy adjustment engine 32. The fuzzy logic rule base 31 stores a series of fuzzy logic rules constructed based on expert experience and historical clinical data, including: a first fuzzy logic rule, a second fuzzy logic rule, a third fuzzy logic rule, a fourth fuzzy logic rule, and a fifth fuzzy logic rule. Specifically, the first fuzzy logic rule is used for fuzzy inference regarding the maintenance traction force during the maintenance phase; the second fuzzy logic rule is used for fuzzy inference regarding the adaptive percentage; the third fuzzy logic rule is used for fuzzy inference regarding the tolerance traction force level of the traction object; the fourth fuzzy logic rule is used for fuzzy inference regarding the linear adaptive rate during the creep phase; and the fifth fuzzy rule is used for fuzzy inference regarding the adaptive height change threshold during the creep phase. The fuzzy logic rule base 31 is permanently stored in the non-volatile memory of the self-learning fuzzy logic controller 3 in the form of software programs or firmware, or dynamically loaded into its random access memory during system operation.
[0035] In a preferred embodiment of this application, the construction and initialization of the fuzzy logic rule base 31 is based on the rule-based transformation of expert experience knowledge and the rule mining and optimization of historical clinical data. Specifically, the clinical diagnostic and treatment experience of multiple senior rehabilitation physicians and / or orthopedic physicians is collected and converted into fuzzy language rules using knowledge engineering methods. For example, a clinician's experience statement, "If the electromyographic signal of the object to be tractioned suddenly increases, it means that the muscle is about to spasm and the traction force needs to be reduced immediately," is quantified as "IF (the rate of change of electromyographic signal amplitude is 'rapidly increasing') AND (the current electromyographic signal amplitude is 'high') THEN (the traction force adjustment is 'negatively large')." Expert experience rules constitute the initial skeleton of the rule base, ensuring that the system has basic intelligent reasoning capabilities that conform to clinical consensus from the outset. Based on expert experience rules, a rule mining mechanism based on big data is introduced. By performing association rule analysis or fuzzy clustering on a large amount of historical clinical treatment data, potential patterns hidden in the data that transcend individual physician experience are discovered. Historical clinical treatment data includes tension curves, electromyographic response curves, and final efficacy evaluations of theoretical subjects of different ages, genders, weights, disease types, and disease stages during the treatment process. Data-driven patterns are transformed into supplementary rules, which are then validated, revised, and enriched to construct a hybrid fuzzy rule base that combines clinical consensus with data insights.
[0036] In a preferred embodiment of this application, during the first physiotherapy session for the subject to be traction, the adaptive percentage in the preloading phase is obtained through fuzzy inference based on the second fuzzy logic rule. In subsequent physiotherapy cycles, it is dynamically adjusted based on the subject's tolerance level to traction in previous physiotherapy cycles. During the first physiotherapy session, the initial adaptive percentage is generated based on the subject's physiological characteristic parameters according to the second fuzzy logic rule. These physiological characteristic parameters include: age, gender, body mass index (BMI), disease type, disease stage, and initial baseline electromyography value. The domain of age is [18, 80], in years, and the fuzzy subset is {young, middle-aged, elderly}. The older the age, the lower the tissue viscoelasticity, and the rate should be appropriately reduced. The domain of body mass index (BMI) is [18, 35], in kg / m², and the fuzzy subset is {underweight, normal, overweight, obese}. Those with a high BMI experience greater soft tissue resistance, and the rate should not be too fast. Disease type is a categorical variable, and disease stage is a qualitative variable; neither can be directly used as numerical input for fuzzy reasoning. This application uses expert experience combined with historical clinical data statistics to directly map disease type to a fuzzy subset {low risk, low-to-medium risk, medium risk, medium-to-high risk, high risk}. The quantitative mapping of disease stage in this application is based on disease severity scores, using expert scoring combined with historical clinical data statistics to establish a mapping table from disease stage to quantitative indicators. According to the quantitative indicators, the domain of discourse for disease stage in this application is [0, 10], and the fuzzy subset is {mild, mild-to-moderate, moderate, moderate-to-high, high}. The domain of discourse for the initial electromyography baseline value is [0, 10]. ], in μV Based on clinical data, the fuzzy subset is set as {low, medium, high}. The initial electromyography baseline value reflects the muscle tension level of the subject in a resting state. A high baseline indicates muscle tension, requiring a reduction in the loading rate. The output variable of the second fuzzy logic rule is the initial adaptive percentage, with a domain of [3%-8%] and a fuzzy subset of {negative large, negative small, zero, positive small, positive large}.
[0037] The second fuzzy logic rule uses triangular functions as membership functions for both input and output variables, and the parameters of the membership functions are calibrated using historical clinical treatment data.
[0038] The second fuzzy logic rule of this application includes multiple IF-THEN rules for the comprehensive evaluation of the initial adaptive percentage. An example of the second fuzzy logic rule is as follows: Rule R2-1: IF (Age: Older) AND (BMI: Obese) AND (Electronic tension index: High) AND (Disease: Herniated disc), THEN (First adaptive percentage: Negative).
[0039] Rule R2-2: IF (Age: Young) AND (BMI: Normal) AND (Electronic tension index: Low) AND (Disease: Herniated disc), THEN (First adaptive percentage: Positive).
[0040] Rule R2-3: IF (Disease: Spinal stenosis), regardless of other conditions, THEN (First Adaptive Percentage: Negative).
[0041] Rule R2-4: IF (Disease: Degenerative Change) AND (Age: Older), THEN (First Adaptive Percentage: Negative).
[0042] The precise first adaptive percentage threshold is obtained by defuzzifying the output of the second fuzzy logic rule using a triangular membership function.
[0043] Based on the initial adaptive percentage, during subsequent physiotherapy sessions, the self-learning fuzzy logic controller 3 dynamically adjusts the adaptive percentage according to the traction tolerance level obtained by the subject at the end of the previous physiotherapy cycle, adapting to the actual response of the subject during the physiotherapy process. The traction tolerance level is obtained through the tolerance assessment module 34 built into the self-learning fuzzy logic controller 3. The tolerance assessment module 34 acquires electromyographic (EMG) signal data of the subject during the previous physiotherapy cycle and uses a pre-trained signal recognition network to extract body tolerance characteristics from the EMG signal data. Based on the body tolerance characteristics, the pain level signal actively fed back by the subject, and the traction convergence value during the maintenance phase, the tolerance traction level is generated based on the third fuzzy logic rule in the fuzzy logic rule base 31. The body tolerance characteristics include at least: initial tolerance threshold, peak EMG signal, spasm duration, and EMG signal fluctuation intensity index. The initial tolerance threshold represents the EMG characteristic value corresponding to the first obvious muscle response in the subject at the initial stage of traction increase. The peak EMG signal represents the maximum intensity of muscle activity throughout the entire physiotherapy cycle. The duration of spasm characterizes the total time for the amplitude of the electromyographic signal to exceed a preset safety threshold. The intensity of the electromyographic signal fluctuation characterizes the muscle stability, relaxation level, and stress level of the autonomic nervous system of the subject to be traction, and is defined as a comprehensive measure of the rate of change of the electromyographic signal amplitude and the irregularity of the waveform within a unit time window.
[0044] In a preferred embodiment of this application, the signal recognition network uses a convolutional neural network and a long short-term memory network as its basic architecture. The convolutional neural network is used to extract local waveform features of the original electromyographic signal, and the long short-term memory network is used to capture long-term dependencies.
[0045] Furthermore, the input variables of the third fuzzy logic rule in this application are the body's tolerance characteristic vector, the pain level signal, and the traction force convergence value of the previous physiotherapy cycle. The traction force convergence value is the traction force value that remained stable during the maintenance phase of the previous physiotherapy cycle. The output variable of the third fuzzy logic rule is the tolerance traction force level.
[0046] Furthermore, a triangular membership function is used to fuzzify the input and output variables of the third fuzzy logic rule. For example, the initial tolerance threshold has a universe of discourse of [0, ...]. The unit is Newton, and the fuzzy subset includes {low, medium, high}; the peak value of the electromyographic signal has a universe of discourse of [0, ..., 0; ..., ..., ... The unit is microvolts, and the fuzzy subset includes {low, medium, high}; the duration of spasm, with its universe of discourse [0, ..., ...]. The unit is seconds, and the fuzzy subset includes {short, medium, long}; the pain level signal has a universe of discourse of [0, 10], is dimensionless, and the fuzzy subset includes {low, medium, high}; the traction force convergence value has a universe of discourse of [0, 10]. The unit is Newton, and the fuzzy subset includes {low, medium, high}; the universe of discourse for the tolerance traction level is [0, 100], which is dimensionless. The higher the value, the better the tolerance. The fuzzy subset includes {low, medium-low, medium, medium-high, high}.
[0047] The third fuzzy logic rule in this application also includes multiple IF-THEN rules for the comprehensive assessment of traction tolerance levels. An example of the third fuzzy logic rule is as follows: Rule R3-1: IF (Initial tolerance threshold: high) AND (EMG peak value: low) AND (Spasm duration: short) AND (Pain level signal: high) AND (Traction convergence value: high), THEN (Tolerance traction level: high).
[0048] Explanation: Well tolerated, can be treated aggressively.
[0049] Rule R3-2: IF (Initial tolerance threshold: low) AND (EMG peak value: high) AND (Spasm duration: long) AND (Pain level signal: low) AND (Traction convergence value: low), THEN (Tolerance traction level: low).
[0050] Explanation: Severe intolerance requires urgent intervention.
[0051] Rule R3-3: IF (duration of spasm: moderate) AND (pain level signal: moderate) AND (other conditions are arbitrary), THEN (tolerance level of traction: moderate).
[0052] Explanation: The situation is ambiguous, so a conservative strategy should be adopted.
[0053] Rule R3-4: IF (Fluctuation Intensity Index: Low) AND (EMG Signal Peak: Medium) AND (Pain Grade Signal: Medium), THEN (Tolerance Traction Level: Medium-High).
[0054] Explanation: Although there is some muscle activation, the signal is stable, indicating that the subject is in a stable tolerance state.
[0055] Rule R3-5: IF (Fluctuation Intensity Index: High) AND (Spasm Duration: Long), THEN (Tolerance to Traction Level: Low), regardless of other inputs.
[0056] Explanation: High volatility combined with persistent spasms indicates that the subject is clearly intolerant to traction and the traction force needs to be reduced immediately.
[0057] In practical applications, after the current physiotherapy cycle ends, raw electromyographic (EMG) signals, pain level signals, and traction force convergence values are collected. The raw EMG signals are input into a pre-trained signal recognition network, which automatically outputs the initial tolerance threshold, peak EMG signal value, and spasm duration. Based on the initial tolerance threshold, peak EMG signal value, spasm duration, pain level signal, and traction force convergence value, fuzzy reasoning and defuzzification are performed according to the third fuzzy logic rule to obtain the precise tolerance traction force level. The tolerance traction force level is sent to the self-learning fuzzy logic controller 3, which also includes a personalized parameter adjustment module 33, used to dynamically adjust the adaptive percentage in subsequent physiotherapy cycles based on the tolerance traction force level of the previous physiotherapy cycle and the initial adaptive percentage.
[0058] In a preferred embodiment of this application, the adaptive percentage for the next treatment cycle is dynamically updated based on the adaptive percentage for the current treatment cycle and the level of tolerable traction force, using a first exponential smoothing model with limited amplitude, as shown below: in, This indicates the adaptive percentage for the next physiotherapy cycle. This indicates the adaptive percentage of the current treatment cycle. This represents the initial learning rate, controlling the speed at which the adaptive percentage is adjusted. The target adaptive percentage, representing the traction tolerance level transformed via the S-shaped mapping function, is expressed as follows: in, , These represent the minimum and maximum allowed values for the adaptive percentage, which are preset values. This indicates the level of traction tolerance assessed at the end of the current physical therapy cycle. This represents the midpoint of the mapping curve. This represents the kurtosis coefficient of the first curve.
[0059] The S-shaped mapping function makes the adaptive percentage change most sensitive when the tolerance traction level is at a median level, while the change tends to be gradual when the tolerance level is extremely high or low, which is consistent with the law of diminishing marginal effects in clinical treatment.
[0060] In a preferred embodiment of this application, the first learning rate can be dynamically adjusted according to the treatment stage. A smaller learning rate is used in the early stage of treatment to ensure a smooth transition in the treatment process. In the middle and late stages of treatment, when the system has fully learned the individual characteristics of the object to be traction, the learning rate can be appropriately increased to accelerate the convergence speed of the individualized parameters.
[0061] By repeatedly performing tolerance assessment and adaptive percentage adjustment after each treatment cycle, the self-learning fuzzy logic controller 3 forms a closed loop of monitoring, assessment, and adjustment, realizing personalized dynamic iteration of the traction plan.
[0062] In a preferred embodiment of this application, the adaptive rate of linear increase in traction force during the creep phase is also dynamically adjusted based on the traction tolerance level of the subject in the previous physiotherapy cycle. During the first physiotherapy session, the initial adaptive rate is generated using the fourth fuzzy logic rule based on the physiological characteristic parameters of the subject. These physiological characteristic parameters include age, sex, body mass index, disease type, disease stage, and initial electromyography baseline value. The universe of discourse and fuzzy subset follow the settings of the second fuzzy logic rule. The universe of discourse for the initial adaptive rate is [0, ...]. The unit is N / s. Based on equipment safety limits and clinical guidelines, for example, 0.5 N / s, the fuzzy subset is {very slow, slow, medium speed, fast}.
[0063] The fourth fuzzy logic rule in this application also includes multiple IF-THEN rules for the comprehensive evaluation of the initial adaptive rate. An example of the fourth fuzzy logic rule is as follows: Rule 4-1: IF (Age: Older) AND (BMI: Obese) AND (Electronic tension index: High), THEN (Initial adaptive rate: Slow).
[0064] Explanation: Elderly patients experience decreased tissue elasticity and weakened repair capabilities; obese patients experience high spinal load and poor soft tissue compliance; high electromyographic tension index indicates that muscles are in a state of sustained high tension, which can easily induce spasms. When all three conditions are met simultaneously, the risk of discomfort or tissue damage during traction treatment is significantly increased. Therefore, the system determines that the initial adaptive rate should be slow, i.e., the traction force should be increased gradually at a low rate of change to ensure the safety of the treatment process and avoid adverse reactions caused by excessively rapid increases.
[0065] Rule 4-2: IF (Age: Young) AND (BMI: Normal) AND (Electromyography: Low) and (Disease: Herniated Disc), THEN (Initial Adaptive Rate: Fast).
[0066] Explanation: Young patients undergoing traction have good tissue elasticity and strong repair capabilities; a normal BMI indicates moderate spinal load; a low electromyographic tension index reflects a relaxed muscle state; and herniated discs, a typical indication for traction therapy, usually respond positively to traction force and have good tolerance. All these conditions indicate that the patient has a good foundation for tolerance. Based on this, the system determines that the initial adaptive rate should be rapid, meaning that the traction force should be increased at a high rate of change to achieve effective treatment requirements as quickly as possible while ensuring safety, thereby improving treatment efficiency.
[0067] Rule 4-3: IF (Disease: Spinal stenosis), THEN (Initial adaptive rate: Very slow), regardless of other conditions.
[0068] Explanation: In patients with spinal stenosis requiring traction, the effective volume of the spinal canal is reduced, and the spinal cord or nerve roots are under relative compression. During traction therapy, if the rate of change in traction force is too rapid, it may cause a sudden change in pressure within the spinal canal, inducing nerve root irritation symptoms or even aggravating nerve damage. Therefore, regardless of the patient's age, BMI, electromyographic tension index, or other parameters, as long as the disease is diagnosed as spinal stenosis, the initial adaptive rate is forcibly set to an extremely slow level, with the traction force increased at the most conservative rate to maximize the safety of the neural structures. This rule has the highest priority, reflecting the system's safety fallback design for high-risk diseases.
[0069] Rule R4-4: IF (Disease: Degenerative) AND (Age: Older), THEN (First Adaptation Rate: Very Slow).
[0070] Explanation: Degenerative changes mean that the spinal structure has undergone varying degrees of wear, hyperplasia, or decreased elasticity; aging further amplifies the fragility of these tissues. When these two conditions combine, the biomechanical stability of the spine decreases, making it more sensitive to traction forces. Rapidly increasing traction force may lead to the accumulation of micro-damage or induce pain. Therefore, the system determines that the initial adaptation rate should be extremely slow, applying force in a gradually increasing manner to provide sufficient adaptation time for the degenerated tissues, balancing efficacy and safety.
[0071] Based on the third fuzzy logic rule, a triangular membership function is used to obtain the accurate first adaptive rate after defuzzification.
[0072] In a preferred embodiment of this application, based on the tolerable traction force level and the initial adaptive rate, a limited second exponential smoothing model is used to dynamically adjust the adaptive rate in subsequent physiotherapy cycles. The second exponential smoothing model is expressed as follows: in, This indicates the adaptive rate for the next physiotherapy cycle. This indicates the adaptive rate of the current physiotherapy cycle. This represents the second learning rate, which controls the adjustment speed of the adaptive rate. The target adaptive rate, obtained by transforming the traction tolerance level through an S-shaped mapping function, is expressed as follows: in, , These represent the minimum and maximum allowable values for the adaptive rate, which are preset values. This represents the steepness coefficient of the second curve.
[0073] The S-shaped mapping function makes the adaptive rate change most sensitive when the tolerance traction level is at a median, while the change tends to be gradual when the tolerance level is extremely high or low, which is consistent with the law of diminishing marginal effects in clinical treatment.
[0074] In a preferred embodiment of this application, the second learning rate is dynamically adjusted according to the treatment stage. A smaller learning rate is used in the early stage of treatment to ensure a smooth transition in the treatment process. In the middle and late stages of treatment, when the system has fully learned the individual characteristics of the object to be traction, the learning rate can be appropriately increased to accelerate the convergence speed of the individualized parameters.
[0075] In this application, during the initial physical therapy, the threshold for the first adaptive displacement change rate during the creep phase is generated using the fifth fuzzy logic rule based on the physiological characteristic parameters of the object to be tractioned. The universe of discourse for the first adaptive displacement change rate threshold is [0, ...]. The unit is mm / s. Based on clinical guidelines, for example, 3 mm / s, the fuzzy subset is {very low, low, medium, high, very high}. The fifth fuzzy logic rule also consists of several IF-THEN fuzzy rules, each corresponding to an adaptive displacement change rate threshold adjustment strategy under a combination of physiological characteristics. An example of the fifth fuzzy logic rule is as follows: Rule 5-1: IF (Age: Older) AND (Disease Type Stage: Advanced / Severe) AND (Initial Electromyography Baseline Value: High), THEN (Initial Adaptive Displacement Rate Threshold: Extremely High).
[0076] Explanation: Because the resistance experienced by the traction bed increases from small to large during the entire traction therapy process, the displacement change rate between the movable bed board and the fixed bed body of the traction bed decreases from large to small. Therefore, for elderly patients, patients in the late stages of disease, and patients with high muscle tension, an extremely high initial adaptive displacement change rate threshold is set to avoid overstretching that may induce spasms or injury.
[0077] Rule 5-2: IF (Age: Young Adult) AND (Weight: Standard) AND (Disease Type Stage: Early / Mild) AND (Sex: Male), THEN (First Adaptive Displacement Rate Threshold: Low).
[0078] Explanation: For young, standard-weight male subjects in the early stages of disease, their tissue tolerance is better, allowing for a lower initial adaptive displacement rate threshold.
[0079] Rule 5-3: IF (Initial EMG baseline: Low) AND (Disease type stage: Intermediate / Moderate), THEN (First adaptive displacement rate threshold: Moderate).
[0080] Explanation: When the initial muscle tone of the object to be traction is low, even if the disease is moderate, a moderate initial adaptive rate of displacement threshold can be used.
[0081] Rule 5-4: IF (weight: overweight) AND (age: elderly), THEN (first adaptive displacement rate threshold: high).
[0082] Explanation: For elderly or overweight individuals requiring traction, a higher threshold for the initial adaptive displacement change rate is set.
[0083] In a preferred embodiment of this application, based on the tolerable traction force level and the initial adaptive displacement change rate threshold, a limited third exponential smoothing model is used to dynamically adjust the adaptive displacement change rate threshold in subsequent physiotherapy cycles. The third exponential smoothing model is expressed as follows: in, This represents the adaptive displacement change rate threshold for the next physiotherapy cycle. This represents the adaptive displacement change rate threshold for the current physiotherapy cycle. This represents the third learning rate, which controls the adjustment speed of the adaptive displacement change rate threshold. The target adaptive displacement change rate threshold, obtained by transforming the traction resistance level using an S-shaped mapping function, is expressed as follows: in, , These represent the minimum and maximum allowable values for the adaptive rate, which are preset values. This represents the steepness coefficient of the third curve.
[0084] In a preferred embodiment of this application, for the first physiotherapy cycle, the second, fourth, and fifth fuzzy logic rules are used to automatically fuzzily infer the threshold values for the initial adaptive percentage, initial adaptive rate, and initial adaptive displacement change rate based on physiological characteristics such as the age, gender, BMI, disease type, disease stage, and initial electromyography baseline value of the subject. This allows the initial traction force, loading speed, and unlocking conditions to automatically match the physical condition and pathological state of the subject, achieving an initial balance between safe initiation and therapeutic effect. By introducing a first, second, and third exponential smoothing model with amplitude limits, the tolerance traction force level of the previous physiotherapy cycle is used as feedback to dynamically correct the adaptive percentage, adaptive rate, and adaptive displacement change rate threshold values for subsequent physiotherapy cycles. Doctors do not need to reset complex mechanical parameters for each treatment; the system automatically optimizes parameters based on historical data and physiological feedback. As the number of physiotherapy sessions increases, the tolerance of the subject usually improves. The exponential smoothing model allows the traction parameters to increase smoothly with the improvement of the subject's tolerance level, avoiding excessive parameter jumps that could cause damage. By using a limit setting, even if sensor noise or extreme feedback causes abnormalities in tolerance assessment, parameter adjustments are limited to preset safe physiological boundaries, preventing system malfunction and ensuring mechanical safety during long-term treatment.
[0085] In a preferred embodiment of this application, during the maintenance phase of the step-by-step traction strategy, the adaptive fuzzy adjustment engine 32 employs a fuzzy parameter adjustment model based on neural networks and fuzzy inference. It dynamically generates traction force control commands by updating the shape parameters and rule weights of the membership functions of the first fuzzy logic rules in the fuzzy logic rule library 31. Specifically, the fuzzy logic rule library 31 stores first fuzzy logic rules for dynamically adjusting the traction force during the maintenance phase to counteract muscle spasms. The first fuzzy logic rule uses the electromyography (EMG) signal deviation (the difference between the current EMG signal and the resting baseline), the rate of change of EMG signal amplitude, and the pain level signal of the object to be tractioned as three-dimensional inputs, and the traction force correction amount as the output. The domain of the EMG signal deviation is [0, ...]. ], in μV Based on clinical electromyography (EMG) data, for example, 1000 μV, the fuzzy subset is {small, medium-small, medium, medium-large, large}. The universe of discourse for the rate of change of EMG signal amplitude is [0, ..., ... ], in μV / s Based on the temporal characteristics of electromyographic signals and clinical data, for example, 200 μV / s, the fuzzy subset is {negative large, negative small, zero, positive small, positive large}. The domain of discourse for the pain level signal of the subject to be traction is [0, 10], dimensionless, with a fuzzy subset of {very small, small, moderate, painful, extremely painful}. The domain of discourse for the traction force correction is set according to safety requirements, with a correction range of [...]. , ],recommend The amount of a single correction shall not exceed this value, and the fuzzy subset is {negative large, negative medium, negative small, zero, positive small, positive medium, positive large}.
[0086] The first fuzzy logic rule specifically includes, but is not limited to, the following fuzzy rules: Rule 1-1 (Spasmodic Burst Inhibition): IF (EMG signal deviation: large, indicating that the EMG signal is far above the baseline) AND (EMG signal amplitude change rate: large positive, indicating that the EMG signal is rising sharply) AND (Pain level signal: pain, indicating the presence of muscle tension), THEN (Traction force correction: large negative, indicating a significant reduction in traction force to quickly avoid an impending intense spasm).
[0087] Rule 1-2 (Intervention for Spastic Precursors): (EMG signal deviation: small to medium, indicating a slight increase in EMG signal) AND (EMG signal amplitude change rate: small positive, indicating an upward trend) AND (Pain level signal: moderate, indicating moderate muscle tension), THEN (Traction force correction: small negative, indicating a slight reduction in traction force for preventative intervention).
[0088] Rule 1-3 (Maintaining a Stable State): IF (EMG signal deviation: Medium, indicating stable EMG signal) AND (EMG signal amplitude change rate: Zero, indicating the change rate approaches zero) AND (Pain level signal: Moderate, indicating moderate muscle tension) THEN (Traction force correction: Zero, indicating maintaining the current traction force).
[0089] Rule 1-4 (Relaxation State Trial): IF (EMG signal deviation: small, indicating EMG signal is below baseline) AND (EMG signal amplitude change rate: small negative, indicating the signal tends to be quieter) AND (Pain level signal: very small, indicating muscle relaxation) THEN (Traction force correction: small positive, indicating that a small increase in traction force can be tried to explore a better treatment intensity).
[0090] Those skilled in the art will understand that the above rules are merely examples, and in practical applications, the number of rules can be increased or the language value division of the rules can be adjusted according to more refined control requirements.
[0091] Each rule employs a first-order Sugeno-type fuzzy inference structure. To facilitate subsequent adaptive adjustment, each rule is associated with an adjustable rule weight. The first fuzzy logic rule is stored in the form of a structured data table, where each record contains a rule number, preconditions, conclusion, rule weight, and activation flag. Default rules are loaded during system initialization, and the rule weights and shape parameters of the membership function of the first fuzzy logic rule are dynamically updated during the treatment process using the adaptive fuzzy adjustment engine 32. Therefore, the adaptive fuzzy adjustment engine 32 of this application includes a fuzzy inference module and a parameter learning module. The fuzzy inference module is responsible for performing fuzzification, rule inference, aggregation, and defuzzification operations, mapping the current input to a precise control output. The parameter learning module is responsible for adjusting the adjustable parameters in the rule base online according to the control effect evaluation index to achieve adaptability.
[0092] The fuzzy parameter adjustment model in this application is a six-layer feedforward network, consisting of an input layer, a fuzzification layer, a rule layer, a normalization layer, a conclusion layer, and an output layer. The input layer receives real-time traction force values, electromyographic (EMG) signal data, and pain level signals, and extracts the EMG signal deviation and the rate of change of EMG signal amplitude from the EMG signal data.
[0093] The fuzzification layer assigns a set of differentiable membership functions to each input variable. In this embodiment, a Gaussian membership function is used, as shown below: in, This indicates the index of the input variable in the fuzzy parameter adjustment model. Indicates a fuzzy subset index. Indicates the first The exact value of each input variable. Indicates the first The first input variable corresponding to the first A fuzzy subset Indicates the first The precise values of each input variable Belongs to the a fuzzy subset The degree of [0, 1] is determined by the value [0, 1]. Indicates the first The input variable of the first The center value of a fuzzy subset determines the position of the membership function in the universe of discourse. Indicates the first The input variable of the first The width of each fuzzy subset determines the steepness of the membership function; the center value and width are stored as trainable parameters in the rule base. This represents an exponential function.
[0094] Each node in the rule layer corresponds to one rule in the first fuzzy logic rules, and the trigger strength of the rule is calculated using the multiplicative T-norm. The trigger strength of each rule is calculated from the T-norm of the membership degree of the input variable, and the calculation formula is as follows: in, Representation rules Trigger strength, Indicates the rule index. This represents the total number of input variables for the fuzzy parameter adjustment model. Each rule corresponds to a set of fuzzy subset combinations. Indicates the first The input variable in the th... The corresponding fuzzy subset in the rule, Indicates the first Input variables for fuzzy sets The degree of membership.
[0095] In this application, the effective trigger strength of each rule in the rule weighting is modified as follows: in, Representation rules Corrected effective trigger strength Representation rules The rule weights are stored as trainable parameters in the rule base.
[0096] The normalization layer calculates the normalized value of the effective trigger strength of each rule. The rule weights are directly applied to the normalized value, and the calculation formula is as follows: in, This represents the sum of the effective trigger strengths of all rules, used as a normalization factor.
[0097] In a preferred embodiment of this application, rule weights directly affect the contribution percentage of each rule in the final output. A larger rule weight gives the rule greater influence in the input space; even if its premise matching degree is low, it may still occupy a larger proportion after normalization. Conversely, a smaller rule weight suppresses the rule's influence. For rules built based on insufficient data or experience, their negative impact can be reduced by lowering their weights, suppressing low-quality rules. For validated, highly reliable rules, higher weights can be assigned to ensure their dominant position in decision-making, achieving adaptive optimization of the rule base. When the input variable is in the overlapping region of multiple rules, the weights can adjust the smoothness of the contribution of each rule.
[0098] Each rule in the conclusion layer corresponds to a conclusion, and the conclusion of a rule is a linear function of the input variables: in, Indicates the first The output conclusion of each rule No. Rule number 1 The conclusion parameters corresponding to each input variable are trainable parameters.
[0099] The output layer performs a weighted sum of the conclusions from each rule to obtain the final traction adjustment amount: in, This indicates the amount of traction adjustment.
[0100] The parameter learning module dynamically updates the center value, width, conclusion parameters, and rule weights of the fuzzy rules. Based on the treatment goals, the parameter learning module constructs a loss function that includes traction control accuracy, safety, and pain level indicators, as shown below: in, This represents the target traction force, which is the sum of the maintaining traction force and the traction adjustment amount. Indicates real-time traction force. This indicates the traction control accuracy index. Indicates the duration of spasm. Indicates safety indicators, Indicates pain level signal, Indicators representing pain levels , and This represents the weighting coefficient and is also used for dimension adjustment.
[0101] During the maintenance phase, the partial derivatives of the loss function with respect to the shape parameters of the membership function of the fuzzy layer, the conclusion parameters of the conclusion layer, and the rule weights of each rule are calculated. Gradient descent is then used to update the parameter values. The updated membership function shape parameters, conclusion parameters, and rule weights are written back to their corresponding storage locations in the fuzzy logic rule base 31, enabling online evolution of the rule base. As the treatment progresses, the rule base continuously adapts to changes in the individual characteristics and rehabilitation status of the patient.
[0102] The self-learning fuzzy logic controller 3 superimposes the traction force correction amount derived from fuzzy inference with the traction force from the previous moment to generate a real-time target traction force command. This command drives the traction bed drive module 2 to complete minute increases or decreases in traction force within milliseconds, achieving precise countermeasures against muscle spasms. The maintenance phase lasts for a second preset time, completing one full treatment session.
[0103] In a preferred embodiment of this application, a six-layer structure—input layer, fuzzification layer, rule layer, normalization layer, conclusion layer, and output layer—fully parameterizes the fuzzy logic reasoning process into a differentiable neural network. The fuzzification layer employs a Gaussian membership function, the rule layer uses a multiplicative T-norm, and the conclusion layer uses a linear function. All layer operations are continuously differentiable, enabling the entire fuzzy parameter adjustment model to be trained end-to-end via backpropagation, achieving adaptive optimization of the fuzzy rule base. Although this application uses a neural network computational framework, the physical meaning of each layer is clear, making the system's decision-making process have a transparent logical chain. Clinicians can understand the contribution of each rule, facilitating clinical verification and intervention. The input layer not only receives real-time traction force values but also extracts the electromyographic signal deviation and amplitude change rate relative to the resting baseline, transforming the raw electromyographic signals into features with clear physiological meaning. Simultaneously, pain level signals are introduced, achieving an organic fusion of objective physiological signals and subjective sensory information, providing a more comprehensive decision-making basis for traction force adjustment. The rule layer, based on the calculated rule trigger strength, introduces rule weights to correct the trigger strength, resulting in an effective trigger strength. As trainable parameters, the rule weights can dynamically adjust the importance of each rule in the inference process based on clinical feedback. This suppresses low-quality rules and strengthens the contribution of high-reliability rules, significantly enhancing the model's expressive power and robustness. The loss function integrates three dimensions: traction control accuracy, safety, and pain level. The traction control accuracy reflects the deviation between the target traction force value and the real-time traction force value; the safety reflects the duration of spasm; and the pain level reflects the pain intensity signal. By adjusting the weight coefficients of each indicator, the system can achieve a dynamic balance between control accuracy, safety, and comfort, meeting multiple clinical treatment objectives.
[0104] This application uses rule weights as learnable parameters, allowing for a certain degree of redundancy or low-quality rules in the rule base. The rule weights suppress their negative impact, enhancing the system's robustness. Different subjects respond differently to the same rule, and the rule weights can converge to their individual optimal values during treatment, achieving personalized needs. The weight values directly reflect the actual contribution of each rule to the current subject's treatment, facilitating clinical understanding and intervention, and enhancing the interpretability of the output results.
[0105] In a preferred embodiment of this application, the spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback also includes a human-computer interaction interface for receiving real-time pain level signals from the object to be tractioned. The self-learning fuzzy logic controller 3 uses the pain level signal as a constraint condition. When the pain level reported by the object to be tractioned exceeds a preset threshold, the traction force is forcibly reduced or traction is suspended.
[0106] In a specific embodiment of this application, the spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback was applied to a 45-year-old male patient with lumbar disc herniation, weighing 75 kg, with a BMI of 24.5 kg / m², diagnosed with L4-L5 segment lumbar disc herniation, grade III. The initial average electromyographic value of the surface electromyographic signal of the erector spinae muscles was 6.5±1.2 μV, the initial integrated electromyographic value was 3.8±0.5 μmV·s, and the initial median frequency was 75±6.3 Hz. During the treatment, based on the second, fourth, and fifth fuzzy logic rules, the following inferences were derived: the initial adaptive percentage was 6%, the initial adaptive rate was 2.5 N / s, and the initial adaptive displacement change rate threshold was 0.15 mm / min.
[0107] During the preloading phase, the initial target traction force was 75 kg × 6% = 45 N. This lasted for 2 minutes, during which electromyographic signals and traction force were monitored in real time.
[0108] During the creep stage, the traction force is increased linearly at a rate of 2.5 N / s, while the displacement of the movable bed board relative to the fixed bed is monitored to obtain the displacement change rate. When the traction force increases to approximately 620 N, the adaptive displacement change rate reaches the threshold of 0.15 mm / min, at which point the linear increase is stopped, and the current traction force is used as the maintenance traction force.
[0109] During the maintenance phase, the self-learning fuzzy logic controller dynamically adjusts the traction force based on real-time traction force values, electromyographic (EMG) signal data, and pain level signals. Throughout the treatment, the patient's EMG signal data remained around 800 μV, indicating that muscle spasms subsided and the traction force converged to 635 N.
[0110] In the second physiotherapy cycle, a signal recognition network was used to extract the initial tolerance threshold of 50 μV, the peak value of the electromyography (EMG) signal of 1000 μV, and the duration of spasm of 1.8 s from the EMG signal data of the first physiotherapy cycle. After the first physiotherapy, the subject reported a pain level of 6. Based on the initial tolerance threshold, the peak value of the EMG signal, the duration of spasm, and the convergence value of the traction force signal according to the pain level, the fuzzy result of the tolerance traction force level was determined to be "medium". .
[0111] Furthermore, the threshold values for the initial adaptive percentage, initial adaptive rate, and initial adaptive displacement change rate are dynamically adjusted using the tolerance traction level. The learning rate of the first exponential smoothing model with limiting is set to 0.35, the kurtosis coefficient of the first curve is set to 0.15, and the universe of discourse for the tolerance traction level is [0, 100]. Therefore, the midpoint of the mapping curve is set to 50, the minimum allowable value for the adaptive percentage is set to 3%, and the maximum allowable value is set to 15%. The calculated adaptive percentage for the second physiotherapy cycle is 7.4%. The learning rate of the second exponential smoothing model with limiting is set to 0.35, the kurtosis coefficient of the second curve is set to 0.15, the minimum allowable value for the adaptive rate is set to 1.0 N / s, and the maximum allowable value for the adaptive rate is set to 5.0 N / s. The calculated adaptive rate for the second physiotherapy cycle is 2.68 N / s. The learning rate of the limited-amplitude third exponential smoothing model was set to 0.35, the kurtosis coefficient of the third curve was set to 0.15, the minimum allowable value of the adaptive displacement change rate threshold was set to 0.05 mm / min, and the maximum allowable value of the adaptive displacement change rate threshold was set to 0.3 mm / min. The calculated adaptive displacement change rate threshold for the second physiotherapy cycle was 0.159 mm / min. Since the tolerance assessment after the first treatment was moderate, the spinal stepwise quantitative traction physiotherapy control system based on closed-loop force feedback in this application judged that its current tolerance level was significantly higher than the conservative parameters used in the first treatment. Therefore, in the second physiotherapy cycle, the adaptive percentage threshold was increased from 6% to 7.4%, the adaptive rate threshold was increased from 2.5 N / s to 2.68 N / s, and the displacement change rate threshold was slightly relaxed.
[0112] The treatment results after 12 treatments are shown in Table 1. Table 1 In Table 1, the VAS pain score, or Visual Analogue Scale, is one of the most commonly used and sensitive pain assessment tools in clinical practice. It quantifies pain intensity by converting the patient's subjective pain perception into a point on a line. The Oswestry Disability Index (ODI) is the gold standard tool for assessing the impact of low back pain on a patient's daily living functions; a higher score indicates more severe functional impairment. Surface electromyography (EMG) signals of the erector spinae muscles are key objective indicators for assessing the effectiveness of spinal traction therapy and muscle function. In spinal traction research and clinical practice, the most commonly used indicators for analyzing EMG signals of the erector spinae muscles are mean EMG value, integrated EMG value, and median frequency. An effective treatment typically reduces mean EMG value and integrated EMG value, meaning that the total amount of muscle activity decreases over the same or longer period, reducing fatigue; while an increase in median frequency enhances muscle fatigue resistance and reduces fatigue. The JOA score, short for Japanese Orthopaedic Association scores, is one of the core indicators for clinically assessing the functional status of spinal diseases. The JOA score is mainly assessed from the perspective of neurological function, and the assessment content is very specific. Doctors score patients based on their performance. The lower the JOA score, the more severe the neurological dysfunction.
[0113] As shown in Table 1, the lumbar disc herniation of the patients under traction was effectively treated, thus proving the effectiveness of the spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback in this application.
[0114] In a specific embodiment of this application, a comparative experiment was designed to compare a traditional open-loop control traction device, a closed-loop traction device with traction force and electromyographic signal feedback, and the spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback of this application. An experimental group, a first control group, and a second control group were set up, with 84 participants in each group. The age distribution, gender distribution, body mass index distribution, disease type distribution, and disease stage distribution were identical in the experimental group, the first control group, and the second control group. The spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback of this application was applied to the experimental group. The first control group used a traditional open-loop control traction device. The control mode was open-loop control, without traction force feedback, electromyographic signal feedback, or feedback on the rate of displacement change between the movable bed board and the fixed bed body of the traction bed. The control parameters were manually set, and no distributed traction strategy was adopted. The second control group used a closed-loop traction device with traction force and electromyographic signal feedback. The control mode was closed-loop control, with traction force feedback and electromyographic signal feedback, but no feedback on the rate of displacement change between the movable bed board and the fixed bed body of the traction bed. The control parameters used PID logic control and a step-by-step traction strategy, but the traction force application used fixed parameters. Each group received traction treatment for 4 weeks, 3 times a week, for a total of 12 sessions.
[0115] The clinical effects of traction therapy in the first control group, the second control group, and the experimental group were evaluated, and the evaluation results are shown in Table 2.
[0116] Table 2 As shown in Table 2, the spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback proposed in this application significantly outperforms traditional open-loop control traction devices and closed-loop traction devices that rely on traction force and electromyographic signal feedback in all core indicators. Particularly noteworthy are the VAS improvement rate (71.3%), which is 86.6% higher than that of traditional open-loop control traction devices (38.2%); the ODI improvement rate (64.8%), which is 99.4% higher than that of traditional open-loop control traction devices (32.5%); and the 3-month recurrence rate (7.5%), which is 72.7% lower than that of traditional open-loop control traction devices (27.5%). The data in Table 2 fully validate the clinical superiority of the proposed technical solution, which combines a step-by-step traction strategy based on closed-loop force feedback, electromyographic signal data feedback, displacement change feedback, and parameter adaptation with self-learning fuzzy control.
[0117] The comparison results of traction safety indicators of the first control group, the second control group and the experimental group are shown in Table 3.
[0118] Table 3 As shown in Table 3, the spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback of this application, through multimodal signal fusion and self-learning fuzzy control, pre-adjusts before the traction force approaches the threshold, achieving pre-emptive prevention. Compared with the second control group, the number of traction force over-limit alarms decreased by 88%, treatment interruption decreased by 73%, and muscle spasm aggravation decreased by 73%, completely avoiding iatrogenic primary tissue damage. This indicates that the spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback of this application has high control precision, strong pre-adjustment capability, and intervenes in the early stage of spasm formation, thereby significantly reducing the risk of spasm aggravation and controlling the instantaneous over-limit peak within a safe range, achieving a clinical result of zero damage.
[0119] This embodiment provides a step-by-step quantitative traction physiotherapy control system for the spine based on closed-loop force feedback, which is used to solve the technical problems of inaccurate traction force control, inability to adapt to individual differences of the subject to traction, and inability to perceive the physiological state of the subject to traction in real time during spinal traction therapy. The system includes: a data acquisition module for acquiring real-time traction force values output by the traction bed drive module, electromyographic signal data of the object to be tractioned, and the displacement change rate between the movable bed board and the fixed bed body of the traction bed; a traction bed drive module for receiving traction force control commands output by a self-learning fuzzy logic controller and performing traction actions on the traction bed body according to the traction force control commands; a self-learning fuzzy logic controller electrically connected to the data acquisition module and the traction bed drive module, for generating traction force control commands based on real-time traction force values, electromyographic signal data, displacement change rate, and pain level signals fed back by the object to be tractioned, using a preset step-by-step traction strategy; the self-learning fuzzy logic controller includes: a fuzzy logic rule base and an adaptive fuzzy adjustment engine; the fuzzy logic rule base stores first fuzzy logic rules constructed based on expert experience and historical clinical data; the adaptive fuzzy adjustment engine is configured to dynamically generate traction force control commands by updating the shape parameters of the membership function of the first fuzzy logic rules. The spinal step-by-step quantitative traction physiotherapy control system provided in this application adopts a three-stage step-by-step traction strategy, which not only conforms to the viscoelastic mechanical properties of spinal tissue under traction, but also makes the control target more specific, improving the precision of treatment. In the creep stage, the rate of displacement change is used as the criterion for stopping linear increases, rather than simply relying on time or traction force thresholds. The rate of displacement change directly reflects the real-time response of the spinal structure under traction. When the rate of displacement change reaches a preset adaptive threshold, the increase stops, effectively avoiding soft tissue damage, muscle spasms, or nerve root stimulation caused by excessive traction speed or amplitude. Simultaneously, in the maintenance stage, the traction force is dynamically adjusted through a self-learning fuzzy logic controller, which can respond in real-time to changes in the electromyographic signals and subjective pain level feedback of the subject being tractioned. The traction force is promptly reduced when pre-spasm symptoms appear, significantly improving the safety of the treatment process. The fuzzy parameter adjustment model is set as a six-layer structure consisting of an input layer, a fuzzification layer, a rule layer, a normalization layer, a conclusion layer, and an output layer, fully parameterizing the fuzzy logic reasoning process into a differentiable neural network. This enables the entire fuzzy parameter adjustment model to be trained end-to-end via backpropagation, achieving adaptive optimization of the fuzzy rule base. The physical meaning of each layer of the neural network is clear, giving the system's decision-making process a transparent logical chain. Clinicians can understand the contribution of each rule, facilitating clinical validation and intervention.
[0120] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.
[0121] The above-described embodiments are merely preferred embodiments of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention. Therefore, the scope of protection of this invention should be determined by the scope of the claims.
Claims
1. A step-by-step quantitative traction physiotherapy control system for the spine based on closed-loop force feedback, characterized in that, include: The data acquisition module is used to acquire the real-time traction force value output by the traction bed drive module, the electromyographic signal data of the object to be tractioned, and the displacement change rate between the movable bed board and the fixed bed body of the traction bed; The traction bed drive module is used to receive traction force control commands output by the self-learning fuzzy logic controller, and to perform traction actions on the traction bed body according to the traction force control commands. The self-learning fuzzy logic controller is electrically connected to the data acquisition module and the traction bed drive module, respectively, and is used to generate the traction force control command based on the real-time traction force value, the electromyographic signal data, the displacement change rate and the pain level signal fed back by the object to be traction, using a preset step-by-step traction strategy. The self-learning fuzzy logic controller includes: a fuzzy logic rule base and an adaptive fuzzy adjustment engine; The fuzzy logic rule base stores the first fuzzy logic rules constructed based on expert experience and historical clinical data; the adaptive fuzzy adjustment engine is configured to use a fuzzy parameter adjustment model constructed based on neural networks and fuzzy inference to dynamically generate the traction control command by updating the shape parameters of the membership function of the first fuzzy logic rules.
2. The spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback according to claim 1, characterized in that, The step-by-step traction strategy includes: During the preloading phase, an adaptive percentage of the weight of the object to be traction is used as the initial target traction force, which is maintained for a first predetermined time. During the creep stage, the traction force is increased linearly at an adaptive rate. When the displacement change rate reaches the adaptive displacement change rate threshold, the linear increase stops, and the traction force at the current moment is used as the maintenance traction force. During the maintenance phase, the maintenance traction force is dynamically adjusted by the self-learning fuzzy logic controller based on the real-time traction force value, the electromyographic signal data, and the pain level signal, and is maintained for a second preset time. The adaptive percentage, the adaptive rate, and the adaptive displacement change rate threshold are all determined based on the physiological characteristic parameters of the object to be tractioned and the tolerance level of the traction force in the previous treatment cycle.
3. The spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback according to claim 2, characterized in that, The self-learning fuzzy logic controller also includes: The tolerance assessment module is used to acquire the electromyographic signal data of the subject to be traction during the previous physical therapy cycle, and to extract the body tolerance characterization features from the electromyographic signal data using a pre-trained signal recognition network; based on the body tolerance characterization features, the pain level signal, and the traction force convergence value of the maintenance phase, the tolerance traction force level is generated based on the third fuzzy logic rule stored in the fuzzy logic rule base. The body tolerance characterization features include at least: initial tolerance threshold, peak electromyographic signal, and spasm duration.
4. A spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback according to claim 2, characterized in that, The fuzzy logic rule base also stores a second fuzzy logic rule, which is used to perform fuzzy reasoning on the adaptive percentage of the first physiotherapy cycle based on the physiological characteristic parameters of the object to be tractioned, so as to obtain the first adaptive percentage. The physiological characteristics parameters include at least: age, sex, body mass index, disease type, disease stage, and initial baseline electromyography value.
5. A spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback according to claim 4, characterized in that, The fuzzy logic rule base also stores a fourth fuzzy logic rule, which is used to perform fuzzy reasoning on the adaptive rate of the first physiotherapy cycle based on the physiological characteristic parameters to obtain the first adaptive rate.
6. A spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback according to claim 4, characterized in that, The fuzzy logic rule base also stores a fifth fuzzy logic rule, which is used to perform fuzzy reasoning on the adaptive displacement change rate threshold of the first physiotherapy cycle based on the physiological characteristic parameters, so as to obtain the first adaptive displacement change rate threshold.
7. A spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback according to any one of claims 4-6, characterized in that, The self-learning fuzzy logic controller also includes: The personalized parameter adjustment module is used to dynamically adjust the adaptive percentage in subsequent treatment cycles using a first exponential smoothing model with limited amplitude, based on the tolerance traction level and the initial adaptive percentage of the previous treatment cycle, to obtain the adaptive percentage of the subsequent treatment cycle; to dynamically adjust the adaptive rate in subsequent treatment cycles using a second exponential smoothing model with limited amplitude, based on the tolerance traction level and the initial adaptive rate of the previous treatment cycle, to obtain the adaptive rate of the subsequent treatment cycle; and to dynamically adjust the adaptive displacement change rate threshold in subsequent treatment cycles using a third exponential smoothing model with limited amplitude, based on the tolerance traction level and the initial adaptive displacement change rate threshold of the previous treatment cycle, to obtain the adaptive displacement change rate threshold of the subsequent treatment cycle.
8. A spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback according to claim 1, characterized in that, The fuzzy parameter adjustment model includes: an input layer, a fuzzification layer, a rule layer, a normalization layer, a conclusion layer, and an output layer; The input layer is used to receive the real-time traction force value, the electromyographic signal data, and the pain level signal, and to extract the electromyographic signal deviation of the electromyographic signal data relative to the resting baseline and the rate of change of the electromyographic signal amplitude of the electromyographic signal data. The blurring layer is used to blur the real-time traction force value, the electromyographic signal deviation, the real-time electromyographic signal amplitude change rate, and the pain level signal using a Gaussian membership function; The rule layer is used to calculate the trigger strength of each rule in the first fuzzy logic rule using the multiplicative T-norm, and to correct the trigger strength based on the rule weight set for each rule in the first fuzzy logic rule to obtain an effective trigger strength; The normalization layer is used to calculate the normalized value of the effective trigger strength of each rule in the first fuzzy logic rule; The conclusion layer is used to calculate the output conclusion of each rule in the first fuzzy logic rule using a linear function; The output layer is used to obtain the traction force adjustment amount based on the normalized value of the effective trigger intensity and the output conclusion.
9. A spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback according to claim 8, characterized in that, The loss function of the fuzzy parameter adjustment model includes: a traction control accuracy index, a safety index, and a pain level index; the traction control accuracy index is set to reflect the deviation between the target traction force value and the real-time traction force value, the safety index is set to reflect the duration of spasm, and the pain level index is set to reflect the pain level signal; the loss function is used to dynamically update the shape parameters of the membership function of the first fuzzy logic rule, the conclusion parameters of the conclusion layer, and the rule weights.
10. A spinal step-by-step quantitative traction physiotherapy control system based on closed-loop force feedback according to claim 1, characterized in that, The system also includes: The human-computer interaction interface is used to receive the pain level signal and use the pain level signal as a constraint condition. When the pain level signal exceeds a preset pain threshold, the traction force is forcibly reduced or the traction is suspended.
