Soft tissue assessment method and device implemented based on massage robot

Abstract

The invention discloses a soft tissue assessment method and device implemented based on a massage robot. The method comprises the steps that the massage robot is controlled to conduct finger pressure testing and finger kneading testing on different soft tissues in a target area; obtaining the massage strength and height of the massage fingers in the test process, and forming a strength-height curve, a strength-time curve and a height-time curve; respectively extracting curve characteristic values according to the two curves; comprise wave crest and wave trough difference values, waveforms, tangent slopes, secant slopes, frequency distribution and the like; directly carrying out hardness calculation by utilizing a formula or inputting original data into a hardness detection model constructed based on machine learning by utilizing the curve characteristic values extracted by the method, calculating hardness values of corresponding soft tissues according to the detection data, and classifying the hardness values; and integrating the classification result and the hardness value corresponding to the target area to form a soft tissue health assessment report.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical devices, and particularly relates to a soft tissue assessment method and device implemented based on a massage robot. Background Art

[0002] In the fields of medical diagnosis and physical rehabilitation, accurate assessment of muscle conditions is crucial. Traditional assessment methods mainly rely on subjective palpation by medical professionals, that is, judging characteristics such as muscle hardness and tension through the sense of touch. However, this method has great subjectivity and uncertainty, and it is difficult to ensure the objectivity and accuracy of the assessment results.

[0003] To overcome this limitation, modern research has introduced a variety of advanced technical means aimed at achieving objective and precise measurement of muscle conditions. Among them, mechanical devices such as hardness meters generate muscle hardness curves through pressure-displacement sensors, thereby realizing objective quantification of muscle hardness. The Myoton device, as a handheld tool, can measure muscle tension, elasticity, and other biomechanical characteristics. Its working principle is to apply a brief mechanical impact on the muscle and measure the tissue's response to calculate its mechanical properties. However, the Myoton device relies on handheld operation and is easily affected by natural tremors and unintentional movements of the human body, resulting in limited measurement stability and accuracy.

[0004] Ultrasonic elastography techniques, including strain and shear wave imaging, have gradually become important means for assessing muscle stiffness. This technique can minimize subjective bias and improve the accuracy of measurement, especially in detecting muscle stiffness changes related to pain. However, the effectiveness of ultrasonic technology largely depends on the skills and expertise of the operator, and professional operating skills are required to correctly interpret the images and obtain accurate readings. This limits its wide application in various clinical settings, especially in situations where resources are limited or professional skills are insufficient.

[0005] In addition, the high cost is also a key factor restricting the wide application of ultrasonic elastography technology. Many individuals, clinics, or research institutions with limited budgets find it difficult to afford the cost of this technology, thus limiting its popularization and application in muscle diagnosis.

[0006] In summary, although existing muscle condition assessment devices have made certain progress in technology and application, there are still many limitations in terms of accuracy, usability, and accessibility. Therefore, it is particularly important to develop a soft tissue assessment method and device that can synchronize the detection and treatment process, have deep flexibility and adaptability, and are cost-effective. Summary of the Invention

[0007] Therefore, the object of the present invention is to provide a soft tissue evaluation method and device implemented based on a massage robot. First, the robot finger is used to perform finger pressure tests and finger kneading tests on different soft tissues in the target area, and then data such as massage force and height are obtained. The data obtained is analyzed and processed by a hardness detection model constructed based on machine learning, and then the hardness value and classification result of the measured soft tissue are obtained.

[0008] To achieve the above object, a soft tissue evaluation method implemented based on a massage robot of the present invention further includes the following steps:

[0009] S1. Control the massage robot to massage different soft tissues in the target area and perform finger pressure tests and finger kneading tests;

[0010] S2. Obtain the massage force and height of the massage finger during the test to form a force-height curve, a force-time curve, and a height-time curve;

[0011] S3. Extract curve feature values according to the force-height curve, the force-time curve, and the height-time curve respectively; the curve feature values include: peak-to-valley difference, waveform, tangent slope, secant slope, and frequency distribution;

[0012] S4. Use the original data or the extracted curve feature values, calculate using a formula or input them into a hardness detection model constructed based on machine learning to calculate the hardness value of the corresponding soft tissue and perform classification;

[0013] S5. Integrate the classification result and hardness value corresponding to the target area to form a soft tissue health assessment report.

[0014] Further preferably, the massage robot drives a lead screw through a motor, and uses the lead screw to control the slide table to drive the massage finger to move up and down; according to the rotation direction and number of turns of the motor encoder, the finger height is calculated; according to the pressure value detected by the top pressure sensor inside the massage finger, the current massage force is determined.

[0015] Further preferably, in the hardness detection model, the peak-to-valley difference, waveform, tangent slope, secant slope, and frequency distribution are associated with the tissue hardness standard value to establish a hardness prediction function;

[0016] A soft tissue hardness calculation and classification model is constructed by using machine learning methods through a large number of sample data.

[0017] Further preferably, in S4, the process of extracting the curve feature values includes:

[0018] Peak-to-valley difference: Extract the maximum force value and the minimum force value within one period according to the force-time curve and the height-time curve, and calculate the difference;

[0019] Slope: For the force - height curve of the target area, calculate the rate of change of force with depth;

[0020] Frequency distribution: Use Fourier transform on the force - time curve and the height - time curve to extract the frequency characteristics of curve fluctuations;

[0021] Amplitude: Extract the maximum amplitude of the force - height curve to characterize the deformation ability of soft tissues.

[0022] Further preferably, in S4, the hardness detection model constructed based on machine learning includes constructing diverse samples according to the collected force, height, and time; and annotating the corresponding true hardness value for each data sample during model training.

[0023] Normalize the force, depth, and time data;

[0024] Apply a filter to remove random noise in the sensor acquisition;

[0025] Eliminate abnormal data points caused by operation errors or malfunctions;

[0026] Use a formula or a machine learning model to predict the category and hardness value of the detection site.

[0027] Further preferably, in S4, during prediction, it further includes: constructing a cross - entropy loss function to evaluate the prediction loss for classifying different soft tissues; and using the mean square error to analyze the deviation between the predicted value and the true value for estimating the hardness value recognition of the test site.

[0028] Further preferably, it further includes:

[0029] During model training, use the gradient descent method to optimize the model parameters;

[0030] Through batch training, improve the calculation efficiency and dynamically adjust the learning rate;

[0031] Use K - fold cross - validation to ensure the generalization ability of the model;

[0032] By calculating the accuracy, precision, recall rate, and F1 - score, ensure the classification accuracy;

[0033] Based on the hardness detection result, assist in diagnosing nodules, fibrosis, or other soft tissue diseases;

[0034] Dynamically monitor the change of soft tissue hardness to evaluate the treatment effect;

[0035] Establish a soft tissue hardness file to track the long - term health status.

[0036] The present invention also provides a soft tissue evaluation device, based on a massage robot, to implement the steps of the above - mentioned soft tissue evaluation method;

[0037] The massage robot includes a lead screw motor module, a finger kneading massage module, a data analysis module, and a data acquisition module arranged in the lead screw motor and the finger kneading massage;

[0038] The lead screw motor module is used to drive the finger kneading massage module to perform finger pressure test and finger kneading test on the soft tissue to be measured;

[0039] The data acquisition module is used to collect the displacement when the lead screw motor module drives the finger kneading massage module, and the pressure when the finger kneading massage module performs finger pressure or finger kneading;

[0040] The data analysis module is used to obtain the massage force and height of the massage finger during the test process, and form a force-height curve, a force-time curve, and a height-time curve;

[0041] Extract curve characteristic values according to the force-height curve, the force-time curve, and the height-time curve; the curve characteristic values include: peak-to-valley difference, waveform, tangent slope, secant slope, and frequency distribution, etc.;

[0042] Use the curve characteristic values extracted from the original data to perform formula calculations or input them into a hardness detection model constructed based on machine learning to calculate the hardness values of the corresponding soft tissues and classify them.

[0043] The present invention provides a massage robot device, including:

[0044] A memory storing computer program instructions;

[0045] A processor, when the computer program instructions are executed by the processor, implementing the steps of the soft tissue evaluation method implemented based on the massage robot as described above.

[0046] The soft tissue evaluation method and device implemented based on the massage robot disclosed in this application use the robot finger to perform a series of repetitive pressing and finger kneading operations on the tissue to be measured. While these operations are being performed, the present invention collects the corresponding pressure and height data. The collection of these data is for subsequent detailed analysis. Neural network technology is used to train these data in order to obtain objective and accurate hardness data and classification results. Through this training, the present invention successfully constructs an optimized neural network model. Finally, this trained neural network model can be integrated into the system of the massage robot, so that the massage robot can use this model to improve the accuracy and efficiency of its massage operations. Description of the Drawings

[0047] Figure 1 It is a flowchart of a preferred embodiment of the present invention;

[0048] Figure 2 Schematic diagram of the massage device in the preferred embodiment of the present invention;

[0049] Figure 3 Schematic diagram of the soft tissue evaluation device in the preferred embodiment of the present invention;

[0050] Figure 4 Schematic diagram of the lead screw motor module in the preferred embodiment of the present invention;

[0051] Figure 5 Schematic diagram of the finger kneading massage module in the preferred embodiment of the present invention.

[0052] In the figure:

[0053] 1. Lead screw motor module; 2. Finger kneading massage module; 101. Encoder; 102. Motor; 103. Synchronous pulley; 104. Timing belt; 105. Lead screw; 106. Slide table; 201. Massage motor; 202. Electric coupling; 203. Driving gear; 204. Driven gear; 205. Crank; 206. Pressure sensor; 207. Pressure transmission rod; 208. Linear bearing. Detailed implementation manners

[0054] The present invention will be further described in detail below with reference to the accompanying drawings and specific implementation manners.

[0055] Please refer to Figures 1 to 5 : In the following embodiments, the hardness obtained through the robot finger test is a physical quantity that is positively correlated with muscle tension. That is, in this application, the muscle tension is indirectly reflected by the hardness.

[0056] As Figure 1 shown, the soft tissue evaluation method implemented based on a massage robot provided by an embodiment of the present invention includes the following steps:

[0057] S1. Control the massage robot to massage different soft tissues in the target area and perform finger pressure tests and finger kneading tests; among them, the structural schematic diagram of the massage robot is as Figure 2 shown; the massage robot drives the lead screw through a motor, and uses the lead screw to control the slide table to drive the massage finger to move up and down; according to the rotation direction and number of turns of the motor encoder, the finger height is calculated; according to the pressure value detected by the top pressure sensor inside the massage finger, the current massage force is determined.

[0058] S2. Obtain the massage force and height of the massage finger during the test to form a force-height curve, a force-time curve, and a height-time curve;

[0059] S3. Extract the curve characteristic values according to the force-height curve, force-time curve, and height-time curve; the curve characteristic values include: the difference between the peak and valley, waveform, tangent slope, secant slope, and frequency distribution, etc.

[0060] More preferably, in this application, two kinds of silicone blocks simulating human tissues are used to simulate human soft tissues. Inputting the data obtained from the silicone experiments into the hardness detection model constructed based on machine learning can help select a suitable machine learning model and obtain more optimized model parameters. Using the optimized hardness detection model, it can be used to predict and classify the actual human soft tissue massage data.

[0061] The two kinds of silicone blocks simulating human tissues include: type I silicone block and type II silicone block; among them, the type I silicone block is used to simulate ordinary human soft tissues; the type II silicone block is used to simulate human soft tissues with cord-like fibers; it should be noted that the hardness of the type I silicone block and the type II silicone block needs to be detected and the data recorded by a standard Shore hardness tester to evaluate the data performance under different hardnesses and obtain the training and test data sets of the machine learning model corresponding to the hardness.

[0062] The type I silicone block is a cuboid structure of uniform material. During the pressing process of the first type of silicone block, experiments have shown that the slope of the force-height curve can be used to distinguish different hardnesses and establish the relationship between the slope and the hardness. Further preferably, the second silicone block is a cuboid structure, and a cylindrical cord is embedded inside the second silicone block. The second silicone block is used to simulate the characteristics of human tissue fibers, and its unique texture and elasticity provide a model close to real human tissues for research and testing. In the grid areas divided by the type II silicone block containing cords, apply pressure to a certain force value to each grid area, and calculate the slope according to the force-height curve generated by each grid; locate the area of the cord.

[0063] In another experiment, we continuously applied a certain massage force to the silicone for finger kneading. We observed that the force-time curve and the height-time curve showed a wavy pattern due to factors such as friction. As the stiffness increased, the fluctuations of the curve became more obvious. The experimental results show that there is a correlation between material stiffness and parameters such as the difference between the maximum and minimum values of the massage finger pressure and height. This relationship enables the use of functions based on these parameters to effectively distinguish different hardnesses and determine the hardness value.

[0064] When using a massager to perform finger kneading on soft tissue, the changes in force and height over time during the operation can be recorded in real time and plotted as a curve. This curve can intuitively reflect the differences in the characteristics of soft tissues at different locations, especially when there are nodules or spines. Due to the different hardness of these areas, the curve will show unique waveform characteristics. Therefore, the process of extracting the curve feature value includes:

[0065] Peak-to-valley difference: Extract the maximum and minimum values ​​within a cycle based on the force-time curve and height-time curve, and calculate the difference.

[0066] Slope: For the force curve of the target area, calculate the rate at which the force changes with height; the slope of the curve can reflect the rate of change of force and height, and carefully characterize the hardness. By analyzing the slope of the curve at a specific point, the hardness characteristics of the soft tissue can be further inferred. Specifically: Tangent slope: The local hardness of the soft tissue can be accurately measured by using the tangent slope of the curve at a specific point. A larger slope indicates a higher hardness, while a smaller slope corresponds to a softer tissue. Secant slope: For larger area analysis, the two-point secant slope on the curve can be used to measure the hardness change trend over a wider range. This method is particularly effective for evaluating the hardness characteristics of non-uniform soft tissue. These slopes can be used to obtain a function related to hardness for calculating hardness, or the softness and hardness can be directly distinguished by the difference in slope to distinguish different parts, such as nodules, spine, etc.

[0067] Frequency distribution: Fourier transform is used to extract the fluctuation frequency characteristics of the force-time curve and height-time curve; by analyzing the fluctuation frequency of the curve, different types of soft tissue characteristics can be further classified.

[0068] Amplitude: The maximum deformation range of soft tissue when subjected to force is extracted from the force-height curve to characterize the deformation capacity of soft tissue.

[0069] Waveform characteristics: The overall shape of the curve, the distribution of peaks and troughs, and the frequency of fluctuations can all reflect the elasticity, hardness, and deformation capacity of soft tissues. Curves in different hardness areas show different characteristics.

[0070] S4. Utilize the original data or the extracted curve characteristic values, use a formula to calculate or input into a hardness detection model built based on machine learning, calculate the hardness value of the corresponding soft tissue, and classify it;

[0071] In the hardness detection model, the peak-to-trough difference, waveform, tangent slope, secant slope and frequency distribution are associated with the standard value of tissue hardness to establish a hardness prediction function; through a large amount of sample data, a soft tissue hardness calculation and classification model is constructed using machine learning methods.

[0072] In S4, the hardness detection model constructed based on machine learning further includes constructing diverse samples according to the collected force, height, and time; annotating the corresponding true category and hardness value for each data sample during model training; normalizing the force, height, and time data; applying a filter to remove random noise in the sensor acquisition; removing abnormal data points caused by operation errors or faults; using a multi-layer perceptron, convolutional neural network, deep neural network, support vector machine, gradient boosting tree, etc. to predict the category and hardness of soft tissues based on the original data or curve feature values of force and height; and during prediction, it further includes: constructing a cross-entropy loss function to evaluate the prediction loss; using the mean square error to analyze the deviation between the predicted value and the true value to identify the weak areas of the model.

[0073] During model training, the gradient descent method is used to optimize the model parameters; through batch training, the computing efficiency is improved, and the learning rate is dynamically adjusted; K-fold cross-validation is used to ensure the generalization ability of the model; by calculating the accuracy, precision, recall rate, and F1 score, the classification accuracy is ensured; based on the hardness detection results, it is used to assist in diagnosing nodules, fibrosis, or other soft tissue diseases; dynamically monitor the change of soft tissue hardness to evaluate the treatment effect; establish a soft tissue hardness file to track the long-term health status.

[0074] S5. Integrate the hardness value and classification result corresponding to the target area to form a soft tissue health assessment report.

[0075] This application also provides a soft tissue assessment device, which is based on a massage robot and implements the steps of the above soft tissue assessment method; as Figure 3 shown, the massage robot includes a lead screw motor module 1, a finger kneading massage module 2, a data analysis module, and a data acquisition module arranged in the lead screw motor and the finger kneading massage;

[0076] The lead screw motor module 1 is used to drive the finger kneading massage module 2 to perform finger pressure tests and finger kneading tests on the soft tissue to be measured; as Figure 4 shown, the lead screw motor module is used to drive the finger kneading module to massage the soft tissue; wherein, the lead screw motor module includes an encoder 101, a motor 102, a synchronous pulley 103, a synchronous belt 104, a lead screw 105, and a slide table 106; wherein, the (forward and reverse) rotation of the motor 102 is transmitted to the lead screw 105 through the synchronous pulley 103 and the synchronous belt 104; the rotation of the lead screw 105 controls the lifting of the slide table 106; the encoder 101 reads the rotation parameters (forward and reverse rotation, number of turns) of the motor 102 in real time, so as to obtain the height parameter of the lifting of the slide table 106.

[0077] As Figure 5As shown in the figure, the finger-kneading massage module 2 includes a massage motor 201, an electric coupling 202, a driving gear 203, a driven gear 204, a crank 205, a pressure sensor 206, a pressure transmission rod 207, and a linear bearing 208, which are arranged from top to bottom. The massage motor 201 is connected to the driving gear 203 through the electric coupling 202 to drive the driven gear 204 to rotate. The driven gear is connected through the crank 205, so that the finger-kneading massager makes a finger-kneading action. In the finger-kneading massager, it is fixed by the pressure transmission rod 207 and makes a small displacement in the linear bearing 208. By pressing the pressure sensor 206, the sensor can obtain the pressure received by the finger-kneading massager in real time.

[0078] The data acquisition module is used to acquire the displacement when the lead screw motor module drives the finger-kneading massage module, and the pressure when the finger-kneading massage module performs finger pressure or finger-kneading.

[0079] The data analysis module is used to obtain the massage strength and height of the massage finger during the test process to form a strength-height curve.

[0080] Extract curve characteristic values according to the strength-height curve, strength-time curve, and height-time curve. The curve characteristic values include: the difference between the peak and valley, waveform, tangent slope, secant slope, and frequency distribution.

[0081] Use the original data or the extracted curve characteristic values, input them into the hardness detection model constructed based on machine learning, calculate the corresponding hardness values, and classify them.

[0082] Embodiment 2. For the above process, the present application proposes the following embodiments for detailed description.

[0083] Data acquisition: First, it is necessary to conduct finger pressure or finger-kneading tests on soft tissues through multiple experiments, and collect relevant data such as strength and height during the test process. These data will be used as input features of the machine learning model. Secondly, the data needs to cover the typical characteristics of different soft and hard tissues, including soft tissues, nodular areas, sclerotic areas, etc., to ensure the diversity and representativeness of the sample data.

[0084] Specifically include:

[0085] Initial calibration: Ensure the accuracy of the pressure sensor reading.

[0086] Pressing operation: The massager gradually increases the pressing force, and the force value and position change are collected synchronously. The force data is recorded in real time by the built-in force sensor, and the position change is monitored by the encoder for the lead screw movement.

[0087] Finger height calculation: The change in finger height is a direct reflection of the change in the position of the lead screw and the deformation of the soft tissue.

[0088] Data Processing: 1) The system can automatically or semi-automatically extract the characteristics of the force-height curve, such as peaks, valleys, slopes, frequencies, amplitude of fluctuations, etc. 2) Data preprocessing such as normalization, noise reduction, and outlier filtering to ensure the consistency and accuracy of the features before inputting them into the model.

[0089] Specifically include:

[0090] 1. Basic Hardness Calculation: Divide the change in the applied force over a period of time by the change in height.

[0091] Where H is the hardness, ΔF is the applied force, and Δd is the change in height value.

[0092] Curve Fitting and Analysis: Plot the collected force F and position change d data into a curve. Analyze the hardness change of soft tissues under different pressure and displacement conditions according to the curve shape. Horizontal axis (X-axis): Screw position, representing the height of the massage finger. Vertical axis (Y-axis): Applied pressure, indicating the acting force during pressing.

[0093] 2. Slope Analysis, this parameter is calculated using the force-height curve

[0094] Tangent Slope: A larger tangent slope indicates a higher hardness (such as near bones or fibrous tissues), and a smaller tangent slope indicates a softer area (such as muscles or fat layers).

[0095] Formula:

[0096] Where H is the hardness and F(a) is the force value at height a

[0097] Secant Slope (Overall Hardness Trend):

[0098] Over a large range, select the secant slope of two points on the curve to measure the average hardness change trend within the area. The secant slope is used to analyze non-uniform tissues (such as areas containing nodules or fat layers).

[0099] Formula:

[0100] 3. Peak-Valley Difference (this parameter is calculated using the force-time curve and height-time curve): The peak represents the maximum value of force or height; the valley represents the minimum value of force or height.

[0101] Formula: ΔF = F max - F min Δd = d max - d min

[0102] Where ΔF and Δd are the peak-valley differences, F max , dmax and F min ,d min are the maximum and minimum values within a certain period, respectively.

[0103] 4. Waveform characteristics (this parameter is calculated using the force-time curve and height-time curve):

[0104] The smoothness, steepness, frequency, and distribution characteristics of the curve shape.

[0105] 5. Frequency distribution (this parameter is calculated using the force-time curve and height-time curve): The fluctuation frequency of the curve represents the response characteristics of different tissue regions. Perform Fourier transform on the curve data to extract frequency distribution characteristics.

[0106] 6. Elastic and deformation characteristics (this parameter is calculated using the force-height curve); Combine the force-height relationship curve to calculate the elastic modulus of the tissue, reflecting the recovery ability of soft tissue within different force ranges. Record the curve area during the loading and unloading processes to analyze the viscoelastic characteristics of the tissue.

[0107] By performing modeling analysis on the above quantitative indicators, a prediction function for soft tissue hardness can be established:

[0108] 1. Data association: Associate the peak-valley difference, slope, and frequency characteristics with known tissue hardness standard values (such as elastic modulus) to train a regression model.

[0109] 2. The hardness value can be calculated using a formula that constitutes the input with force and height-related data. For example, use the peak-valley difference of force, the peak-valley difference of height, the standard deviation of force, the standard deviation of height, and the correlation coefficient of force and height, etc.

[0110] 3. Using the collected sample data, use machine learning methods (such as linear regression, neural network) to constitute a model. The input is the original data or curve feature values, and the output is the category or hardness value of the soft tissue.

[0111] Model selection and target definition.

[0112] 1) Classification problem: If the goal is to distinguish tissue regions with different hardnesses, such as soft tissues, hard nodules, etc., the model output is multiple categories, and each category corresponds to a tissue type. Commonly used models include support vector machine (SVM), random forest, or neural network models based on deep learning.

[0113] 2) Regression problem: If the goal is to directly calculate the hardness value, the model output is a continuous variable (estimated hardness value). Applicable models include multi-layer perceptron (MLP), convolutional neural network (CNN), or ensemble learning models (such as gradient boosting tree).

[0114] 3) The two models can also be combined so that the model outputs both the soft tissue category and the hardness value simultaneously.

[0115] Loss Function Design

[0116] 1) For classification problems, the Cross-Entropy Loss function is used to measure the difference between the model's predicted class and the actual class.

[0117] 2) For regression problems, the Mean Squared Error (MSE) is used as the loss function to minimize the deviation between the model's predicted value and the true hardness value.

[0118] Model Training and Optimization

[0119] 1) Use the gradient descent method (such as the Adam optimization algorithm) to train the model parameters and minimize the loss function through multiple iterations.

[0120] 2) The cross-validation method can be used to evaluate the model performance, which helps to identify and reduce the risk of overfitting and optimize the model's generalization ability.

[0121] Model Validation and Application

[0122] 1) Use unseen test data to validate the model and evaluate its classification accuracy or the mean squared error of hardness value prediction.

[0123] 2) Deploy the trained model into the system to achieve the real-time hardness detection function.

[0124] By using machine learning methods, the limitations of manually designed features can be overcome, key information can be automatically extracted from complex data, and the accuracy and efficiency of hardness detection can be improved. Especially with the support of large-scale sample data, deep learning models can capture higher-dimensional feature correlations and are suitable for the hardness assessment of complex soft tissues. This method is widely used in fields such as medical diagnosis, rehabilitation evaluation, and intelligent massage devices, providing an intelligent solution for soft tissue health monitoring.

[0125] It can also achieve 1) dynamic control and real-time adjustment. The system combines dynamic control algorithms to automatically adjust the pressing force and finger height. This includes: dynamic force adjustment: According to the real-time pressure feedback, precisely control the output of the force sensor to avoid the influence of excessive or insufficient force on the evaluation accuracy. Dynamic height adjustment: Combine the relationship between the screw position and the force value to dynamically change the finger height and improve the data acquisition efficiency.

[0126] 2) Multi-parameter analysis combined with mechanical models: Analyze the elasticity, viscoelasticity, and hysteresis of soft tissues using pressure-height curves. Based on the data during the deformation-recovery process, evaluate the following characteristics: Elastic modulus: Reflects the hardness and deformation ability of the tissue. Stress relaxation characteristics: Used to judge the viscoelastic properties of the tissue.

[0127] 3) Data visualization, real-time plotting of force-height curves, providing intuitive feedback to users. Including hardness mapping: Construct a three-dimensional hardness map to show the hardness distribution of different parts in the target area. Anomaly detection: Quickly locate abnormal areas through comparison of hardness data, such as nodules or high-hardness points near the spine.

[0128] 4) Technical support for health assessment and disease diagnosis, including:

[0129] Classification of soft and hard tissues: According to the hardness data, distinguish soft tissue types such as fat, muscle, and connective tissue. Provide basic data for health assessment, such as muscle tension or fibrosis conditions.

[0130] Localization of abnormal areas: High-hardness areas may correspond to pathological tissues (such as nodules, myofibrositis, scar tissue). Abnormal shapes of the force-height curve provide diagnostic clues.

[0131] Evaluation of deformation characteristics: Through dynamic analysis of the curve (loading and unloading processes), study the elastic recovery ability and deformation hysteresis characteristics of the tissue.

[0132] Obviously, the above embodiments are only examples for clear illustration and not limitations on the implementation manners. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation manners here. And the obvious changes or modifications derived therefrom are still within the protection scope of the present invention.

Claims

1. A soft tissue assessment method based on a massage robot, characterized in that: The following steps are involved: S1, controlling the massage robot to massage different soft tissues in the target area and perform finger pressure test and finger kneading test; S2, obtaining the massage force and height of the massage finger during the test, forming a force-height curve, a force-time curve, and a height-time curve; S3, extracting curve characteristic values ​​according to the force-height curve, the force-time curve and the height-time curve respectively; the curve characteristic values ​​include: peak-to-valley difference, waveform, tangent slope, secant slope and frequency distribution; S4. Utilize the original data or the extracted curve characteristic values, use a formula to calculate or input into a hardness detection model built based on machine learning, calculate the hardness value of the corresponding soft tissue, and classify it; S5. Integrate the classification results and hardness values ​​corresponding to the target area to form a soft tissue health assessment report.

2. The soft tissue assessment method based on the massage robot according to claim 1, characterized in that: The massage robot drives the screw rod through a motor, and uses the screw rod to control the slide table to drive the massage fingers to move up and down; calculates the displacement height according to the rotation direction and number of turns of the motor encoder; and determines the current massage intensity according to the pressure value detected by the top pressure sensor inside the massage finger.

3. The soft tissue assessment method based on the massage robot according to claim 1 is characterized in that: In the hardness detection model, the peak-to-valley difference, waveform, tangent slope, secant slope and frequency distribution are associated with the tissue hardness standard value to establish a hardness prediction function; Through a large amount of sample data, a soft tissue hardness calculation and classification model is constructed using machine learning methods.

4. The soft tissue assessment method based on the massage robot according to claim 1 is characterized in that: In S4, the process of extracting the curve characteristic value includes: Peak-to-valley difference: Extract the maximum and minimum values ​​within a cycle based on the force-time curve and height-time curve, and calculate the difference; Slope: For the force-height curve of the target area, calculate the rate at which force changes with height; Frequency distribution: Use Fourier transform to extract the frequency characteristics of the force-time curve and height-time curve; Amplitude: The maximum amplitude of the force-height curve is extracted to characterize the deformation capacity of soft tissue.

5. The soft tissue assessment method based on the massage robot according to claim 1, characterized in that: In S4, the hardness detection model constructed based on machine learning includes constructing diversified samples according to the collected force, height and time; marking the corresponding real hardness value for each data sample during model training, Normalize the force, height, and time data; Apply filters to remove random noise from sensor acquisition; Eliminate abnormal data points caused by operational errors or failures; Use formulas or machine learning models to predict the type and hardness value of the test area.

6. The soft tissue assessment method based on the massage robot according to claim 1, characterized in that: In S4, the prediction also includes: constructing a cross entropy loss function to evaluate the prediction loss to classify different soft tissues; using the mean square error to analyze the deviation between the predicted value and the true value to estimate the hardness value identification of the test site.

7. The soft tissue assessment method based on the massage robot according to claim 1, characterized in that: Also includes: When training the model, the gradient descent method is used to optimize the model parameters; Improve computational efficiency through batch training and dynamically adjust learning rate; Use K-fold cross validation to ensure the generalization ability of the model; Ensure classification accuracy by calculating accuracy, precision, recall, and F1 score; According to the hardness test results, assist in the diagnosis of nodules, fibrosis or other soft tissue diseases; Dynamically monitor changes in soft tissue hardness and evaluate treatment effects; Create a soft tissue firmness profile to track long-term health.

8. A soft tissue assessment device, characterized in that: Based on the massage robot, the steps of the soft tissue assessment method described in any one of claims 1 to 7 are implemented; The massage robot comprises a screw motor module, a finger kneading massage module, a data analysis module and a data acquisition module arranged in the screw motor and the finger kneading massage module; The screw motor module is used to drive the finger kneading massage module to massage the soft tissue under test and perform finger pressure test and finger kneading test; The data acquisition module is used to collect the displacement of the finger-kneading massage module when the screw motor module drives the finger-kneading massage module, and the pressure of the finger-kneading massage module when performing finger pressing or finger kneading; The data analysis module is used to obtain the massage force and height of the massage fingers during the test, and form a force-height curve or a force-time curve and a height-time curve; Extracting curve characteristic values ​​according to the force-height curve or the force-time curve and the height-time curve, wherein the characteristic values ​​include: peak-to-valley difference, waveform, tangent slope, secant slope and frequency distribution; The original data or extracted curve characteristic values ​​are used to perform formula calculations or input into a hardness detection model built based on machine learning to calculate the hardness value of the corresponding soft tissue and classify it.

9. A massage robot device, characterized in that: include: a memory storing computer program instructions; A processor, when the computer program instructions are executed by the processor, implements the steps of the soft tissue assessment method based on the massage robot as described in any one of claims 1 to 7.

Images