Data processing method, medium and device in knee arthroplasty

Abstract

The application provides a data processing method, medium and device in knee arthroplasty, and belongs to the technical field of medical electronics. The method comprises the following steps: acquiring pressure distribution data of an interarticular cavity in a full cycle of joint movement measured by an array sensor; calculating a pressure center at each flexion degree in the full cycle of joint movement based on the pressure distribution data, and forming a corresponding pressure center trajectory; calculating shear stress distribution data in the full cycle of joint movement based on the pressure center trajectory data; calculating a dynamic pressure envelope line in the full cycle of joint movement based on the pressure distribution data; and generating an intraoperative correction suggestion for knee arthroplasty based on the pressure center trajectory, the shear stress distribution data and the dynamic pressure envelope line. The correction suggestion given by the application is more accurate for abnormal situations in knee arthroplasty.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical electronics, and in particular to a data processing method in knee arthroplasty, a medium and an equipment. BACKGROUND

[0002] Knee arthroplasty is the core means for treating end-stage knee disease, including total knee arthroplasty (TKA) and unicompartmental arthroplasty, and its core goal is to reconstruct the joint mechanical balance by implanting artificial prosthesis to relieve pain and restore function. The accurate intraoperative assessment of the joint biomechanical state of knee arthroplasty is the key to the success of the operation, which directly affects the stability of the prosthesis, postoperative function and service life. However, the existing intraoperative assessment technology of knee arthroplasty still has many defects, which is difficult to meet the increasingly strict clinical needs.

[0003] Taking unicompartmental arthroplasty as an example, in the traditional intraoperative data detection and evaluation process of unicompartmental arthroplasty, mechanical measurement usually needs to rely on gap measurement blocks of different thicknesses or meniscus pads of different thicknesses, and the passive movement of the knee given by the doctor is used to evaluate the intercondylar pressure and joint stability. The pressure detection device used is usually a rigid sensor, which can only provide static measurement, cannot fit the joint surface and interferes with the natural motion trajectory, resulting in inaccurate pressure data, and it is also difficult to accurately measure or deduce related mechanical indicators other than pressure to assist in intraoperative evaluation.

[0004] For TKA intraoperative assessment technology, it is currently mainly divided into traditional mechanical measurement evaluation and electronic pressure sensor technology evaluation. For the traditional mechanical measurement evaluation technology, it is similar to unicompartmental arthroplasty, and it is also very dependent on gap measurement instruments, tension meters and other tools. By physically measuring the gap parameters of the knee joint in flexion, extension and other static positions, the joint stability and soft tissue balance degree are evaluated in combination with the doctor's sense of touch. However, it has the following significant limitations: 1) strong subjectivity, the evaluation of soft tissue balance degree depends on the doctor's sense of touch, lacks quantitative pressure data support, and the results of different doctors differ greatly, which is easy to cause evaluation deviation; 2) static measurement limitation, only the gap value at a specific static position (such as flexion 0°, 90°) can be obtained, and the dynamic pressure change of the knee joint from flexion 0° to 120° full motion cycle cannot be captured, making it difficult to reflect the mechanical characteristics under the real motion state; 3) limited detection range, the contact area of the measurement tool is usually less than 20 , which cannot cover the edge area of the prosthesis, and it is difficult to identify the micro-scale stress concentration area with an area less than 10 , which is an important inducement for early wear and loosening of the prosthesis.

[0005] For electronic pressure sensing technology, it can realize quantitative measurement of contact pressure between femoral and tibial prostheses, and obtain pressure distribution data at each point between femoral and tibial prostheses. However, in the existing process of intraoperative guidance based on pressure distribution data, only the pressure distribution center is calculated based on the pressure distribution data, and the correction suggestions such as adjustment of osteotomy angle are evaluated and guided according to the pressure distribution center. However, this adjustment and correction suggestion has the defect of single data reference dimension, and cannot comprehensively evaluate intraoperative biomechanics, leading to inaccurate intraoperative biomechanical evaluation, easy to cause postoperative complications (such as prosthesis loosening, accelerated wear and tear, joint instability, etc.), reduce the postoperative life quality of patients, and increase the risk of revision surgery.

[0006] Therefore, there is an urgent need for a multi-dimensional data processing system in knee arthroplasty to overcome the defects of the prior art and improve the accuracy and reliability of intraoperative data detection and evaluation in knee arthroplasty. SUMMARY

[0007] The purpose of the present application is to provide a data processing method, medium and equipment in knee arthroplasty to solve at least one of the above technical problems.

[0008] In a first aspect, the present application provides a data processing method in knee arthroplasty, the method comprising:

[0009] obtaining pressure distribution data of an interarticular cavity in a full cycle of joint motion measured by an array sensor;

[0010] calculating a pressure center at each flexion degree in the full cycle of joint motion based on the pressure distribution data, and forming a corresponding pressure center trajectory;

[0011] calculating shear stress distribution data in the full cycle of joint motion based on the pressure center trajectory data;

[0012] calculating a dynamic pressure envelope in the full cycle of joint motion based on the pressure distribution data;

[0013] generating intraoperative correction suggestions for knee arthroplasty based on the pressure center trajectory, shear stress distribution data and dynamic pressure envelope.

[0014] Optionally, the method further comprises:

[0015] calling a preset convolutional neural network model to extract spatial features and temporal features from the pressure distribution data to obtain gradient distribution data of medial-lateral pressure ratio and phase relationship data of pressure change rate and joint angular velocity;

[0016] generate the dynamic pressure envelope line based on the gradient distribution data and the phase relationship data.

[0017] Optionally, the preset convolutional neural network model is called to extract spatial features and temporal features from the pressure distribution data to obtain gradient distribution data of medial-lateral pressure ratio and phase relationship data of pressure change rate and joint angular velocity, which comprises:

[0018] Based on the pressure distribution data, the pressure ratio of the knee joint medial-lateral at each time step is calculated to obtain the medial-lateral pressure ratio data;

[0019] The medial-lateral pressure ratio data is input into the convolutional neural network model, and the spatial relationship in the medial-lateral pressure ratio data is captured through the convolutional layer and the pooling layer in the convolutional neural network model to obtain the gradient distribution data;

[0020] Based on the pressure distribution data, the pressure change rate data and the joint angular velocity data at each time step are calculated;

[0021] The pressure change rate data and the joint angular velocity data are input into the network structure containing the recurrent layer in the convolutional neural network model, and the phase relationship between the pressure change rate data and the joint angular velocity data is captured through the recurrent layer to output the phase relationship data.

[0022] Optionally, the method further comprises:

[0023] Obtain a set of pressure distribution training data, preprocess the pressure distribution training data in the set of pressure distribution training data to form a preprocessed data set conforming to the standard normal distribution;

[0024] Iteratively train the pre-trained model based on the preprocessed data set, calculate the loss value based on the feature vector output by each iteration, adjust the parameters in the pre-trained model based on the calculated loss value, and end the iteration training when the number of iterations reaches a preset number of times threshold or the latest loss value is less than a preset loss threshold, and output the trained convolutional neural network model;

[0025] Wherein, the process of each iteration comprises sequentially tensor processing, convolution processing, normalization processing, activation processing and pooling processing of the preprocessed data set.

[0026] Optionally, the pressure distribution data comprises pressure values measured by each sensing unit in the array sensor at each time; and the pressure center in each flexion degree in the joint movement full cycle is calculated based on the pressure distribution data to form a corresponding pressure center trajectory, which comprises:

[0027] The pressure center at the corresponding time is calculated based on the pressure value of each sensing unit at the same time and the position coordinates of the corresponding sensing unit; and the pressure center trajectory is formed based on the pressure center at each time.

[0028] Optionally, the shear stress distribution data in the entire cycle of the joint movement is calculated based on the pressure center trajectory data, including:

[0029] The height of the action point of the friction force is determined according to the surface friction layer elevation of the array sensor;

[0030] The moving distance in the tangential direction is calculated based on the pressure center trajectory;

[0031] The shear stress at the corresponding time is calculated based on the height of the action point, the moving distance and the pressure center at the corresponding time.

[0032] Optionally, the intraoperative correction suggestion for knee arthroplasty is generated based on the pressure center trajectory, the shear stress distribution data and the dynamic pressure envelope, including:

[0033] A transfer function of the bone cutting angle adjustment amount and the pressure distribution change amount is established, the pressure distribution change amount being calculated based on the offset amount of the pressure center trajectory and the gradient change of the shear stress distribution data;

[0034] An optimization suggestion vector of the bone cutting angle adjustment is generated according to the transfer function and the normal range of the dynamic pressure envelope.

[0035] Optionally, the transfer function of the bone cutting angle adjustment amount and the pressure distribution change amount includes: training an intelligent agent through reinforcement learning, taking the bone cutting angle adjustment amount as the action input, taking the pressure distribution change amount and the feedback signal obtained after the action is performed as the reward feedback, and iteratively optimizing to obtain a mapping relationship, and converting the mapping relationship into the transfer function.

[0036] In a second aspect, a computer readable storage medium is provided, and the computer readable storage medium stores executable instructions. When the executable instructions are executed by a processor, the processor performs the method in any one of the embodiments of the present application.

[0037] In a third aspect, an electronic device is provided, and the electronic device includes one or more processors, and a memory storing one or more programs, which when executed by the one or more processors, cause the one or more processors to perform the method in any one of the embodiments of the present application.

[0038] The data processing method, medium and device in knee arthroplasty in the application cover the edges of the prosthesis and the full movement cycle through the array sensor, so that the information of the obtained pressure distribution data is more abundant, and the doctor's subjective hand feeling is converted into digital data (such as pressure center offset, shear stress value) by obtaining the pressure center trajectory, shear stress based on the pressure distribution data; and the dynamic pressure envelope is further calculated according to the pressure distribution data, which provides an objective judgment standard for intraoperative evaluation, avoids misjudgment caused by relying on "empirical threshold"; through the fusion analysis of multi-dimensional data such as pressure center trajectory, shear stress, dynamic pressure envelope, the correction suggestion given for the abnormal situation in knee arthroplasty such as total knee arthroplasty and unicompartmental arthroplasty is more accurate and comprehensive. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the application, and therefore should not be regarded as limiting the scope of the application.

[0040] Figure 1 A flowchart of the data processing method in knee arthroplasty in an embodiment;

[0041] Figure 2 A structural diagram of the array sensor in an embodiment;

[0042] Figure 3 A structural diagram of the array sensor in another embodiment;

[0043] Figure 4 A pressure distribution cloud chart in an embodiment;

[0044] Figure 5 A flowchart of calling a preset convolutional neural network model to extract spatial features and time features from pressure distribution data, obtain gradient distribution data of medial-lateral pressure ratio, and phase relationship data of pressure change rate and joint angular velocity in an embodiment;

[0045] Figure 6 A structural diagram of an electronic device in an embodiment. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the application more clear and understandable, the following will further describe the application in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application, and are not used to limit the application.

[0047] All terms used herein, including technical and scientific terms, have the meanings commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the use of certain terms herein, such as those listed below, should not be interpreted to imply any particular significance of importance unless otherwise indicated.

[0048] For example, the terms "first", "second", and the like as used herein merely denote differentiating descriptions for similar objects, and distinguish a first object from another object, rather than describing a specific order or sequence, nor can it be understood as indicating or implying relative importance.

[0049] The present application proposes a data processing method in knee arthroplasty, as shown in Figure 1 The method comprises:

[0050] In step 110, pressure distribution data of the intercondylar compartment during the whole cycle of joint movement is obtained by measuring through an array sensor.

[0051] In this embodiment, the knee arthroplasty can include total knee arthroplasty and unicompartmental arthroplasty, etc. The array sensor can be placed between the tibias after osteotomy during the operation for pressure measurement. The array sensor is matched with the shape of the knee joint and can cover the load-bearing surface of the tibial prosthesis. Specifically, the array sensor can cover the load-bearing surface of the tibial prosthesis and its edge extension area, such as covering its edge 5mm extension area, and can be attached to the surface of the tibial prosthesis, deforming adaptively with the shape of the prosthesis during knee flexion and extension, to ensure measurement stability. The array sensor can be a flexible composite sensor array, such as an array capacitive sensor, which realizes distributed measurement of intercondylar pressure through pressure-electric signal conversion.

[0052] The intercondylar compartment refers to the space between the femur and tibia, and between the femur and patella in the knee joint, including the medial compartment (between the medial femoral condyle and tibial plateau in the knee joint) and the lateral compartment (between the lateral femoral condyle and tibial plateau in the knee joint). The intercondylar compartment is the core area of pressure transmission, and its pressure distribution directly reflects the mechanical balance state of the knee joint. The pressure distribution data refers to the set of pressure values measured by each sensing unit in the array sensor at different times. The pressure distribution data can be presented in the form of a two-dimensional matrix or a one-dimensional vector, and the elements therein reflect the pressure size and distribution characteristics (spatial resolution <10 micro-scale stress concentration area).

[0053] The whole cycle of joint movement refers to the continuous movement process of the knee joint from complete extension (such as 0°) to maximum flexion (such as 120°), covering the flexion angles commonly seen in daily activities.

[0054] In the intraoperative evaluation stage of knee arthroplasty, the array sensor is attached to the load-bearing surface of the tibial prosthesis, ensuring that the array sensor completely covers the joint contact area. When the knee joint completes continuous flexion movement from 0° to 120° under the drive of a mechanical auxiliary device, the sensor array collects the pressure values of each sensing unit in real time at a preset acquisition frequency (such as 100 Hz or higher), generating pressure distribution data at each moment.

[0055] As shown in Figure 2 , it shows one form of the array sensor which matches the surface of the tibial prosthesis in total knee arthroplasty. The array sensor can be an array capacitive sensor, covering the load-bearing surface of the tibial prosthesis and a 5mm extension area on the edge, with multiple sensing units 201 (only the sensing units on the left half are shown in Figure 2 ) uniformly distributed on it. The sensing units can be pressure sensing units, which can measure the pressure at the corresponding position. The corresponding pressure center 202 can be calculated from the pressure at each position, and combined with the change of time, the corresponding pressure center trajectory 203 is formed.

[0056] As shown in Figure 3 , it shows another form of the array sensor which matches the surface of the tibial prosthesis in unicompartmental knee arthroplasty. The array sensor is also an array capacitive sensor, covering the load-bearing surface of the tibial prosthesis and a 5mm extension area on the edge, with multiple sensing units 301 (only part of the sensing units are shown in Figure 3 ) uniformly distributed on it. The sensing units can be pressure sensing units, which can measure the pressure at the corresponding position. The corresponding pressure center 302 can be calculated from the pressure at each position, and combined with the change of time, the corresponding pressure center trajectory 303 is formed.

[0057] The pressure distribution data can be reflected by a pressure cloud chart, as shown in Figure 4 , which shows the distribution of pressure of the array sensor at different positions. The closer the color is to red, the greater the pressure is; the closer the color is to blue, the smaller the pressure is.

[0058] In step 120, the pressure center at each flexion degree in the whole cycle of joint movement is calculated based on the pressure distribution data, and the corresponding pressure center trajectory is formed.

[0059] The pressure center refers to the "point of action of the resultant force" of the pressure distribution in the joint compartment, which is the moment balance point of the pressure field, similar to the center of mass of an object. The pressure center trajectory refers to the continuous path of the pressure center changing with the flexion angle in the whole cycle of joint movement (0° to 120° flexion), reflecting the trend of the center of gravity of dynamic pressure distribution. The pressure center can be calculated comprehensively according to the pressure value of each sensing unit and the position of the sensing unit.

[0060] Specifically, the pressure distribution data includes the pressure values measured by each sensing unit in the array sensor at each time point. Step 120 includes: calculating the pressure center at the corresponding time point based on the pressure values of each sensing unit at the same time point and the position coordinates of the corresponding sensing unit; and forming a pressure center trajectory based on the pressure centers at each time point.

[0061] The electronic device can establish a corresponding spatial coordinate system based on the pressure sensor or the joint, map each sensing unit in the pressure sensor to the spatial coordinate system, and obtain the coordinate position of the center point in each sensing unit in the spatial coordinate system. Taking the sensing units in the pressure sensor as an m x n rectangular sensing unit distribution as an example, let P ij represent the measured pressure value of the sensing unit in the ith row and the jth column; (x ij , y ij ) represents the coordinate position of the center point of the sensing unit in the coordinate system; and the pressure center coordinates on the corresponding two-dimensional pressure field are marked as (X CoP , Y CoP ). Wherein, , .

[0062] Connect the pressure center coordinates corresponding to all flexion angles in the full cycle in order of angle to form a pressure center trajectory.

[0063] By tracking the real-time pressure center during the operation, the best prosthesis rotation angle of the tibial plateau is determined, the prosthesis position is dynamically adjusted to make the postoperative pressure center trajectory close to the natural knee joint characteristics, and at the same time, errors caused by traditional dependence on anatomical landmarks can be avoided (such as avoiding abnormal patellar trajectory caused by excessive internal rotation of the tibial prosthesis), the balance effect is displayed digitally, and the subjectivity of traditional "hand feeling evaluation" is reduced.

[0064] Step 130, based on the pressure center trajectory data, calculate the shear stress distribution data in the full cycle of joint movement.

[0065] Shear stress distribution data refers to tangential stress distribution data parallel to the joint contact surface. Shear stress reflects the friction mechanics characteristics of the knee joint during shear, sliding or torsional movement, and its unit is MPa. In actual contact, in addition to the normal pressure, there is also a tangential friction force (shear stress), which is a manifestation of external shear, sliding or torsional force. Since the sensor cannot directly sense the tangential force, we infer the existence and direction of the shear stress indirectly through the dynamic change of the pressure center, and then form the shear stress distribution data.

[0066] Based on the pressure center trajectory, the offset amount (Δx, Δy, i.e. the difference between the current coordinate and the baseline coordinate) of the pressure center under different flexion angles is analyzed, the shear stress is calculated according to the moment balance principle and / or Hooke's law, and the shear stress distribution data is formed.

[0067] In one embodiment, step 130 comprises: determining the action point height of the friction force according to the surface friction layer elevation of the array sensor; calculating the moving distance in the tangential direction based on the pressure center trajectory; and calculating the shear stress at the corresponding moment based on the action point height, the moving distance, and the pressure center at the corresponding moment.

[0068] The surface friction layer elevation refers to the thickness of the friction layer in the array sensor that directly contacts the articular cartilage or the surface of the prosthesis, which is usually made of medical-grade silicone or polyurethane material, and the elevation determines the position of the action point of the friction force. The surface friction layer elevation parameter is usually provided by the sensor manufacturer and stored in the device database, and the surface friction layer elevation parameter can be obtained from the device database. The action point height of the friction force refers to the vertical height of the actual action position of the friction force between the joint contact surface and the sensor friction layer.

[0069] The moving distance in the tangential direction refers to the displacement amount of the pressure center in the tangential direction (i.e. parallel to the direction of the joint contact surface) between two adjacent moments (or flexion angles), reflecting the dynamic changes of the joint during sliding or torsion. In terms of the coordinates of the pressure centers of the two adjacent moments (or flexion angles) as (X CoP1 , Y CoP1 ) and (X CoP2 , Y CoP2 ), the moving distance in the tangential direction of the pressure centers of the two adjacent moments (or flexion angles) can be expressed as the following formula 1.

[0070] Formula 1

[0071] Shear stress can be calculated according to the following formula 2.

[0072] Formula 2

[0073] This formula expresses that when the contact state is constant and the friction model is linear, the shear stress is proportional to the moving distance of the pressure center, the greater the shear stress, the easier it is to produce a large tangential friction force, but the shear stress corresponding to the unit moving distance of the pressure center also increases. Wherein, h represents the equivalent friction force action point height (approximately the thickness of the material contact layer or the surface friction layer elevation of the sensor), represents the offset amount of the pressure center in the tangential direction, which is approximately equal to the moving distance​ , Pn is the normal total pressure.

[0074] At step 140, a dynamic pressure envelope line in the whole cycle of joint movement is calculated based on the pressure distribution data.

[0075] The dynamic pressure envelope line refers to the boundary of the normal pressure range generated based on the pressure distribution data of a large number of normal knee replacement surgeries. Specifically, the dynamic pressure envelope line is expressed as a mean value (μ) ± 2 times a standard deviation (2σ), and covers the pressure distribution characteristics in the whole cycle of 0°-120° flexion.

[0076] The envelope line is used to determine whether the intraoperative pressure distribution is normal, wherein the upper limit (μ+2σ) and the lower limit (μ-2σ) correspond to the maximum and minimum values of the normal pressure, respectively. For example, the dynamic pressure envelope line of the medial compartment at 0° flexion is 3.0-5.0 MPa, and if the measured value during the surgery is 5.8 MPa, it exceeds the upper limit, which can prompt an abnormality.

[0077] Specifically, the spatial features and the temporal features can be extracted from the obtained pressure distribution data, and the dynamic pressure envelope line in the whole cycle of joint movement is generated based on the spatial features and the temporal features. The spatial features can be the gradient distribution of the medial-lateral pressure ratio, and the temporal features can be the phase relationship between the pressure change rate and the joint angular velocity.

[0078] At step 150, an intraoperative correction suggestion for the knee replacement surgery is generated based on the pressure center trajectory, the shear stress distribution data, and the dynamic pressure envelope line.

[0079] In this embodiment, after obtaining the pressure center trajectory, the shear stress distribution data, and the dynamic pressure envelope line, it can be determined based on these data whether the current knee replacement surgery is normal during the surgery, and when an abnormality occurs, a corresponding correction suggestion is generated.

[0080] Specifically, it can be detected whether the pressure center trajectory deviation exceeds the upper limit or the lower limit of the dynamic pressure envelope line, and when it exceeds the upper limit or the lower limit, it indicates that there is an abnormality, and a corresponding correction suggestion is given according to the specific situation. It is detected whether the shear stress exceeds the corresponding shear stress threshold value, and if it exceeds the shear stress threshold value, a corresponding correction suggestion is also given to reduce the shear stress.

[0081] For example, if the pressure center trajectory deviation exceeds 1.5 mm (such as a medial deviation of 2.0 mm), and the medial pressure exceeds the upper limit of the envelope line (6.2 MPa>5.0 MPa), it can be prompted that the medial collateral ligament is too tight, and it is suggested to release the medial ligament by 0.5 mm; if the shear stress is 7 MPa at 90° flexion (the corresponding shear stress threshold value is 5 MPa), and the pressure center deviates backward, it is prompted that the posterior joint capsule is tight, and it is suggested to adjust the posterior inclination angle of the tibial prosthesis by 1°.

[0082] In one embodiment, when there is a shear stress region greater than a shear stress threshold value in the shear stress distribution data, a correction suggestion is generated in combination with the coordinate offset of the pressure center trajectory in the region, and the correction suggestion can be specifically a prosthesis rotation angle adjustment suggestion, wherein the shear stress threshold value is determined based on a shear stress safety range of the dynamic pressure envelope.

[0083] The shear stress distribution data of the entire cycle of joint movement can be monitored in real time, and it is determined whether there is a region greater than the shear stress threshold value (such as 5 MPa) by point-by-point comparison. The corresponding coordinate of the region exceeding the shear stress threshold value in the pressure center trajectory is located, and the difference (offset) between the coordinate and the normal pressure center coordinate in the dynamic pressure envelope is calculated. According to the position and offset of the region exceeding the shear stress threshold value, a corresponding correction suggestion is generated in combination with a preset "shear stress-rotation angle" mapping relationship (such as a transfer function).

[0084] For example, the mapping relationship indicates that for every 1 MPa of medial shear stress exceeding the threshold value, and 1 mm of medial offset of the pressure center, 1° of external rotation of the femoral prosthesis is required. If the current situation is that the shear stress exceeds the shear stress threshold value by 0.8 MPa, and the offset is 2.0 mm, the generated correction suggestion is: suggest 3° of external rotation of the femoral prosthesis.

[0085] The correction suggestion can be presented in the form of one or a combination of more of local highlighting of a three-dimensional cloud map, voice prompt, correction suggestion pop-up window, etc., to realize intraoperative warning prompt. For example, when it is detected that the local pressure of the relevant part exceeds the preset pressure threshold value (such as 2.5 Mpa) and the duration exceeds the preset duration (such as 2 seconds), a three-dimensional cloud map is generated, and in the cloud map, local highlighting is performed for the part where the local pressure exceeds 2.5 Mpa and the duration exceeds 2 seconds. In addition, when it is detected that the medial-lateral pressure difference exceeds the preset percentage threshold value (such as 35%) and the trajectory offset exceeds the preset offset threshold value (such as 1.5 mm), the feedback form can be voice prompt + correction suggestion pop-up window.

[0086] In one embodiment, the pressure distribution data and the upper and lower limits of the dynamic pressure envelope are compared in real time, and when the pressure distribution data exceeds the upper and lower limits, the offset direction and offset amount of the pressure center trajectory at the corresponding timestamp are extracted, and a corresponding correction suggestion is generated in combination with the directional features of the shear stress distribution data. The correction suggestion can be specifically a suggestion of soft tissue release or pad thickness adjustment.

[0087] The upper and lower limits of the dynamic pressure envelope represent the normal pressure range boundary of the dynamic pressure envelope, and the upper and lower limits constitute the "normal interval" of the pressure distribution. The offset direction refers to the deviation direction of the pressure center relative to the normal position (envelope center) (such as inside, outside, front, back); the offset amount refers to the distance of the deviation. The direction feature of the shear stress distribution data refers to the main direction of action of the shear stress (such as forward, backward, inward, outward), which is determined by the moving direction of the pressure center trajectory.

[0088] The soft tissue release suggestion refers to the intraoperative release scheme proposed for soft tissue tension (such as ligament contracture), including the release site (such as the medial collateral ligament, the posterior joint capsule) and the release degree (such as the release length).

[0089] The spacer thickness adjustment suggestion refers to a scheme for adjusting the joint gap size by replacing polyethylene spacers of different thicknesses to balance the pressure distribution.

[0090] Specifically, the pressure distribution data is compared with the upper and lower limits of the dynamic pressure envelope, and the areas exceeding the range are marked. The coordinates of the pressure center corresponding to the areas exceeding the range are located, and the deviation from the normal coordinates (the center of the envelope) is calculated. At the same time, the direction of the shear stress corresponding to the areas exceeding the range is analyzed, the sliding trend is judged in combination with the offset direction of the pressure center, and the correction suggestions are generated according to the offset amount, the direction of the shear stress, and the pressure exceeding the upper limit amplitude.

[0091] For example, if the pressure exceeding the upper limit amplitude is ≤1.0 MPa, and the offset amount is ≤2 mm: it is suggested to release the tight soft tissue (such as the lateral collateral ligament), for example, when the offset is 2.5 mm, it is suggested to release 0.5 mm; if the pressure exceeding the upper limit amplitude is >1.0 MPa, and the offset amount is >2 mm: it is suggested to adjust the spacer thickness at the same time, for example, replacing the original 2 mm spacer with a 2.5 mm spacer, increasing the lateral gap, and reducing the pressure.

[0092] The data processing method in knee arthroplasty in the present application covers the edges of the prosthesis and the full motion cycle through the array sensor, so that the information of the obtained pressure distribution data is more abundant. The doctor's subjective hand feeling is converted into digital data (such as pressure center offset amount, shear stress value) by obtaining the pressure center trajectory, shear stress based on the pressure distribution data. And according to the pressure distribution data, the dynamic pressure envelope is further calculated, which provides an objective judgment standard for intraoperative evaluation, avoiding misjudgment caused by relying on "empirical threshold". Through the fusion analysis of multiple dimensions of pressure center trajectory, shear stress, dynamic pressure envelope and other data, the accuracy of identifying abnormal conditions in total knee arthroplasty, unicompartmental arthroplasty and other knee arthroplasty can be improved, and the rationality of the correction suggestions given for abnormal conditions can be improved.

[0093] In one embodiment, step 140 comprises: calling a preset convolutional neural network model to perform spatial feature and time feature extraction from the pressure distribution data, to obtain gradient distribution data of the medial-lateral pressure ratio and phase relationship data of the pressure change rate and the joint angular velocity; and generating a dynamic pressure envelope based on the gradient distribution data and the phase relationship data.

[0094] In this embodiment, the convolutional neural network model comprises multiple layers of structures such as convolutional layers and pooling layers, which are used to perform spatial feature and time feature extraction on the pressure distribution data. The spatial feature represents the difference and trend of the pressure distribution in the spatial dimension, focusing on the uniformity of the pressure distribution in the knee joint anatomical region; the time feature reflects the dynamic characteristics of the pressure distribution changing with time (or joint movement angle), focusing on the synchronization of pressure change and joint movement. Through spatial feature extraction, gradient distribution data of the medial-lateral pressure ratio can be obtained; through time feature extraction, phase relationship data of the pressure change rate and the joint angular velocity can be obtained.

[0095] The gradient distribution data of the medial-lateral pressure ratio refers to the rate of change of the medial-lateral pressure ratio (ML Ratio) in space, reflecting the transition trend of the pressure from the medial side to the lateral side. The gradient of the medial-lateral pressure ratio (ML Ratio = medial pressure mean / lateral pressure mean) represents the change amount of the ML Ratio per unit distance, and the greater the absolute value of the gradient, the more significant the difference between the medial and lateral pressures. The phase relationship data of the pressure change rate and the joint angular velocity refers to the time difference (or phase difference) data between the pressure change rate (dP / dt) and the joint flexion angular velocity (dθ / dt), reflecting the synchronization of the two. By calculating the angle difference (or phase difference) between the pressure peak time and the angular velocity peak time, it can be determined whether the pressure change matches the movement rhythm. The normal phase difference is usually within ±15°, and exceeding it indicates dynamic imbalance.

[0096] Specifically, the acquired joint movement full-cycle (0°-120° flexion) pressure distribution data can be preprocessed first, which includes missing value filling, outlier removal, data format conversion, and data standardization, etc. After data format conversion, the pressure distribution data can be converted into a tensor format, including sample number, time step, sensor number (or pressure and movement information combined dimension), etc. The time step corresponds to the flexion angle (e.g. 1200 steps, 0.1° per step); the feature dimension includes the pressure value of each sensing unit (100 units) and the joint angular velocity at the corresponding time (extracted from the movement trajectory data). For example, for the first sample, the time step is 1s, the sensor value length recorded in 1s is n, and the coordinates of the pressure center are (x, y), then the corresponding preprocessed tensor format dimension is [1, 1, n+2].

[0097] In the data standardization process, the pressure value and the angular velocity are standardized to convert them into standard normal distribution data with a mean of 0 and a standard deviation of 1. The following formula 3 can be used for processing:

[0098] Formula 3

[0099] Where μ and σ are the mean and standard deviation of the data, respectively, x is the original data, and x' is the data after standardization. For example, the range of the pressure value after standardization is [-1, 1], and the range of the angular velocity after standardization is [-0.5, 0.5].

[0100] The convolutional neural network model can be a 16-layer network structure, where the first layer is the input layer, the second to twelfth layers are the convolutional layer group, which includes four convolutional layers, three activation layers, and four batch normalization layers; the thirteenth to fourteenth layers are the pooling layer group, which includes two pooling layers; the fifteenth layer is the fully connected layer, and the sixteenth layer is the output layer. It can be understood that the convolutional neural network model can also be any other suitable network structure.

[0101] In the spatial feature extraction process, according to the distribution of each sensing unit in the array sensor inside and outside the knee joint (such as the medial compartment corresponding to sensing units 1-50 and the lateral side corresponding to sensing units 51-100), the medial pressure mean (denoted as Pmedial) and the lateral pressure mean (denoted as Plateral) of each time step are calculated. Based on the medial pressure mean and the lateral pressure mean, the medial-lateral pressure ratio (ML Ratio = Pmedial / Plateral) can be calculated, and then the gradient distribution data of the medial-lateral pressure ratio can be obtained by extracting the change rate of the medial-lateral pressure ratio in space through the convolutional layer of the convolutional neural network model, i.e., the change amount of ML Ratio per millimeter distance from the medial side to the lateral side.

[0102] In the time feature extraction process, the pressure distribution data is time-differentiated to obtain the pressure change rate (unit: MPa / s) of each time step; according to the joint motion trajectory, the flexion angle change rate of each time step is calculated. The phase relationship data of the pressure change rate and the joint angular velocity can be obtained by calculating the peak time difference of the pressure change rate and the joint angular velocity through the convolutional neural network model.

[0103] After extracting spatial and temporal features, the gradient distribution data and phase relationship data are integrated according to the time step (buckling angle) to form a comprehensive feature representation. For example, the two feature vectors can be concatenated or combined in other suitable ways. Based on the integrated features, the dynamic pressure envelope is calculated using statistical methods. The upper and lower limits of the pressure distribution at each angle can be determined based on the mean μ and standard deviation σ under normal conditions; for example, the upper limit is μ+2σ and the lower limit is μ-2σ.

[0104] The generated dynamic pressure envelope is presented in a visual form (such as a colored area on the intraoperative display screen, with green indicating the normal range and red indicating abnormality), allowing for real-time comparison with the pressure distribution data of the current surgery to determine if it is within the range.

[0105] In one embodiment, such as Figure 5 As shown, a pre-defined convolutional neural network model is invoked to extract spatial and temporal features from the pressure distribution data, obtaining gradient distribution data of the inner and outer pressure ratio and phase relationship data between the pressure change rate and the joint angular velocity, including:

[0106] Step 510: Calculate the pressure ratio between the medial and lateral sides of the knee joint at each time step based on the pressure distribution data to obtain the medial and lateral pressure ratio data.

[0107] For time step division, for example, the entire cycle of joint movement (0°-120° flexion) can be divided into 1200 time steps at a sampling frequency of 100Hz. Each step corresponds to 0.01 seconds and 0.1° flexion angle, and each step corresponds to one pressure distribution data point, such as data formed from the pressure values ​​of 10×10 sensing units. Medial and lateral region division: Based on the anatomical positioning of the array sensors, for example, sensing units 1-50 can be preset for the medial compartment (covering the medial tibial plateau), and sensing units 51-100 can be preset for the lateral compartment (covering the lateral tibial plateau).

[0108] For each time step, the mean inner pressure (P_inner) and the mean outer pressure (P_outer) are calculated. Based on these mean inner and outer pressures, the corresponding inner-outer pressure ratio can be calculated.

[0109] Step 520: Input the inner and outer pressure ratio data into the convolutional neural network model. The convolutional and pooling layers in the convolutional neural network model capture the spatial relationship in the inner and outer pressure ratio data to obtain gradient distribution data.

[0110] First, the inner and outer pressure ratio data is converted into the input format of a convolutional neural network model. Then, gradient distribution data is obtained through processes such as capturing local spatial features through convolutional layers, extracting key features through pooling layers, and generating gradient distribution data.

[0111] Specifically, the medial-lateral pressure ratio data can be converted into a 1200x1 time sequence matrix, where the rows correspond to time steps and the columns correspond to pressure ratios, and expanded into a two-dimensional feature map (1200x1x1) to adapt to the convolution operation.

[0112] For the convolution layer to capture local spatial features, the first convolution layer uses a 3x1 convolution kernel (covering 3 consecutive time steps) with 16 numbers to perform sliding calculation on the pressure ratio data, capturing local change patterns (such as the increasing trend of the ratio from 1.18 to 1.19 to 1.20 in the t1-t3 step). Activation and normalization: introducing nonlinearity through the ReLU activation layer, and stabilizing training through the batch normalization layer, outputting 16 feature maps (each 1198x1, truncated at the edge). The second convolution layer uses a 5x1 convolution kernel (covering 5 time steps) with 32 numbers to further capture more complex spatial relationships (such as the gradient change of the ratio in the consecutive 5 steps).

[0113] For the pooling layer to extract key features, the first pooling layer uses 2x1 max pooling with a step size of 2 to downsample the feature maps output by the convolution layer, retaining the maximum value in each 2-step window (such as retaining the peak value of the ratio change), outputting 32 feature maps (599x1). The second pooling layer uses 2x1 max pooling with a step size of 2 to further simplify the data, outputting 32 feature maps (300x1).

[0114] For the generation of gradient distribution data, the outputs of the convolution layer and the pooling layer are integrated through a fully connected layer to generate gradient distribution data of the medial-lateral pressure ratio, i.e., the rate of change of the pressure ratio from the medial side to the lateral side at each time step.

[0115] In step 530, the pressure change rate data and joint angular velocity data at each time step are calculated based on the pressure distribution data.

[0116] In step 540, the pressure change rate data and joint angular velocity data are input into the network structure containing the recurrent layer in the convolutional neural network model, and the phase relationship data is output by capturing the phase relationship between the pressure change rate data and the joint angular velocity data through the recurrent layer.

[0117] The recurrent layer can use an LSTM (Long Short-Term Memory Network) structure with 64 memory units to capture the long-term dependence of time series data (the sequential relationship between pressure change rate and angular velocity). The pressure change rate data (such as a 1200x1 time sequence matrix) and the joint angular velocity data (such as a 1200x1 time sequence matrix) are aligned by time step to form a two-dimensional input matrix (such as a 1200x2 time sequence matrix), which is input into the LSTM layer.

[0118] The LSTM stores historical information (such as the angular velocity at t-10 steps) through the memory unit, and compares the current pressure change rate with the historical angular velocity to determine the synchronization. The time step corresponding to the peak of the pressure change rate (t_p) and the time step corresponding to the peak of the angular velocity (t_θ) are identified, and the phase difference φ=(t_p-t_θ)×0.1° (step corresponding to angle) is calculated.

[0119] The cycle layer outputs the phase difference array of the full cycle, i.e., the phase relationship data, reflecting the pressure-motion synchronization of each key movement stage (such as accelerated flexion, decelerated flexion).

[0120] The execution order between the above steps 510-520 and steps 530-540 can not be limited, for example, steps 510-520 and steps 530-540 can be executed in parallel.

[0121] In an embodiment, the above method further includes a model training process for the convolutional neural network, which includes: obtaining a pressure distribution training data set, preprocessing the pressure distribution training data in the pressure distribution training data set to form a preprocessed data set conforming to a standard normal distribution; iteratively training the preprocessed data set on the pre-trained model, calculating the loss value based on the feature vector output by each iteration, adjusting the parameters in the pre-trained model based on the calculated loss value, and ending the iteration training when the number of iterations reaches a preset number threshold or the latest loss value is less than a preset loss threshold, and outputting the trained convolutional neural network model; wherein the process of each iteration includes sequentially performing tensor processing, convolution processing, normalization processing, activation processing, and pooling processing on the preprocessed data set.

[0122] In this embodiment, the pressure distribution training data set contains multiple pressure distribution training data, which can be the pressure distribution collected by the array sensor of the knee shape during the historical knee replacement surgery.

[0123] The preprocessing process mainly includes data cleaning and data standardization. Data cleaning includes checking whether there are missing values and abnormal values in the data. For missing values, interpolation filling can be performed using the pressure values at adjacent time points or similar movement angles. For abnormal values, they can be identified and removed according to the 3σ principle in statistics. The 3σ range almost contains almost all (99.73%) data. Therefore, data outside the [μ-3σ, μ+3σ] interval in the pressure distribution training data set is considered to be an extreme value or an abnormal value, and is removed.

[0124] The pressure data and movement trajectory data are standardized to convert them into standard normal distribution data with a mean of 0 and a standard deviation of 1. This helps to improve the efficiency and stability of subsequent model training.

[0125] The first layer in the convolutional neural network model is an input layer, in which the training data containing pressure values and motion trajectory information is arranged in a tensor form according to the characteristics of the preprocessed training data.

[0126] The second to twelfth layers in the convolutional neural network model are a convolutional layer group, which specifically includes four convolutional layers, three activation layers and four batch normalization layers. In the convolutional layer group, the local features in the data, such as specific patterns of pressure distribution and local trends of motion trajectory, are automatically extracted by sliding the convolutional kernel on the input data and performing element multiplication and summation operations, thereby reducing the workload of manually extracting features and better capturing the internal structure of the data.

[0127] The convolutional layers are distributed in the second, fifth, eighth and eleventh layers. In the second layer, a 3x3 convolutional kernel is used, 16 convolutional kernels are set, and the local features are extracted from the input data by convolution operation; then a batch normalization layer is added to normalize the convolution output. In the fifth layer, the number of convolutional kernels is adjusted to 32, the size of the convolutional kernel is kept at 3x3, and the convolution operation is continued to extract more complex features; similarly, a batch normalization layer is added. In the eighth layer, the number of convolutional kernels is increased to 64, and the size of the convolutional kernel is changed to 5x5, further extracting deep-level features; followed by a batch normalization layer. In the eleventh layer, the number of convolutional kernels is increased to 128, and the 5x5 convolutional kernel is used to deepen the feature extraction; finally, a batch normalization layer is added to accelerate the training convergence and reduce the gradient problem. Finally, the feature tensor processed by convolution and batch normalization is output.

[0128] An activation layer group is arranged immediately after each batch normalization layer, and the activation function used in the activation layer group can be one or more of ReLU, Sigmoid, Tanh, etc.

[0129] The thirteenth and fourteenth layers of the convolutional neural network model are pooling layers. In the thirteenth layer, a 2x2 maximum pooling window is used with a step size of 2 to downsample the activation output data of the thirteenth layer. In the fourteenth layer, a 2x2 maximum pooling window is also used with a step size of 2 to further downsample the data after the thirteenth layer of pooling, thereby reducing the data dimension, reducing the calculation amount and preserving the main features.

[0130] The fifteenth layer of the convolutional neural network model is a fully connected layer, which sets the number of neurons according to the task requirements and model complexity, integrates the features extracted by the previous convolution and pooling (or recurrent layer), and outputs the integrated feature vector.

[0131] The fifteenth layer of the convolutional neural network model is an output layer, which designs the output layer structure according to the feature extraction target and outputs the feature vector result related to the corresponding final task.

[0132] After the network architecture of the convolutional neural network model is built, the parameters in the network are initialized. The initialization process can use Xavier initialization or Kaiming initialization to assign appropriate initial values to the parameters such as the weights of the convolutional layer and the weights of the fully connected layer, avoiding problems such as gradient vanishing or gradient explosion in the training process of the network, and ensuring the normal training and convergence of the network.

[0133] During the data training process, the pressure distribution training data set is divided into a training set, a validation set, and a test set. Generally, it is divided according to a certain proportion, for example, according to a 7:1:2 ratio, 70% of the pressure distribution training data is used to train the model, 10% of the pressure distribution training data is used to verify the performance of the model during the training process, and adjust the hyperparameters, and 20% of the pressure distribution training data is used to evaluate the generalization ability and accuracy of the model.

[0134] For the loss function, the Mean Squared Error (MSE) loss function can be used to calculate the error between the predicted feature value of the model and the true feature value, and the model parameters are optimized by minimizing the error. The calculation formula of the loss function MES is as follows:

[0135] Formula 4

[0136] wherein, is the number of samples, is the true feature value, is the predicted feature value of the model.

[0137] The related parameters in the convolutional neural network model can be updated using the Stochastic Gradient Descent (SGD) optimizer, Adam optimizer, etc. Through the optimizer, the learning rate can be automatically adjusted during the training process, and the convergence speed of the model can be accelerated. Taking the Adam optimizer selected in the model training as an example, the initial learning rate is set to 0.001, and as the training progresses, the learning rate can be adjusted according to the performance changes of the validation set.

[0138] During the model training process, the divided training set data is input into the built convolutional neural network model, and training is performed in batches. In each batch training, the output of the model is calculated by forward propagation, the loss value is calculated according to the defined loss function, and then the gradient is calculated by the back propagation algorithm, and the parameters of the network are updated using the optimizer. During the training process, the performance of the model is evaluated regularly using the validation set data, and the loss value and related indicators (such as the accuracy of feature extraction) on the validation set are recorded. According to the performance change of the validation set, the hyperparameters of the network are adjusted, such as the learning rate, the number of convolution kernels, the number of network layers, etc., to prevent the model from overfitting or underfitting. Continue to train the model until the performance on the validation set reaches a stable state or meets the pre-set training stopping conditions, which can be one or more of the following: the number of training rounds reaches a certain value, the loss value on the validation set no longer decreases significantly, etc.

[0139] In one embodiment, step 150 includes: establishing a transfer function of the osteotomy angle adjustment amount and the pressure distribution change amount, the pressure distribution change amount being calculated based on the offset amount of the pressure center trajectory and the gradient change of the shear stress distribution data; generating an optimization suggestion vector of the osteotomy angle adjustment according to the transfer function and the normal range of the dynamic pressure envelope.

[0140] In this embodiment, the osteotomy angle adjustment amount refers to the correction value (unit: degrees, °) of the femoral or tibial osteotomy surface angle during the operation, which is used to adjust the installation and positioning of the prosthesis (such as the femoral distal external rotation angle, the tibial platform retroversion angle). The pressure distribution change amount refers to the change amount (unit: MPa or percentage) of the interarticular compartment pressure distribution after the osteotomy angle adjustment, reflecting the improvement degree of pressure uniformity.

[0141] The transfer function refers to a mathematical model describing the quantitative relationship between the osteotomy angle adjustment amount (denoted as Δθ) and the pressure distribution change amount (denoted as ΔP), which is usually expressed as ΔP=f(Δθ). The optimization suggestion vector refers to a multi-dimensional suggestion data containing specific osteotomy angle adjustment value and expected pressure improvement effect generated based on the transfer function and the normal range of the dynamic pressure envelope. The parameters in the optimization suggestion vector can include adjustment amount, expected effect and target range, for example, the optimization suggestion vector is [Δθ=+1.7°, expected ΔP=-1.2MPa, target pressure=5.0MPa].

[0142] The pressure distribution change amount ΔP can be calculated based on the offset amount of the pressure center trajectory (denoted as Δx) and the gradient change of the shear stress distribution data (denoted as ∇τ), for example, the two can be weighted and summed to obtain the pressure distribution change amount. For example, ΔP=k1×|Δx|+k2×∇τ, where k1 and k2 are weight coefficients, which can be set according to relevant experience.

[0143] The transfer function can be obtained by one or a combination of experimental measurement, finite element analysis, and reinforcement learning.

[0144] For experimental measurement, the transfer function can be fitted by measuring the corresponding pressure distribution changes under different osteotomy angles on a bone model or a living body. For finite element analysis, a three-dimensional model of the knee joint can be constructed using computer simulation software (such as ANSYS, Abaqus) to simulate the mechanical response under different osteotomy angles and obtain the relationship between Δθ and ΔP, thereby obtaining the transfer function. For reinforcement learning, the goal is to learn the optimal strategy by maximizing the cumulative reward.

[0145] For example, by adjusting the external rotation angle of 100 samples, recording Δθ (-5° to +5°) and the corresponding ΔP, and fitting the transfer function by linear regression: ΔP = -3.33 x Δθ, the negative sign indicates that the medial pressure decreases as the external rotation angle increases. For example, for every 1° increase in osteotomy external rotation angle, the medial pressure of the knee joint will decrease by 3.33%. If the osteotomy angle is adjusted to Δθ = +3°, the pressure distribution change ΔP can be calculated according to the transfer function as -10%, i.e. the medial pressure of the knee joint decreases by 10%.

[0146] In one embodiment, the transfer function between the osteotomy angle adjustment and the pressure distribution change is established, including: training an agent through reinforcement learning, taking the osteotomy angle adjustment as the action input, taking the pressure distribution change and the feedback signal obtained after executing the action as the reward feedback, and iteratively optimizing to obtain the mapping relationship, and converting the mapping relationship into the transfer function.

[0147] In this embodiment, reinforcement learning is a machine learning method that learns the optimal strategy through continuous interaction between the agent and the environment in a "trial and error" manner, and the goal is to find the best mapping from state to action by maximizing the cumulative reward. The agent refers to an algorithm entity that performs actions, receives environmental feedback, and optimizes strategies in reinforcement learning, responsible for deciding the osteotomy angle adjustment and updating the behavior pattern according to the feedback. The action input refers to the operation that the agent can perform in a specific state, which in this scheme is the osteotomy angle adjustment (Δθ), including the adjustment direction (internal rotation / external rotation, anteversion / posterior tilt) and the specific angle value. The reward feedback refers to the evaluation signal (numerical value) of the environment for the agent's action, which is determined by the pressure distribution change and surgical trauma, and positive reward encourages effective actions, while negative reward suppresses ineffective or harmful actions. The mapping relationship refers to the "state-action" correspondence learned by the agent after reinforcement learning, i.e. given the current pressure distribution state (such as pressure center offset, shear stress), the function relationship of the optimal osteotomy angle adjustment is output.

[0148] The core elements of reinforcement learning include state, action, reward, and policy.

[0149] State represents the environment state that the agent is in, such as the current knee joint pressure distribution, bone geometry parameters, patient weight, etc. The state contains the current key biomechanical parameters of the knee joint, the offset of the pressure center trajectory (such as Δx medial / lateral, Δy anterior / posterior), the gradient maximum of the shear stress distribution data, and the deviation of the current pressure distribution from the dynamic pressure envelope. For example, a state S=(Δx=2mm, Δy=0.5mm,▽τ_max=0.6MPa / mm, ΔP_total=1.8MPa) represents a 2mm medial offset, high shear stress gradient, and overall pressure exceeding the normal range by 1.8MPa.

[0150] Action represents the operations that the agent can perform in the current state, such as different osteotomy angle adjustment schemes (e.g. external rotation +3°, internal rotation -2°); reward represents the feedback signal obtained after performing the action, such as the improvement of pressure distribution (e.g. 10% reduction in medial pressure reward +10 points), and the size of surgical trauma (e.g. 1% increase in trauma, then penalty -5 points); policy represents the mapping relationship from state to action, and the best osteotomy angle adjustment scheme is selected according to the current state.

[0151] By constructing a knee replacement surgery digital simulation environment, inputting patient knee three-dimensional model and pressure distribution training data, simulating pressure changes under different osteotomy angles (e.g. 3° external rotation, 1.5MPa reduction in medial pressure), and obtaining the corresponding transfer function.

[0152] The agent iterative training process is as follows: the agent randomly selects an action (e.g. Δθ=+2°), the environment feedbacks the pressure change (e.g. ΔP=-1.0MPa) and the reward (R=-22), the agent records "state S-action A-reward R", and adjusts the policy (reduces the selection probability of the action).

[0153] When the average reward of consecutive multiple iterations stabilizes within the preset range, the policy is determined to be converged, and the training stops.

[0154] After the iteration is completed, the mapping relationship can be extracted and converted into a transfer function. The "state-optimal action" mapping relationship can be extracted from the trained agent policy, for example:

[0155] When the state S=(Δx=2mm,▽τ_max=0.6MPa / mm) is optimal, the action A=+3°;

[0156] When the state S = (Δx = 1mm,▽τ_max = 0.3MPa / mm), the optimal action A = +1.5°.

[0157] Transfer function transformation: the transfer function is obtained by curve fitting of Δθ and corresponding ΔP in the mapping relationship; for example, the transfer function can be fitted as: ΔP = -3.33 x Δθ.

[0158] In this embodiment, the transfer function is generated by reinforcement learning, which can improve the accuracy of the osteotomy angle adjustment suggestion, and provides key technical support for the intelligent optimization of knee replacement surgery.

[0159] In one embodiment, a computer readable storage medium is provided, which stores executable instructions, and the instructions, when executed by a processor, cause the processor to perform the steps in the above method embodiments.

[0160] In one embodiment, an electronic device is also provided, which includes one or more processors; and a memory, the memory storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors perform the steps in the above method embodiments.

[0161] In one embodiment, as shown in FIG. 6, Figure 6 The electronic device includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 602 or programs loaded from a storage portion 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The CPU 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0162] The following components are connected to the I / O interface 605: an input portion 606 including a keyboard, a mouse, and the like; an output portion 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), and the like, and a speaker, and the like; a storage portion 608 including a hard disk, and the like; and a communication portion 609 including a network interface card such as a LAN card, a modem, and the like. The communication portion 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as necessary. A removable medium 611 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like is attached to the drive 610 as necessary, so that a computer program read therefrom is installed in the storage portion 608 as necessary.

[0163] In particular, the processes described above with reference to the flow charts can be implemented as a computer software program in accordance with embodiments of the present application. For example, embodiments of the present application include a computer program product which includes a computer readable medium bearing instmctions which, in such an embodiment, can be downloaded and installed from a network via the communication portion 609 and / or installed from the removable media 611. When the instructions are executed by the central processing unit (CPU) 601, the various method steps described in the present application are performed.

[0164] Finally, it should be noted that the above-described embodiments are merely intended to illustrate the technical solutions of the present application, rather than limit the scope of the present application; although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions recorded in the above embodiments can be modified or equivalent replacements can be made to some or all of the technical features; and such modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0165] Furthermore, those skilled in the art will appreciate that a combination of features of different embodiments can mean that a combination of features is within the scope of the present application and forms a different embodiment. For example, all of the above embodiments can be used in any combination. The information disclosed in this Background section is only intended to enhance the understanding of the general background of the present application and should not be taken as an acknowledgement or any form of suggestion that this information forms the prior art that is already known to those skilled in the art.

Claims

1. A data processing method for knee replacement surgery, characterized in that, The method includes: Acquire pressure distribution data of the interarticular compartments throughout the entire joint movement cycle, measured by an array sensor; Based on the pressure distribution data, the pressure center at each degree of flexion during the entire cycle of joint movement is calculated, and the corresponding pressure center trajectory is formed. Based on the pressure center trajectory data, the shear stress distribution data under the entire cycle of joint movement is calculated; The dynamic pressure envelope of the joint movement throughout the entire cycle is calculated based on the pressure distribution data. Intraoperative correction suggestions for knee arthroplasty are generated based on the pressure center trajectory, shear stress distribution data, and dynamic pressure envelope.

2. The method according to claim 1, characterized in that, The calculation of the dynamic pressure envelope throughout the entire joint movement cycle based on the pressure distribution data includes: A preset convolutional neural network model is invoked to extract spatial and temporal features from the pressure distribution data, thereby obtaining gradient distribution data of the inner and outer pressure ratio and phase relationship data between the pressure change rate and the joint angular velocity. The dynamic pressure envelope is generated based on the gradient distribution data and the phase relationship data.

3. The method according to claim 2, characterized in that, The process involves using a pre-defined convolutional neural network model to extract spatial and temporal features from the pressure distribution data, obtaining gradient distribution data of the inner and outer pressure ratio and phase relationship data between the pressure change rate and joint angular velocity, including: Based on the pressure distribution data, the pressure ratio between the inner and outer sides of the knee joint at each time step is calculated to obtain the inner and outer side pressure ratio data. The inner and outer pressure ratio data is input into the convolutional neural network model, and the spatial relationship in the inner and outer pressure ratio data is captured by the convolutional layer and pooling layer in the convolutional neural network model to obtain the gradient distribution data; Based on the pressure distribution data, the pressure change rate data and joint angular velocity data at each time step are calculated; The pressure change rate data and the joint angular velocity data are used as inputs to the network structure containing recurrent layers in the convolutional neural network model. The recurrent layers capture the phase relationship between the pressure change rate data and the joint angular velocity data, and output the phase relationship data.

4. The method according to claim 2, characterized in that, The method further includes: Obtain a set of training data on stress distribution, and preprocess the training data on stress distribution in the set of training data to form a preprocessed dataset that conforms to a standard normal distribution. The pre-processed dataset is used to iteratively train the pre-trained model. The loss value is calculated based on the feature vector output by each iteration. The parameters in the pre-trained model are adjusted based on the calculated loss value until the number of iterations reaches a preset threshold or the latest loss value is less than the preset loss threshold. Then the iterative training ends and the trained convolutional neural network model is output. Each iteration of the training process includes sequentially performing tensor processing, convolution processing, normalization processing, activation processing, and pooling processing on the preprocessed dataset.

5. The method according to claim 1, characterized in that, The pressure distribution data includes the pressure values ​​measured by each sensing unit in the array sensor at various times; the calculation of the pressure center at each flexion degree throughout the entire joint movement cycle based on the pressure distribution data, forming the corresponding pressure center trajectory, includes: The pressure center at the corresponding moment is calculated based on the pressure value of each sensing unit at the same time and the position coordinates of the corresponding sensing unit. The pressure center trajectory is formed based on the pressure center at each moment.

6. The method according to claim 1, characterized in that, The calculation of shear stress distribution data throughout the entire joint motion cycle based on the pressure center trajectory data includes: The height of the point of application of the friction force is determined based on the elevation of the surface friction layer of the array sensor; The tangential movement distance is calculated based on the trajectory of the pressure center. The shear stress at the corresponding moment is calculated based on the height of the point of application, the distance traveled, and the pressure center at the corresponding moment.

7. The method according to any one of claims 1 to 6, characterized in that, The intraoperative correction suggestions for knee arthroplasty generated based on the pressure center trajectory, shear stress distribution data, and dynamic pressure envelope include: A transfer function is established between the osteotomy angle adjustment and the pressure distribution change, wherein the pressure distribution change is calculated based on the offset of the pressure center trajectory and the gradient change of the shear stress distribution data; Based on the transfer function and the normal range of the dynamic pressure envelope, an optimized suggestion vector for adjusting the osteotomy angle is generated.

8. The method according to claim 7, characterized in that, The establishment of the transfer function between the osteotomy angle adjustment and the pressure distribution change includes: The agent is trained through reinforcement learning, with the osteotomy angle adjustment as the action input and the pressure distribution change and the feedback signal obtained after the action as the reward feedback. The mapping relationship is obtained through iterative optimization and then transformed into the transfer function.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores executable instructions that, when executed by a processor, cause the processor to perform the method as described in any one of claims 1 to 8.

10. An electronic device, characterized in that, include: One or more processors; A memory for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to perform the method as described in any one of claims 1 to 8.

Images