CN-121997672-A - Deformation error compensation method for spreader in total knee replacement operation

CN121997672ACN 121997672 ACN121997672 ACN 121997672ACN-121997672-A

Abstract

The invention discloses a deformation error compensation method of a spreader for total knee replacement operation, and aims to solve the problem of measurement error caused by cantilever structure deformation. Firstly, carrying out parameterized dynamics modeling and simulation on an external execution mechanism through finite element software, calibrating a model by combining physical prototype actual measurement data, and constructing a high-precision deformation data set. And then, training a BP neural network optimized by a genetic algorithm by using the data set, wherein the BP neural network takes the force and displacement measured values acquired in real time as input and outputs the compensated real displacement value. The trained lightweight network model is finally deployed on an embedded processing unit of the spreader, so that real-time error compensation is realized. The invention effectively compensates complex nonlinear deformation errors, the average absolute error is reduced to 0.056mm, the real-time response delay is less than or equal to 5ms, the method is obviously superior to the traditional linear method, more accurate clearance measurement is provided for the operation, the method can be further used for forming closed-loop control, and the intellectualization and the accuracy of the operation are improved.

Inventors

ZHANG ENKUI
ZHANG KAI
JIN YU
HAN TAILIN
ZHU CUNXING
Fan Boxuan
XU HE
JU MINGCHI
LIU XUAN
WANG YINGZHI
GUO ZIXI

Assignees

长春理工大学
吉林雷普光电科技开发有限公司

Dates

Publication Date: 20260508
Application Date: 20260212

Claims (8)

1. The deformation error compensation method of the spreader for the total knee replacement operation is characterized by comprising the following steps of: s1, carrying out dynamic modeling and simulation on an external execution mechanism of the spreader for the total knee replacement operation through finite element software to obtain simulated displacement amounts under different cantilever lengths and different applied forces; S2, dividing a data set obtained by the finite element simulation after calibration into a training set and a testing set, and carrying out normalization processing on the data; s3, constructing a BP neural network as an error compensation model, wherein the BP neural network is input by a force application and spreader displacement measured value and output as a compensated real displacement value; S4, training the BP neural network model by using the training set, wherein the training process comprises forward propagation calculation of a predicted value, error calculation through a loss function, backward propagation updating of network parameters, and iteration until the error meets the requirement; and S5, deploying the trained error compensation model to an embedded processing unit of the spreader, and performing error compensation on the sensor data acquired in real time to obtain a corrected displacement value.
2. The method according to claim 1, wherein the step S1 comprises: S101, establishing a model of the external execution mechanism in finite element software, and carrying out parameterization simulation by taking the cantilever length L 1 and the applied force F as variables, wherein the range of F is 1N to 100N, the interval 1N, the range of L 1 is 1mm to 20mm, and the interval 1mm, so as to obtain a simulation displacement data set; S102, carrying out actual measurement on a physical prototype under the selected cantilever length and the applied force to obtain an actual measurement displacement and calculating an actual measurement deformation value; and S103, if the relative error is more than 5%, adjusting the material parameters of the simulation model for calibration, and repeating the steps S101 and S102 until the relative error is less than 5%.
3. The method according to claim 2, wherein the measured deformation value δz is represented by the formula Calculation of wherein For the theoretical displacement to be a function of the theoretical displacement, Is the actual displacement, the relative error RE passes through the formula And (5) calculating.
4. The method according to claim 1, wherein in the step S2, the formula for normalizing the data is: ; Wherein, the For the data to be normalized, And Respectively minimum and maximum values in the dataset, The normalized result is obtained.
5. The method according to claim 1, wherein in the step S3, the constructing of the BP neural network includes: s301, setting a network topology structure, wherein an input layer is 2 neurons and corresponds to a force application and a displacement measurement value respectively; s302, adopting a Tanh function as an activation function, wherein the specific formula is as follows: ;; Wherein, the Is the input of the function, and the output range of the function is fixed between (-1, 1); s303, initializing network weight by adopting an Xavier method, wherein the specific formula is as follows: ; Regularization is carried out by adopting a Dropout method; S304, adopting a variance cost function as a loss function, wherein the function expression is as follows: ; Wherein the method comprises the steps of Is an output that is expected to be present, For the actual output of the neuron, Is a function value; and S305, optimizing the super parameters of the BP neural network by adopting a genetic algorithm, wherein the super parameters comprise a hidden layer structure, a learning rate and regularization strength.
6. The method according to claim 5, wherein the genetic algorithm optimization step comprises encoding the super parameters to generate an initial population, evaluating individuals with the performance on the test set as fitness, iteratively optimizing by selecting, crossing, and mutating operations to finally obtain an optimal super parameter combination and constructing the BP neural network.
7. The method according to claim 1, wherein the step S5 specifically includes: S501, loading the trained error compensation model to the embedded processing unit, and calibrating a displacement sensor and a pressure sensor; S502, acquiring displacement measurement values and applied forces in real time, and carrying out normalization processing; S503, inputting the normalized data into the error compensation model to obtain a normalized compensation displacement value, and performing inverse normalization to obtain a final true displacement value; S504, outputting the real displacement value to a display interface.
8. The method of claim 7, further comprising the step of generating a control signal to adjust a drive motor of the distractor to form a closed loop control based on the deviation of the true displacement value from a target clearance.

Description

Deformation error compensation method for spreader in total knee replacement operation Technical Field The invention relates to the technical field of data processing, in particular to a deformation error compensation method of a spreader for total knee replacement operation. Background Total knee arthroplasty is a technique for replacing a damaged joint with an artificial prosthesis, and is commonly used for treating arthritis, trauma, rheumatoid and other diseases. In the treatment process, a rectangular joint gap is required to be kept when the knee joint is bent and straightened, two main osteotomy techniques can achieve the aim, namely an equivalent osteotomy method is measured, wherein the osteotomy position and the osteotomy amount are determined according to the anatomical structure of the femur and the tibia, and a gap balance osteotomy method is needed to be unfolded at a certain distance when the knee joint is straightened, so that the soft tissue is balanced and then subjected to osteotomy, the knee joint is unfolded with the same unfolding force, and the femur condyles are subjected to osteotomy after being unfolded so as to achieve the balance of the soft tissue tension at the straightening position and the buckling position. Along with the progress of the technology, a novel spreader for total knee joint replacement operation is proposed in patent CN 114847887a, which consists of an upper bottom plate 1, an extension rod 2, a flange 3, a motion mechanism 4, a motor 5 and a displacement sensor as shown in fig. 1, wherein for the subsequent study, the combination of the upper bottom plate and the extension rod beyond the flange is called an external actuator, and the simplified model and the stress analysis of the knee joint spreader are shown in fig. 2 and 3. The following formula is obtained by mechanical analysis: Wherein, the Is the deformation of the backsheet on the spreader, F is the applied force,Is the length of the upper bottom plate,Is the modulus of elasticity of the upper backsheet,Is the section moment of inertia of the upper bottom plate; It is the length of the movement mechanism beyond the flange, Is the modulus of elasticity thereof, and the elastic modulus,Is its cross-sectional moment of inertia. From the above formula, the deformation of the upper plate of the spreader is mainly affected by the applied force and the length of the movement mechanism beyond the flange, and the effect is nonlinear. Patent CN117442396a mentions a linear regression error compensation method, but this method has the following technical drawbacks: 1. the model assumption is not consistent with the actual structure. The technology establishes corresponding spreader stress and deformation models under different force values by utilizing the assumption that stress values are in an elastic deformation range, does not consider that an upper bottom plate and a long rod structure are not in rigid connection, and because a gap is arranged between a flange and a motion mechanism, the motion mechanism is slightly inclined under the action of a bending moment, so that the downward displacement of a stress end of the upper bottom plate is larger, and the gap parameter measurement of the total knee joint replacement operation is inaccurate. 2. Deformation influencing factors are not fully considered. The technology only considers the deformation of the actuating mechanism and the stress value, but does not consider the influence of the change of the length of the long rod structure of the spreader in the use process of the total knee replacement operation. Disclosure of Invention The technical scheme for solving the technical problems is that the invention provides a deformation error compensation method of a spreader for total knee replacement surgery, which comprises the following steps: S1, carrying out dynamic modeling and simulation on an external execution mechanism of the spreader for the total knee replacement operation through finite element software to obtain simulation displacement amounts under different cantilever lengths and different applied forces; S2, dividing a data set obtained by the finite element simulation after calibration into a training set and a testing set, and carrying out normalization processing on the data; s3, constructing a BP neural network as an error compensation model, wherein the BP neural network is input by a force application and spreader displacement measured value and output as a compensated real displacement value; s4, training the BP neural network model by using the training set, wherein the training process comprises forward propagation calculation of a predicted value, error calculation through a loss function, backward propagation updating of network parameters, and iteration until the error meets the requirement; and S5, deploying the trained error compensation model to an embedded processing unit of the spreader, and performing error compensation on the sens