Method for analyzing influence of suture boundary on artificial cornea optical imaging

Abstract

The invention discloses a method for analyzing influence of a suture boundary on keratoprosthesis optical imaging, which adopts a method of combining human eye finite element model simulation and human eye optical model simulation, analyzes the effect of the suture boundary on the keratoprosthesis through a finite element model, and can obtain stress response and configuration change of the keratoprosthesis in real time with high efficiency and low cost. And the final imaging effect of the operation can be simulated in real time by combining a human eye optical model. The mechanical-optical double-path hybrid network of the artificial cornea is established, personalized parameter adjustment is provided for an operation according to the postoperative requirements of a patient, the operation effect is optimized, and the postoperative visual recovery quality and precision of the patient can be finally improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial cornea stress and configuration analysis, in particular to an analysis method for suture boundary affecting optical imaging of artificial cornea. BACKGROUND

[0002] Corneal blindness is a serious visual impairment caused by corneal opacity, and corneal transplantation technology is the only means to treat corneal blindness. Suture is the most important skill for corneal surgeons, and the suture boundary of the cornea is affected by the penetration depth of the suture line and the suture length of the suture line. Effective suture placement can also reduce postoperative astigmatism and maintain long-term stability.

[0003] The existing artificial cornea research has the following problems:

[0004] 1. The existing suture research of artificial cornea ignores the dynamic evolution of corneal stress and configuration over time;

[0005] 2. The suture process of the existing artificial cornea lacks standards, and the tightness of the suture line is determined only by the experience and feel of the doctor;

[0006] 3. The existing non-contact artificial cornea measurement method such as corneal topography can only measure the configuration of the upper surface of the artificial cornea, while the imaging of the artificial cornea is related to the configuration of the front and back surfaces of the artificial cornea;

[0007] 4. Postoperative imaging is one of the evaluation indicators of the effect of artificial cornea transplantation surgery, and the existing research cannot predict postoperative imaging to make individual adjustments to intraoperative suturing.

[0008] In summary, the suture boundary has a complex effect on the imaging of the artificial cornea, but the existing artificial cornea stress and configuration research does not consider the effect of the suture boundary, and a mechanical-optical dual-path hybrid network combining finite element and optical simulation is needed to meet the research needs. SUMMARY

[0009] The present application aims to overcome the shortcomings and deficiencies of the prior art, and provides an analysis method for suture boundary affecting optical imaging of artificial cornea, which combines human eye finite element model simulation and human eye optical model simulation, analyzes the effect of suture boundary parameters on artificial cornea through human eye finite element model, can obtain the stress response and configuration change of artificial cornea in real time, efficiently and at low cost, combines human eye optical model simulation, can simulate the final imaging effect of the surgery in real time, builds a mechanical-optical dual-path hybrid network of artificial cornea, provides individualized parameter adjustment for the surgery according to the postoperative needs of the patient, optimizes the surgical effect, and ultimately improves the quality and precision of postoperative visual recovery of the patient.

[0010] To achieve the above object, the technical scheme provided by the present application is as follows: an analysis method for suture boundary affecting optical imaging of artificial cornea, comprising the following steps:

[0011] S1: design suture boundary parameters, establish a finite element model of human eye containing sutures, and perform stress and configuration analysis to obtain stress distribution, strain distribution and configuration change data of the finite element model of human eye;

[0012] S2: extract the inner surface node coordinates and outer surface node coordinates of the artificial cornea in the finite element model of human eye, respectively perform surface fitting on the inner surface node coordinates and outer surface node coordinates to obtain the inner surface fitting surface and the outer surface fitting surface of the artificial cornea; based on the inner surface fitting surface and the outer surface fitting surface of the artificial cornea, establish an optical model of human eye to obtain optical imaging data of the optical model of human eye corresponding to the finite element model of human eye under the design suture boundary parameters;

[0013] S3: adjust the suture boundary parameters and the structure and mechanical parameters of the artificial cornea, repeat steps S1 and S2 to obtain the stress distribution, strain distribution, configuration change data of the finite element model of human eye and the optical imaging data of the optical model of human eye under different suture boundary parameters and structure and mechanical parameters of the artificial cornea, and establish a suture boundary parameter and optical imaging database;

[0014] S4: based on the suture boundary parameter and optical imaging database, build and train an artificial cornea mechanical-optical dual-path hybrid network, the artificial cornea mechanical-optical dual-path hybrid network is composed of a forward prediction module and a reverse optimization module, and works collaboratively through a weight sharing layer; the forward prediction module adopts multi-scale feature fusion and physical constraint design, inputs the structure and mechanical parameters of the artificial cornea and the suture boundary parameters into an input layer, transfers the preprocessed data of the input layer to a parameter embedding layer, transmits the fusion feature vector output by the parameter embedding layer to a physical constraint convolution layer, and finally transmits the data output by the physical constraint convolution layer to a mechanical feature head and an optical feature head, maps the data into mechanical features and optical indexes, the mechanical features include stress distribution, strain distribution and configuration change data of the finite element model of human eye, and the optical indexes are optical imaging data of the optical model of human eye; the reverse optimization module is constructed based on a conditional generative adversarial network, and includes a generator, a discriminator and a physical constraint layer, the reverse optimization module accepts target optical indexes and structure parameters of the artificial cornea, transmits them to the generator, outputs preliminary suture boundary parameters, then transmits them to the discriminator to evaluate clinical rationality, and transmits the suture boundary parameters that pass the verification to the conditional generative adversarial training to output optimal suture boundary parameters;

[0015] S5: Based on the required imaging needs and personalized human eye parameters, input the optical indicators and artificial corneal structural parameters into the trained artificial corneal mechanical-optical dual-path hybrid network to obtain the optimal suture boundary parameters for artificial corneal transplantation surgery.

[0016] Furthermore, in step S1, the finite element model of the human eye includes a simplified eyeball model and a scratching plate model simulating the eyelid. The simplified eyeball model includes a corneal implant bed, sutures, artificial cornea, limbus, and sclera. The scratching plate model simulating the eyelid has a plate curvature consistent with the outer surface of the cornea, and its edges are chamfered.

[0017] Furthermore, step S1 includes the following steps:

[0018] S11: Design the suture boundary parameters and establish a finite element model of the eyeball containing the suture using finite element modeling software;

[0019] S12: A shell-shaped flexible plate conforming to the shape of the cornea is created using finite element modeling software, and the boundary of the inner surface of the shell-shaped flexible plate in contact with the eyeball is chamfered.

[0020] S13: Based on the material parameters of the standard human eye and the material parameters of the artificial cornea, input the material parameters and contact friction coefficients of each part of the finite element model of the eyeball into the finite element modeling software;

[0021] S14: Controls the movement of the flexible plate to simulate the scraping process of the eyelid against the artificial cornea;

[0022] S15: Extract the node coordinates of the inner and outer surfaces of the artificial cornea, and perform polynomial fitting on the node coordinates of the inner and outer surfaces of the artificial cornea to obtain the fitted surfaces of the inner and outer surfaces of the artificial cornea.

[0023] Further, in step S2, the optical imaging data includes the refractive power of the artificial cornea, visual acuity level, and modulation transfer function (MTF) curve; the human eye optical model includes the inner surface of the artificial cornea, the outer surface of the artificial cornea, the inner surface of the lens, the virtual plane, the outer surface of the lens, and the retina.

[0024] Furthermore, step S2 includes the following steps:

[0025] S21: Establish a standard human eye optical model based on the Liou human eye optical model and standard human eye parameters;

[0026] S22: Replace the standard human eye optical model with the fitted surfaces of the inner and outer surfaces of the artificial cornea before and after scraping with the soft plate to obtain the required human eye optical model.

[0027] S23: Imaging simulation using optical software;

[0028] S24: Extract the visual acuity results and imaging effect diagrams of the human eye optical model before and after scratching.

[0029] Furthermore, in step S3, the suture boundary parameters include: suture penetration depth, suture length, number of sutures, and suture force; the artificial corneal structural parameters include: artificial corneal central thickness, anterior and posterior surface curvature, and transverse diameter; the artificial corneal mechanical parameters include: elastic modulus, Poisson's ratio, and relaxation curve of the artificial corneal material.

[0030] Furthermore, in step S4, the forward prediction module achieves optical performance prediction through multi-stage feature extraction and physical constraints, including the following steps:

[0031] S411: The structural parameters, suture boundary parameters, and time parameters of the artificial cornea are normalized and transformed into standardized feature vectors;

[0032] S412: The interaction weights between standardized feature vectors are calculated through a cross-attention mechanism. Based on the weights, each standardized feature vector is weighted and fused to generate a joint feature vector, while preserving the non-linear coupling relationship between parameters.

[0033] S413 passes the joint feature vector to a physically constrained convolutional layer, which is embedded with Hooke's law to ensure that the output conforms to the laws of mechanics.

[0034] S414: Passes the output of the physically constrained convolutional layer to the mechanical feature head, outputting the stress and strain distribution of the artificial cornea;

[0035] S415: Input the stress distribution and strain distribution into the optical feature head to predict optical performance and output the refractive power and modulation transfer function (MTF) curve of the artificial cornea.

[0036] Furthermore, in step S4, the inverse optimization module is responsible for generating optimal suture boundary parameters that meet clinical feasibility based on the given optical target and artificial corneal structure parameters. The core of the inverse optimization module is a conditional generative adversarial network (cGAN), which is combined with physical constraints for optimization, including the following steps:

[0037] S421: Input information, including optical target t, artificial corneal structural parameters c, and random noise z;

[0038] S422: The generator receives the optical target t and the artificial corneal structure parameter c, and generates the suture boundary parameter G(c,t,z);

[0039] S423: The discriminator evaluates the suture boundary parameter G(c,t,z) output by the generator to determine whether it is close to the real clinical data, and outputs a scalar D(G(c,t,z)|c,t)∈[0,1] to represent the probability of data authenticity;

[0040] S424: The suture boundary parameter G(c,t,z) passed through the discriminator is passed to the forward prediction module to obtain optical performance prediction data. The joint loss function result of the artificial corneal mechanical-optical dual-path hybrid network is judged. If it is greater than the set threshold, the joint loss function result is passed to the inverse optimization module to adjust the generator in reverse. Steps S422-S424 are repeated to perform joint iterative training of the inverse optimization module and the forward prediction module.

[0041] S425: When the decrease in the joint loss function is less than the set threshold, output the optimal stitching boundary parameters.

[0042] Furthermore, in step S4, during the training phase, a constraint loss is applied to the mechanical output using limited metadata from the human eye to ensure that the network prediction results follow mechanical laws and material properties. The joint loss function of the artificial cornea mechanical-optical dual-path hybrid network is:

[0043]

[0044] In the formula, For joint losses, For constitutive relation loss, For optical loss, α and β are the loss weights for constitutive loss and optical loss, respectively, where:

[0045]

[0046] In the formula, N node σ represents the total number of nodes, e represents the cell number of the artificial cornea mesh, and σ represents the total number of nodes. e Predict the stress tensor for the nodes, ε e The strain tensor is calculated from artificial corneal morphology data. For the stiffness matrix of artificial corneal materials, w is the square of the Frobenius norm. n The weight of the region where the node is located;

[0047]

[0048] In the formula, n is the equivalent refractive index of the cornea and aqueous humor, and R... eff D is the effective radius of curvature calculated from the predicted topography. pred κ is the diopter value directly output by the network, ΔR is the smoothing weight coefficient, and ΔR is the Laplace operator of the curvature field, used to characterize the degree of local fluctuation of curvature.

[0049] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0050] 1. This invention uses finite element simulation combined with a human eye optical model to simulate the effect of suture boundaries on artificial corneal imaging. Based on the adjustment of human eye structure and mechanical parameters, it can intuitively, efficiently, in real time and at low cost obtain the stress response and configuration changes of individual artificial corneas.

[0051] 2. This invention uses sutures to connect the artificial cornea to the corneal implant bed to simulate the effect of clinical single-needle intermittent suturing on the cornea. Individual sutures can be adjusted individually, and the suture boundary parameters can be flexibly adjusted according to the patient's needs.

[0052] 3. This invention constructs a hybrid network of mechanical and optical paths for artificial cornea, which can provide precise parameter adjustments for the surgery based on the final imaging effect, optimize the surgical effect, and ultimately improve the quality and accuracy of the patient's postoperative visual recovery. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the structure of a finite element model of a human eye including sutures.

[0054] Figure 2 This is a schematic diagram of the channel-suture connection model at the contact boundary between the artificial cornea and the corneal implant bed.

[0055] Figure 3 This is a schematic diagram of the optical model of the human eye.

[0056] Figure 4 This is a diagram of the architecture of a hybrid mechanical-optical network for artificial cornea.

[0057] Figure 5 This is the architecture diagram of the positive prediction module.

[0058] Figure 6 This is the architecture diagram of the reverse optimization module. Detailed Implementation

[0059] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0060] This embodiment discloses an analysis method for the impact of suture boundaries on the optical imaging of artificial corneas, the specific details of which are as follows:

[0061] Step S1: Design the suture boundary parameters, establish a finite element model of the human eye including the suture, and perform stress and configuration analysis on it to obtain the stress distribution, strain distribution and configuration change data of the human eye finite element model;

[0062] Imaging simulation of artificial corneas includes finite element simulation of the human eye and optical imaging simulation of the human eye, such as... Figure 1 and Figure 2 As shown, the simplified human eye model includes 1. the sclera, with the lower edge of 1. the sclera bound to the upper edge of 2. the limbus, and 3. the upper edge of the corneal implant bed bound to the lower edge of 2. The corneal implant bed and 4. the artificial cornea are connected by sutures. A coefficient of friction is set at the contact surface between the artificial cornea and the implant bed. The binding connection between 1, 2, and 3 seals the inside of the eyeball, and intraocular pressure is simulated by applying a fluid field. 5. A scraping soft plate simulating the eyelid conforms to the outer surface of the cornea, with chamfered edges.

[0063] Specifically, step S1 includes the following steps:

[0064] S11: Design the suture boundary parameters and establish a finite element model of the eyeball containing the suture using finite element modeling software;

[0065] S12: A shell-shaped flexible plate conforming to the shape of the cornea is created using finite element modeling software, and the boundary of the inner surface of the shell-shaped flexible plate in contact with the eyeball is chamfered.

[0066] S13: Based on the material parameters of the standard human eye and the material parameters of the artificial cornea, input the material parameters and contact friction coefficients of each part of the finite element model of the eyeball into the finite element modeling software;

[0067] S14: Controls the movement of the flexible plate to simulate the scraping process of the eyelid against the artificial cornea;

[0068] S15: Extract the node coordinates of the inner and outer surfaces of the artificial cornea, and perform polynomial fitting on the node coordinates of the inner and outer surfaces of the artificial cornea to obtain the fitted surfaces of the inner and outer surfaces of the artificial cornea.

[0069] Step S2: Extract the coordinates of the inner and outer surface nodes of the artificial cornea from the finite element model of the human eye. Perform surface fitting on the inner and outer surface node coordinates to obtain the inner and outer surface fitted surfaces of the artificial cornea. Based on the inner and outer surface fitted surfaces of the artificial cornea, establish the optical model of the human eye to obtain the optical imaging data of the human eye optical model corresponding to the finite element model of the human eye under the designed suture boundary parameters. The optical imaging data includes the refractive power, visual acuity, and modulation transfer function (MTF) curve of the artificial cornea.

[0070] like Figure 3 As shown, the human eye optical imaging simulation model includes the outer surface of the artificial cornea, the inner surface of the artificial cornea, the outer surface of the lens, a virtual plane, the inner surface of the lens, and the retina. A human eye optical simulation model with a normal cornea is then established.

[0071] Specifically, step S2 includes the following steps:

[0072] S21: Establish a standard human eye optical model based on the Liou human eye optical model and standard human eye parameters;

[0073] S22: Replace the standard human eye optical model with the fitted surfaces of the inner and outer surfaces of the artificial cornea before and after scraping with the soft plate to obtain the required human eye optical model.

[0074] S23: Imaging simulation using optical software;

[0075] S24: Extract the visual acuity results and imaging effect diagrams of the human eye optical model before and after scratching.

[0076] Step S3: Adjust the suture boundary parameters and the structural and mechanical parameters of the artificial cornea. Repeat steps S1 and S2 to obtain the stress distribution, strain distribution, configuration change data of the human eye finite element model and the optical imaging data of the human eye optical model under different suture boundary parameters and the structural and mechanical parameters of the artificial cornea. Establish a database of suture boundary parameters and optical imaging.

[0077] The suture boundary parameters include: suture penetration depth, suture length, number of sutures, and suture force; the artificial cornea structural parameters include: artificial cornea central thickness, anterior and posterior surface curvature, and transverse diameter; the artificial cornea mechanical parameters include: elastic modulus, Poisson's ratio, and relaxation curve of the artificial cornea material.

[0078] Step S4: Based on the suture boundary parameters and optical imaging database, construct and train a hybrid network of mechanical-optical dual paths for the artificial cornea; such as... Figure 4 As shown, the artificial corneal mechanical-optical dual-path hybrid network consists of a forward prediction module and a reverse optimization module, which work together through a weight sharing layer.

[0079] like Figure 5 As shown, the forward prediction module employs multi-scale feature fusion and physical constraint design. The module inputs the artificial corneal structure, mechanical parameters, and suture boundary parameters into the input layer, and then transfers the preprocessed data from the input layer to the parameter embedding layer. The fused feature vector output from the parameter embedding layer is then passed to the physical constraint convolutional layer. Finally, the data output from the physical constraint convolutional layer is passed to the mechanical feature head and optical feature head, mapping the data into mechanical features and optical indicators. The mechanical features include stress distribution, strain distribution, and configuration change data from the human eye finite element model, while the optical indicators are the optical imaging data from the human eye optical model.

[0080] like Figure 6As shown, the inverse optimization module is constructed based on a conditional generative adversarial network and includes a generator G, a discriminator, and a physical constraint layer. The inverse optimization module accepts the target optical parameters and the structural parameters of the artificial cornea, passes them to the generator G, and outputs the preliminary suture boundary parameters. Then, it passes them to the discriminator to evaluate the clinical rationality, and passes the validated suture boundary parameters to the conditional generative adversarial training to output the optimal suture boundary parameters.

[0081] Specifically, the forward prediction module predicts optical performance through multi-stage feature extraction and physical constraints, including the following steps:

[0082] S411: The structural parameters, suture boundary parameters, and time parameters of the artificial cornea are normalized and transformed into standardized feature vectors;

[0083] S412: The interaction weights between standardized feature vectors are calculated through a cross-attention mechanism. Based on the weights, each standardized feature vector is weighted and fused to generate a joint feature vector, while preserving the non-linear coupling relationship between parameters.

[0084] S413: Pass the joint feature vector to a physically constrained convolutional layer, which is embedded with Hooke's law to ensure that the output conforms to the laws of mechanics.

[0085] S414: Passes the output of the physically constrained convolutional layer to the mechanical feature head, outputting the stress and strain distribution of the artificial cornea;

[0086] S415: Input the stress distribution and strain distribution into the optical feature head to predict optical performance and output the refractive power and modulation transfer function (MTF) curve of the artificial cornea.

[0087] Specifically, the inverse optimization module is responsible for generating optimal suture boundary parameters that meet clinical feasibility based on given optical targets and artificial corneal structure parameters. The core of the inverse optimization module is a conditional generative adversarial network (cGAN), which is combined with physical constraints for optimization, including the following steps:

[0088] S421: Input information, including optical target t, artificial corneal structural parameters c, and random noise z;

[0089] S422: The generator receives the optical target t and the artificial corneal structure parameter c, and generates the suture boundary parameter G(c,t,z);

[0090] S423: The discriminator evaluates the suture boundary parameter G(c,t,z) output by the generator to determine whether it is close to the real clinical data, and outputs a scalar D(G(c,t,z)|c,t)∈[0,1] to represent the probability of data authenticity;

[0091] S424: The suture boundary parameter G(c,t,z) passed through the discriminator is passed to the forward prediction module to obtain optical performance prediction data. The joint loss function result of the artificial corneal mechanical-optical dual-path hybrid network is judged. If it is greater than the set threshold, the joint loss function result is passed to the inverse optimization module to adjust the generator in reverse. Steps S422-S424 are repeated to perform joint iterative training of the inverse optimization module and the forward prediction module.

[0092] S425: When the decrease in the joint loss function is less than the set threshold, output the optimal stitching boundary parameters.

[0093] Specifically, during the training phase, a constraint loss is applied to the mechanical output using limited metadata from the human eye to ensure that the network prediction results follow mechanical laws and material properties. The joint loss function of the artificial cornea mechanical-optical dual-path hybrid network is:

[0094]

[0095] In the formula, For joint losses, For constitutive relation loss, For optical loss, α and β are the loss weights for constitutive loss and optical loss, respectively, where:

[0096]

[0097] In the formula, N node σ represents the total number of nodes, e represents the cell number of the artificial cornea mesh, and σ represents the total number of nodes. e Predict the stress tensor for the nodes, ε e The strain tensor is calculated from artificial corneal morphology data. For the stiffness matrix of artificial corneal materials, w is the square of the Frobenius norm. n The weight of the region where the node is located;

[0098]

[0099] In the formula, n is the equivalent refractive index of the cornea and aqueous humor, and R... eff D is the effective radius of curvature calculated from the predicted topography. pred κ is the diopter value directly output by the network, ΔR is the smoothing weight coefficient, and ΔR is the Laplace operator of the curvature field, used to characterize the degree of local fluctuation of curvature.

[0100] Step S5: Based on the required imaging needs and personalized human eye parameters, input the optical indicators and artificial corneal structural parameters into the trained artificial corneal mechanical-optical dual-path hybrid network to obtain the optimal suture boundary parameters for artificial corneal transplantation surgery.

[0101] In summary, this invention employs a method combining human eye finite element model simulation and human eye optical model simulation. By analyzing the effect of suture boundary parameters on the artificial cornea using the human eye finite element model, the stress response and conformational changes of the artificial cornea can be obtained in real time, efficiently, and at low cost. Combined with the human eye optical model simulation, the final imaging effect of the surgery can be simulated in real time. A hybrid mechanical-optical dual-path network for the artificial cornea is constructed, providing personalized parameter adjustments based on the patient's postoperative needs, optimizing surgical outcomes, and ultimately improving the quality and accuracy of postoperative visual recovery.

[0102] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. An analytical method for analyzing the impact of suture boundaries on optical imaging of artificial corneas, characterized in that, Includes the following steps: S1: Design the suture boundary parameters, establish a finite element model of the human eye including the suture, and perform stress and configuration analysis on it to obtain the stress distribution, strain distribution and configuration change data of the human eye finite element model; S2: Extract the coordinates of the inner and outer surface nodes of the artificial cornea in the finite element model of the human eye, and perform surface fitting on the inner and outer surface node coordinates to obtain the inner and outer surface fitted surfaces of the artificial cornea. An optical model of the human eye is established based on the inner and outer surface fitting surfaces of the artificial cornea, and the optical imaging data of the human eye optical model corresponding to the finite element model of the human eye under the designed suture boundary parameters are obtained. S3: Adjust the suture boundary parameters and the structural and mechanical parameters of the artificial cornea. Repeat steps S1 and S2 to obtain the stress distribution, strain distribution, configuration change data of the human eye finite element model and the optical imaging data of the human eye optical model under different suture boundary parameters and the structural and mechanical parameters of the artificial cornea. Establish a database of suture boundary parameters and optical imaging. S4: Based on the suture boundary parameters and optical imaging database, a hybrid network of mechanical and optical paths for artificial cornea is constructed and trained. This hybrid network consists of a forward prediction module and an inverse optimization module, which work collaboratively through a weight-sharing layer. The forward prediction module employs multi-scale feature fusion and physical constraint design. It inputs the artificial cornea structure, mechanical parameters, and suture boundary parameters into the input layer, transfers the preprocessed data from the input layer to the parameter embedding layer, then passes the fused feature vector output from the parameter embedding layer to the physical constraint convolutional layer, and finally passes the data output from the physical constraint convolutional layer to the mechanical feature head. The system uses an optical feature head to map data into mechanical features and optical indices. The mechanical features include stress distribution, strain distribution, and configuration change data of the human eye finite element model, while the optical indices are optical imaging data of the human eye optical model. The inverse optimization module is built based on a conditional generative adversarial network and includes a generator, a discriminator, and a physical constraint layer. The inverse optimization module accepts the target optical indices and the structural parameters of the artificial cornea, passes them to the generator, outputs preliminary suture boundary parameters, and then passes them to the discriminator to evaluate clinical rationality. The validated suture boundary parameters are then passed to the conditional generative adversarial training to output the optimal suture boundary parameters. S5: Based on the required imaging needs and personalized human eye parameters, input the optical indicators and artificial corneal structural parameters into the trained artificial corneal mechanical-optical dual-path hybrid network to obtain the optimal suture boundary parameters for artificial corneal transplantation surgery.

2. The analytical method for analyzing the impact of suture boundaries on optical imaging of artificial cornea according to claim 1, characterized in that, In step S1, the finite element model of the human eye includes a simplified eyeball model and a scratching plate model simulating the eyelid. The simplified eyeball model includes a corneal implant bed, sutures, artificial cornea, limbus, and sclera. The scratching plate model simulating the eyelid has a plate curvature consistent with the outer surface of the cornea, and its edges are chamfered.

3. The analytical method for analyzing the impact of suture boundaries on artificial corneal optical imaging according to claim 2, characterized in that, Step S1 includes the following steps: S11: Design the suture boundary parameters and establish a finite element model of the eyeball containing the suture using finite element modeling software; S12: A shell-shaped flexible plate conforming to the shape of the cornea is created using finite element modeling software, and the boundary of the inner surface of the shell-shaped flexible plate in contact with the eyeball is chamfered. S13: Based on the material parameters of the standard human eye and the material parameters of the artificial cornea, input the material parameters and contact friction coefficients of each part of the finite element model of the eyeball into the finite element modeling software; S14: Controls the movement of the flexible plate to simulate the scraping process of the eyelid against the artificial cornea; S15: Extract the node coordinates of the inner and outer surfaces of the artificial cornea, and perform polynomial fitting on the node coordinates of the inner and outer surfaces of the artificial cornea to obtain the fitted surfaces of the inner and outer surfaces of the artificial cornea.

4. The analytical method for analyzing the impact of suture boundaries on optical imaging of artificial cornea according to claim 3, characterized in that, In step S2, the optical imaging data includes the refractive power of the artificial cornea, visual acuity level, and modulation transfer function (MTF) curve; the human eye optical model includes the inner surface of the artificial cornea, the outer surface of the artificial cornea, the inner surface of the lens, the virtual plane, the outer surface of the lens, and the retina.

5. The analytical method for analyzing the impact of suture boundaries on optical imaging of artificial cornea according to claim 4, characterized in that, Step S2 includes the following steps: S21: Establish a standard human eye optical model based on the Liou human eye optical model and standard human eye parameters; S22: Replace the standard human eye optical model with the fitted surfaces of the inner and outer surfaces of the artificial cornea before and after scraping with the soft plate to obtain the required human eye optical model. S23: Imaging simulation using optical software; S24: Extract the visual acuity results and imaging effect diagrams of the human eye optical model before and after scratching.

6. The analytical method for analyzing the impact of suture boundaries on optical imaging of artificial cornea according to claim 5, characterized in that, In step S3, the suture boundary parameters include: suture penetration depth, suture length, number of sutures, and suture force; the artificial corneal structural parameters include: artificial corneal central thickness, anterior and posterior surface curvature, and transverse diameter; the artificial corneal mechanical parameters include: elastic modulus, Poisson's ratio, and relaxation curve of the artificial corneal material.

7. The analytical method for analyzing the impact of suture boundaries on optical imaging of artificial cornea according to claim 6, characterized in that, In step S4, the forward prediction module predicts optical performance through multi-stage feature extraction and physical constraints, including the following steps: S411: The structural parameters, suture boundary parameters, and time parameters of the artificial cornea are normalized and transformed into standardized feature vectors; S412: The interaction weights between standardized feature vectors are calculated through a cross-attention mechanism. Based on the weights, each standardized feature vector is weighted and fused to generate a joint feature vector, while preserving the non-linear coupling relationship between parameters. S413: Pass the joint feature vector to a physically constrained convolutional layer, which is embedded with Hooke's law to ensure that the output conforms to the laws of mechanics. S414: Passes the output of the physically constrained convolutional layer to the mechanical feature head, outputting the stress and strain distribution of the artificial cornea; S415: Input the stress distribution and strain distribution into the optical feature head to predict optical performance and output the refractive power and modulation transfer function (MTF) curve of the artificial cornea.

8. The analytical method for analyzing the impact of suture boundaries on optical imaging of artificial cornea according to claim 7, characterized in that, In step S4, the inverse optimization module is responsible for generating optimal suture boundary parameters that meet clinical feasibility based on the given optical target and artificial corneal structure parameters. The core of the inverse optimization module is a conditional generative adversarial network (cGAN), which is combined with physical constraints for optimization, including the following steps: S421: Input information, including optical target t, artificial corneal structural parameters c, and random noise z; S422: The generator receives the optical target t and the artificial corneal structure parameter c, and generates the suture boundary parameter G(c,t,z); S423: The discriminator evaluates the suture boundary parameter G(c,t,z) output by the generator to determine whether it is close to the real clinical data, and outputs a scalar D(G(c,t,z)|c,t)∈[0,1] to represent the probability of data authenticity; S424: The suture boundary parameter G(c,t,z) passed through the discriminator is passed to the forward prediction module to obtain optical performance prediction data. The joint loss function result of the artificial corneal mechanical-optical dual-path hybrid network is judged. If it is greater than the set threshold, the joint loss function result is passed to the inverse optimization module to adjust the generator in reverse. Steps S422-S424 are repeated to perform joint iterative training of the inverse optimization module and the forward prediction module. S425: When the decrease in the joint loss function is less than the set threshold, output the optimal stitching boundary parameters.

9. The analytical method for analyzing the impact of suture boundaries on optical imaging of artificial cornea according to claim 8, characterized in that, In step S4, during the training phase, a constraint loss is applied to the mechanical output using limited metadata from the human eye to ensure that the network prediction results follow mechanical laws and material properties. The joint loss function of the artificial cornea mechanical-optical dual-path hybrid network is: In the formula, For joint losses, For constitutive relation loss, For optical loss, α and β are the loss weights for constitutive loss and optical loss, respectively, where: In the formula, N node σ represents the total number of nodes, e represents the cell number of the artificial cornea mesh, and σ represents the total number of nodes. e Predict the stress tensor for the nodes, ε e The strain tensor is calculated from artificial corneal morphology data. For the stiffness matrix of artificial corneal materials, w is the square of the Frobenius norm. n The weight of the region where the node is located; In the formula, n is the equivalent refractive index of the cornea and aqueous humor, and R... eff D is the effective radius of curvature calculated from the predicted topography. pred κ is the diopter value directly output by the network, ΔR is the smoothing weight coefficient, and ΔR is the Laplace operator of the curvature field, used to characterize the degree of local fluctuation of curvature.

Images