Ophthalmology simulation teaching equipment and method

Abstract

The invention discloses an ophthalmology simulation teaching device and method, and the device comprises a disease generation module which is used for generating an ophthalmology three-dimensional model in a specific disease state through a preset ophthalmology disease generation model; the model demonstration module is used for demonstrating the ophthalmology three-dimensional model through interactive display equipment; the model printing module is used for selecting corresponding materials for different material layers in the ophthalmology three-dimensional model through 3D printing equipment for printing and film covering, and an ophthalmology entity model is obtained; and the operation training platform is used for collecting operation information of the user on the ophthalmology entity model and analyzing the operation information to generate an evaluation result of operation correctness. According to the method, automatic generation of the eyeball disease model and entity printing are combined, automatic generation of diseases, teaching demonstration and training, operation simulation operation and scoring are integrated, the efficiency of ophthalmology teaching is improved, the training cost is reduced, and meanwhile the risk in actual operation training is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical intelligent teaching, and in particular relates to an ophthalmology simulation teaching device and method. Background Art

[0002] In modern medical education and clinical training, the cultivation of ophthalmic surgical skills is crucial to the professional development of doctors. However, traditional teaching methods are often limited by the scarcity of practical operation opportunities and the limitations of training materials. Actual surgical training is high-risk, costly, and resource-limited, while traditional models and teaching tools are difficult to provide sufficiently realistic operation experience. Therefore, it is necessary to develop a new teaching device and method to provide a safe, efficient, and economical training platform that allows doctors and medical students to practice and improve surgical skills in a simulated real surgical environment.

[0003] The technical background shows that with the development of computer graphics, 3D modeling technology, 3D printing technology and artificial intelligence, it has become possible to simulate real organs and surgical operations. In particular, as a specialty in the medical field that requires extremely high precision operations, ophthalmology has an urgent need for simulation training. The complexity of ophthalmic surgery requires doctors to have a high degree of spatial perception and hand-eye coordination, and the cultivation of these skills requires a lot of practice and feedback.

[0004] The ophthalmic simulation teaching device and method proposed in the present invention are based on such technical background and needs. The device and method automatically generate an ophthalmic three-dimensional model under a specific disease state through a preset ophthalmic disease generation model, use an interactive display device for teaching demonstration, and produce a realistic ophthalmic entity model through 3D printing technology. In addition, the system can also collect the user's operation information on the entity model, analyze this information and generate an evaluation result of the correctness of the operation, thereby realizing the training and evaluation of surgical skills. This method not only improves the efficiency of learning, but also reduces the cost of training, while reducing the risks in actual surgical training, providing an innovative solution for ophthalmic education and clinical training. Summary of the invention

[0005] In view of this, the present invention proposes an ophthalmic simulation teaching device and method, which combines the automatic generation and physical printing of eye disease models, integrates disease self-generation, teaching demonstration and training, surgical simulation operation and scoring, and helps to improve learning efficiency.

[0006] In order to achieve the above object, the present invention provides the following technical solutions: The present invention provides an ophthalmology simulation teaching device, comprising: A disease generation module, used to generate an ophthalmic three-dimensional model under a specific disease state through a preset ophthalmic disease generation model; Model demonstration module, used to demonstrate ophthalmic three-dimensional models through interactive display devices; A model printing module is used to select corresponding materials for different material layers in the ophthalmic three-dimensional model through a 3D printing device for printing and laminating to obtain an ophthalmic physical model; The surgical training platform is used to collect the user's operation information on the ophthalmic entity model, and analyze the operation information to generate evaluation results on the correctness of the operation.

[0007] Preferably, the ophthalmic disease generation model is trained by: Obtain a large amount of fundus lesion images and their corresponding diagnostic result texts in advance; The fundus images were standardized, and the text of the diagnosis results was cleaned and standardized; The two-level diffusion large model architecture SD1V5 was selected, and it was trained and verified based on the diagnosis result text and fundus lesion image structure and pre-marked multi-scale conditional constraints to obtain the initial ophthalmic disease generation model; The output end of the initial ophthalmic disease generation model is connected to the three-dimensional modeling unit to obtain the ophthalmic disease generation model; Among them, the three-dimensional modeling unit is used to extract key features in the image output results of the initial ophthalmic disease generation model, and generate the key features correspondingly to the preset standard three-dimensional eyeball white model to obtain an ophthalmic three-dimensional model.

[0008] Preferably, the two-stage diffusion large model architecture SD1V5 includes an autoencoder and a UNet denoiser, wherein: The autoencoder consists of an encoder and decoder , in the first stage of its training, the encoder Mapping the fundus lesion image to a vector in the latent space, the decoder Decode the latent vector into a pixel image; In the second stage, SD1V5 trains the UNet denoiser to perform denoising directly in the latent space, and the optimization process is based on the formula: ; in, is the loss function of the diffusion model, are the parameters of the model, including the weights and biases of the autoencoder and UNet denoiser, is the true value of the added noise, is the time step, For the encoder, For conditional restrictions, To predict each time step The noise estimate is , and each time step is trained The noise estimate of is getting closer to the actual noise value added. represents the standard normal distribution, is the original fundus lesion image, E represents the expected value, represents the square of the L2 norm, is a latent variable, where: ; in, For a predetermined variance schedule, based on as well as Predict the noise estimate at each time step, conditional on Medium Conditions The model generation is controlled by input text, pictures or preset prior conditions. After a preset number of iterations, the existing noise is subtracted and the model is generated. The latent features are fed into the decoder of the autoencoder for image generation.

[0009] Preferably, the two-stage diffusion model architecture SD1V5 uses the CLIP text encoder to embed text into conditional constraints and introduce a specific encoder Will Projected to the intermediate representation, the cross-attention layer is then implemented and mapped to the intermediate layer of UNet through the following formula group: ; in, It means that UNet embeds text prompt features into image features. , As two learnable embeddings, is the learnable projection matrix, is the query matrix representing the text prompt features, K is the key matrix representing the image characteristics, V is the value matrix representing the image features, and d is the dimension of the key vector.

[0010] Preferably, the model demonstration module includes: An image display for demonstrating ophthalmic three-dimensional models; An interactive device, used to obtain interactive instructions from users; The interactive processing unit is used to analyze the interactive instructions and, based on the analysis results, to enlarge, reduce, move in various directions, rotate at various angles, select, mark, section and explode the ophthalmic three-dimensional model or its individual components in the image display.

[0011] Preferably, the model printing module performs the following operations: Determine the distribution of the disease on the eyeball corresponding to the disease according to the ophthalmic three-dimensional model, and determine the preset printing materials, colors and printing transparency selection results corresponding to the normal part and the diseased part on the diseased eyeball respectively; Selecting a soft material with corresponding elasticity, softness, color and transparency similar to real eyeball tissue according to the selection result, wherein the soft material includes silicone, flexible resin or corresponding 3D printing polymer to simulate the elasticity, softness, color and transparency of the eyeball; The ophthalmic solid model is printed using a high-precision multi-nozzle 3D printer through layered printing technology to obtain an ophthalmic solid model.

[0012] Preferably, the surgical training platform performs the following operations: The operation video stream of the user during the operation is obtained through the image acquisition device preset on the surgical simulation platform; Analyze the images in the operation video stream frame by frame to determine the relative position relationship between the surgical instrument, the user's hand and the ophthalmic physical model; Determine the type of surgical instruments used in each surgical stage, the angle between the surgical instruments and the ophthalmic physical model, the way the hands hold the surgical instruments, and the movement trajectory of the surgical instruments; Based on the specific disease state corresponding to the ophthalmic three-dimensional model, the corresponding surgical type is determined, wherein the surgical categories include one or more combinations of vitrectomy, fundus laser surgery, posterior vitreous detachment, internal limiting membrane peeling, cataract phacoemulsification surgery, and retinal rhegmatogenous detachment surgery; Determine the corresponding surgical process based on the determined type of surgery, as well as the technical requirements for each surgical stage, where the technical requirements include the type of surgical instrument, the angle between the surgical instrument and the ophthalmic physical model, the way the hand holds the surgical instrument, and the movement trajectory of the surgical instrument; The degree of compatibility between the determined user operation technique and the technique requirements of each surgical stage is judged, and an evaluation result of the correctness of the operation is generated.

[0013] To achieve the above objectives, the present invention also provides an ophthalmology simulation teaching method, comprising: Generate an ophthalmic three-dimensional model under a specific disease state through a preset ophthalmic disease generation model; Demonstrate ophthalmic 3D models through interactive display devices; Select corresponding materials for different material layers in the ophthalmic three-dimensional model through 3D printing equipment for printing and lamination to obtain an ophthalmic physical model; The user's operation information on the ophthalmic entity model is collected, and the operation information is analyzed to generate an evaluation result of the correctness of the operation.

[0014] The present invention has achieved at least the following beneficial effects: 1. Automatically generate eye disease models and expand the scope of example presentation in ophthalmology teaching.

[0015] 2. It not only improves the efficiency and safety of surgical skills training, but also enhances the interactivity and practicality of teaching.

[0016] 3. Combined with the automatic generation and physical printing of eye disease models, it integrates disease self-generation, teaching demonstration and training, surgical simulation operation and scoring, which helps to improve learning efficiency.

[0017] Other advantages, objectives and features of the present invention will be described in the following description and will be apparent to those skilled in the art to some extent, or those skilled in the art may be taught from the practice of the present invention. The objectives and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to make the purpose, technical solution and beneficial effects of the present invention clearer, the present invention provides the following drawings for illustration: Figure 1 A simplified structural diagram of an ophthalmology simulation teaching device in an embodiment of the present invention; Figure 2 It is a structural diagram of the two-stage diffusion large model architecture SD1V5 in an embodiment of the present invention; Figure 3 This is a flow chart of the steps executed by the surgical training platform in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0020] The present invention provides an ophthalmology simulation teaching device, referring to Figure 1 ,include: A disease generation module, used to generate an ophthalmic three-dimensional model under a specific disease state through a preset ophthalmic disease generation model; Model demonstration module, used to demonstrate ophthalmic three-dimensional models through interactive display devices; A model printing module is used to select corresponding materials for different material layers in the ophthalmic three-dimensional model through a 3D printing device for printing and laminating to obtain an ophthalmic physical model; The surgical training platform is used to collect the user's operation information on the ophthalmic entity model, and analyze the operation information to generate evaluation results on the correctness of the operation.

[0021] The working principle and beneficial effects of the above technical solution are as follows: through the integrated disease generation module, model demonstration module, model printing module and surgical training platform, a comprehensive teaching solution is provided for medical education and clinical training. The disease generation module generates an ophthalmic three-dimensional model under a specific disease state based on the preset ophthalmic disease generation model and the semantic analysis results of the guiding text formulated by the user for generating an ophthalmic model of a specific disease, thereby providing a large number of non-repetitive and demand-satisfactory ophthalmic three-dimensional models, which can accurately simulate the morphology and structure of the real ophthalmic disease that the user wants to achieve. The model demonstration module uses high-resolution, interactive display devices to enable users to observe and learn these three-dimensional models from different angles, enhancing the interactivity and intuitiveness of teaching. The model printing module uses advanced 3D printing technology to select corresponding materials for different material layers in the ophthalmic three-dimensional model for printing, and performs lamination treatment to obtain an ophthalmic entity model with a real eyeball touch and appearance. The surgical training platform collects the user's operation information on the ophthalmic entity model, analyzes it, and generates an evaluation result on the correctness of the operation, thereby providing instant feedback to help users improve their surgical skills. This integrated device not only improves the efficiency and safety of surgical skills training, but also enhances the interactivity and practicality of teaching, which is of great significance for improving the professional skills of medical students and doctors. By simulating a real surgical environment, users can practice surgical skills in a risk-free environment, reducing the risk to real patients, and also providing a new and efficient teaching tool for medical education.

[0022] In a preferred embodiment, the ophthalmic disease generation model is trained in the following manner: Obtain a large amount of fundus lesion images and their corresponding diagnostic result texts in advance; The fundus images were standardized, and the text of the diagnosis results was cleaned and standardized; The two-level diffusion large model architecture SD1V5 was selected, and it was trained and verified based on the diagnosis result text and fundus lesion image structure and pre-marked multi-scale conditional constraints to obtain the initial ophthalmic disease generation model; The output end of the initial ophthalmic disease generation model is connected to the three-dimensional modeling unit to obtain the ophthalmic disease generation model; Among them, the three-dimensional modeling unit is used to extract key features in the image output results of the initial ophthalmic disease generation model, and generate the key features correspondingly to the preset standard three-dimensional eyeball white model to obtain an ophthalmic three-dimensional model.

[0023] The working principle and beneficial effects of the above technical solution are as follows: First, a large number of fundus lesion images and their corresponding diagnostic result texts are collected and sorted. Then, these images are standardized, including resizing, contrast enhancement and denoising to improve image quality, and the diagnostic result text is cleaned and standardized to ensure the consistency and accuracy of the information. Then, the two-level diffusion large model architecture SD1V5 is selected, and the model is trained and verified using the structure of the diagnostic result text and fundus lesion image, combined with the pre-set multi-scale conditional restrictions, so as to obtain the initial ophthalmic disease generation model. The output end of the model is further connected to the three-dimensional modeling unit, which is responsible for extracting the key features in the model image output results, and generating these features correspondingly to the preset standard three-dimensional eyeball white model, and finally obtaining an accurate ophthalmic three-dimensional model. This technical solution not only improves the accuracy and realism of ophthalmic disease simulation, but also greatly improves the efficiency and effect of ophthalmic teaching and training through the automated model generation and feature extraction process, providing an innovative tool for medical education and clinical practice. In this way, users can practice and improve their surgical skills in a risk-free environment. At the same time, the device also provides a realistic simulation platform for the research and teaching of ophthalmic diseases.

[0024] In a specific embodiment, the code implementation of the ophthalmic disease generation model is as follows: # Import necessary libraries import numpy as np import pandas as pd from torchvision import transforms from torch.utils.data import Dataset, DataLoader import torch import torch.nn as nn from diffusion_model import SD1V5 # Assume this is a predefined SD1V5 model class from three_d_modeller import ThreeDModeller # Assume this is a predefined 3D modeling unit class # Data preprocessing class RetinaDataset(Dataset): def __init__(self, images, texts, transforms=None): self.images = images self.texts = texts self.transforms = transforms def __len__(self): return len(self.images) def __getitem__(self, idx): image = self.images[idx] text = self.texts[idx] if self.transforms: image = self.transforms(image) return image, text # Standardization and data loading def standardize_data(images, texts): # Normalize the image standardized_images = [preprocess_image(image) for image inimages] # Clean and standardize the text standardized_texts = [preprocess_text(text) for text in texts] return standardized_images, standardized_texts # Train a two-level diffusion model def train_model(model, dataloader, epochs): for epoch in range(epochs): for images, texts in dataloader: # Input images and text into the model outputs = model(images, texts) # Calculate loss and backpropagate loss = calculate_loss(outputs) loss.backward() optimizer.step() optimizer.zero_grad() print(f"Epoch {epoch}, Loss: {loss.item()}") # Generate ophthalmic 3D model def generate_3d_models(model, three_d_modeller, dataloader): 3d_models = [] for images, texts in dataloader: # Use the model to generate images of eye diseases generated_images = model.generate(images, texts) # Extract key features and generate 3D models for image in generated_images: features = three_d_modeller.extract_features(image) # Apply the features to the standard 3D eyeball model _3d_model = three_d_modeller.generate_3d_model(features) 3d_models.append(_3d_model) return 3d_models # Main function def main(): # Load data images, texts = load_data() # Data preprocessing standardized_images, standardized_texts = standardize_data(images, texts) # Create dataset and data loader dataset = RetinaDataset(standardized_images, standardized_texts,transforms=transform) dataloader = DataLoader(dataset, batch_size=32, shuffle=True) # Initialize the model and 3D modeling unit model = SD1V5() three_d_modeller = ThreeDModeller() # Train the model train_model(model, dataloader, epochs=10) # Generate 3D model _3d_models = generate_3d_models(model, three_d_modeller,dataloader) if __name__ == "__main__": main() In a preferred embodiment, the two-stage diffusion model architecture SD1V5 includes an autoencoder and a UNet denoiser, referring to Figure 2 ,in: The autoencoder consists of an encoder and decoder In the first stage of its training, the encoder Mapping the fundus lesion image to a vector in the latent space, the decoder Decode the latent vector into a pixel image; In the second stage, SD1V5 trains the UNet denoiser to perform denoising directly in the latent space, and the optimization process is based on the formula: ; in, is the loss function of the diffusion model, are the parameters of the model, including the weights and biases of the autoencoder and UNet denoiser, is the true value of the added noise, is the time step, For the encoder, For conditional restrictions, To predict each time step The noise estimate is , and each time step is trained The noise estimate of is getting closer to the actual noise value added. represents the standard normal distribution, is the original fundus lesion image, E represents the expected value, represents the square of the L2 norm, is a latent variable, where: ; in, For a predetermined variance schedule, based on as well as Predict the noise estimate at each time step, conditional on Medium Conditions The model generation is controlled by input text, pictures or preset prior conditions. After a preset number of iterations, the existing noise is subtracted and the model is generated. The latent features are fed into the decoder of the autoencoder for image generation.

[0025] The working principle and beneficial effects of the above technical solution are as follows: The two-stage diffusion large model architecture SD1V5 includes an autoencoder and a UNet denoiser, and its training process is divided into two stages. In the first stage, the encoder of the autoencoder maps the fundus lesion image to a vector in the latent space, and the decoder decodes these latent vectors back to pixel images. This process aims to accurately reconstruct the input image. In the second stage, SD1V5 trains the UNet denoiser so that it can directly denoise in the latent space. The optimization process is achieved by minimizing the loss function of the diffusion model, which measures the difference between the noise predicted by the model and the actual noise added. In this process, the parameters of the model, including the weights and biases of the autoencoder and the UNet denoiser, are adjusted to make the noise estimate at each time step constantly approach the actual added noise value. The setting of the time step t is based on a predetermined variance schedule that guides the model to predict the noise estimate at each time step. The conditional constraints are controlled by input text, pictures or preset prior conditional constraints to control model generation, where the prior conditional constraints are used to train ophthalmic diseases that meet specific semantic descriptions, thereby meeting the generation requirements of ophthalmic three-dimensional models. After a preset number of iterations, the model subtracts the existing noise and sends the latent features to the decoder of the autoencoder for image generation. The beneficial effect of this technical solution is that it can generate high-quality fundus lesion images while maintaining the efficiency and controllability of the generation process, providing a powerful tool for ophthalmic teaching and research. In this way, realistic fundus lesions can be simulated for medical education, surgical simulation, and clinical diagnosis training, thereby improving doctors' professional skills and diagnostic accuracy.

[0026] In a preferred embodiment, the two-stage diffusion model architecture SD1V5 uses the CLIP text encoder to embed text into conditional constraints. and introduce a specific encoder Will Projected to the intermediate representation, the cross-attention layer is then implemented and mapped to the intermediate layer of UNet through the following formula group: ; in, It means that UNet embeds text prompt features into image features. , As two learnable embeddings, is the learnable projection matrix, is the query matrix representing the text prompt features, K is the key matrix representing the image characteristics, V is the value matrix representing the image features, and d is the dimension of the key vector.

[0027] In a preferred embodiment, the model demonstration module includes: An image display for demonstrating ophthalmic three-dimensional models; An interactive device, used to obtain interactive instructions from users; The interactive processing unit is used to analyze the interactive instructions and, based on the analysis results, to enlarge, reduce, move in various directions, rotate at various angles, select, mark, section and explode the ophthalmic three-dimensional model or its individual components in the image display.

[0028] The working principle and beneficial effects of the above technical solution are as follows: The model demonstration module is designed to provide a highly interactive and intuitive demonstration platform for displaying ophthalmic three-dimensional models. The module consists of three core parts: an image display, an interactive device, and an interactive processing unit. The image display is responsible for clearly displaying the ophthalmic three-dimensional model, which can be a high-resolution screen or a virtual reality (VR) device to provide users with a realistic visual experience. The interactive device, such as a touch screen, a mouse, a keyboard, or a dedicated gesture recognition device, is used to capture the user's interactive instructions, including clicks, drags, zooms, rotations, and other operations. These instructions are then analyzed by the interactive processing unit, which is the brain of the module and is responsible for parsing the user's input and converting it into specific operation commands. Based on these commands, the interactive processing unit can perform operations such as zooming in, zooming out, moving, rotating, etc. on the ophthalmic three-dimensional model or its individual components in the image display, as well as advanced functions such as model selection, marking, section display, and exploded view. The beneficial effect of this technical solution is that it greatly enhances the interactivity of teaching and learning, allowing users to deeply understand the details of the ophthalmic model from different angles and levels, and improves teaching efficiency and learning effects. Furthermore, this modular and highly customizable presentation module provides a flexible tool for ophthalmic education and clinical training, which can be adapted and optimized for different teaching needs and scenarios.

[0029] In a preferred embodiment, the model printing module performs the following operations: Determine the distribution of the disease on the eyeball corresponding to the disease according to the ophthalmic three-dimensional model, and determine the preset printing materials, colors and printing transparency selection results corresponding to the normal part and the diseased part on the diseased eyeball respectively; Selecting a soft material with corresponding elasticity, softness, color and transparency similar to real eyeball tissue according to the selection result, wherein the soft material includes silicone, flexible resin or corresponding 3D printing polymer to simulate the elasticity, softness, color and transparency of the eyeball; The ophthalmic solid model is printed using a high-precision multi-nozzle 3D printer through layered printing technology to obtain an ophthalmic solid model.

[0030] The working principle and beneficial effects of the above technical solution are as follows: The design goal of the model printing module is to accurately convert the digital ophthalmic three-dimensional model into a physical model for use in ophthalmic medical education and clinical practice. The module first determines the distribution of the disease on the eyeball based on the ophthalmic three-dimensional model, and distinguishes the parts of normal tissue and diseased tissue, and then presets the corresponding printing materials, colors and transparency for these different areas. Soft materials such as silicone, flexible resin or specific 3D printing polymers that are similar to real eyeball tissue in elasticity, softness, color and transparency are selected to simulate the physical properties of the real eyeball. Then, a high-precision multi-nozzle 3D printer is used to perform detailed printing and lamination printing on the ophthalmic physical model through layered printing technology to ensure that every detail is accurately replicated. This process can not only generate an ophthalmic model that looks and feels close to the real thing, but also adjust the materials and printing parameters according to different disease requirements to meet specific teaching or training purposes. The beneficial effect of this technical solution is that it provides an efficient, flexible and cost-effective method for manufacturing solid models, which greatly enhances the realism and interactivity of ophthalmic teaching and surgical training, helps medical professionals better understand and master ophthalmic surgical skills, and provides a valuable tool for ophthalmic research.

[0031] In a preferred embodiment, referring to Figure 3 The surgical training platform performs the following operations: The operation video stream of the user during the operation is obtained through the image acquisition device preset on the surgical simulation platform; Analyze the images in the operation video stream frame by frame to determine the relative position relationship between the surgical instrument, the user's hand and the ophthalmic physical model; Determine the type of surgical instruments used in each surgical stage, the angle between the surgical instruments and the ophthalmic physical model, the way the hands hold the surgical instruments, and the movement trajectory of the surgical instruments; Based on the specific disease state corresponding to the ophthalmic three-dimensional model, the corresponding surgical type is determined, wherein the surgical categories include one or more combinations of vitrectomy, fundus laser surgery, posterior vitreous detachment, internal limiting membrane peeling, cataract phacoemulsification and retinal rhegmatogenous detachment surgery; based on the determined surgical type, the corresponding surgical process and the technical requirements of each surgical stage are determined, wherein the technical requirements include the selection of surgical instrument types, the angle between the surgical instrument and the ophthalmic physical model, the way the hand holds the surgical instrument and the movement trajectory of the surgical instrument; The degree of compatibility between the determined user operation technique and the technique requirements of each surgical stage is judged, and an evaluation result of the correctness of the operation is generated.

[0032] The working principle and beneficial effects of the above technical solution are as follows: The design of the surgical training platform provides a comprehensive surgical skills training and evaluation system. The platform obtains the user's operation video stream during the operation in real time through a preset image acquisition device, and analyzes the images in the video stream frame by frame to determine the relative position relationship between the surgical instrument, the user's hand and the ophthalmic entity model. Through this high-precision analysis, the platform can identify the type of surgical instrument used in each surgical stage, the angle between the instrument and the model, the way the hand holds the instrument, and the movement trajectory of the instrument. Combined with the specific disease state corresponding to the ophthalmic three-dimensional model, the platform can determine the corresponding type of surgery, including but not limited to vitrectomy, fundus laser surgery, posterior vitreous detachment, internal limiting membrane peeling, cataract phacoemulsification surgery, and retinal rhegmatogenous detachment surgery. Based on the determined type of surgery, the platform further determines the corresponding surgical process and the technique requirements for each stage, which involve the selection, angle, holding method and movement trajectory of surgical instruments. Finally, the platform compares the user's operation technique with the standard technique requirements for each surgical stage, judges the degree of fit, and generates an evaluation result of the correctness of the operation. This real-time feedback mechanism can help users understand their performance instantly and guide them on how to improve their surgical skills. The technical solution provides a highly simulated and interactive surgical training environment, allowing doctors and medical students to practice and improve surgical skills without risk. Through precise surgical operation analysis and evaluation, the platform can not only improve the efficiency of surgical skills training, but also ensure the standardization and safety of surgical operations. In addition, this system can also provide personalized training plans based on the different levels of users, thereby more effectively improving the user's surgical skills and the quality of surgical results. In short, the surgical training platform of the present invention provides an innovative tool for ophthalmic surgery education and clinical practice, which helps to train the next generation of ophthalmologists.

[0033] To achieve the above objectives, the present invention also provides an ophthalmology simulation teaching method, comprising: Generate an ophthalmic three-dimensional model under a specific disease state through a preset ophthalmic disease generation model; Demonstrate ophthalmic 3D models through interactive display devices; Select corresponding materials for different material layers in the ophthalmic three-dimensional model through 3D printing equipment for printing and lamination to obtain an ophthalmic physical model; The user's operation information on the ophthalmic entity model is collected, and the operation information is analyzed to generate an evaluation result of the correctness of the operation.

[0034] The working principle and beneficial effects of the above technical solution are as follows: based on the preset ophthalmic disease generation model, on the basis of the semantic analysis results of the guiding text formulated by the user for generating an ophthalmic model of a specific disease, an ophthalmic three-dimensional model under a specific disease state is generated, thereby providing a large number of non-repetitive and demand-satisfactory ophthalmic three-dimensional models, which can accurately simulate the morphology and structure of the real ophthalmic disease that the user wants to achieve. The use of high-resolution, interactive display devices enables users to observe and learn these three-dimensional models from different angles, enhancing the interactivity and intuitiveness of teaching. Advanced 3D printing technology is used to select corresponding materials for printing for different material layers in the ophthalmic three-dimensional model, and the coating is performed to obtain an ophthalmic entity model with a real eyeball touch and appearance. The user's operation information on the ophthalmic entity model is collected and analyzed to generate an evaluation result of the correctness of the operation, thereby providing instant feedback to help users improve their surgical skills. This method not only improves the efficiency and safety of surgical skill training, but also enhances the interactivity and practicality of teaching, which is of great significance for improving the professional skills of medical students and doctors. By simulating a real surgical environment, users can practice surgical skills in a risk-free environment, reducing risks to real patients, while also providing a new and efficient teaching tool for medical education.

[0035] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. An ophthalmic simulation teaching device, characterized in that: include: A disease generation module, used to generate an ophthalmic three-dimensional model under a specific disease state through a preset ophthalmic disease generation model; Model demonstration module, used to demonstrate ophthalmic three-dimensional models through interactive display devices; A model printing module is used to select corresponding materials for different material layers in the ophthalmic three-dimensional model through a 3D printing device for printing and laminating to obtain an ophthalmic physical model; The surgical training platform is used to collect the user's operation information on the ophthalmic entity model, and analyze the operation information to generate evaluation results on the correctness of the operation.

2. The ophthalmic simulation teaching device according to claim 1, characterized in that: The ophthalmic disease generative model was trained in the following way: Obtain a large amount of fundus lesion images and their corresponding diagnostic result texts in advance; The fundus images were standardized, and the text of the diagnosis results was cleaned and standardized; The two-level diffusion large model architecture SD1V5 was selected, and it was trained and verified based on the diagnosis result text and fundus lesion image structure and pre-marked multi-scale conditional constraints to obtain the initial ophthalmic disease generation model; The output end of the initial ophthalmic disease generation model is connected to the three-dimensional modeling unit to obtain the ophthalmic disease generation model; Among them, the three-dimensional modeling unit is used to extract key features in the image output results of the initial ophthalmic disease generation model, and generate the key features correspondingly to the preset standard three-dimensional eyeball white model to obtain an ophthalmic three-dimensional model.

3. The ophthalmic simulation teaching device according to claim 2, characterized in that: The two-stage diffusion model architecture SD1V5 includes an autoencoder and a UNet denoiser, where: The autoencoder consists of an encoder and decoder , in the first stage of its training, the encoder Mapping the fundus lesion image to a vector in the latent space, the decoder Decode the latent vector into a pixel image; In the second stage, SD1V5 trains the UNet denoiser to perform denoising directly in the latent space, and the optimization process is based on the formula: ； in, is the loss function of the diffusion model, are the parameters of the model, including the weights and biases of the autoencoder and UNet denoiser, is the true value of the added noise, is the time step, For the encoder, For conditional restrictions, To predict each time step The noise estimate is , and each time step is trained The noise estimate of is getting closer to the actual noise value added. represents the standard normal distribution, is the original fundus lesion image, E represents the expected value, represents the square of the L2 norm, is a latent variable, where: ； in, For a predetermined variance schedule, based on as well as Predict the noise estimate at each time step, conditional on Medium Conditions The model generation is controlled by input text, pictures or preset prior conditions. After a preset number of iterations, the existing noise is subtracted and the model is generated. The latent features are fed into the decoder of the autoencoder for image generation.

4. The ophthalmic simulation teaching device according to claim 3, characterized in that: The two-stage diffusion model architecture SD1V5 uses the CLIP text encoder to embed text into conditional constraints and introduce a specific encoder Will Projected to the intermediate representation, the cross-attention layer is then implemented and mapped to the intermediate layer of UNet through the following formula group: ； in, It means that UNet embeds text prompt features into image features. , As two learnable embeddings, is the learnable projection matrix, is the query matrix representing the text prompt features, K is the key matrix representing the image characteristics, V is the value matrix representing the image features, and d is the dimension of the key vector.

5. The ophthalmic simulation teaching device according to claim 1, characterized in that: The model demonstration module includes: An image display for demonstrating ophthalmic three-dimensional models; An interactive device, used to obtain interactive instructions from users; The interactive processing unit is used to analyze the interactive instructions and, based on the analysis results, to enlarge, reduce, move in various directions, rotate at various angles, select, mark, section and explode the ophthalmic three-dimensional model or its individual components in the image display.

6. The ophthalmic simulation teaching device according to claim 1, characterized in that: The model printing module performs the following operations: Determine the distribution of the disease on the eyeball corresponding to the disease according to the ophthalmic three-dimensional model, and determine the preset printing materials, colors and printing transparency selection results corresponding to the normal part and the diseased part on the diseased eyeball respectively; Selecting a soft material with corresponding elasticity, softness, color and transparency similar to real eyeball tissue according to the selection result, wherein the soft material includes silicone, flexible resin or corresponding 3D printing polymer to simulate the elasticity, softness, color and transparency of the eyeball; The ophthalmic solid model is printed using a high-precision multi-nozzle 3D printer through layered printing technology to obtain an ophthalmic solid model.

7. The ophthalmic simulation teaching device according to claim 1, characterized in that: The surgical training platform performs the following operations: The operation video stream of the user during the operation is obtained through the image acquisition device preset on the surgical simulation platform; Analyze the images in the operation video stream frame by frame to determine the relative position relationship between the surgical instrument, the user's hand and the ophthalmic physical model; Determine the type of surgical instruments used in each surgical stage, the angle between the surgical instruments and the ophthalmic physical model, the way the hands hold the surgical instruments, and the movement trajectory of the surgical instruments; Based on the specific disease state corresponding to the ophthalmic three-dimensional model, the corresponding surgical type is determined, wherein the surgical categories include one or more combinations of vitrectomy, fundus laser surgery, posterior vitreous detachment, internal limiting membrane peeling, cataract phacoemulsification surgery, and retinal rhegmatogenous detachment surgery; Determine the corresponding surgical process based on the determined type of surgery, as well as the technical requirements for each surgical stage, where the technical requirements include the type of surgical instrument, the angle between the surgical instrument and the ophthalmic physical model, the way the hand holds the surgical instrument, and the movement trajectory of the surgical instrument; The degree of compatibility between the determined user operation technique and the technique requirements of each surgical stage is judged, and an evaluation result of the correctness of the operation is generated.

8. An ophthalmology simulation teaching method, characterized in that: include: Generate an ophthalmic three-dimensional model under a specific disease state through a preset ophthalmic disease generation model; Demonstrate ophthalmic 3D models through interactive display devices; Select corresponding materials for different material layers in the ophthalmic three-dimensional model through 3D printing equipment for printing and lamination to obtain an ophthalmic physical model; The user's operation information on the ophthalmic entity model is collected, and the operation information is analyzed to generate an evaluation result of the correctness of the operation.

Images