Accurate advanced UWF-FA image generation method based on LPUWF-LDM model
Through the LPUWF-LDM model, combined with the gated convolutional encoder, low-frequency noise and differential loss function across time regions, the problem of generating high-quality late UWF-FA images is solved, which improves diagnostic accuracy and reduces the fluorescein dyes. rely.
Patent Information
- Application Number
- CN202411146497.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-20
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-08-20
AI Technical Summary
It is difficult for the prior art to generate high-quality late-stage UWF-FA images, especially in the case of a limited paired data set, and the modal differences between UWF-SLO and UWF-FA images make it difficult to effectively fusion and retain lesion characteristics.
Using the precise late UWF-FA image generation method based on the LPUWF-LDM model, high-resolution, detailed preservation details are generated through VAE module with gated convolutional encoder, low-frequency enhanced noise addition module and cross-time region differential loss function. Featured late UWF-FA images.
It significantly improves image quality and diagnostic accuracy, can more accurately reflect the time development of the disease, reduces patients' dependence on fluorescein dyes, and provides new medical image processing ideas.
Smart Images

Figure CN118982599A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of cross-modal image conversion and generation, and in particular to an accurate late UWF-FA image generation method based on a LPUWF-LDM model. Background Art
[0002] Ultra-Wide-Field Fluorescein Angiography (UWF-FA) is a key dynamic imaging technology used for the diagnosis and treatment of fundus-related diseases. The process involves injecting fluorescein dye into the patient's vein, which then flows to the fundus, which may cause nausea or vomiting. There are also risks for patients with serious heart diseases. In contrast, Ultra-Wide-Field Scanning Laser Ophthalmoscopy (UWF-SLO) uses laser imaging to quickly scan the retina without adverse effects on the patient. However, the vascular images produced by UWF-SLO are less detailed. UWF-FA is mainly divided into two stages: early and late, and doctors will interpret the images at specific times depending on the condition. Early images are particularly important for retinal vascular problems, such as central retinal artery occlusion and retinal vein occlusion; late images are more critical when evaluating retinal structure or barrier problems, such as central serous chorioretinopathy. In addition, some features of the condition become more apparent in advanced stages, such as the 'nut sign' in late stages of central serous choroidal disease and tiny hemorrhages in diabetic retinopathy.
[0003] Previous studies, such as VTGAN and Reg-GAN, have attempted to use cross-modal generation techniques to convert UWF-SLO images into UWF-FA images to address this challenge. However, these methods only rely on supervised learning of a single UWF-SLO and early-stage UWF-FA images, thereby ignoring the dynamic characteristics of UWF-FA images over time. For the assessment of diseases involving retinal structure, such as central serous chorioretinopathy, key information is often reflected in late-stage UWF-FA images. However, due to the scarcity of high-quality and paired UWF-SLO, early-stage and late-stage UWF-FA datasets, generating high-quality late-stage UWF-FA images from UWF-SLO images has become a daunting challenge.
[0004] In general, the current technical problems are as follows:
[0005] (1) Generate high-quality late-stage UWF-FA with detailed lesion information to assist diagnosis; (2) Maintain the quality of the generated late-stage UWF-FA with a limited paired dataset; (3) There are significant differences between UWF-SLO and UWF-FA images in imaging principles, information density, and lesion manifestations. It is necessary to effectively fuse the information of these two modalities to retain key lesion features and improve image quality during the conversion process. Summary of the invention
[0006] In view of the shortcomings of the prior art, the present invention proposes an accurate late UWF-FA image generation method based on the LPUWF-LDM model. The method uses the input UWF-SLO image as the data source and automatically synthesizes a high-resolution, high-definition and detailed late UWF-FA image through an advanced algorithm, thereby significantly improving the image quality and diagnostic accuracy. In order to solve the above technical problems, the technical solution of the present invention is:
[0007] A method for generating an accurate late UWF-FA image based on the LPUWF-LDM model comprises the following steps:
[0008] S1. Collect image data and preprocess them, wherein one set of image data includes paired UWF-SL0 images and early UWF-FA images, and the other set of image data includes paired UWF-SLO images, early UWF-FA images, and late UWF-FA images;
[0009] Regarding data preprocessing, in order to ensure data consistency, the image resolution of the two datasets was adjusted to 853x682 pixels, and various enhancement measures were applied, including random rotation within [-5°, 5°], cropping to 512x512 pixels, and random flipping.
[0010] S2. Build an accurate late UWF-FA image generation network model based on the LPUWF-LDM model. The overall framework of the accurate late UWF-FA image generation network model based on the LPUWF-LDM model mainly consists of three parts: a VAE module with a gated convolutional encoder, a noise addition module using low-frequency enhancement noise, and a noise prediction backbone trained by a cross-temporal regional difference loss (CTRD loss) function.
[0011] The VAE module with gated convolutional encoder consists of a VAE backbone and a gated convolutional encoder. The VAE backbone contains an encoder and a decoder, as well as a discriminator to improve the quality of the generated image. First, the late UWF-FA image (y i) is fed into the encoder. It first passes through a convolutional layer and then through three downsampling layers, each of which consists of two residual blocks and one downsampling block. The output of the encoder provides the mean and variance in the latent space, from which the latent vector is sampled The latent vector is then fed into the decoder, which consists of three upsampling blocks. Since the decoder also receives additional spatial information, each upsampling block checks this additional information and merges it into the vector of the current layer if available. The vector then passes through two residual blocks and a self-attention block, and finally it is reconstructed through a group normalization layer and a convolutional layer. and transform it into the final output. It can be expressed as:
[0012]
[0013] Where σ represents nonlinear computation, Conv represents convolution operation, and · represents element-wise multiplication. The training of the entire VAE consists of two steps. First, transfer learning is performed from the stable diffusion AutoencoderKL, and then it is fine-tuned on the dataset. The loss function used is:
[0014]
[0015] Where D is the discriminator, Perc is the perceptual loss, and KL is the Kullback-Leibler loss. The loss function of the D discriminator is:
[0016]
[0017] When training the gated convolutional encoder, only set its parameters to be trainable and remove the KL loss from the loss function while keeping other components unchanged. Continue training until the model is well fitted.
[0018] Regarding the noise addition module that utilizes low-frequency enhancement noise, this is achieved by adding a low-frequency noise component to the original Gaussian noise, which follows a standard normal distribution. To generate this low-frequency noise, a value a is first sampled from N(0,1) and a scaling factor β is sampled from N(0,0.5) to control the degree of low-frequency noise. Then, a and β are multiplied and the resulting product is replicated in the height and width dimensions of the image, creating a low-frequency noise term ∈ L By adding this low-frequency noise term to the original noise, the new noise vector z t The same distribution assumption is maintained. Therefore, the new z t It can be expressed as:
[0019]
[0020] Among them ∈ and ∈ L Represent pure noise and low-frequency noise respectively.
[0021] In the above technical solution, under the common noise adding strategy, the low frequency still maintains a relatively clear texture, while the strategy of the present invention makes the damage of high-frequency and low-frequency noise more balanced.
[0022] Regarding the design of the cross-time regional difference loss function. The overall learning goal based on the diffusion model is:
[0023]
[0024] Where α is a hyperparameter representing the weight of this loss term. In the experiment, α gradually increases from 0.25 to 1 in the first half of the training cycle and remains at 1 in the second half of the training cycle. w represents the above heat map, which is related to the early and late UWF-FA. Represents a noisy image, where t represents the current step size of adding noise, x i Enter the condition.
[0025] S3, using the preprocessed UWF-SLO images and early and late UWF-FA as input to repeatedly train the network model, optimize the network parameters, and continuously perform iterative optimization;
[0026] S4. The preprocessed UWF-SLO image is input into the trained accurate late UWF-FA image generation network model based on the LPUWF-LDM model, and finally a high-quality late UWF-FA image is generated.
[0027] The present invention has the following characteristics and beneficial effects:
[0028] The innovative technical solution of the present invention truly realizes the medical demand of automatically generating high-quality late ultra-wide-angle fluorescein angiography images based on ultra-wide-angle fundus color photos, and has the following characteristics: 1) The present invention introduces a cross-time regional difference loss module, which can capture the dynamic change characteristics of UWF-FA images from early to late stages. This design enables the model to more accurately reflect the actual development of the disease over time when generating late UWF-FA images, especially in the lesion area and fine details. 2) In the diffusion forward modeling process, the present invention adopts a low-frequency enhancement noise strategy. This strategy helps to improve the realism of the generated image, making the generated late UWF-FA image closer to the actual image in visual effect, and enhancing the reliability of medical diagnosis. 3) The present invention uses a gated convolutional encoder to extract additional information from conditional images (such as UWF-SLO images). The gating mechanism can adaptively control the information flow and improve the mapping ability of the encoder on a limited data set, thereby generating higher quality late UWF-FA images. The beneficial effects of the present invention are as follows: 1) By generating high-quality late-stage UWF-FA images, the present invention helps doctors to more accurately evaluate retinal structure and vascular lesions, thereby improving the diagnostic accuracy of fundus-related diseases. 2) Traditional UWF-FA requires injection of dye, which may cause adverse reactions to patients. The present invention reduces the need for dye injection and reduces the risk to patients by generating UWF-FA images from UWF-SLO images. 3) Due to the scarcity of existing high-quality and paired UWF-SLO and UWF-FA data sets, the successful application of the present invention provides new ideas and methods for similar medical image processing problems, and is expected to be promoted and applied in other fields. 4) The present invention has made important progress in cross-modal generation technology, latent diffusion models, and medical image processing, providing a solid foundation and reference for future research and application. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0030] Figure 1 It is a schematic diagram of the process of the present invention;
[0031] Figure 2 The overall architecture of the accurate late UWF-FA image generation network model based on the LPUWF-LDM model of the present invention;
[0032] Figure 3The architecture diagram of the VAE module with a gated convolutional encoder.
[0033] Figure 4 (a), (b), and (c) are the comparison results of the real UWF-SLO image, the real late UWF-FA image, and the late UWF-FA image generated by the model. DETAILED DESCRIPTION
[0034] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.
[0035] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.
[0036] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.
[0037] The present invention provides an accurate late UWF-FA image generation method based on the LPUWF-LDM model, such as Figure 1 As shown, the specific operations are as follows:
[0038] Step 1: Data acquisition and preprocessing. This embodiment uses data sets from two cooperative hospitals. The data set of the first hospital mainly includes paired UWF-SL0 images and early UWF-FA images. In this embodiment, these data must first be strictly screened to exclude images that may affect clinical diagnosis. The specific screening requirements include: 1) The capture interval must not exceed three months; 2) Visible fresh bleeding; 3) Severe eyelash occlusion; 4) Poorly focused images. Finally, 304 pairs of high-quality images were selected from the hospital. The data set of the second hospital provides paired UWF-SLO images, early UWF-FA images, and late UWF-FA images. After the same screening, 387 groups of images were selected from the hospital, each of which contains all three image types. These images are then divided into 309 pairs as training sets and 78 pairs as test sets.
[0039] It should be noted that in this embodiment, the early UWF-FA images and the late UWF-FA images are classified according to the doctor's definition. The early UWF-FA images are 0-15 seconds after the patient is injected with fluorescein dye, and the late UWF-FA images are 10 minutes after the patient is injected with fluorescein dye. Further, with regard to data preprocessing, in order to ensure the consistency of the data, the resolution of all images is adjusted to 853x682 pixels, and various enhancement measures are applied. Including random rotation within [-5°, 5°], cropping to 512x512 pixels and random flipping. Specifically, in this embodiment, these enhancements effectively expand the original training data set from 304 pairs to 19,456 pairs, and expand 309 groups to 6,192 groups, totaling 25,648 UWF-SL0 images, 25,648 early UWF-FA images and 6,192 late UWF-FA images. Finally, all the early UWF-FA and matching UWF-SLO images in the datasets provided by the two hospitals are used as the first dataset (a total of 25,648 UWF-SL0 images and 25,648 early UWF-FA images). This dataset will be used for extensive model training so that the model can first learn the structural features of the images. All the late UWF-FA and matching UWF-SLO images in the datasets provided by the two hospitals are used as the second dataset (a total of 6,192 UWF-SL0 images and 6,192 early UWF-FA images). The second dataset is used for training the model to generate late UWF-FA images.
[0040] Step 2: If Figure 2 As shown in Figure 1, a precise late UWF-FA image generation network model based on the LPUWF-LDM model is built. Specifically, the overall framework of the precise late UWF-FA image generation network model based on the LPUWF-LDM model mainly consists of three parts: a VAE module with a gated convolutional encoder (see Figure 3 ), a noise addition module that utilizes low-frequency enhancement noise, and a noise prediction backbone trained with a cross-temporal region difference loss function.
[0041] Specifically, a VAE module with a gated convolutional encoder, such as Figure 3 As shown in Figure 1, it consists of a VAE backbone and a gated convolutional encoder. The VAE backbone contains an encoder and a decoder, as well as a discriminator to improve the quality of the generated image. First, the late UWF-FA image (y i ) is fed into the encoder. It first passes through a convolutional layer and then through three downsampling layers, each of which consists of two residual blocks and one downsampling block. The output of the encoder provides the mean and variance in the latent space, from which the latent vector is sampled The latent vector is then fed into the decoder, which consists of three upsampling blocks. Since the decoder also receives additional spatial information, each upsampling block checks this additional information and merges it into the vector of the current layer if available. The vector then passes through two residual blocks and a self-attention block, and finally it is reconstructed through a group normalization layer and a convolutional layer. and transform it into the final output. It can be expressed as:
[0042]
[0043] Where σ represents nonlinear computation, Conv represents convolution operation, and · represents element-wise multiplication. The training of the entire VAE consists of two steps. First, transfer learning is performed from the stable diffusion AutoencoderKL, and then it is fine-tuned on the dataset. The loss function used is:
[0044]
[0045] Where D is the discriminator, Perc is the perceptual loss, and KL is the Kullback-Leibler loss. The loss function of the D discriminator is:
[0046]
[0047] When training the gated convolutional encoder, only its parameters are set to be trainable and the KL loss is removed from the loss function while keeping the other components unchanged. Continue training until the model is well fitted. Regarding the noise addition module that utilizes low-frequency enhancement noise, the traditional approach is to add the same distribution to the image. Gaussian noise affects high-frequency and low-frequency information differently, and when the noisy UWF-FA image is decomposed into these components, a significant difference in noise injection between high-frequency and low-frequency regions is observed. Specifically, the low-frequency region is less perturbed by noise than the high-frequency region. This difference impairs the model's ability to recover the rich low-frequency details in medical images. To address this issue, the present embodiment adds noise perturbations to the low-frequency information. This is achieved by adding a low-frequency noise component to the original Gaussian noise, which follows a standard normal distribution.
[0048] Specifically, to generate this low-frequency noise, first sample a value a from N(0,1) and a scaling factor β from N(0,0.5) to control the degree of low-frequency noise. Then, multiply a and β and replicate the resulting product in the height and width dimensions of the image to create a low-frequency noise term ∈ L Under the common noise adding strategy, the low frequency still maintains a relatively clear texture, while the strategy of this embodiment makes the damage of high-frequency and low-frequency noise more balanced. By adding this low-frequency noise term to the original noise, the new noise vector z t The same distribution assumption is maintained. Therefore, the new z t It can be expressed as:
[0049]
[0050] Among them ∈ and ∈ L Represent pure noise and low-frequency noise respectively.
[0051] Regarding the cross-temporal regional difference loss function. It should be noted that the early UWF-FA is very clear in terms of structural information, such as the locations of arteries, veins, and optic discs. Therefore, the model can better learn the mapping from UWF-SLO to the early UWF-FA. However, although structural information can guide the realism of image generation, it is not conducive to generating detailed information related to lesions, because more lesion information is actually reflected in the later UWF-FA. Therefore, the designed function mainly solves two problems: 1) How to locate the lesion area without manual annotation; 2) How to make the diffusion model pay more attention to these areas.
[0052] Therefore, in this embodiment, a cross-time regional difference loss function is proposed. Specifically, the early and late UWF-FA images are first registered, and then the absolute difference pixel by pixel is calculated, and the result is normalized to the range of (0,1) to obtain a heat map with the same shape and size as the UWF-FA image. Areas with high values indicate that the difference between the early and late stages is large, and areas with low values indicate that the difference is small. After statistics, the average difference induced by fluorescence leakage is 0.87, and the average difference induced by neovascularization is 0.71. Positions with higher values indicate that the model should focus more on generating these areas, forming an unsupervised attention to the lesion area. Since diffusion is trained in the latent space, the heat map is resized to match the same size as the latent space. Therefore, the overall learning goal based on the diffusion model is:
[0053]
[0054] Where α is a hyperparameter representing the weight of this loss term. In this embodiment, α gradually increases from 0.25 to 1 in the first half of the training cycle and remains at 1 in the second half of the training cycle. w represents the above heat map, which is related to the early and late UWF-FA. Represents a noisy image, where t represents the current step size of adding noise, x i Enter the condition.
[0055] Step 3: In this embodiment, the Pytorch 2.0 and cuda 11.8 frameworks are used to build an accurate late UWF-FA image generation network model based on the LPUWF-LDM model, and two NVIDIA A100 GPUs are used to train it. The training process includes the VAE component and the noise prediction backbone. First, the noise prediction backbone is extensively trained on the first dataset, with 1000 training rounds, a batch size of 32, and 56 hours of training. Then, the training is performed on the second dataset, and both the VAE component and the noise prediction backbone are trained for 1000 rounds. The batch size used by the VAE component is 8, and the training is 52 hours. The batch size used by the noise prediction backbone is 32, and the training is 33 hours. The Adam algorithm is used for optimization during the training process, and the learning rate of the VAE component is set to 0.0002, and the learning rate of the noise prediction backbone is set to 0.0001.
[0056] Step 4: In the model inference stage, the test set is applied and the UWF-SLO image is input. After the UWF-SLO has undergone the preprocessing stage, the trained accurate late UWF-FA image generation network model based on the LPUWF-LDM model is used to finally generate a high-quality late UWF-FA image. Figure 4From left to right, the comparison effect diagrams of the real UWF-SLO image, the real late UWF-FA image and the late UWF-FA image generated by the model are shown respectively; Figure 4 From left to right, the ultra-wide-angle fundus color photograph, the real late ultra-wide-angle fluorescein angiography image, and the model-generated late ultra-wide-angle fluorescein angiography image are compared. The first two rows show the normal conversion from UWF-SLO image to late UWF-FA image, while the last row shows the conversion when obvious lesions are present. The method of this embodiment performs well in generating blood vessels and optic disc structures, with more continuous blood vessels and blood vessel thickness closer to the real situation. In addition, in the third row, our method is more successful than other methods in generating laser scars, and our method can correctly identify and convert lesions even with a small data set. This shows that our model has a particular advantage in generating tasks with limited data.
[0057] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments including components are made without departing from the principles and spirit of the present invention, and still fall within the scope of protection of the present invention.
Claims
1. An accurate late UWF-FA image generation method based on the LPUWF-LDM model, characterized in that: The steps include: S1. Collect and preprocess two sets of image data, one set of image data includes paired UWF-SL0 images and early UWF-FA images, and the other set of image data includes paired UWF-SLO images, early UWF-FA images, and late UWF-FA images; S2. Building an accurate late UWF-FA image generation network model based on the LPUWF-LDM model, wherein the accurate late UWF-FA image generation network model based on the LPUWF-LDM model includes a VAE module with a gated convolutional encoder, a noise addition module using low-frequency enhancement noise, and a noise prediction backbone trained by a cross-temporal region difference loss function; S3, using the preprocessed UWF-SLO image and early and late UWF-FA as input to repeatedly train the accurate late UWF-FA image generation network model based on the LPUWF-LDM model, optimize the network parameters, and continuously perform iterative optimization; S4. The preprocessed UWF-SLO image is input into the trained accurate late UWF-FA image generation network model based on the LPUWF-LDM model, and finally a high-quality late UWF-FA image is generated.
2. According to claim 1, a method for generating an accurate late UWF-FA image based on the LPUWF-LDM model, characterized in that: The preprocessing method is as follows: the image resolution of the two datasets is adjusted to 853x682 pixels, and data enhancement is performed by randomly rotating within [-5°, 5°], cropping to 512x512 pixels, and randomly flipping operations.
3. According to claim 1, a method for generating an accurate late UWF-FA image based on the LPUWF-LDM model, characterized in that: The VAE module with a gated convolutional encoder includes a VAE backbone and a gated convolutional encoder, wherein the VAE backbone includes an encoder and a decoder, as well as a discriminator to improve the quality of the generated image.
4. The method for generating an accurate late UWF-FA image based on the LPUWF-LDM model according to claim 3, characterized in that: The implementation method of the VAE module with a gated convolutional encoder is as follows: The later UWF-FA image y i The encoder is fed into a convolutional layer and then three downsampling layers. Each downsampling layer consists of two residual blocks and one downsampling block. The output of the encoder provides the mean and variance in the latent space, from which the latent vector is sampled. The latent vector is then fed into the decoder, which consists of three upsampling blocks. Since the decoder also receives additional spatial information, each upsampling block checks this additional information and merges it into the vector of the current layer if available. The vector then passes through two residual blocks and a self-attention block. Finally, it is reconstructed through a group normalization layer and a convolutional layer. and transform it into the final output, It can be expressed as: Where σ represents nonlinear computation, Conv represents convolution operation, and · represents element-wise multiplication.
5. The method for generating an accurate late UWF-FA image based on the LPUWF-LDM model according to claim 4, characterized in that: The training method of the VAE module with the gated convolutional encoder includes: First, transfer learning is performed from the stable diffusion AutoencoderKL, which is then fine-tuned on the dataset using the loss function: Where D is the discriminator, Perc is the perceptual loss, and KL is the Kullback-Leibler loss; When training the gated convolutional encoder, only its parameters are set to be trainable and the KL loss is removed from the loss function while keeping other components unchanged and continuing training until the model is well fitted.
6. The method for generating an accurate late UWF-FA image based on the LPUWF-LDM model according to claim 5, characterized in that: The loss function of the D discriminator is:
7. The method for generating an accurate late UWF-FA image based on the LPUWF-LDM model according to claim 1, characterized in that: In step S2, the method of adding noise using the noise adding module of low-frequency enhanced noise is: First, a value a is sampled from N(0,1) and a proportional factor β is sampled from N(0,0.5) to control the degree of low-frequency noise; Then, a and β are multiplied and the resulting product is replicated across the height and width dimensions of the image, creating a low-frequency noise term ∈ L ; By adding this low-frequency noise term to the original noise, the new noise vector z t The same distribution assumption is maintained, so the new z t It can be expressed as: Among them ∈ and ∈ L Represent pure noise and low-frequency noise respectively.
8. The method for generating an accurate late UWF-FA image based on the LPUWF-LDM model according to claim 1, characterized in that: In step S3, the cross-time regional difference loss function is: Where α is a hyperparameter representing the weight of this loss term, and w represents the heat map, which is related to the early and late UWF-FA. Represents a noisy image, where t represents the current step size of adding noise, x i Enter the condition.
Citation Information
Patent Citations
Visual question and answer enhancement method based on graph convolution
CN110399518A
Restorated image detection method based on multistage gating unit
CN116777855A
Dynamic ultra-wide-angle fluorescein contrast synthesis method based on conditional diffusion model
CN118072921A
