A training method and application method of a microscopic image generation model and related systems

Abstract

The application discloses a training method and application method of a microscopic image generation model and related systems, relates to the technical field of microscopic image generation, and comprises the following steps: acquiring an original microscopic image; performing Gaussian noise adding processing on the original microscopic image to obtain a first microscopic image; generating a second microscopic image by using a generator in the microscopic image generation model according to the first microscopic image; performing Gaussian noise adding processing on the second microscopic image to obtain a third microscopic image; calculating the probability that the third microscopic image is a reasonable de-noising distribution of the first microscopic image by using a discriminator in the microscopic image generation model according to the third microscopic image and the first microscopic image; and training the microscopic image generation model according to the probability that the third microscopic image is a reasonable de-noising distribution of the first microscopic image and an adversarial loss function, so as to obtain a trained microscopic image generation model. The application can improve the quality and efficiency of generated microscopic images.

Description

Technical Field

[0001] This application relates to the field of microscopic image generation technology, and in particular to a training method, application method and related system for a microscopic image generation model. Background Technology

[0002] In modern medical diagnostics, microscopic images, as a crucial type of medical imaging data, play an irreplaceable role in clinical diagnosis and pathological research. Particularly in the diagnosis of parasitic diseases, physicians often rely on microscopic images of parasites to identify pathogens and observe pathological changes. However, high-quality parasite microscopic images are difficult to obtain due to the complexity of the parasite life cycle, the scarcity of infection cases, and limitations in sampling conditions. These images often require acquisition under specific experimental conditions by experienced professionals operating microscopes, and the integrity and representativeness of the samples must be ensured during acquisition, making the process extremely challenging and time-consuming. Therefore, developing a method and system capable of efficiently generating high-quality parasite microscopic images is of great significance for advancing medical research, improving diagnostic efficiency, and advancing medical education.

[0003] Although deep learning technology is now widely used in the generation of microscopic images, existing methods still have some limitations in practice. The most significant problem is the long generation time, and the quality of the generated microscopic images cannot fully meet the needs of high-precision medical diagnosis. Especially in the context of the subtle, complex, and variable features of microscopic images, existing methods often struggle to capture sufficient detail, failing to generate high-quality microscopic images and thus affecting their application value in clinical medicine.

[0004] Therefore, how to generate higher quality microscopic images more efficiently and improve the quality and efficiency of generating microscopic images has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] The purpose of this application is to provide a training method, application method, and related system for a microscopic image generation model, which can improve the quality and efficiency of generated microscopic images.

[0006] To achieve the above objectives, this application provides the following solution:

[0007] Firstly, this application proposes a training method for a microscopic image generation model, comprising:

[0008] Obtain the raw microscopic images.

[0009] Gaussian noise was added to the original microscopic image to obtain the first microscopic image.

[0010] Based on the first microscopic image, a second microscopic image is generated using the generator in the microscopic image generation model; the microscopic image generation model is a model composed of a diffusion model and a generative adversarial model, and the generative adversarial model includes a generator and a discriminator.

[0011] Gaussian noise is added to the second microscopic image to obtain a third microscopic image; the Gaussian noise added to the third microscopic image and the first microscopic image are both multi-step Gaussian noise, and the Gaussian noise added to the first microscopic image is t-step Gaussian noise, while the Gaussian noise added to the third microscopic image is t-1-step Gaussian noise.

[0012] Based on the third microscopic image and the first microscopic image, the probability that the third microscopic image is a reasonable denoised distribution of the first microscopic image is calculated using the discriminator in the microscopic image generation model.

[0013] Based on the probability of using the third microscopic image as a reasonable denoising distribution of the first microscopic image and the adversarial loss function, the microscopic image generation model is trained to obtain a trained microscopic image generation model.

[0014] Optionally, the original microscopic image is processed by adding Gaussian noise to obtain a first microscopic image, specifically represented as follows:

[0015] X t =U(X0,t);

[0016] Among them, X t Let X0 represent the first microscopic image, U(·) represent the original microscopic image, t represent the Gaussian noise addition operation, and t represent the noise addition step number, where t∈[1,2,…,T].

[0017] Optionally, based on the first microscopic image, a second microscopic image is generated using the generator in the microscopic image generation model, specifically as follows:

[0018]

[0019] in, X represents the second microscopic image, X0 represents the original microscopic image, G(·) represents the generator, and X t Let Z represent the first microscopic image, Z represent a random Gaussian variable, and t represent the number of noise addition steps, where t∈[1,2,…,T];

[0020] Gaussian noise was added to the second microscopic image to obtain the third microscopic image, which is specifically represented as follows:

[0021]

[0022] in, The third microscopic image is represented by U(·), which indicates the Gaussian noise addition operation, and t-1 indicates the number of noise addition steps applied.

[0023] Optionally, the adversarial loss function is expressed as:

[0024]

[0025] in, This represents the discriminator loss function. Represents the generator loss function. This represents the third microscopic image, X. t-1 This represents the photomicrograph of the previous noise-adding step, where D(·) represents the discriminator, and X... t Let represent the first microscopic image, and t represent the noise addition step number, t∈[1,2,…,T].

[0026] Secondly, this application proposes an application method for a microscopic image generation model, including:

[0027] Determine the target image to be generated;

[0028] Random Gaussian noise with the same shape as the target image matrix to be generated is input into a trained microscopic image generation model to generate a new high-quality microscopic image; the trained microscopic image generation model is a model trained according to the training method of a microscopic image generation model described in the first aspect.

[0029] Thirdly, this application proposes a training system for a microscopic image generation model, comprising:

[0030] The raw microscopic image acquisition module is used to acquire raw microscopic images.

[0031] The first noise-adding module is used to add Gaussian noise to the original microscopic image to obtain the first microscopic image.

[0032] The generation module is used to generate a second microscopic image based on the first microscopic image using a generator in a microscopic image generation model; the microscopic image generation model is a model composed of a diffusion model and a generative adversarial model, and the generative adversarial model includes a generator and a discriminator.

[0033] The second noise-adding module is used to add Gaussian noise to the second microscopic image to obtain a third microscopic image; the Gaussian noise added to the third microscopic image and the first microscopic image are both multi-step Gaussian noise, and the Gaussian noise added to the first microscopic image is t-step Gaussian noise, while the Gaussian noise added to the third microscopic image is t-1-step Gaussian noise.

[0034] The discrimination module is used to calculate the probability that the third microscopic image is a reasonable denoised distribution of the first microscopic image, based on the third microscopic image and the first microscopic image, using the discriminator in the microscopic image generation model.

[0035] The model training module is used to train the microscopic image generation model based on the probability of the third microscopic image serving as a reasonable denoising distribution of the first microscopic image and the adversarial loss function, so as to obtain a trained microscopic image generation model.

[0036] Optionally, the first noise-adding module adds Gaussian noise to the original microscopic image to obtain a first microscopic image, specifically represented as follows:

[0037] X t =U(X0,t);

[0038] Among them, X t Let X0 represent the first microscopic image, U(·) represent the original microscopic image, t represent the Gaussian noise addition operation, and t represent the noise addition step number, where t∈[1,2,…,T].

[0039] Optionally, the generation module generates a second microscopic image based on the first microscopic image using the generator in the microscopic image generation model, specifically as follows:

[0040]

[0041] in, X represents the second microscopic image, X0 represents the original microscopic image, G(·) represents the generator, and X t Let Z represent the first microscopic image, Z represent a random Gaussian variable, and t represent the number of noise addition steps, where t∈[1,2,…,T];

[0042] Gaussian noise was added to the second microscopic image to obtain the third microscopic image, which is specifically represented as follows:

[0043]

[0044] in, The third microscopic image is represented by U(·), which indicates the Gaussian noise addition operation, and t-1 indicates the number of noise addition steps applied.

[0045] Optionally, the adversarial loss function is expressed as:

[0046]

[0047] in, This represents the discriminator loss function. Represents the generator loss function. This represents the third microscopic image, X. t-1 This represents the photomicrograph of the previous noise-adding step, where D(·) represents the discriminator, and X... t Let represent the first microscopic image, and t represent the noise addition step number, t∈[1,2,…,T].

[0048] Fourthly, this application proposes an application system for a microscopic image generation model, including:

[0049] The target image determination module is used to determine the target image to be generated;

[0050] The model application module is used to input random Gaussian noise with the same shape as the target image matrix to be generated into the trained microscopic image generation model to generate a new high-quality microscopic image; the trained microscopic image generation model is a model trained according to the training method of a microscopic image generation model described in the first aspect.

[0051] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0052] This application provides a training method, application method, and related system for a microscopic image generation model. Gaussian noise is introduced into the original microscopic image to enrich its representation. Using a generator in the microscopic image generation model, a second microscopic image is generated based on a noisy first microscopic image. The generator achieves denoising of the noisy first microscopic image, outputting a large number of high-quality second microscopic images. Then, using the same Gaussian noise processing method, t-1 steps of Gaussian noise are applied to the second microscopic image, resulting in a noisy third microscopic image. A discriminator calculates the probability that the third microscopic image is a reasonable denoising distribution for the first microscopic image based on the first and third microscopic images, thereby enabling the discriminator to evaluate the denoising effect of the microscopic images output by the generator. In training the microscopic image generation model, the process of denoising noisy microscopic images by the generator and evaluating the denoising effect by the discriminator employs an adversarial loss optimization network to enhance the model's generation and discrimination capabilities. This facilitates the large-scale generation of high-quality microscopic image data of the same category using the trained microscopic image generation model, effectively improving the quality and efficiency of the generated microscopic images. This application enables the generation of a large number of microscopic images efficiently using only a small number of real microscopic image samples without compromising image category information, which is of great significance for clinical medical diagnosis and research. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a schematic flowchart illustrating a training method for a microscopic image generation model provided in an embodiment of this application.

[0055] Figure 2 This is a schematic diagram of the structure of a microscopic image generation model provided in an embodiment of this application.

[0056] Figure 3 This is a schematic diagram illustrating the principle of a training method for a microscopic image generation model provided in an embodiment of this application.

[0057] Figure 4 This is a structural block diagram of a training system for a microscopic image generation model provided in one embodiment of this application.

[0058] Figure 5 This is a structural block diagram of an application system for a microscopic image generation model provided in one embodiment of this application. Detailed Implementation

[0059] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0060] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0061] Example 1

[0062] like Figure 1 As shown, this embodiment proposes a training method for a microscopic image generation model. It combines a diffusion model with a generative adversarial network (GAN) to form a diffusion generative adversarial model (DGAM), which serves as the microscopic image generation model. In other words, the microscopic image generation model in this embodiment uses a diffusion generative adversarial model. Figure 2As shown, this method aims to efficiently generate high-quality microscopic images. Through this diffusion-generative adversarial model, key features of microscopic images, such as cell morphology and structural details, can be rapidly and accurately reproduced, effectively supporting medical research and diagnostic processes and providing strong technical support for clinical and laboratory research. The training method specifically includes the following steps:

[0063] Step S1: Obtain the original microscopic image.

[0064] Step S2: Add Gaussian noise to the original microscopic image to obtain the first microscopic image.

[0065] Step S3: Based on the first microscopic image, generate a second microscopic image using the generator in the microscopic image generation model. The microscopic image generation model is a model composed of a diffusion model and a generative adversarial model, wherein the generative adversarial model includes a generator and a discriminator. The diffusion model is used to learn the information attenuation pattern caused by noise, and then uses the learned denoising distribution to help the generative adversarial model recover the original microscopic image. The generative adversarial model constructs an adversarial training framework using a generator and a discriminator, and improves the performance of the generative adversarial model through an adversarial training process between the generator and the discriminator.

[0066] Step S4: Add Gaussian noise to the second microscopic image to obtain a third microscopic image; that is, apply t-1 steps of Gaussian noise to the generated second microscopic image to obtain the third microscopic image. The Gaussian noise added to both the third and first microscopic images is multi-step Gaussian noise, with t-step Gaussian noise added to the first microscopic image and t-1 steps added to the third microscopic image.

[0067] Step S5: Based on the third microscopic image and the first microscopic image, use the discriminator in the microscopic image generation model to calculate the probability that the third microscopic image is a reasonable denoised distribution of the first microscopic image.

[0068] Step S6: Based on the probability of the third microscopic image serving as a reasonable denoising distribution for the first microscopic image and the adversarial loss function, train the microscopic image generation model to obtain a trained microscopic image generation model. The trained microscopic image generation model is used to efficiently and in large quantities generate high-quality microscopic images with a style similar to the original microscopic images after inputting random Gaussian noise. The adversarial loss function includes a term for the reasonable denoising probability of the third microscopic image serving as the first microscopic image and a term for the reasonable denoising probability distribution of the microscopic image from the previous denoising step serving as the first microscopic image.

[0069] Figure 3The principle of the training method for the microscopic image generation model is shown, such as... Figure 3 As shown, the main technical principles of this embodiment in training the microscopic image generation model are as follows:

[0070] Step 101: Perform multi-step Gaussian noise injection on the original microscopic image to obtain the first microscopic image. The first microscopic image is the original microscopic image after noise addition.

[0071] In this embodiment, a multi-step noise-adding process based on Gaussian distribution is performed on the original microscopic image X0 to generate a noisy image X. t and X t-1 The noise-adding process can be expressed as equation (1):

[0072] X t =U(X0,t) (1).

[0073] Among them, X t Let X0 represent the original microscopic image, U(·) represent the Gaussian noise addition operation, and t represent the noise addition step number, t∈[1,2,…,T]. In each training epoch, the algorithm randomly selects a noise addition time step t and adds noise to the microscopic image according to a predetermined Gaussian distribution, thus obtaining the noisy first microscopic image. That is, t is randomly selected from [1,2,…T] in each epoch to assist training. Specifically, given an original microscopic image x0∈X0, the noise addition steps can also be expressed by the following formula:

[0074]

[0075] Where, x t This represents the microscopic image after adding noise at time step t, β t The hyperparameter is set to a sequence that increases linearly with time step between [0,1], and is used to control the amount of noise added to the micrograph at each noise-adding time step t.

[0076] Step 102: The generator outputs a denoised image to obtain a second microscopic image. The second microscopic image is the microscopic image obtained after denoising the first microscopic image using the generator.

[0077] In this embodiment, the microscopic image generation model can be divided into two modules: a generator G(·) and a discriminator D(·). In this step, the input to the generator G(·) includes the noisy time step t and random latent variables. And the first microscopic image X tThe generator can adopt a U-Net structure or a Transformer structure, etc., and this embodiment preferably uses a U-Net structure with multiple ResNet blocks. Meanwhile, the generator in this embodiment borrows design concepts from StyleGAN, introducing latent variables sampled from a random Gaussian distribution. Furthermore, all Group Normalization (GN) layers in the network are replaced with Adaptive Group Normalization (AdaGN) layers to finely enhance the diversity of generated images at different levels, resulting in images that are not only of high quality but also rich and diverse in style. The generator's computation process can be represented as follows:

[0078]

[0079] in, This represents the output microscopic image of the generator, i.e., the second microscopic image; X0 represents the original microscopic image; G(·) represents the generator; X t This represents the first microscopic image, and Z represents a random Gaussian variable. t represents the number of noise-adding steps, t∈[1,2,…,T].

[0080] In this embodiment, when the generator is embedded in the network, the random variable First, the model is transformed into an embedding vector W through a linear layer. Then, the model feature map and the embedding vector W are input into the AdaGN layer, and the detail representation in the image generation process is dynamically adjusted through the following formula to enhance the diversity and realism of the generated images:

[0081]

[0082] Where, m i This represents the i-th channel of the model feature map input to the AdaGN layer, where μ(·) and σ(·) are used to calculate the mean and variance, respectively. s,i ,y b,i )∈Y represents the scaling and translation parameters used to transform the feature map, and the feature map transformation parameters. Learning is performed by a single linear layer, where N l This represents the total number of channels in the feature map of the input model.

[0083] In this embodiment, to ensure that the generator G(·) of the model can fully consider the impact of the noise addition step t on image generation, sinusoidal position embedding is used to transform t into a time-coded vector to embed the generator G(·). The specific formula is as follows:

[0084]

[0085] Where PE(·) represents the function that maps time step t to a time-coded vector, j represents the position index of the encoding, and d is a hyperparameter representing the total length of the time-coded vector.

[0086] Step 103: Apply t-1 step Gaussian noise to the generated second microscopic image to obtain the third microscopic image. The third microscopic image is the second microscopic image after noise addition.

[0087] In this embodiment, the noise-adding function U(·) mentioned in step 101 is used as the second microscopic image output by the generator. By adding Gaussian noise at step t-1, a third microscopic image is obtained. The noise addition formula is similar to that in equation (1), as follows:

[0088]

[0089] in, This represents the third microscopic image. U represents the second microscopic image, U(·) represents the Gaussian noise addition operation, t represents the noise addition step, and t-1 represents the number of noise addition steps applied (t-1 steps). Here, t-1 steps of Gaussian noise with the same distribution as the original image are applied to the generated second microscopic image.

[0090] Step 104: The discriminator evaluates the denoising effect of the microscopic image.

[0091] In this embodiment, both the noisy image and the denoised image are input into the discriminator D(·). After feature extraction using a multi-layer CNN (Convolutional Neural Network), the output features are concatenated and then fed into a classifier composed of multiple fully connected layers for evaluation. The probability that the denoised image is a reasonable denoised distribution of the noisy image is calculated. The discriminator D(·) can adopt a CNN network structure or a Vision Transformer model structure, etc. In this embodiment, the discriminator D(·) preferably uses a CNN network structure with residual connections similar to the generator G(·), and t is embedded into the discriminator D(·) using the same sinusoidal encoding processing method as the generator. The discriminator D(·) is then gradually optimized through a loss function, thereby comprehensively analyzing the above inputs and improving its discrimination ability.

[0092] Step 105: Optimize the DGAM model with adversarial loss.

[0093] In this embodiment, the adversarial loss consists of two parts: discriminator loss and generator loss. The discriminator D(·) aims to identify X as many times as possible. t A reasonable denoising distribution. Therefore, the discriminator loss function The discriminator D(·) is encouraged to recognize X obtained from the processing of the original microscopic image. t-1 To reasonably denoise the distribution, and simultaneously identify the noise generated by generator G(·) This is due to an unreasonable noise distribution. Therefore, the discriminator loss function... Defined as:

[0094]

[0095] in, This represents the discriminator loss function. This represents the third microscopic image, X. t-1 This represents the photomicrograph of the previous noise-adding step, where D(·) represents the discriminator, and X... t Let represent the first microscopic image, and t represent the noise addition step number, t∈[1,2,…,T].

[0096] In this embodiment, the goal of the generator G(·) is to generate an image that retains as many features as possible from the original microscopic image. The generator loss function is... The generator G(·) is encouraged to create images that are difficult for the discriminator D(·) to distinguish. Therefore, the generator loss function... Defined as:

[0097]

[0098] in, Represents the generator loss function. This represents the third microscopic image, X. t Let represent the first microscopic image, and t represent the noise addition step number, t∈[1,2,…,T].

[0099] It should be noted that the formula for adversarial loss is similar to that of generative adversarial loss in GAN networks. Both the generator and the discriminator need to be trained on parameters based on the output of the discriminator. The training objective of the generator is opposite to that of the discriminator.

[0100] Ultimately, the overall adversarial loss function is expressed as a combination of the generator loss and the discriminator loss. During model training, the discriminator and generator are trained alternately using their respective loss functions until a preset number of iterations is reached.

[0101] In this embodiment, the probability values ​​obtained by the discriminator are input into the loss function for model training. During training, the discriminator should be input with two sets of samples: the first microscopic image X. t The noise sample X from the previous step of the first microscopic image t-1 As the first set of samples; the first microscopic image X t and third microscopic image As the second set of samples. First, the noise sample X from the previous step of the first microscopic image.t-1 As the first microscopic image X t The probability value of a reasonable denoised distribution is given by D(t,X) in the loss function. t-1 ,X t () indicates; secondly, the third microscopic image. As the first microscopic image X t The probability value of a reasonable denoised distribution is given by the loss function. The former can be compared to the probability obtained by inputting a real image into the discriminator in the original GAN ​​network. The discriminator should consider it a reasonable denoised distribution, and the higher the probability value, the better. The latter can be compared to the probability obtained by inputting a fake image into the discriminator in the original GAN ​​network. The discriminator should consider it an unreasonable denoised distribution, and the lower the probability value, the better. In this way, as training progresses, the discriminator will gain a stronger ability to distinguish whether the denoised distribution is reasonable; conversely, the generator's loss encourages it to generate images that are more likely to confuse the discriminator.

[0102] This embodiment optimizes the adversarial loss function, which not only improves the quality of generated microscopic images but also ensures the authenticity and reliability of the generated microscopic images.

[0103] In this embodiment, after step 105, which optimizes the DGAM model using adversarial loss and obtains the trained DGAM model, the following steps may also be included:

[0104] Step 106: Generate a large amount of microscopic image data using the trained DGAM model.

[0105] In this embodiment, the trained DGAM model has the ability to efficiently generate microscopic images, producing a large amount of high-quality microscopic image data in a short time. This microscopic image data can not only be used for in-depth research in clinical pathology, but also provide strong data support for disease diagnosis and treatment.

[0106] In this embodiment, given a set of noise O sampled from a random Gaussian distribution with the same size as the generated image, and a fully trained generator G(·) in the DGAM model, high-quality microscopic images with a style similar to the original microscopic images can be generated in batches using the following formula:

[0107]

[0108] Where H(·) represents the cyclic denoising generation process, This represents a set of generated high-quality microscopic images, where T is a pre-defined hyperparameter and O represents the random Gaussian noise input during image generation.

[0109] It should be noted that, in this embodiment, to prevent problems such as instability and mode collapse in the DGAM model, Instead of being generated step-by-step by the generator G(·), the process involves gradually denoising the random Gaussian noise through a series of iterative steps. Specifically, the iterative denoising process represented by H(·) in formula (9) is as follows: Given random Gaussian noise O, a set of reconstructed images is first generated using a well-trained generator G(·). Then, use formula (1) from step 101 to add T-1 steps of noise to the image to obtain the denoised image. Then, repeat the above steps to obtain the denoised image. Wait until the image is gradually restored to a noise-free image. By using this cyclic denoising method to generate images, the DGAM model can decompose the generation process into a series of denoising steps, each of which is relatively simple, ensuring that the model has better stability and pattern coverage.

[0110] In this embodiment, during training, parasite or cell images were obtained at 1000x magnification using a bright light microscope (Olympus IX53), totaling 5741 Toxoplasma gondii samples, 6981 erythrocyte samples, and 5758 Plasmodium samples. Using the DGAM model, batch generation of microscopic images of different categories, including Toxoplasma gondii, erythrocytes, and Plasmodium, was successfully achieved. During generation, the model not only accurately captured key shape and texture information of the microscopic images but also ensured the accuracy of category attributes. The microscopic image generation technology based on the DGAM model guarantees high-quality images while preserving biological significance. Therefore, the model can not only effectively generate microscopic images in batches but also provide strong technical support for medical research, especially in situations where microscopic image samples are scarce. It offers researchers a new data acquisition method with high application value and broad development prospects.

[0111] In this embodiment, the FID (Fréchet Inception Distance) metric was used as a tool to quantify the quality of the generated images. The time required to generate a batch (100 images) of microscopic images on the A100 GPU was recorded to evaluate the model's generation efficiency. Experimental results are shown in Table 1. The results demonstrate that the microscopic image generation method and system based on the microscopic image generation model in this embodiment can efficiently generate hundreds of high-quality microscopic images in less than one second. The generated images are not only visually highly similar to the original microscopic images but also exhibit extremely high consistency in key features. Therefore, the DGAM algorithm has significant practical implications and application value for advancing medical research.

[0112] Table 1. Experimental results of the microscopic image generation effect of the DGAM model.

[0113] Microscopic image categories FID Time (s) Red blood cells 51.58 0.72

[0114] This embodiment proposes a diffusion-generative adversarial model by introducing a novel diffusion process and a targeted adversarial training strategy, which can efficiently generate high-quality microscopic images. This model not only significantly reduces the time cost of image generation but also improves image clarity and detail realism, better preserving the biological characteristics of the original microscopic images. It has the potential to contribute to promoting the development of clinical medicine and improving the accuracy and efficiency of medical image analysis.

[0115] Example 2

[0116] This embodiment provides an application method for a microscopic image generation model. This embodiment is actually a method for the practical application of the model based on the training method of Embodiment 1. The application method includes the following steps:

[0117] Determine the target images to be generated, including their size and number;

[0118] Random Gaussian noise with the same shape as the target image matrix to be generated is input into the trained microscopic image generation model to generate a new high-quality microscopic image; the trained microscopic image generation model is a model trained according to the training method of a microscopic image generation model described in Example 1.

[0119] Example 3

[0120] This embodiment provides a training system for a microscopic image generation model. This embodiment is actually the system corresponding to the training method in Embodiment 1, such as... Figure 4 As shown, the training system specifically includes:

[0121] The raw microscopic image acquisition module 201 is used to acquire raw microscopic images.

[0122] The first noise-adding module 202 is used to add Gaussian noise to the original microscopic image to obtain the first microscopic image.

[0123] The generation module 203 is used to generate a second microscopic image based on the first microscopic image using a generator in a microscopic image generation model; the microscopic image generation model is a model composed of a diffusion model and a generative adversarial model, and the generative adversarial model includes a generator and a discriminator.

[0124] The second noise-adding module 204 is used to add Gaussian noise to the second microscopic image to obtain a third microscopic image; the Gaussian noise added to the third microscopic image and the first microscopic image are both multi-step Gaussian noise, and the Gaussian noise added to the first microscopic image is t-step Gaussian noise, and the Gaussian noise added to the third microscopic image is t-1-step Gaussian noise.

[0125] The discrimination module 205 is used to calculate the probability that the third microscopic image is a reasonable denoised distribution of the first microscopic image based on the third microscopic image and the first microscopic image, using the discriminator in the microscopic image generation model.

[0126] The model training module 206 is used to train the microscopic image generation model based on the probability that the third microscopic image is a reasonable denoising distribution of the first microscopic image and an adversarial loss function, to obtain a trained microscopic image generation model. The adversarial loss function includes a term for the reasonable denoising probability of the third microscopic image as the first microscopic image and a term for the reasonable denoising probability of the microscopic image from the previous denoising step as the first microscopic image.

[0127] This embodiment, through the coordinated operation of the various modules described above, constitutes a highly efficient and reliable microscopic image generation system. Gaussian noise is added to the original microscopic image data to provide necessary data perturbation for the generation process. Then, based on the noise-processed data, a deep learning algorithm is used to generate denoised microscopic images, simulating the characteristics and details of real microscopic images. The image denoising effect is then evaluated to ensure the rationality and accuracy of the denoising process. Next, the DGAM model is trained using an adversarial loss function, enabling it to generate high-quality microscopic images. Finally, the fully trained DGAM model is used to generate a large number of microscopic images in batches to meet the needs of clinical and research applications. This not only improves the speed of image generation but also ensures the high quality and application value of the generated images.

[0128] Example 4

[0129] This embodiment provides an application system for a microscopic image generation model. This embodiment is actually the system corresponding to the application method in Embodiment 2, such as... Figure 5 As shown, the application system specifically includes:

[0130] The target image determination module 207 is used to determine the target image to be generated.

[0131] The model application module 208 is used to input random Gaussian noise with the same shape as the target image matrix to be generated into the trained microscopic image generation model to generate a new high-quality microscopic image; the trained microscopic image generation model is a model trained according to the training method of a microscopic image generation model described in Example 1.

[0132] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0133] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0134] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A training method for a microscopic image generation model, characterized in that, include: Acquire raw microscopic images; Gaussian noise was added to the original microscopic image to obtain the first microscopic image; Based on the first microscopic image, a second microscopic image is generated using the generator in the microscopic image generation model; the microscopic image generation model is a model composed of a diffusion model and a generative adversarial model, and the generative adversarial model includes a generator and a discriminator. Gaussian noise is added to the second microscopic image to obtain a third microscopic image; the Gaussian noise added to the third microscopic image and the first microscopic image are both multi-step Gaussian noise, and the Gaussian noise added to the first microscopic image is t-step Gaussian noise, while the Gaussian noise added to the third microscopic image is t-1-step Gaussian noise. Based on the third microscopic image and the first microscopic image, the probability that the third microscopic image is a reasonable denoised distribution of the first microscopic image is calculated using the discriminator in the microscopic image generation model. Based on the probability of using the third microscopic image as a reasonable denoising distribution of the first microscopic image and the adversarial loss function, the microscopic image generation model is trained to obtain a trained microscopic image generation model.

2. The training method for the microscopic image generation model according to claim 1, characterized in that, The original microscopic image is processed by adding Gaussian noise to obtain the first microscopic image, which is specifically represented as follows: X t = U(X0,t); Among them, X t Let X0 represent the first microscopic image, U(·) represent the original microscopic image, t represent the Gaussian noise addition operation, and t represent the noise addition step number, where t∈[1,2,…,T].

3. The training method for the microscopic image generation model according to claim 1, characterized in that, Based on the first microscopic image, a second microscopic image is generated using the generator in the microscopic image generation model, specifically as follows: in, X represents the second microscopic image, X0 represents the original microscopic image, G(·) represents the generator, and X t Let Z represent the first microscopic image, Z represent a random Gaussian variable, and t represent the number of noise addition steps, where t∈[1,2,…,T]; Gaussian noise was added to the second microscopic image to obtain the third microscopic image, which is specifically represented as follows: in, The third microscopic image is represented by U(·), which indicates the Gaussian noise addition operation, and t-1 indicates the number of noise addition steps applied.

4. The training method for the microscopic image generation model according to claim 1, characterized in that, The adversarial loss function is expressed as: in, This represents the discriminator loss function. Represents the generator loss function. This represents the third microscopic image, X. t-1 This represents the photomicrograph of the previous noise-adding step, where D(·) represents the discriminator, and X... t Let represent the first microscopic image, and t represent the noise addition step number, t∈[1,2,…,T].

5. An application method for a microscopic image generation model, characterized in that, include: Determine the target image to be generated; Random Gaussian noise with the same shape as the target image matrix to be generated is input into the generator of the trained microscopic image generation model to generate a new high-quality microscopic image; the trained microscopic image generation model is a model trained by the training method of a microscopic image generation model according to any one of claims 1-4.

6. A training system for a microscopic image generation model, characterized in that, include: The raw microscopic image acquisition module is used to acquire raw microscopic images; The first noise-adding module is used to add Gaussian noise to the original microscopic image to obtain the first microscopic image. The generation module is used to generate a second microscopic image based on the first microscopic image using a generator in a microscopic image generation model; the microscopic image generation model is a model composed of a diffusion model and a generative adversarial model, and the generative adversarial model includes a generator and a discriminator. The second noise-adding module is used to add Gaussian noise to the second microscopic image to obtain a third microscopic image; the Gaussian noise added to the third microscopic image and the first microscopic image are both multi-step Gaussian noise, and the Gaussian noise added to the first microscopic image is t-step Gaussian noise, and the Gaussian noise added to the third microscopic image is t-1-step Gaussian noise. The discrimination module is used to calculate the probability that the third microscopic image is a reasonable denoised distribution of the first microscopic image based on the third microscopic image and the first microscopic image, using the discriminator in the microscopic image generation model. The model training module is used to train the microscopic image generation model based on the probability of the third microscopic image serving as a reasonable denoising distribution of the first microscopic image and the adversarial loss function, so as to obtain a trained microscopic image generation model.

7. The training system for the microscopic image generation model according to claim 6, characterized in that, The first noise-adding module adds Gaussian noise to the original microscopic image to obtain a first microscopic image, specifically represented as follows: X t = U(X0,t); Among them, X t Let X0 represent the first microscopic image, U(·) represent the original microscopic image, t represent the Gaussian noise addition operation, and t represent the noise addition step number, where t∈[1,2,…,T].

8. The training system for the microscopic image generation model according to claim 6, characterized in that, The generation module generates a second microscopic image based on the first microscopic image using the generator in the microscopic image generation model, specifically as follows: in, X represents the second microscopic image, X0 represents the original microscopic image, G(·) represents the generator, and X t Let Z represent the first microscopic image, Z represent a random Gaussian variable, and t represent the number of noise addition steps, where t∈[1,2,…,T]; The second noise-adding module adds Gaussian noise to the second microscopic image to obtain a third microscopic image, specifically as follows: in, The third microscopic image is represented by U(·), which indicates the Gaussian noise addition operation, and t-1 indicates the number of noise addition steps applied.

9. The training system for the microscopic image generation model according to claim 6, characterized in that, The adversarial loss function is expressed as: in, This represents the discriminator loss function. Represents the generator loss function. This represents the third microscopic image, X. t-1 This represents the photomicrograph of the previous noise-adding step, where D(·) represents the discriminator, and X... t Let represent the first microscopic image, and t represent the noise addition step number, t∈[1,2,…,T].

10. An application system for a microscopic image generation model, characterized in that, include: The target image determination module is used to determine the target image to be generated; The model application module is used to input random Gaussian noise with the same shape as the target image matrix to be generated into the trained microscopic image generation model to generate a new high-quality microscopic image; the trained microscopic image generation model is a model trained by the training method of a microscopic image generation model according to any one of claims 1-4.

Images