Diffusion three-dimensional medical image generation method for mask generation and image decoupling

Abstract

The invention discloses a mask generation and image decoupling diffusion three-dimensional medical image generation method, relates to the technical field of image processing and analysis, and solves the problems of poor generalization ability, low calculation efficiency and high dependence on training data during image conversion in the prior art. The method comprises the following steps: acquiring a mask image, and processing the mask image by using a multi-condition diffusion probability model to obtain a multi-label mask sequence; screening the three-dimensional medical image data according to a screening condition, determining an informed slice set, and inputting the multi-label mask sequence and the informed slice set into a volume diffusion generation model to obtain a three-dimensional composite image; random noise is added to the three-dimensional composite image, a diffusion semantic refinement generation model is utilized, the three-dimensional composite image added with the random noise is refined, and a refined three-dimensional image is obtained; according to the invention, data decoupling and model generation volume consistency are realized, and a three-dimensional medical image with higher quality can be generated.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing and analysis, and in particular to a diffusion three-dimensional medical image generation method for decoupling a generated mask from an image. Background Art

[0002] With the advancement of medical imaging technologies, such as computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET), the volume and complexity of medical image data continue to increase, posing new challenges to data processing and analysis. Traditional image processing methods are often limited by computing power and storage space when processing large-scale three-dimensional data, and it is difficult to ensure processing efficiency and accuracy of results.

[0003] In the existing technology, the image conversion from CT to PET involves complex changes in modal characteristics, and the method has insufficient generalization ability when processing different types or stages of lung cancer, especially for tumors with atypical or rare characteristics; although the method mentions the extraction of local and global information, how to balance the two to generate images that are both detailed and consistent with overall pathological characteristics remains a challenge; although the existing technologies provide new technical approaches in the field of medical image generation, they may need further optimization and improvement in terms of generalization ability, computational efficiency, training data dependence, and complexity of clinical applications. Summary of the invention

[0004] The present invention provides a diffusion 3D medical image generation method in which a generation mask is decoupled from an image, thereby solving the problems of poor generalization ability, low computational efficiency, and high dependence on training data when performing image conversion in the prior art, and realizing a decoupled training process to solve the problem of inconsistent generation model volumes, thereby generating higher quality 3D medical images.

[0005] The present invention provides a diffusion three-dimensional medical image generation method for decoupling a generation mask from an image, the method comprising:

[0006] Acquire a mask image, and process the mask image using a multi-conditional diffusion probability model to obtain a multi-label mask sequence;

[0007] Filtering the three-dimensional medical image data according to the filtering conditions to determine an informed slice set, and inputting the multi-label mask sequence and the informed slice set into a volume diffusion generation model to obtain a three-dimensional synthetic image;

[0008] Random noise is added to the three-dimensional synthetic image, and the three-dimensional synthetic image after the random noise is added is refined by using a diffusion semantic refinement generation model to obtain a refined three-dimensional image.

[0009] In a possible implementation, the using a multi-conditional diffusion probability model to process the mask image to obtain a multi-label mask sequence includes:

[0010] The mask image is cropped and filled to obtain multiple images of uniform size, and the multiple images are represented as two-dimensional sequences respectively;

[0011] Selecting a part of the sequence from the two-dimensional sequence as a two-dimensional subsequence, and unconditionally generating an initial subsequence of continuous slices from random noise; wherein the length of the two-dimensional subsequence is equal to the length of the initial subsequence;

[0012] The initial subsequence is added to the two-dimensional subsequence, and the two-dimensional subsequence after the initial subsequence is added is forwardly and backwardly extended by an autoregressive method to obtain a multi-label mask sequence.

[0013] In a possible implementation, the two-dimensional subsequence after adding the initial subsequence is forward and backward extended by an autoregressive method to obtain a multi-label mask sequence, including:

[0014] Determining generation parameters of the multi-label mask sequence;

[0015] The multi-conditional diffusion probability model performs forward and backward expansion on the two-dimensional subsequence after adding the initial subsequence according to the generation parameters and the autoregressive method to obtain a multi-label mask sequence.

[0016] In a possible implementation, screening the three-dimensional medical image data according to the screening condition to determine the informed slice set includes:

[0017] Determine the extraction starting position according to the screening condition, select a plurality of different informed slices in the three-dimensional medical image through resampling technology and the extraction starting position, and form the informed slice set with the plurality of different informed slices.

[0018] In a possible implementation, the multi-conditional diffusion probability model is expressed as:

[0019]

[0020] Where ε represents the noise sampled from the standard normal distribution; ε θ represents the noise predicted by the model; represents the part of the subsequence to be generated after the time step t, which is disturbed by noise; X C Indicates that in each iteration, the model determines the conditional slice according to probability; z represents the relative position condition of the subsequence z~=z / D; t represents the time step in the diffusion process; Indicates that the loss function obeys the probability distribution of multiple conditions.

[0021] In a possible implementation, the use of the diffusion semantic thinning generation model to thin the three-dimensional synthetic image after adding random noise to obtain a thinned three-dimensional image includes:

[0022] Reading the three-dimensional synthetic image after adding random noise according to different dimensions to obtain informed slices of corresponding dimensions; wherein the different dimensions include: coronal slice mode, sagittal slice mode and axial slice mode;

[0023] The informed slices of different dimensions are respectively input into the corresponding diffusion semantic refinement generation model to generate the corresponding refined images of different dimensions;

[0024] The mean of the refined images corresponding to different dimensions is calculated, and a refined three-dimensional image is obtained based on the mean.

[0025] In a possible implementation, the loss function of the volume diffusion generation model is expressed as:

[0026]

[0027] Among them, E x0,c,t,ε～N(0,1) represents the probability distribution of the loss function subject to various conditions; ε represents the noise sampled from the standard normal distribution; ε θ represents the noise predicted by the model; It represents the product of the denoising coefficients accumulated step by step from time step 1 to t, which represents the denoising strength after multiple time steps. ; t represents the time step; x0 represents the original image or data point without adding noise; c represents the conditional slice determined by the model according to probability in each iteration.

[0028] In a possible implementation, the informed slices during training of the volume diffusion generation model are extracted from a three-dimensional medical image window.

[0029] One or more technical solutions provided in the present invention have at least the following technical effects or advantages:

[0030] (1) The present invention significantly enhances the quality and fidelity of image generation by adding a multi-conditional diffusion probability model and a diffusion semantic refinement generation model to the volume diffusion generation model and decoupling the multi-label mask sequence from the mask image.

[0031] (2) The present invention receives informed slices as input through a volume diffusion generation model, and combines a multi-label mask sequence as conditional information to generate a three-dimensional synthetic image through an iterative denoising process, which not only improves the image quality but also enhances the controllability of the generation process;

[0032] (3) The present invention allows users to adjust the features of generated images according to specific medical application scenarios by combining the conditional information of multi-label mask sequences and informed slices in the volumetric diffusion generation model;

[0033] (4) The present invention optimizes computational efficiency and resource consumption while maintaining the quality of generated images. By improving the network architecture and parameter settings, the demand for computing resources is reduced, making the model more suitable for resource-constrained environments;

[0034] (5) The present invention better protects the privacy of patients through the use of informed slices, complies with ethical standards and privacy protection requirements, which is particularly important in the field of medical image analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A flowchart of the steps of a method for generating a diffuse three-dimensional medical image with decoupling of generating a mask and an image provided by an embodiment of the present invention;

[0036] Figure 2 A schematic diagram of a training process for generating a multi-label mask sequence through a multi-conditional diffusion probability model provided by an embodiment of the present invention;

[0037] Figure 3 A schematic diagram of the reasoning process of generating a multi-label mask sequence through a multi-conditional diffusion probability model provided by an embodiment of the present invention;

[0038] Figure 4 A schematic diagram of the use of the volume diffusion generation model provided in an embodiment of the present invention;

[0039] Figure 5 A schematic diagram of the reasoning process of the volume diffusion generation model provided in an embodiment of the present invention;

[0040] Figure 6 A schematic diagram of the network architecture of the volume diffusion generation model and semantic refiner provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] The present invention provides a diffusion three-dimensional medical image generation method that decouples the generation mask from the image. Figure 1 As shown, the method includes the following steps S101 to S103.

[0043] S101, obtaining a mask image, and processing the mask image using a multi-conditional diffusion probability model to obtain a multi-label mask sequence.

[0044] Specifically, in step S101, the mask image is processed using a multi-conditional diffusion probability model to obtain a multi-label mask sequence, including:

[0045] The mask image is cropped and padded to obtain multiple images of uniform size, and the multiple images are represented as two-dimensional sequences respectively;

[0046] Selecting a part of the sequence from the two-dimensional sequence as a two-dimensional subsequence, and unconditionally generating an initial subsequence of continuous slices from random noise; wherein the length of the two-dimensional subsequence is equal to the length of the initial subsequence;

[0047] The initial subsequences are added to the two-dimensional subsequences respectively, and the two-dimensional subsequences after adding the initial subsequences are forward and backward extended by the autoregressive method to obtain a multi-label mask sequence.

[0048] Here, the two-dimensional subsequence after adding the initial subsequence is forward and backward extended by the autoregressive method to obtain a multi-label mask sequence, including:

[0049] (1) Determine the generation parameters of the multi-label mask sequence;

[0050] (2) The multi-conditional diffusion probability model forward and backward expands the two-dimensional subsequence after adding the initial subsequence according to the generation parameters and the autoregressive method to obtain a multi-label mask sequence.

[0051] Specifically, in step S101, the multi-conditional diffusion probability model is expressed as:

[0052]

[0053] Where ε represents the random variable used to simulate noise in the diffusion model; ε θ represents the noise predicted by the model; represents the part of the subsequence to be generated after the time step t, which is disturbed by noise; X C represents the part of the conditional slice; z represents the relative position condition of the subsequence z~=z / D; t represents; Indicates that the loss function obeys the probability distribution of multiple conditions.

[0054] For example, a mask generation framework is designed to crop and pad the mask image to a uniform size of 96×320×320, represented as a two-dimensional sequence. An initial subsequence of m consecutive slices is unconditionally generated from random noise. Figure 2 and Figure 3As shown, after the cropped and padded mask image is represented as a two-dimensional subsequence, the subsequence is extended forward and backward in an autoregressive manner, conditioned on the existing slices, and a multi-conditional diffusion probability model is set to generate a multi-label mask sequence. The generation parameters of the mask sequence include but are not limited to sequence length, conditional probability, etc.

[0055] For a given 3D multi-label mask M∈R D×H×W , select a subsequence of m consecutive slices as {M z ,M z+1 ,...,M z+(m-1)}, where z is a randomly selected starting index. For each subsequence, we select the first n or the last n slices, or no slice, based on probability p C ∈{p Forward ,p Backward ,p Uncondition} to determine the conditional slice The goal of the multi-conditional diffusion probability model is to generate the remaining slices. In order to incorporate the position condition, the relative position of the subsequence is used where z is the index of the starting slice of the subsequence. The positional conditional embedding is then concatenated with the temporal embedding to help the generation process. A binary indicator is used for each slice in the subsequence to indicate the presence of the conditional slice. The joint distribution of the inverse diffusion process (RDP) of the conditional slice XC can be written as:

[0056]

[0057] During inference, the multi-conditional diffusion probability model first generates a subsequence of m slices from random noise given a random position z. Then a multi-label mask sequence can be generated by autoregressively extending in two directions, conditioned on the existing slices.

[0058] S102, screening the three-dimensional medical image data according to the screening conditions, determining an informed slice set, and inputting the multi-label mask sequence and the informed slice set into a volume diffusion generation model to obtain a three-dimensional synthetic image.

[0059] Specifically, in step S102, the three-dimensional medical image data is screened according to the screening conditions to determine the informed slice set, including: determining the extraction starting position according to the screening conditions, selecting multiple different informed slices in the three-dimensional medical image through resampling technology and the extraction starting position, and forming the multiple different informed slices into an informed slice set.

[0060] Exemplarily, during training, informed slices are extracted from a 3D medical image window. Reconstructing a 3D medical image from a mask image involves a one-to-many mapping. To alleviate the problems of volume inconsistency and poor generation fidelity and improve generation quality, additional information, namely informed slices, are separated from the 3D image, which indicates the patient's anatomical appearance, physical position, and other scanning modes.

[0061] During inference, informed slices are initially generated by the model or randomly selected in accessible 3D medical images, and then autoregressively assigned synthetic ones, e.g. Figure 5 In addition to guiding the generation process, the informed slice conditional generation also achieves generation enhancement in the dataset through resampling. Therefore, the input condition c in the volume diffusion generation model includes the multi-label mask sequence M i,j and Informed Slice I i,j , for the j-th slice of the i-th 3D medical image.

[0062] The selection of informed slices is crucial to the generation process because it carries the patient’s anatomical prior information. Figure 4 As shown, informed slices and 3D mask images from sample 3D medical images are used. Due to the generalization ability, informed slices can be sampled from different 3D medical images by simple random selection to drive conditional generation given the mask volume and introduce information from other patients' 3D medical images. To increase the variation, a simple way to sample informed slices is to extract an initial slice I from a randomly selected 3D medical image at a random starting position. i,j , that is, i is sampled from {1,2,...,N}, and j is sampled from {1,2,...,Z i}, where N represents the number of samples in the dataset, and Z i Represents the number of slices of the i-th 3D medical image.

[0063] To achieve this, the position information of the slices in the volume is normalized to ensure that the position information of different samples is physically aligned.

[0064] S103, adding random noise to the three-dimensional synthetic image, and using the diffusion semantic refinement generation model to refine the three-dimensional synthetic image after the random noise is added to obtain a refined three-dimensional image.

[0065] Specifically, in step S103, the three-dimensional synthetic image after adding random noise is refined by using the diffusion semantic refinement generation model to obtain a refined three-dimensional image, including:

[0066] (1) reading the three-dimensional synthetic image after adding random noise according to different dimensions to obtain informed slices of the corresponding dimensions; wherein the different dimensions include: coronal slice mode, sagittal slice mode and axial slice mode;

[0067] (2) Inputting the informed slices of different dimensions into the corresponding diffusion semantic refinement generation model to generate the corresponding refined images of different dimensions;

[0068] (3) Calculate the mean of the refined images corresponding to different dimensions, and obtain the refined three-dimensional image based on the mean.

[0069] Specifically, the loss function of the volume diffusion generation model is expressed as:

[0070]

[0071] Among them, E x0,c,t,ε～N(0,1) represents the probability distribution of the loss function subject to various conditions; ε represents the noise sampled from the standard normal distribution; ε θ represents the noise predicted by the model; It represents the product of the denoising coefficients accumulated step by step from time step 1 to t, which represents the denoising strength after multiple time steps. ; t represents the time step; x0 represents the original image or data point without adding noise; c represents the conditional slice determined by the model according to probability in each iteration.

[0072] Specifically, the informed slices during training of the volumetric diffusion generative model are extracted from 3D medical image windows.

[0073] For example, the goal of the volume diffusion generative model is to learn the inverse process, denoising the noisy data samples from the noise and restoring them to follow the true data distribution x0~q(x). In the inverse process, the denoising network ε θ (·) Predict the corrupted data x at time step t∈{1,2,...,T} t noise and construct a parameterized Gaussian transfer The denoising network is built on top of U-Net and trained with mean squared error:

[0074]

[0075] Among them, E x0,c,t,ε～N(0,1) represents the probability distribution of the loss function subject to various conditions; ε represents the noise sampled from the standard normal distribution; ε θ represents the noise predicted by the model; It represents the product of the denoising coefficients accumulated step by step from time step 1 to t, which represents the denoising strength after multiple time steps. ; t represents the time step; x0 represents the original image or data point without adding noise; c represents the conditional slice determined by the model according to probability in each iteration.

[0076] 3D synthetic images and set up a diffusion semantic refinement generation model, such as Figure 6 As shown, the multi-label mask sequence in S101 and the informed slice set in S102 are combined as conditional inputs to a volume diffusion generation model to generate a three-dimensional medical image.

[0077] Although the volumetric diffusion generative model achieves high cross-slice consistency and spatial continuity, problems such as blocking, blurring, and suboptimal texture generation still exist. Therefore, a diffusion semantic refinement generative model is used to refine the previously generated images.

[0078] For three different views, specifically, axial view, coronal view and sagittal view, a two-dimensional mask and an image with added noise are taken as input to generate an image aligned with the input mask.

[0079] During inference, a small amount of noise is applied to the generated image to preserve the overall anatomical structure and spatial continuity. Afterwards, the final refined 3D image is the average result from the 3 views using a pre-trained diffuse semantic refinement generative model.

[0080] Compared with the prior art, the present invention significantly improves the quality of 3D medical image generation by introducing a training process that decouples mask generation from image generation, and fine-tunes the generated structure through a diffusion semantic refinement generative model (SDM), solving the shortcomings of the volumetric diffusion generative model in capturing medical image details. Through the use of informed slices, the present invention enhances the controllability of the generation process, allowing users to adjust the features of the generated image according to specific needs.

[0081] The present invention reduces the demand for computing resources while maintaining computational efficiency, making the model more suitable for resource-constrained environments. Through an improved generation framework, the present invention alleviates the problem of volume inconsistency and improves the fidelity and practicality of three-dimensional medical images.

[0082] In a specific use embodiment provided by the present invention, it includes: step 1, generating a multi-label mask sequence; step 2, determining an informed slice set and generating a three-dimensional synthetic image; step 3, using diffusion semantics to refine the generation model to obtain a refined three-dimensional image.

[0083] In step 1, the mask image is cropped and padded to a uniform size and represented as a two-dimensional sequence. An initial subsequence of m consecutive slices is unconditionally generated from random noise. Then, the initial subsequence is extended forward and backward in an autoregressive manner, conditioned on the existing slices, and a multi-conditional volume diffusion generation model is set to generate a multi-label mask sequence. The generation parameters of the multi-label mask sequence include but are not limited to sequence length, conditional probability, etc.

[0084] During inference, the multi-conditional diffusion probability model first generates a subsequence of 6 slices from random noise given a random position z. Then a multi-label mask sequence can be generated by autoregressively expanding in both directions, conditioned on the existing slices.

[0085] Specifically, in the first step, a starting index z is randomly selected from the 3D multi-label mask dataset.

[0086] In the second step, random noise is used to unconditionally generate an initial subsequence of m consecutive slices.

[0087] In the third step, based on the existing slices, the subsequences are expanded in a backward and forward autoregressive manner, using a multi-conditional diffusion probability model.

[0088] The fourth step is to use the probability p C Select conditional slice X C , which may be the first n or last n slices of the subsequence, or no slice.

[0089] In the fifth step, the relative position z~ of the subsequence is used as the condition,embedding and combined with the temporal embedding, while a binary indicator is used to mark the conditional slices.

[0090] Step 6: Train the multi-conditional diffusion probability model to minimize the loss function and generate the remaining slices X P .

[0091] Step 7. In the inference phase, a subsequence of m slices is generated starting from a random position z.

[0092] In the eighth step, we autoregressively extend the subsequence to the entire mask sequence, conditioned on the existing slices.

[0093] In the ninth step, the mask sequence is generated using the detailed process and network structure in the supplementary material.

[0094] Differences from the prior art: The traditional method does not use autoregressive method and multi-conditional diffusion probability model to generate mask sequence. The present invention improves the generation quality and accuracy of mask sequence through volume diffusion generation model and relative position embedding.

[0095] Effect achieved: The generated mask sequences are of higher quality and diversity, improving the realism and practicality of 3D medical images.

[0096] In the second step, informed slices are extracted from image windows during training; during inference, they are generated by existing models or randomly selected from accessible images, and then autoregressively integrated into the synthesis process.

[0097] In subsequent operations, different informed slices are selected through resampling techniques. In the volume diffusion generation model, the condition c of the volume diffusion generation model is composed of a multi-label mask sequence M i,j and Informed Slice I i,j The denoising network in the volume diffusion generative model is responsible for predicting the noise of the data at each time step and building a Gaussian transfer model based on it. The volume diffusion generative model is built based on the U-Net architecture and optimized by minimizing the mean square error.

[0098] Specifically, the specific steps for selecting and generating informed slices are as follows:

[0099] In the first step, during the training phase, informed slices are randomly selected from the 3D medical image window or generated based on the model. These slices contain information such as the patient's anatomical appearance, physical position, and scanning mode.

[0100] In the second step, during the inference phase, the initial informed slices can be generated by the model or randomly selected from the accessible volume and synthesized through an autoregressive process.

[0101] In the third step, due to memory limitations, the entire volume cannot be fed into the model, so windows are randomly sampled from the original volume to introduce variations in different conditions and starting positions for training and sampling.

[0102] The fourth step is to provide a window V of the i-th volume for each data batch. i，j：j+n , including n slices starting from the jth slice.

[0103] Step 5: Randomly select informed slice I in the training unit i，j：j+n , and compare it with the multi-label mask sequence M i，j：j+n The volume diffusion generative model is injected together for denoising.

[0104] In the sixth step, during the autoregressive volume sampling process, the positional relationship of the informed slices at the window endpoints is ensured and processed by diffusion in-painting.

[0105] In the seventh step, the conditional information, including informed slices and mask volumes, are integrated in the latent space through connections.

[0106] In the eighth step, the conditional latent features are combined with the noisy target latent features through a simple connection method to achieve control injection.

[0107] In the ninth step, we train a position ID guided slice generation model using the normalized position ID p and embedding the sinusoidal embedding as a cross attention condition.

[0108] In the tenth step, a cascaded volume diffusion generative model is constructed, which first samples the informed slice conditions and then samples the volume during the inference process to achieve generative enhancement within the dataset.

[0109] Differences from the prior art: The prior art does not utilize the role of informed slices in the generation process. The present invention enhances the controllability of generated images and meets the requirements of medical application scenarios through the use of informed slices.

[0110] Effect achieved: The generated images are more in line with the needs of medical application scenarios, improving the quality and practicality of the images.

[0111] In step three, based on the volumetric diffusion generation, the diffusion semantic refinement model (SDM) further refines the image through a diffusion probability model. During the refinement process, a limited number of noise steps are applied to the image to maintain the anatomical structure and spatial continuity, and then the image is optimized by the pre-trained model. By integrating the refinement results of three different views, a high-quality 3D image is obtained.

[0112] Specifically, the diffusion semantic thinning generation model is used to thin the three-dimensional synthetic image after adding random noise to obtain a thinned three-dimensional image, including:

[0113] The first step is to use convolutional networks to generate preliminary 3D medical images with high cross-slice consistency and spatial continuity.

[0114] Step 2: For each different view, such as axial view, coronal view and sagittal view, a diffuse semantic refinement generative model (SDM) is trained to accept multi-label mask sequences and informed slices as input.

[0115] Step 3: During the inference phase, a small amount of noise is applied to the initially generated image to preserve the overall anatomical structure and spatial continuity of the image.

[0116] Step 4: Use the pre-trained SDM to refine the slightly noisy image to improve image quality and reduce blur and texture issues.

[0117] Step 5: Calculate the average of the refined image results obtained from three different views to obtain the final 3D refined image.

[0118] Difference from the existing technology: The existing technology does not use semantic diffusion refiner to improve the quality of the initially generated image. The new method reduces blur and texture problems and improves the detail performance of the image through the refinement process.

[0119] The effect achieved: The refined 3D image has higher quality, reduces blur and texture problems, and improves image clarity and realism.

[0120] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. All or part of the present invention can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, mobile communication terminals, multi-processor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0121] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the present invention.

Claims

1. A diffusion three-dimensional medical image generation method with decoupling of mask generation and image generation, characterized in that: include: Acquire a mask image, and process the mask image using a multi-conditional diffusion probability model to obtain a multi-label mask sequence; Filtering the three-dimensional medical image data according to the filtering conditions to determine an informed slice set, and inputting the multi-label mask sequence and the informed slice set into a volume diffusion generation model to obtain a three-dimensional synthetic image; Random noise is added to the three-dimensional synthetic image, and the three-dimensional synthetic image after the random noise is added is refined by using a diffusion semantic refinement generation model to obtain a refined three-dimensional image.

2. The method for generating diffuse 3D medical images by decoupling the generation mask from the image according to claim 1, characterized in that: The method of processing the mask image by using a multi-conditional diffusion probability model to obtain a multi-label mask sequence includes: The mask image is cropped and filled to obtain multiple images of uniform size, and the multiple images are represented as two-dimensional sequences respectively; Selecting a part of the sequence from the two-dimensional sequence as a two-dimensional subsequence, and unconditionally generating an initial subsequence of continuous slices from random noise; wherein the length of the two-dimensional subsequence is equal to the length of the initial subsequence; The initial subsequence is added to the two-dimensional subsequence, and the two-dimensional subsequence after the initial subsequence is added is forwardly and backward extended by an autoregressive method to obtain a multi-label mask sequence.

3. The method for generating diffuse 3D medical images by decoupling the generation mask from the image according to claim 2, characterized in that: The method of forward and backward expansion of the two-dimensional subsequence after adding the initial subsequence is carried out by the autoregressive method to obtain a multi-label mask sequence, including: Determining generation parameters of the multi-label mask sequence; The multi-conditional diffusion probability model performs forward and backward expansion on the two-dimensional subsequence after adding the initial subsequence according to the generation parameters and the autoregressive method to obtain a multi-label mask sequence.

4. The method for generating diffuse 3D medical images by decoupling the generation mask from the image according to claim 1, characterized in that: The step of screening the three-dimensional medical image data according to the screening conditions to determine the informed slice set includes: Determine the extraction starting position according to the screening condition, select a plurality of different informed slices in the three-dimensional medical image through resampling technology and the extraction starting position, and form the informed slice set with the plurality of different informed slices.

5. The method for generating diffuse 3D medical images with decoupling of mask generation and image generation according to claim 1, characterized in that: The multi-conditional diffusion probability model is expressed as: Where ε represents the random variable used to simulate noise in the diffusion model; ε θ represents the noise predicted by the model; represents the part of the subsequence to be generated after the time step t, which is disturbed by noise; X C represents the part of the conditional slice; z represents the relative position condition of the subsequence, z~=z / D; t represents; Indicates that the loss function obeys the probability distribution of multiple conditions.

6. The method for generating diffuse 3D medical images with decoupling of mask generation and image generation according to claim 1, characterized in that: The method of using the diffusion semantic refinement generation model to refine the three-dimensional synthetic image after adding random noise to obtain a refined three-dimensional image includes: Reading the three-dimensional synthetic image after adding random noise according to different dimensions to obtain informed slices of corresponding dimensions; wherein the different dimensions include: coronal slice mode, sagittal slice mode and axial slice mode; The informed slices of different dimensions are respectively input into the corresponding diffusion semantic refinement generation model to generate the corresponding refined images of different dimensions; The mean of the refined images corresponding to different dimensions is calculated, and a refined three-dimensional image is obtained based on the mean.

7. The method for generating diffuse 3D medical images with decoupling of mask generation and image generation according to claim 1, characterized in that: The loss function of the volume diffusion generation model is expressed as: Among them, E x0,c,t,ε～N(0,1) represents the probability distribution of the loss function subject to various conditions; ε represents the random variable used to simulate noise in the diffusion model; ε θ Represents the noise predicted by the model; α represents the product of the denoising coefficients accumulated step by step from time step 1 to t, which represents the denoising strength after multiple time steps. ; t represents the time step; x0 represents the original image or data point without adding noise; c represents the conditional slice determined by the model according to probability in each iteration.

8. The method for generating diffuse 3D medical images with decoupling of mask generation and image generation according to claim 1, characterized in that: The informed slices during training of the volume diffusion generation model are extracted from a three-dimensional medical image window.

Images