Medical image reconstruction method, device and equipment based on diffusion model

Abstract

The invention belongs to the technical field of physiological information processing, and particularly discloses a medical image reconstruction method, device and equipment based on a diffusion model, and the method comprises the steps: obtaining a first medical image; generating a second medical image from the first medical image according to a preset conversion rule; inputting the second medical image into a measurement diffusion model to obtain a noise measurement value of any time step; and inputting the second medical image into an image diffusion model, and obtaining a third medical image based on the constraint of the noise value of any time step. According to the method, multiple reconstruction tasks such as under-sampling MRI reconstruction, sparse view CT reconstruction and medical image super-resolution can be executed by utilizing arbitrary diffusion generation priori. In order to improve the reconstruction quality of the diffusion model, a measurement diffusion model with a determined trajectory is provided for guiding reconstruction, and a parallel trajectory constraint mode is used for enabling the image diffusion model and the measurement diffusion model to be subjected to parallel evolution, so that the reconstruction quality and the reconstruction precision are greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of physiological information processing, and in particular relates to a medical image reconstruction method, device and equipment based on a diffusion model. Background Art

[0002] Medical inverse problems refer to the problem of inferring physiological or pathological parameters of the human body from medical measurement data. They are widely present in the field of medical imaging, especially in the image reconstruction process of magnetic resonance imaging (MRI) and computed tomography (CT). These image reconstruction techniques usually rely on restoring a complete image from limited or sparse sampling data. With the development of imaging technology, medical imaging plays an increasingly important role in diagnosis, but it also faces the problem of how to obtain clear images in a shorter time and efficiently. In the existing technology, although there are many image reconstruction methods, there are still the following challenges:

[0003] In MRI imaging, long scanning time is a significant problem, especially for large-scale data acquisition in clinical applications. In order to reduce scanning time and improve clinical efficiency, accelerated imaging technology is usually used. However, accelerated imaging often leads to sparse or missing data, making it difficult for traditional reconstruction methods (such as fast Fourier transform, etc.) to effectively restore high-quality images, and image blur or artifacts are prone to occur. CT imaging technology generates three-dimensional tomographic images by projecting and reconstructing X-rays at different angles. However, sparse view CT reconstruction may cause problems of lost details or artifacts in the reconstructed image due to insufficient viewing angle. Existing sparse view reconstruction methods usually rely on regularization techniques to improve the reconstruction quality of the image, but these methods often perform poorly when dealing with high noise or extremely sparse data.

[0004] Super-resolution technology aims to restore higher-resolution images from low-resolution images. It can restore higher-resolution images from low-resolution MRI images through algorithm processing, so as to clearly present the details of tissues and organs. It can also process low-quality sparse view CT reconstructed images and supplement the missing detail information through algorithms to improve image quality. This is crucial for fine medical diagnosis, especially when detecting tiny lesions or tumors. However, traditional super-resolution algorithms usually face difficulties in restoring image details, especially when the original image itself is of low quality. Existing super-resolution methods often result in blurred or distorted details.

[0005] Existing medical image reconstruction techniques based on optimization methods (such as the least squares method, algebraic reconstruction method, etc.) mainly rely on a large amount of calculations and complex mathematical derivations. When dealing with large-scale data sets, these methods have a large computational overhead, low efficiency, and are easily affected by noise, resulting in unstable quality of the reconstructed images. In addition, the existing reconstruction methods mainly propose different solutions for different problems and lack generality. Therefore, there is a need to develop a new, more efficient, and more general image reconstruction method that can overcome the deficiencies of these existing technologies. Summary of the Invention

[0006] The object of the present invention is to provide a medical image reconstruction method, device, and equipment based on a diffusion model, aiming to solve many challenges in existing medical image reconstruction technologies. By utilizing the superior generation ability of the diffusion model and existing fast sampling algorithms, the present invention can accurately restore MRI and CT images and improve the resolution of the images under the condition of sparse sampling, and obtain high-quality reconstructed images.

[0007] In the first aspect of the present invention, the present invention provides a medical image reconstruction method based on a diffusion model, including the following steps:

[0008] Obtain a first medical image;

[0009] Generate a second medical image from the first medical image according to a preset conversion rule;

[0010] Input the second medical image into a measurement diffusion model to obtain a noise measurement value at any time step;

[0011] Input the second medical image into an image diffusion model, and based on the constraint of the noise value at any time step, obtain a third medical image.

[0012] In the second aspect of the present invention, the present invention also provides a medical image reconstruction device based on a diffusion model, and the device includes:

[0013] An acquisition module for obtaining a first medical image;

[0014] An image conversion module for generating a second medical image from the first medical image according to a preset conversion rule;

[0015] An image processing module for inputting the second medical image into a measurement diffusion model to obtain a noise measurement value at any time step;

[0016] An image reconstruction module for inputting the second medical image into an image diffusion model, and based on the constraint of the noise value at any time step, obtain a third medical image.

[0017] In a third aspect of the present invention, the present invention further provides a computer device, including:

[0018] One or more processors;

[0019] A storage device for storing one or more programs;

[0020] When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method as described in the first aspect of the present invention.

[0021] The beneficial effects of the present invention are:

[0022] Aiming at the generality problem ignored by existing medical image reconstruction methods, the present invention improves the sampling process of the diffusion model. Based on the constraint of the noise value at any time step, while ensuring the reconstruction quality and reconstruction accuracy, the generality is guaranteed, so that the present invention can solve various medical image inverse problems and adapt to different acquisition parameters, which is exactly the requirement of clinical applications. Through the technology of the present invention, while ensuring generality, it can effectively improve the reconstruction quality of medical images, reduce the scanning time, and provide more accurate diagnostic basis for doctors, contributing to the development of medical imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flowchart of an embodiment of the medical image reconstruction method based on the diffusion model of the present invention;

[0024] Figure 2 It is a sampling process diagram of the diffusion model in the example of the present invention;

[0025] Figure 3 It is an example diagram of the multi-coil MRI reconstruction result with an acceleration factor of 4 executed by the present invention;

[0026] Figure 4 It is an example diagram of the 4-view CT reconstruction result executed by the present invention.

[0027] Figure 5 It is an example diagram of the 4× downsampled CT image super-resolution result executed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0029] As used in this application and the appended claims, the singular forms "a", "the", and "said" are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0030] In the present invention, unless otherwise specified, the terms "first", "second", etc. are used to describe various elements and are not intended to define the positional relationship, temporal relationship, or relative importance of these elements. Such terms are only used to distinguish one element from another. In some examples, the first information and the second information may refer to the same instance of the information, while in certain cases, based on the context description, they may also refer to different instances.

[0031] In the related art, when performing medical image reconstruction, it mainly relies on a large amount of calculations and complex mathematical derivations. These methods have a large computational overhead, low efficiency, and are easily affected by noise when dealing with large-scale data sets, resulting in unstable quality of the reconstructed images. In addition, the existing reconstruction methods mainly propose different solutions for different problems and lack generality.

[0032] To solve the above problems, the embodiments of the present invention obtain a first medical image; generate a second medical image from the first medical image according to a preset conversion rule; input the second medical image into a measurement diffusion model to obtain a noise measurement value at any time step; input the second medical image into an image diffusion model, and based on the constraint of the noise value at any time step, obtain a third medical image. The embodiments of the present invention use the noise measurement value at any time step output by the measurement diffusion model as the discrete conditional constraint of the image diffusion model, which can perform controllable reconstruction on the third medical image, can be adapted to medical images such as MRI and CT images, effectively improve the reconstruction quality of medical images, and reduce the scanning time.

[0033] According to one aspect of the present invention, there is provided a medical image reconstruction method based on a diffusion model. As Figure 1 shown, the method includes:

[0034] Step 101, obtain a first medical image;

[0035] In the embodiments of the present invention, the first medical image may be an MRI image, may also be a CT image, or may also be a CBCT (cone beam CT), ultrasound image, PET (positron emission tomography) image, etc. The present invention does not limit the type of the specific first medical image.

[0036] Step 102, generate a second medical image from the first medical image according to a preset conversion rule;

[0037] In the embodiments of the present invention, in order to better reconstruct the first medical image, this embodiment converts the first medical image according to some preset conversion rules. For example, a multi-coil undersampled MRI image is converted to the k-space (Fourier space or frequency space), normalized after cropping to an image size of 320×320, and then converted to the image domain; a sparse-view CT image is converted to a sinogram, normalized after cropping to an image size of 256×256, and then converted to the image domain.

[0038] In some preferred embodiments, after generating the second medical image, post-processing such as smoothing and sharpening can also be performed to optimize the effect of the medical image.

[0039] Step 103: Input the second medical image into the measurement diffusion model to obtain the noise measurement value at any time step;

[0040] In the embodiments of the present invention, this embodiment constructs a measurement diffusion model, and uses the measurement diffusion model for generating measurement values to guide the generation of the image diffusion model. From the perspective of the stochastic differential equation (SDE), for any time step, there is a corresponding noise measurement value used to constrain the reconstruction of the image diffusion model, and finally the solution of the inverse problem is realized.

[0041] Specifically, a measurement diffusion model is constructed to simulate the generation process of the measurement value, and its forward SDE and reverse SDE are respectively defined as:

[0042] dy t =f(y t ,t)dt+Ag(t)dw t

[0043]

[0044] where y t represents the noise measurement value at the t-th time step, f(·,t) represents the drift coefficient, A is the degradation process, the generation of the measurement value can be realized through the iteration of the inverse diffusion process, g(t) represents the diffusion coefficient, and dw t is the increment of the Wiener process w t , represents the gradient operator with respect to t t , p t (y t ) represents the probability distribution of y t , represents the standard Wiener process in the reverse time direction.

[0045] By fixing the reverse SDE trajectory corresponding to the measurement diffusion model, since the original measurement value y0 It is known that when the initial value and the increment of the Brownian motion in the reverse time direction are determined, the reverse SDE trajectory can be determined through the substitution of the Tweedie formula, that is, for any time step, the corresponding noise measurement value can be directly obtained, avoiding the computational overhead generated by using a neural network to estimate the score function.

[0046] Specifically, for the generation process of the measurement diffusion model, the trajectory of its corresponding reverse SDE should ultimately point to the corresponding original measurement value. Since the end point of the reverse SDE is known, when the initial value and the increment of the Brownian motion in the reverse time direction are determined, only the score function needs to be solved to obtain the trajectory of the reverse SDE. The Tweedie formula reveals the connection between the score function and the original measurement value. Therefore, the trajectory of the reverse SDE is completely determined and does not require any neural network estimation.

[0047] Based on the above analysis, in the embodiment of the present invention, random Gaussian noise can be used as the initial value, and the second medical image can be used as the target value, and input into the reverse diffusion model of the measurement diffusion model. Based on the increment of the Brownian motion in the reverse time direction, the noise measurement value y at any time step can be obtained. t 。

[0048] Step 104: Input the second medical image into the image diffusion model, and based on the constraint of the noise value at the arbitrary time step, obtain the third medical image.

[0049] In the embodiment of the present invention, an image diffusion model is constructed in this embodiment. The image diffusion model is used to obtain the reconstructed image. In this embodiment, a parallel trajectory constraint is introduced, a linear mapping is introduced into the image diffusion model, and the reverse SDE trajectory of the corresponding parallel measurement diffusion model is derived through Itô's lemma. Then, it is ensured that it is completely consistent with the reverse SDE trajectory obtained by the measurement diffusion model in step 103. By analyzing the drift term and minimizing the difference, the parallel evolution of the two is ensured, realizing accurate reconstruction.

[0050] Specifically, an image diffusion model is constructed to obtain the reconstructed image. Its forward SDE and reverse SDE are defined as:

[0051] dx t =f(x t ,t)dt+g(t)dw t

[0052]

[0053] where x t represents the reconstructed medical image at the t-th time step, represents the gradient operator of x t , p t (x t)Represents the probability distribution of x t of.

[0054] Specifically, based on Itô's lemma and the linear mapping relationship, the reverse SDE of the measurement diffusion model corresponding to the image diffusion model can be obtained. By minimizing the difference between this reverse SDE and the reverse SDE with fixed trajectories obtained in step 103, and modifying the drift term of the reverse diffusion SDE corresponding to the image diffusion model, the generation process of the image diffusion model can be constrained, and more accurate reconstruction can be performed.

[0055] In the embodiment of the present invention, the sampling processes of the two diffusion models are as Figure 2 shown. Where x t represents the reconstructed image at the current time step, and y t represents the noise measurement value corresponding to the measurement value with missing information (undersampled MRI, sparse-view CT, or low-resolution image) at the current time step. As sampling progresses, that is, when t = 0, the reconstructed image x 0 and its corresponding original measurement value y 0 can be obtained. There are two sets of variables in the figure, namely the reconstructed medical image x and the noise measurement value y; throughout the process, there is always a "parallel trajectory constraint" in the processing of the reconstructed medical image x and the noise measurement value y. This means that the conversion processes of the reconstructed medical image x and the noise measurement value y are interrelated and proceed in parallel, and are restricted by this constraint condition at each step of the conversion. Initially, from the noise measurement value y T at the T-th time step to the noise measurement value y t+1 at the (t + 1)-th time step, and from the reconstructed medical image x T at the T-th time step to the reconstructed medical image x t+1 at the (t + 1)-th time step, during the conversion process, intermediate states x t+1 and y t+1 are obtained; the Tweedie formula is applied to the intermediate states x t+1 and y t+1 respectively to obtain the posterior mean and the original measurement value The operation of "correcting the drift term and projecting" is performed on the posterior mean to obtain the reconstructed medical image x t at the t-th time step; finally, from the noise measurement value y t at the t-th time step to the original measurement value From the reconstructed medical image x t at the t-th time step to the posterior mean In this embodiment, through a series of specific operations and constraints, the final result is gradually obtained from the initial state to achieve medical image reconstruction.

[0056] The specific operations are as follows:

[0057] (1) Construct an image diffusion model to obtain a reconstructed image. Its forward SDE and reverse SDE are defined as:

[0058] dx t = f(x t , t)dt + g(t)dw t

[0059]

[0060] (2) Construct a measurement diffusion model to obtain the noise measurement value at any time step. Its forward and reverse SDEs are defined as:

[0061] dy t = f(y t , t)dt + A(g(t))dw t

[0062]

[0063] where y t represents the noise measurement value at the t-th time step, A represents the degradation function, f(·, t) represents the drift coefficient, g(t) represents the diffusion coefficient, represents the gradient operator for y t , p t (y t ) represents the probability distribution of y t , and represents the standard Wiener process in the reverse time direction.

[0064] (3) Since the original measurement value y 0 is known, the logarithmic gradient term in the reverse SDE of the measurement diffusion model is replaced according to the Tweedie formula, and this reverse SDE becomes:

[0065]

[0066] where y t represents the noise measurement value at the t-th time step, A represents the degradation function, f(·, t) represents the drift coefficient, g(t) represents the diffusion coefficient, represents the standard Wiener process in the reverse time direction, represents a function that changes with time t.

[0067] Since the end point of this reverse SDE, i.e., the original measurement, is determined, when the initial value and the increment of the Brownian motion in the reverse time direction are determined, the trajectory of this reverse SDE is also determined. Specifically, for any time t, the noise measurement value at this time can be solved according to the above formula for the constraint of the image diffusion model.

[0068] (4) To ensure that the image diffusion model can evolve parallel to the SDE corresponding to the known measurement diffusion model, it is necessary to analyze the reverse SDE of the image diffusion model. According to the derivation of Itô's lemma, the reverse SDE of the image diffusion model, after introducing the linear mapping y t = Ax t has the following SDE for the corresponding measured values:

[0069]

[0070] where y t represents the noise measurement value at the t-th time step, A represents the degradation function, f(·,t) represents the drift coefficient, g(t) represents the diffusion coefficient, represents the gradient operator for x t , p t (x t ) represents the probability distribution of x t , and represents the time-reversed standard Wiener process.

[0071] Minimizing the difference between the above equation and the reverse SDE of the trajectory-determined measurement diffusion model can ensure their parallel evolution, that is, the following relationship holds at any time t:

[0072]

[0073] y t = Ax t

[0074] (5) In actual operation, the continuous SDE is often discretized and sampling is performed using methods such as DDPM. This also applies to DDIM based on the probability flow ODE because the ODE does not contain random terms and its reverse ODE trajectory is also determined. Similar to the SDE, only the relationship in the above equation needs to be maintained to perform reconstruction. In addition, the method based on the probability flow ODE has higher reconstruction efficiency, which is beneficial for clinical applications. The sampling form based on DDIM is:

[0075]

[0076] where x t represents the reconstructed image at the t-th time step.

[0077] The following constraint is executed at each sampling step until t = 0:

[0078]

[0079] where represents the posterior mean, and y 0represents the original measurement value, x t represents the reconstructed image at the tth time step, s θ (x t+1 ,t) represents the score function of the t+1th time step, represents a function that changes with time t, and A represents a degradation function. Through the above constraints, the universality can be guaranteed under the premise of ensuring the reconstruction quality and reconstruction accuracy, so that the present invention can solve a variety of medical imaging inverse problems and adapt to different acquisition parameters.

[0080] The above process is iterated until t=0 to obtain an accurate reconstructed image. Figure 3 , Figure 4 and Figure 5 They are multi-coil MRI reconstruction with an acceleration factor of 4, sparse view CT reconstruction of 4 views, and CT image super-resolution with 4× downsampling. It can be seen that the medical image reconstruction method, device, and apparatus of the embodiments of the present invention can solve a variety of medical image inverse problems and adapt to different acquisition parameters, effectively improve the reconstruction quality of medical images, reduce scanning time, and provide doctors with more accurate diagnostic basis, thus contributing to the development of medical imaging.

[0081] Based on the same inventive concept as the medical image reconstruction method based on a diffusion model provided in this embodiment, this embodiment also provides a medical image reconstruction device based on a diffusion model, which includes at least one software function module that can be stored in a memory or fixed in an electronic device in the form of software. The processor in the electronic device is used to execute the executable module stored in the memory. For example, the software function module and computer program included in the medical image reconstruction device based on the diffusion model. The device includes:

[0082] An acquisition module, used for acquiring a first medical image;

[0083] An image conversion module, used for converting the first medical image into a second medical image according to a preset conversion rule;

[0084] An image processing module, used for inputting the second medical image into a measurement diffusion model to obtain a noise measurement value at any time step;

[0085] The image reconstruction module is used to input the second medical image into an image diffusion model, and obtain a third medical image based on the constraint of the noise value at any time step.

[0086] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0087] It should also be understood that if the above embodiments are implemented in the form of software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.

[0088] This embodiment also provides a computer device, including one or more processors; a storage device for storing one or more programs; when the program is executed by the processor, it implements the medical image reconstruction method based on the diffusion model provided in this embodiment. Among them, the storage device can be a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc and other various media that can store program codes. In some embodiments, the storage device can be a read-only memory (Read Only Memory, ROM), a programmable read-only memory (Programmable Read-Only Memory, PROM), an erasable programmable read-only memory (Erasable Programmable Read-Only Memory, EPROM), an electrically erasable programmable read-only memory (Electric Erasable Programmable Read-Only Memory, EEPROM), a flash memory, etc.; in some embodiments, the storage drive can be a disk drive, a solid-state drive, any type of storage disk (such as an optical disc, a DVD, etc.), or a similar storage medium, or a combination thereof, etc.

[0089] The conventional technologies and the solutions not described in detail in the above embodiments are well known in the art, so they will not be elaborated here in detail. The above embodiments and / or experimental examples have described in detail the preferred embodiments of the present invention. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, and these simple modifications all fall within the protection scope of the present invention.

[0090] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program, and this program can be stored in a computer-readable storage medium. The storage medium can include: ROM, RAM, magnetic disk, or optical disc, etc.

[0091] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A medical image reconstruction method based on a diffusion model, characterized in that: The following steps are involved: acquiring a first medical image; Generate a second medical image by converting the first medical image according to a preset conversion rule; Inputting the second medical image into a measurement diffusion model to obtain a noise measurement value at any time step; The second medical image is input into an image diffusion model, and a third medical image is obtained based on the constraint of the noise value at any time step.

2. The medical image reconstruction method based on diffusion model according to claim 1, characterized in that: When the first medical image is an MRI image, the preset conversion rule is a frequency space; when the first medical image is a CT image, the preset conversion rule is a sinusoidal graph space.

3. The medical image reconstruction method based on diffusion model according to claim 1, characterized in that: Inputting the second medical image into the measurement diffusion model, obtaining the noise measurement value at any time step includes: The random Gaussian noise is used as the initial value and the second medical image is used as the target value, which are input into the reverse diffusion model of the measurement diffusion model, and the noise measurement value of any time step is obtained based on the time-reverse Brownian motion increment.

4. The medical image reconstruction method based on diffusion model according to claim 3, characterized in that: The calculation formula of the noise measurement value at any time step is expressed as: Among them, y t represents the noise measurement value at the tth time step, α t represents the noise scheduling parameter of the predefined t-th time step, y0 represents the original measurement value, y t+1 represents the noise measurement value at the t+1th time step, α t+1 represents the predefined noise scheduling parameter at the t+1th time step, σ t represents the parameter controlling randomness that changes over time t, ∈ t represents the Gaussian noise at the t-th time step.

5. The medical image reconstruction method based on diffusion model according to claim 1, characterized in that: Inputting the second medical image into the image diffusion model, and obtaining the third medical image based on the constraint of the noise value at any time step comprises: The second medical image is used as an initial value, and the noise value at the arbitrary time step is used as a conditional value, and input into the inverse diffusion model of the image diffusion model to obtain a reconstructed medical image at the corresponding time step; until a third medical image at the final time step is obtained.

6. The medical image reconstruction method based on diffusion model according to claim 5, characterized in that: The calculation formula of the reconstructed medical image at the corresponding time step is expressed as: Among them, x t represents the reconstructed medical image at the tth time step, α t represents the noise scheduling parameters of the predefined t-th time step, represents the posterior mean, corresponding to the expected value of the reconstructed medical image in the entire posterior probability distribution at the 0th sampling step, x t+1 represents the reconstructed medical image at the t+1th time step, α t+1 represents the predefined noise scheduling parameter at the t+1th time step, σ t represents the parameter controlling randomness that varies with time step t, ∈ t represents the Gaussian noise at the t-th time step.

7. The medical image reconstruction method based on diffusion model according to claim 6, characterized in that: The constraints of the sampling step include: Among them, y0 represents the original measurement value, s θ (x t+1 ,t) represents the score function of the t+1th time step, σ t+1 represents the parameter controlling randomness that changes with time step t+1, A represents the degradation function, and y t represents the noisy measurement at the tth time step.

8. A medical image reconstruction device based on a diffusion model, characterized in that: The device comprises: An acquisition module, used for acquiring a first medical image; An image conversion module, used for converting the first medical image into a second medical image according to a preset conversion rule; An image processing module, used for inputting the second medical image into a measurement diffusion model to obtain a noise measurement value at any time step; The image reconstruction module is used to input the second medical image into an image diffusion model, and obtain a third medical image based on the constraint of the noise value at any time step.

9. A computer device, characterized in that: include: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method according to any one of claims 1 to 7.

Images