Image processing method and device, electronic equipment and storage medium

Abstract

The invention provides an image processing method and device, electronic equipment and a storage medium, and relates to the technical field of image processing. The method comprises the following steps: acquiring a plurality of de-noised images; each de-noised image corresponds to different channels, and each de-noised image is a single-channel image which is decoded and does not carry noise; performing coding and decoding processing on each de-noised image by using the target matrix to obtain a plurality of target PD images; each target PD image corresponds to different channels, and each target PD image is a decoded single-channel image carrying noise; determining the difference between the de-noised image corresponding to each channel and the target PD image; and if the difference meets a set condition, determining the target matrix as an optimal matrix, so that a plurality of single-channel images which do not carry noise without noise reduction can be obtained after the acquired original image is decoded according to the optimal matrix. The problem that magnetic resonance imaging is not accurate enough due to the fact that priori knowledge is not accurate enough in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the field of image processing technology. Specifically, the present application relates to an image processing method, device, electronic device and storage medium. Background Art

[0002] Magnetic Resonance Imaging (MRI) is a non-ionizing form of medical imaging technology that is widely used for in vivo diagnosis, characterization and research of diseases. The B1 field, also known as the radiofrequency field (RF field), will produce a standing wave effect in ultra-high field MRI because the wavelength of the radiofrequency electromagnetic wave is close to the size of human tissue, causing the distortion of the radiofrequency field in the human body. These complex interference mechanisms will change the spatial distribution of the original electromagnetic field, which is specifically manifested as a decrease in the uniformity of the B1 field distribution in human tissue. This greatly affects the imaging quality and patient safety, and hinders the widespread application of ultra-high field MRI in clinical practice.

[0003] At present, there are two main methods to obtain a single-channel B1 map: one is to use multi-channel coding technology to map using channel linear combinations (LCs) to solve the single-channel B1 map; the other is to obtain a single absolute (RF excitation field) diagram and a set of relative The images are then linked together through mathematical operations to obtain the accurate B1 map for each channel.

[0004] However, currently the selection of coding matrices mostly relies on prior knowledge. But in fact, the coding matrix obtained by prior knowledge is not necessarily optimal. This means that when relying on the coding matrix obtained by prior knowledge to collect a single-channel B1 image, the accuracy of the B1 image may be limited in certain situations or locations.

[0005] From the above, it can be seen that the problem of inaccurate magnetic resonance imaging caused by uneven B1 field distribution due to inaccurate prior knowledge still needs to be solved. Summary of the invention

[0006] The present application provides an image processing method, device, electronic device and storage medium, which can solve the problem of inaccurate magnetic resonance imaging due to inaccurate prior knowledge in the related art. The technical solution is as follows: According to one aspect of the present application, an image processing method includes: obtaining multiple denoised images; each of the denoised images corresponds to a different channel, and each of the denoised images is a single-channel image that has been decoded and does not carry noise; using a target matrix to encode and decode each of the denoised images to obtain multiple target PD images; each of the target PD images corresponds to a different channel, and each of the target PD images is a single-channel image that has been decoded and carries noise; determining the difference between the denoised image corresponding to each of the channels and the target PD image; if the difference meets the set conditions, determining that the target matrix is ​​an optimal matrix, so that the original image obtained by the acquisition can be decoded according to the optimal matrix to obtain multiple single-channel images that do not carry noise without the need for noise reduction.

[0007] According to one aspect of the present application, an image processing device includes: an image acquisition module, used to acquire multiple denoised images; each of the denoised images corresponds to a different channel, and each of the denoised images is a single-channel image that has been decoded and does not carry noise; a coding and decoding processing module, used to use a target matrix to perform coding and decoding processing on each of the denoised images to obtain multiple target PD images; each of the target PD images corresponds to a different channel, and each of the target PD images is a single-channel image that has been decoded and carries noise; a difference determination module, used to determine the difference between the denoised image corresponding to each of the channels and the target PD image; an optimal matrix output module, used to determine that the target matrix is ​​an optimal matrix if the difference meets a set condition, so that the original image obtained by the acquisition can be a plurality of single-channel images that do not carry noise without the need for noise reduction after being decoded according to the optimal matrix.

[0008] According to one aspect of the present application, an electronic device includes at least one processor and at least one memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the image processing method as described above is implemented.

[0009] According to one aspect of the present application, a storage medium stores a computer program thereon, and when the computer program is executed by one or more processors, the image processing method as described above is implemented.

[0010] According to one aspect of the present application, a computer program product includes a computer program, and when the computer program is executed by one or more processors, the image processing method described above is implemented.

[0011] The beneficial effects of the technical solution provided by this application are: In the above technical solution, whether the target matrix is ​​the optimal matrix is ​​determined based on the denoising image and the target PD image, and the PD image satisfies a linear relationship. In the decoding and reconstruction process, the single-channel PD image can fully reflect the correlation between the channels, thereby avoiding calculation errors caused by the introduction of complex variables (such as the B1 field); in addition, using the optimal matrix to decode the original image can omit the denoising step and directly obtain a nearly noise-free single-channel image with strong noise resistance, thereby effectively solving the problem of inaccurate magnetic resonance imaging due to inaccurate prior knowledge in related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are only some embodiments of the present application, and those skilled in the art can obtain other drawings based on these drawings without creative work.

[0013] Figure 1 It is a schematic diagram of the implementation environment involved in this application; Figure 2 is a hardware structure diagram of a computer device according to an exemplary embodiment; Figure 3 is a flowchart of an image processing method according to an exemplary embodiment; Figure 4 yes Figure 3 The steps after step 370 in the corresponding embodiment are a flowchart of an embodiment; Figure 5a yes Figure 3 A schematic diagram of a flip angle dynamic range of ABO acquisition involved in a corresponding embodiment; Figure 5b yes Figure 3 A schematic diagram of a flip angle dynamic range of an optimal matrix acquisition involved in a corresponding embodiment; Figure 6a yes Figure 3 A B1 field distribution map obtained by performing optimal matrix decoding on a head MPRI image involved in the corresponding embodiment; Figure 6b yes Figure 3 A B1 field distribution map obtained by performing ABO matrix decoding on a head MPRI image involved in the corresponding embodiment; Figure 7 yes Figure 3 A flowchart of an embodiment of the steps after obtaining the optimal matrix in the corresponding embodiment; Figure 8aIt is a schematic diagram of a specific implementation of an image processing method in an application scenario; Figure 8b yes Figure 8a Schematic diagram of the residual difference between a single-channel PD decoded by an ABO matrix after adding noise and a noise-free image in an application scenario; Figure 8c yes Figure 8a A schematic diagram of the residual between a single-channel PD image and a noise-free image after optimal matrix decoding after adding noise in an application scenario; Fig. 9 is a structural block diagram of an image processing device according to an exemplary embodiment; Fig.10 It is a structural block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0014] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be interpreted as limiting the present application.

[0015] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present disclosure refers to the presence of the features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0016] As mentioned above, when a single-channel B1 image is acquired by relying on a matrix obtained based on prior knowledge, the accuracy of the B1 image may be limited in certain situations or locations.

[0017] First of all, it should be noted that the use of parallel transmission (pTx) related technology is a solution to the problem of uneven magnetic field. Before using the related parallel field shim technology, it is necessary to obtain all single-channel B1 images of the transmitter as prior knowledge in order to understand the B1 field distribution of each channel. Finally, by adjusting each field, the composite B1 field can be made as uniform as possible, making imaging more accurate.

[0018] There are two mainstream methods to obtain a single-channel B1 map. One is to use multi-channel coding technology, that is, to solve a single channel by mapping linear combinations (LCs) of channels to a certain extent. The other method is to first obtain a single absolute Figure and a set of relative The graphs are then linked together through certain mathematical operations to finally obtain the accurate B1 graph for each channel.

[0019] About multi-channel encoding technology: In order to reduce the number of sub-channels collected without increasing the acquisition time ( In order to maintain the dynamic range of the field mapping, a linear combination of multi-channel encoding technology is proposed. In order to maintain an approximate flip level in most of the imaging area, the excitation magnetic field map is measured by simultaneously exciting multiple emission channels. When M emission modes are used to combine and excite N emission channels, the hardware encoding is based on a complex matrix E defined with a size of M×N. M coded RF mappings will be obtained, and Represented by the pseudo-inverse of the encoding matrix Decode to get the required sub-channel , the process is shown as follows: ; in, is the Hermitian conjugate matrix of the encoding matrix E, and W is a positive definite diagonal matrix that can be used for weighting To eliminate potential errors that may exist, r represents the position vector in space. When M=N and E^2 is a full-rank matrix, W is usually a unit matrix. For UHF (ultra-high frequency) multi-channel parallel excitation systems, a coding mode of M≥N is usually used to ensure that the equation is positive or overdetermined. For the radio frequency calibration sequence of ultra-high field cardiac magnetic resonance imaging, M=N is usually required to ensure that the acquisition time remains unchanged. Commonly used coding matrices are: (1) The diagonal is 1, and the other positions are 0 (only one channel is opened each time); (2) The diagonal is 0, and the other positions are 1 (only one channel is turned off each time it is stimulated); (3) The diagonal is -1, and the other positions are 1 (each time one of the channels is excited, the phase is reversed); (4) Fourier coding matrix (using discrete Fourier matrix as the coding matrix).

[0020] Because the optimal linear encoding matrix E will be different in different parts and main magnetic field strengths, in the study of 9.4T magnetic resonance imaging system, Fourier transform is a more robust encoding method, and it can increase the number of codes and sacrifice a certain amount of acquisition time to improve the channel The anti-interference ability of the 3.0T study makes the obtained results more accurate. In the 3.0T study, the best linear combination is determined by the coil array design and distribution as well as the load change, and the best linear combination may not always be found. Therefore, in order to make the acquired 5.0T cardiac magnetic resonance channel radio frequency mapping more accurate, it is necessary to find the coding coefficients with the strongest anti-interference ability through systematic experiments. In addition, whether the initial phase based on the statistical optimization results can help to obtain better calibration data remains to be verified, combined with ECG gating, in order to produce stable and robust channel radio frequency mapping in the 5.0T CMR calibration sequence (cardiac magnetic resonance calibration sequence).

[0021] Regarding the hybrid B1 mapping technology, the basic process is as follows: 1. Small flip angle 2DGRE image acquisition: A series of 16 small flip angle 2DGRE images are acquired, where only one channel is transmitted at a time and received on 16 channels simultaneously.

[0022] 2. The actual flip angle (AFI) technique obtains a 3D excitation flip angle map: All channels are transmitted simultaneously to obtain a 3D map of the excitation flip angle. This map is then merged with the previously acquired small flip angle GRE image to calculate the amplitude map of each of the 16 coil elements.

[0023] Hybrid improves the classic As shown below: ; in, represents the amplitude of the B1 field of the kth channel; represents the sum of all channels k; represents the complex B1 field of the kth channel; represents a complex exponent; Represents the relevant parameters of the kth channel at the jth sampling point.

[0024] However, in the multi-channel encoding method, it is crucial to choose a suitable encoding matrix A. Currently, the selection of encoding matrices mostly relies on prior knowledge, such as using a matrix with -1 on the diagonal and 1 in other positions (i.e., stimulating one channel at a time and inverting its phase). But in fact, this encoding matrix is ​​not necessarily the best, because the best transformation matrix A will change with different parts and main magnetic field strengths. This means that relying on the matrix obtained by prior knowledge to collect a single-channel B1 map, the accuracy of the B1 map may be limited in certain situations or parts.

[0025] Based on this, this solution optimizes the B1 mapping technology to make the acquired single-channel B1 image as accurate as possible. When the B1 image is accurate as prior knowledge, the subsequent field shim effect will be improved, ultimately making the magnetic resonance imaging more accurate.

[0026] As can be seen from the above, the related art still has the defect that the magnetic resonance imaging is not accurate enough due to the inaccurate prior knowledge.

[0027] To this end, the image processing method provided in the present application can effectively improve the accuracy of magnetic resonance imaging. Accordingly, the image processing method is suitable for an image processing device, which can be deployed in an electronic device. The electronic device can be a computer device configured with a von Neumann architecture, for example, the computer device includes a desktop computer, a laptop computer, a server, etc.

[0028] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0029] Figure 1 It is a schematic diagram of an implementation environment involved in an image processing method. It should be noted that this implementation environment is only an example adapted to the present invention and cannot be considered as providing any limitation on the scope of use of the present invention.

[0030] The implementation environment includes a collection end 110 and a service end 130 .

[0031] Specifically, the acquisition end 110 can also be considered as an image acquisition device, which has the function of acquiring medical images of human tissues, parts, and organs. For example, the acquisition end 110 is an MRI device.

[0032] The server 130 may be an electronic device such as a desktop computer, a laptop computer, a server, etc., or may be a computer cluster composed of multiple servers, or even a cloud computing center composed of multiple servers. The server 130 is used to provide background services, for example, background services include but are not limited to image processing services, etc.

[0033] The server 130 and the acquisition terminal 110 establish a network communication connection in advance through wired or wireless means, and data transmission between the server 130 and the acquisition terminal 110 is realized through the network communication connection. The transmitted data includes but is not limited to: original images and the like.

[0034] In an application scenario, through the interaction between the acquisition end 110 and the server end 130, the acquisition end 110 acquires an original image of human tissue and uploads the original image to the server end 130 to request the server end 130 to provide an image processing service.

[0035] For the server 130, after receiving the original image uploaded by the acquisition end 110, it calls the image processing service to perform image processing on the original image, so that the original image is decoded into at least one single-channel image without noise, so as to solve the problem of inaccurate magnetic resonance imaging due to inaccurate prior knowledge in related technologies.

[0036] See also Figure 2 , Figure 2 is a hardware structure diagram of a computer device according to an exemplary embodiment. The computer device is suitable for Figure 1 A server 130 in an implementation environment is shown.

[0037] It should be noted that the computer device is only an example adapted for the present application and cannot be considered to provide any limitation on the scope of use of the present application. The computer device cannot be interpreted as requiring dependence on or having Figure 2 One or more components of exemplary computer device 200 are shown.

[0038] The hardware structure of the computer device 200 may vary greatly due to differences in configuration or performance, such as Figure 2 As shown, the computer device 200 includes: a power supply 210 , an interface 230 , at least one memory 250 , and at least one central processing unit (CPU) 270 .

[0039] Specifically, the power supply 210 is used to provide operating voltage for each hardware device on the computer device 200 .

[0040] The interface 230 includes at least one wired or wireless network interface 231 for interacting with external devices. Figure 1 The interaction between the collection end 110 and the service end 130 in the implementation environment is shown.

[0041] Of course, in other examples adapted by the present application, the interface 230 may further include at least one serial-to-parallel conversion interface 233, at least one input-output interface 235, and at least one USB interface 237, etc. Figure 2 As shown, this is not intended to be a specific limitation.

[0042] The memory 250 is a carrier for storing resources, which may be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon include an operating system 251, an application 253 and data 255, etc. The storage method may be temporary storage or permanent storage.

[0043] Among them, the operating system 251 is used to manage and control the hardware devices and application programs 253 on the computer device 200 to enable the central processor 270 to calculate and process the massive data 255 in the memory 250. It can be WindowsServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0044] The application program 253 is a computer program formed by computer-readable instructions based on the operating system 251 to complete at least one specific task, and may include at least one module ( Figure 2 (not shown), each module may include corresponding computer-readable instructions. For example, the image processing apparatus may be regarded as an application 253 deployed on the computer device 200.

[0045] The data 255 may be photos, pictures, etc. stored in a disk, or may be single-channel images, multi-channel images, etc. stored in the memory 250 .

[0046] The central processor 270 may include one or more processors and is configured to communicate with the memory 250 through at least one communication bus to read the computer program stored in the memory 250, thereby realizing the operation and processing of the mass data 255 in the memory 250. For example, the image processing method is completed in the form of the central processor 270 reading the application program 253 stored in the memory 250.

[0047] In addition, the present application can also be implemented through hardware circuits or hardware circuits combined with software. Therefore, the implementation of the present application is not limited to any specific hardware circuits, software, or a combination of the two.

[0048] See also Figure 3 The present application embodiment provides an image processing method, which is applicable to an electronic device, which may be Figure 1 The server 130 in the implementation environment is shown. For example, the server 130 may refer to a computer device, which includes a desktop computer, a laptop computer, a server, etc. The specific hardware structure may be as follows: Figure 2 shown.

[0049] In the following method embodiments, for ease of description, the execution subject of each step of the method is taken as an electronic device as an example for illustration, but this does not constitute a specific limitation.

[0050] like Figure 3 As shown, the method may include the following steps: Step 310: Acquire multiple denoised images.

[0051] Among them, each denoised image corresponds to a different channel, and each denoised image is a single-channel image that has been decoded and does not carry noise.

[0052] Specifically, step 310 may include the following steps: acquiring an original image; decoding the original image according to the ABO matrix to obtain a plurality of single-channel images carrying noise; and performing denoising on the plurality of single-channel images carrying noise to obtain a plurality of denoised images.

[0053] The original image is a multi-channel image that has not been decoded. For example, the original image may refer to a multi-channel MRI image.

[0054] In a possible implementation, an MRI device is used to collect MRI images of human tissue to obtain original images.

[0055] It should be noted that MRI equipment is often equipped with a multi-channel radio frequency coil with multiple independent radio frequency channels. Each channel can transmit or receive signals independently. The multi-channel coil can collect data in parallel, thereby accelerating the imaging process and improving image quality. The original image collected by the multi-channel can be converted into a single-channel image through an appropriate encoding or decoding process.

[0056] Then, after decoding the original image, multiple single-channel images with noise can be obtained, namely the initial noisy PD map. Among them, the PD map can refer to the proton density image, which is an image generated by measuring the proton (hydrogen nucleus) density in the tissue, and can reflect the water content or proton concentration in the tissue. It can be understood that different types of tissues have different proton densities.

[0057] Specifically, the ABO matrix can be used to decode a multi-channel image into a single-channel image. The ABO matrix can refer to a matrix used to describe and analyze electromagnetic fields (e.g., in magnetic resonance imaging (MRI) and radio frequency (RF) technology). For example, if the MRI device is equipped with 8 channels, the decoding process will generate 8 separate single-channel images. It should be understood that due to the influence of sensors, environmental interference or other factors, the single-channel images obtained after decoding will carry noise, which will affect the image quality of the single-channel images.

[0058] In one possible implementation, the ABO matrix is ​​a matrix with -1 on the diagonal and 1 in the rest of the elements.

[0059] Therefore, each single-channel image needs to be denoised to remove the noise it carries. Gaussian filtering, wavelet transform, deep learning denoising and other methods can be used to remove the noise of each single-channel image, which are not limited here.

[0060] In one possible implementation, a BM3D filter is used to remove the noise carried by each single channel image to obtain a corresponding denoised image.

[0061] Then, the single-channel image after removing the noise is called the denoised image. It should be noted that the denoised image can be used as the gold standard image for subsequent steps. The gold standard image refers to the reference image in the field of medical image analysis and image processing. The gold standard image is not necessarily perfect, but it can be regarded as the optimal result under the current conditions. The gold standard image can be used as a benchmark to verify or evaluate the accuracy of other imaging methods.

[0062] Step 330 , using the target matrix to perform encoding and decoding processing on each denoised image to obtain a plurality of target PD images.

[0063] Among them, each target PD map corresponds to a different channel, and each target PD map is a decoded single-channel image carrying noise.

[0064] Specifically, step 330 may include the following steps: encoding each denoised image according to the target matrix to obtain an intermediate image; adding Gaussian noise to the intermediate image to obtain a noisy intermediate image; decoding the noisy intermediate image according to the target matrix to obtain multiple target PD images.

[0065] It can be understood that each denoised image belongs to a single-channel image, so each denoised image can be encoded into a multi-channel image, that is, an intermediate image, through the target matrix. Since the denoised image has been denoised, Gaussian noise is added to the intermediate image to obtain an intermediate image with noise. It can be understood that adding Gaussian noise can simulate the noise effect that may occur during MRI scanning.

[0066] Then, the intermediate image carrying noise can be decoded through the target matrix to obtain multiple target PD images.

[0067] It should be noted that although the ABO matrix can be used to decode a multi-channel image into a single-channel image, the decoded single-channel image still carries noise and needs to be subjected to noise reduction processing. Therefore, the purpose of this scheme is to decode the noisy multi-channel image through the target matrix to obtain a noise-free or less-noise target PD image.

[0068] Step 350 , determining the difference between the denoised image corresponding to each channel and the target PD image.

[0069] Among them, the difference actually reflects the gap between the B1 field distribution of the denoised image corresponding to each channel and the target PD image.

[0070] As mentioned above, the denoised image has been processed with noise reduction, so the noise in the denoised image is less, while the target PD image has not been processed with noise reduction. If the target matrix can decode the multi-channel image with noise and obtain a noise-free or less-noise target PD image, then the difference between the denoised image and the target PD image should be small, because the denoised image and the target PD image are actually obtained based on the same original image after processing.

[0071] The B1 field, or radio frequency field, refers to the alternating magnetic field generated by the radio frequency coils (RF coils) during MRI scanning, which can excite the hydrogen nuclei in the tissue and cause them to resonate. The B1 field affects the excitation efficiency of the radio frequency pulse, which in turn affects the signal intensity of different areas in the image. If the B1 field is weak or uneven, some tissue areas may not be fully excited, resulting in low signal intensity in these areas, which affects the accuracy of the PD map.

[0072] It can be understood that the difference between the denoised image and the target PD image is not only caused by noise, but may also be related to the difference in the distribution of the B1 field. Therefore, if the design of the target matrix meets expectations, the distribution of the B1 field between the target PD image and the denoised image should be relatively small.

[0073] In one possible implementation, the difference is described by the standardized root mean square error.

[0074] Specifically, step 350 may include the following steps: using the Hybrid B1 mapping technology, calculating the standardized root mean square error between the denoised image corresponding to each channel and the target PD image; based on the standardized root mean square error between the denoised image corresponding to each channel and the target PD image, evaluating the difference between the denoised image corresponding to each channel and the target PD image.

[0075] Among them, the HybridB1 mapping technology can estimate the spatial distribution characteristics of the B1 field, thereby determining the B1 field distribution of the denoised image and the target PD image.

[0076] Furthermore, for each channel, the normalized root mean square error can be used to measure the difference between the denoised image and the target PD image. If the normalized root mean square error is low, it can indicate that the target PD image is highly similar to the denoised image, and the decoding effect of the target matrix is ​​good; if the normalized root mean square error is high, it can indicate that the target PD image is significantly different from the denoised image, which may be caused by insufficient design of the target matrix.

[0077] In one possible implementation, the normalized root mean square error NRMSE is calculated as follows: ; in, is the denoised image, is the target PD map.

[0078] If the difference satisfies the set condition, step 370 is executed.

[0079] If the difference does not meet the set condition, then execute steps 380 to 390. The set condition is used to determine whether the design effect of the target matrix meets expectations. The set condition can be a threshold. If the threshold is exceeded, it means that the difference between the target PD map and the denoised map is too large, and the design effect of the target matrix is ​​not good enough. It is necessary to continue to optimize the target matrix, that is, execute steps 380 to 390; otherwise, it means that the design of the target matrix is ​​reasonable, and the difference between the target PD map and the denoised map is small, and it can be determined as the optimal matrix, and step 370 is executed.

[0080] Step 370, determining the target matrix as the optimal matrix, so that after the acquired original image is decoded according to the optimal matrix, a plurality of single-channel images without noise can be obtained without noise reduction.

[0081] Specifically, by decoding the original image (ie, the multi-channel image) using the target matrix, a single-channel image that does not carry noise or can be approximately regarded as not carrying noise can be obtained.

[0082] Through the above process, whether the target matrix is ​​the optimal matrix is ​​determined based on the denoised image and the target PD image. The PD image satisfies the linear relationship. During the decoding and reconstruction process, the single-channel PD image can fully reflect the correlation between the channels, thereby avoiding the calculation error caused by the introduction of complex variables (such as the B1 field). In addition, using the optimal matrix to decode the original image can omit the denoising step and directly obtain a nearly noise-free single-channel image with strong noise resistance.

[0083] Furthermore, if Figure 4 As shown, in an exemplary embodiment, after step 370, the method further includes the following steps: Step 380: Update the target matrix.

[0084] The updating process refers to adjusting each weight coefficient in the target matrix. The weight coefficient in the target matrix can be used to adjust the B1 field distribution of the multi-channel image during decoding and encoding to optimize the difference between the target PD image and the denoised image generated in the end.

[0085] It can be understood that the denoised image can better preserve the distribution of the actual B1 field after denoising, while the B1 field distribution of the target PD image may be affected by the target matrix decoding design, and the B1 field appears uneven. If the difference between the target PD image and the denoised image is large, it can indicate that the design of the target matrix has not yet compensated for the non-uniformity of the B1 field.

[0086] Specifically, step 380 may include the following steps: determining each weight coefficient in the target matrix; and adjusting each weight coefficient in the target matrix based on minimization of the non-uniform distribution of the B1 field to form an updated target matrix.

[0087] As mentioned above, the difference actually reflects the gap between the denoised image corresponding to each channel and the B1 field distribution of the target PD image. In other words, based on the minimization of the non-uniform distribution of the B1 field, adjusting the weight coefficients in the target matrix can make the B1 field distribution of the target PD image decoded using the target matrix more uniform, thereby improving the inaccurate B1 field distribution of the target PD image.

[0088] In one possible implementation, the formula for the target matrix is ​​as follows: ; in, represents the target matrix, the weight coefficient is A, is the element in row i and column k of A, yes By adjusting A, the target matrix can be updated.

[0089] Step 390 , triggering each denoised image to enter an iterative processing phase based on the updated target matrix.

[0090] The iterative processing stage refers to performing multiple encoding and decoding processes on each denoised image until the difference in the iterative processing stage meets the set conditions and the optimal matrix is ​​obtained.

[0091] Specifically, based on the updated target matrix, each denoised image is re-encoded and decoded to obtain a new round of target PD images. Then the difference between the new denoised image and the target PD image is re-determined. If the difference meets the set conditions, the iterative processing stage is terminated, and the target matrix obtained at the end of the iterative processing stage is used as the optimal matrix; if the difference does not meet the set conditions, steps 370 to 390 are continued until the difference meets the set conditions.

[0092] In addition, the setting condition may also be a maximum number of iterations, for example, 10 times, 20 times, etc. After the number of iterations reaches the maximum number of iterations, the optimization process is stopped and the current matrix is ​​output as the optimal matrix.

[0093] It should be noted that by limiting the maximum number of iterations, it is possible to avoid repeated adjustments of the target matrix under conditions of slight differences, reduce computational overhead and the risk of overfitting. In addition, if the standardized root mean square error oscillates within a certain range and is difficult to stabilize, setting the maximum number of iterations can ensure that the optimization process can be completed as planned.

[0094] It is further explained that, in the iterative processing stage, first of all, it is necessary to ensure that the single-channel images decoded using the target matrix are within the linear range of the B1 field calculation method. For example, if the Turbo FLASH (Turbo Fast Low AngleShot) sequence is selected to calculate the B1 field, it is necessary to control the B1 field flip angle to be between 20 and 120°; secondly, when making a safety judgment, the calculation of SAR (Specific Absorption Rate), that is, the special absorption ratio or the specific frequency electromagnetic wave absorption ratio, can be added to the iterative processing stage. The purpose is to ensure that the target matrix obtained by iteration will not cause safety problems.

[0095] Figure 5a A schematic diagram of the dynamic range of flip angle acquired by ABO is shown in FIG. Figure 5b A schematic diagram of the flip angle dynamic range of an optimal matrix acquisition is shown in FIG. Figure 5a and Figure 5b As shown in the figure, after adding the flip angle limit in the iterative processing stage, the collected B1 image can be kept within the linear range of 20° to 120°, ensuring the accuracy of B1; Figure 5a and Figure 5b Tx0 to Tx7 represent different transmission channels of the MRI device.

[0096] Figure 6a FIG. 2 shows a B1 field distribution diagram obtained by optimal matrix decoding of a head MPRI image. Figure 6b FIG. 4 shows a B1 field distribution diagram obtained by ABO matrix decoding of a head MPRI image. Figure 6a and Figure 6b In the B1 field that does not meet the range of 20° to 120°, the Figure 6a and Figure 6b We can see the specific distribution area that does not meet the linear flip angle range. Figure 6a As shown in , the PD map obtained by optimal matrix decoding has only a small number of B1 fields that do not meet the linear flip angle range. However, Figure 6b As shown in the figure, the PD map obtained by ABO matrix decoding has a large number of B1 fields that do not meet the linear flip angle range; Figure 6a and Figure 6b Tx0 to Tx7 represent different transmission channels of the MRI device.

[0097] In summary, through optimal matrix decoding, not only can the noise impact be minimized, but the flip angle of the measured PD image can also be accurately controlled within the range of 20° to 120°.

[0098] It should be noted that due to the complexity of the B1 field distribution (such as low or uneven local B1 field intensity), large measurement errors may occur in some areas. If the B1 field is directly selected as the basis for calculating the channel relationship, these errors may be further amplified during the encoding, decoding and superposition process, resulting in an increase in the overall deviation of the experimental results. Therefore, using the PD map as the iteration basis can avoid the introduction of unnecessary B1 field measurement errors and make the entire process more robust.

[0099] Through the above process, the Hybrid B1 mapping technology can be used to assist in generating the optimal matrix. By dynamically adjusting the weight coefficients in the target matrix, based on the principle of minimizing the non-uniformity of the B1 field distribution, the target PD map generated after decoding the target matrix is ​​closer to the denoised map in terms of the B1 field distribution. This process effectively compensates for the deficiencies in the target matrix design, improves the uniformity of the B1 field distribution, and thus improves the accuracy of the target PD map.

[0100] See also Figure 7 In an exemplary embodiment, after obtaining the optimal matrix, the above method may further include the following steps: Specifically, in step 410, a plurality of optimal matrices are obtained.

[0101] Among them, each optimal matrix corresponds to the target part of different objects.

[0102] As mentioned above, different types of tissues have different proton densities, that is, for different tissues, the weight coefficients need to be adjusted according to the signal characteristics of different tissues to obtain the corresponding optimal matrix. For example, tissues with more water content may need to be moderately compensated when allocating weights to ensure that the B1 field distribution uniformity is consistent with that of other tissues.

[0103] Then, each optimal matrix can be iteratively optimized to reflect a more accurate B1 field distribution for the target part of a specific object (such as the brain, chest, etc.).

[0104] Step 430: Perform statistical processing on each optimal matrix to obtain the best encoding matrix corresponding to the target part.

[0105] For different people and different tissues, considering that each iteration optimization for a specific patient's specific part takes a lot of time, the optimal matrix obtained in the final iteration processing stage of the same part (such as the head) of different people can be statistically processed to obtain a set of optimal coding matrices that can better measure the part of all people. Then, for subsequent scanning of the part, the optimal coding matrix can be directly selected as prior knowledge collection, thereby improving the collection effect without spending extra time.

[0106] For example, 20 groups of original images of the heads of 20 people are collected. These 20 groups of original images satisfy the normal distribution of each group of people. Through steps 310 to 390, 20 groups of different optimal matrices are obtained. The 20 groups of optimal matrices can be statistically processed, and finally an optimal coding matrix that can better represent all people can be obtained. Of course, different parts can also extract different representative optimal coding matrices. Based on the optimal coding matrix corresponding to the target part, it can be applied to the decoding process of the original image of the target part, and finally the original image of the target part can be decoded according to the optimal coding matrix to obtain multiple single-channel images without noise without noise reduction.

[0107] In one possible implementation, statistical processing may include the following process: first, the weight coefficients of each optimal matrix of a specific part may be analyzed one by one, and weighted average or other fusion operations may be performed according to the decoding performance of the specific part, so as to filter out the matrix weights with large deviations in certain areas and retain the parts with stable performance; thus, the matrix with the best encoding and decoding performance may be selected or multiple matrices may be weightedly fused. In addition, the matrix parameters may be locally adjusted according to the specific part (such as fine-tuning the weight of a specific channel) to improve the adaptability to the characteristic part, and no specific restrictions are made here.

[0108] Through the above process, multiple optimal matrices provide targeted optimization for different target parts, improving the adaptability of the optimal matrix. Through statistical processing, the best coding matrix combines the advantages of multiple optimal matrices to ensure the global optimality of the final output results.

[0109] Figure 8a This is a schematic diagram of a specific implementation of an image processing method in an application scenario. In this application scenario, the heart of a patient is collected by an MRI device to obtain an MRI image.

[0110] Through step 801, the MRI image is decoded through the ABO matrix to obtain 8 denoised images carrying noise, and the 8 denoised images carrying noise are subjected to denoising processing to obtain 8 noise-free denoised images.

[0111] Through step 803, the eight denoised images are encoded using the target matrix to obtain an intermediate image.

[0112] Through step 805, Gaussian noise is added to the intermediate image to obtain an intermediate image carrying noise.

[0113] Through step 807, the intermediate image carrying noise is decoded according to the target matrix to obtain 8 target PD images.

[0114] Through step 809 , the difference between the denoised image corresponding to each channel and the target PD image is determined. If the difference meets the set condition, step 811 is executed.

[0115] Through step 811, the target matrix is ​​determined to be the optimal matrix, so that after the acquired MRI image is decoded according to the optimal matrix, a plurality of single-channel images without noise can be obtained without noise reduction.

[0116] If the difference does not meet the set condition, step 813 to step 815 are executed.

[0117] Through step 813, the target matrix is ​​updated.

[0118] Through step 815, each denoised image is triggered to enter the iterative processing stage based on the updated target matrix, and each denoised image is subjected to multiple encoding and decoding processes until the difference in the iterative processing stage meets the set conditions, thereby obtaining the optimal matrix.

[0119] By comparing the improved optimal matrix with the ABO matrix, the noise is significantly reduced, and the residual between the decoded single-channel image and the noise-free image is smaller. Through residual analysis, it can be found that the ABO matrix has a larger residual in the channel with higher noise and is significantly affected by noise, while the residual of the optimal matrix is ​​significantly reduced, and the noise interference is better suppressed, which further verifies the improvement of the optimal matrix in denoising performance.

[0120] Figure 8b A schematic diagram of the residual difference between a single-channel PD decoded by an ABO matrix and a noise-free image is shown in FIG. Figure 8c A schematic diagram of the residual difference between a single-channel PD image after optimal matrix decoding and a noise-free image is shown in FIG. Figure 8b and Figure 8c As shown, the optimal matrix is ​​more noise-resistant under high noise conditions; Figure 8b and Figure 8c Tx0 to Tx7 represent different transmission channels of the MRI device.

[0121] In this application scenario, the improved optimal matrix effectively reduces the impact of noise during the decoding process. Compared with the ABO matrix, the new optimal matrix shows stronger noise resistance in channels with large noise interference. In addition, residual analysis shows that the new optimal matrix can significantly reduce the residual between the single-channel image and the noise-free reference image, thereby improving the image quality.

[0122] Of course, the PD images decoded by the new optimal matrix are particularly significant in areas with high noise, and can more evenly compensate for the uneven distribution of the B1 field, thereby improving the overall clarity and consistency of the MRI images. This technical improvement not only improves the diagnostic value of the image, but also provides more powerful technical support for the clinical application of ultra-high field MRI.

[0123] Compared with related technologies, this scheme can determine whether the target matrix is ​​the optimal matrix based on the denoising image and the target PD image. The PD image satisfies the linear relationship. During the decoding and reconstruction process, the single-channel PD image can fully reflect the correlation between the channels, thereby avoiding the calculation error caused by the introduction of complex variables (such as the B1 field); in addition, using the optimal matrix to decode the original image can omit the denoising step and directly obtain a single-channel image that is close to noise-free and has strong noise resistance.

[0124] Furthermore, this solution can use HybridB1 mapping technology to assist in generating the optimal matrix. By dynamically adjusting the weight coefficients in the target matrix and based on the principle of minimizing the non-uniformity of the B1 field distribution, the target PD map generated after decoding the target matrix is ​​closer to the denoised map in terms of the B1 field distribution. This process effectively compensates for the deficiencies in the target matrix design, improves the uniformity of the B1 field distribution, and thus improves the accuracy of the target PD map.

[0125] In addition, it can also provide targeted optimization for different target parts based on multiple optimal matrices, improving the adaptability of the optimal matrix. Through statistical processing, the optimal coding matrix combines the advantages of multiple optimal matrices to ensure the global optimality of the final output results.

[0126] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0127] The following is an embodiment of the device of the present application, which can be used to execute the image processing method involved in the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the method embodiment of the image processing method involved in the present application.

[0128] See also Fig. 9In an embodiment of the present application, an image processing device 900 is provided, including but not limited to: an image acquisition module 910, a coding and decoding processing module 930, a difference determination module 950, and an optimal matrix output module 970.

[0129] The image acquisition module 910 is used to acquire multiple denoised images; each denoised image corresponds to a different channel, and each denoised image is a single-channel image that has been decoded and does not carry noise.

[0130] The encoding and decoding processing module 930 is used to use the target matrix to perform encoding and decoding processing on each denoised image to obtain multiple target PD images; each target PD image corresponds to a different channel, and each target PD image is a decoded single-channel image carrying noise.

[0131] The difference determination module 950 is used to determine the difference between the denoised image corresponding to each channel and the target PD image.

[0132] The optimal matrix output module 970 is used to determine that the target matrix is ​​the optimal matrix if the difference meets the set conditions, so that the original image obtained by the acquisition can be a plurality of single-channel images without carrying noise without noise reduction after being decoded according to the optimal matrix.

[0133] In an exemplary embodiment, the image acquisition module 910 is also used to acquire an original image; the original image is a multi-channel image that has not been decoded; the original image is decoded according to the ABO matrix to obtain a plurality of single-channel images carrying noise; and the plurality of single-channel images carrying noise are denoised to obtain a plurality of denoised images.

[0134] In an exemplary embodiment, the encoding and decoding processing module 930 is also used to encode each denoised image according to the target matrix to obtain an intermediate image; add Gaussian noise to the intermediate image to obtain a noisy intermediate image; and decode the noisy intermediate image according to the target matrix to obtain multiple target PD images.

[0135] In an exemplary embodiment, the difference determination module 950 is also used to calculate the standardized root mean square error between the denoised image corresponding to each channel and the target PD image using the Hybrid B1 mapping technology; based on the standardized root mean square error between the denoised image corresponding to each channel and the target PD image, the difference between the denoised image corresponding to each channel and the target PD image is evaluated.

[0136] In an exemplary embodiment, the image processing device 900 is also used to update the target matrix if the difference does not meet the set conditions; based on the updated target matrix, each denoising image is triggered to enter the iterative processing stage until the difference meets the set conditions in the iterative processing stage to obtain the optimal matrix.

[0137] In an exemplary embodiment, the image processing device 900 is further used to determine each weight coefficient in the target matrix; based on minimization of the non-uniform distribution of the B1 field, adjust each weight coefficient in the target matrix to form an updated target matrix.

[0138] In an exemplary embodiment, the image processing device 900 is also used to determine that the target matrix is ​​the optimal matrix, and the method also includes: obtaining multiple optimal matrices; each optimal matrix corresponds to a target part of a different object; performing statistical processing on each optimal matrix to obtain an optimal coding matrix corresponding to the target part, so as to apply the optimal coding matrix corresponding to the target part to the decoding processing of the original image of the target part.

[0139] It should be noted that, when the image processing device provided in the above embodiment performs image processing, the division of the above-mentioned functional modules is only used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the image processing device will be divided into different functional modules to complete all or part of the functions described above.

[0140] In addition, the image processing device and the image processing method provided in the above embodiments belong to the same concept, wherein the specific manner in which each module performs the operation has been described in detail in the method embodiment and will not be repeated here.

[0141] See also Fig.10 In an embodiment of the present application, an electronic device 4000 is provided, and the electronic device 4000 may include: a desktop computer, a laptop computer, a server, etc.

[0142] exist Fig.10 In the embodiment, the electronic device 4000 includes at least one processor 4001 and at least one memory 4003.

[0143] The data interaction between the processor 4001 and the memory 4003 can be realized through at least one communication bus 4002. The communication bus 4002 may include a path for transmitting data between the processor 4001 and the memory 4003. The communication bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 4002 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0144] Optionally, the electronic device 4000 may further include a transceiver 4004, which may be used for data interaction between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present application.

[0145] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0146] The memory 4003 can be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store a computer program in the form of instructions or data structures and can be accessed by the electronic device 4000, but is not limited to these.

[0147] The memory 4003 stores a computer program, and the processor 4001 can read the computer program stored in the memory 4003 through the communication bus 4002 .

[0148] The computer program is executed by one or more processors 4001 to implement the image processing method in the above-mentioned embodiments.

[0149] In addition, an embodiment of the present application provides a storage medium on which a computer program is stored. The computer program is executed by one or more processors to implement the image processing method as described above.

[0150] A computer program product is provided in an embodiment of the present application, including a computer program, which is executed by one or more processors to implement the image processing method as described above.

[0151] The above description is only a partial implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. An image processing method, characterized in that: include: Acquire multiple denoised images; each of the denoised images corresponds to a different channel, and each of the denoised images is a decoded single-channel image without noise; The target matrix is ​​used to encode and decode each of the denoised images to obtain a plurality of target PD images; each of the target PD images corresponds to a different channel, and each of the target PD images is a decoded single-channel image carrying noise; Determine the difference between the denoised image corresponding to each of the channels and the target PD image; If the difference satisfies the set condition, the target matrix is ​​determined to be the optimal matrix, so that after the acquired original image is decoded according to the optimal matrix, a plurality of single-channel images without noise can be obtained without noise reduction.

2. The method according to claim 1, characterized in that The obtaining of multiple denoised images comprises: Acquire the original image; the original image is a multi-channel image that has not been decoded; Decoding the original image according to the ABO matrix to obtain a plurality of single-channel images carrying noise; The multiple single-channel images carrying noise are subjected to noise reduction processing to obtain multiple denoised images.

3. The method according to claim 1, characterized in that The target matrix is ​​used to encode and decode each of the denoised images to obtain a plurality of target PD images, including: Performing encoding processing on each of the denoised images according to the target matrix to obtain an intermediate image; Adding Gaussian noise to the intermediate image to obtain the intermediate image carrying noise; The intermediate image carrying noise is decoded according to the target matrix to obtain a plurality of target PD images.

4. The method according to claim 1, characterized in that The determining the difference between the denoised image corresponding to each of the channels and the target PD image comprises: Calculating the normalized root mean square error between the denoised image and the target PD image corresponding to each channel by using a hybrid B1 mapping technique; Based on the normalized root mean square error between the denoised image corresponding to each of the channels and the target PD image, the difference between the denoised image corresponding to each of the channels and the target PD image is evaluated.

5. The method according to any one of claims 1 to 4, characterized in that: After determining the difference between the denoised image corresponding to each of the channels and the target PD image, the method further includes: If the difference does not satisfy the set condition, updating the target matrix; Each denoised image is triggered to enter an iterative processing stage based on the updated target matrix until the difference in the iterative processing stage meets the set condition, thereby obtaining the optimal matrix.

6. The method according to claim 5, characterized in that The updating process of the target matrix includes: Determining each weight coefficient in the target matrix; Based on minimization of the non-uniform distribution of the B1 field, each weight coefficient in the target matrix is ​​adjusted to form the updated target matrix.

7. The method according to any one of claims 1 to 4, characterized in that: After determining that the target matrix is ​​the optimal matrix, the method further includes: Acquire a plurality of the optimal matrices; each of the optimal matrices corresponds to a target part of a different object; Statistical processing is performed on each of the optimal matrices to obtain an optimal coding matrix corresponding to the target part, so that the optimal coding matrix corresponding to the target part is applied to the decoding process of the original image of the target part.

8. An image processing device, characterized in that: include: An image acquisition module, used to acquire a plurality of denoised images; each of the denoised images corresponds to a different channel, and each of the denoised images is a decoded single-channel image without noise; A coding and decoding processing module, used for coding and decoding each of the denoised images using a target matrix to obtain a plurality of target PD images; each of the target PD images corresponds to a different channel, and each of the target PD images is a decoded single-channel image carrying noise; A difference determination module, used to determine the difference between the denoised image corresponding to each of the channels and the target PD image; The optimal matrix output module is used to determine that the target matrix is ​​the optimal matrix if the difference meets the set conditions, so that the original image obtained by the acquisition can be a plurality of single-channel images without carrying noise without noise reduction after being decoded according to the optimal matrix.

9. An electronic device comprising at least one processor and at least one memory, wherein: The memory stores a computer program, wherein the computer program, when executed by the processor, implements the image processing method according to any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by one or more processors, the image processing method according to any one of claims 1 to 7 is implemented.

Images