Image noise reduction method and related device

Abstract

The invention discloses an image noise reduction method and a related device, and relates to the field of machine learning, and the method comprises the steps: obtaining the coil data of a plurality of coils of a target cutting layer obtained by scanning a target object, and the target cutting layer is any one cutting layer; and obtaining an original coil image of each coil based on the coil data of each coil of the target cutting layer. And obtaining an amplitude image of the target cutting layer based on the original coil image of each coil. And acquiring a simulation coil image of each coil based on the amplitude image of the target cutting layer and the sensitivity image and the phase image of each coil. And based on the simulation coil image and the original coil image of each coil, obtaining an estimated noise component of the original coil image of each coil by using a noise estimation model. According to the method, the target amplitude image of the scanned target object is obtained based on the estimated noise component of each coil of each cutting layer and the original coil image, and the high signal-to-noise ratio characteristic of the amplitude image and the Gaussian distribution characteristic of the coil image are utilized at the same time, so that the denoising effect is improved.

Description

Technical Field

[0001] The present application relates to the field of machine learning technology, and in particular to an image denoising method and related devices. Background Art

[0002] MRI (Magnetic Resonance Imaging) technology uses radiofrequency coils to receive magnetic resonance signals from a target object (e.g., human tissue). Each coil unit independently receives the magnetic resonance signals, generating a complex image. The MRI image is then reconstructed from these complex images. Currently, machine learning models are being trained to identify and eliminate noise in MRI images, thereby improving their quality. Summary of the Invention

[0003] In view of the above problems, this application provides an image noise reduction method and related devices to achieve the purpose of improving the image quality of MRI images. The specific solution is as follows:

[0004] The first aspect of the present application provides an image noise reduction method, comprising:

[0005] Acquire coil data of a plurality of coils of a target slice obtained by scanning a target object, wherein the target slice is any slice;

[0006] obtaining an original coil image of each coil based on the coil data of each coil of the target slice;

[0007] acquiring an amplitude image of the target slice based on the original coil images of each coil;

[0008] acquiring a simulated coil image of each coil based on the amplitude image of the target slice and the sensitivity map and phase map of each coil;

[0009] Based on the simulated coil image and the original coil image of each coil, using a noise estimation model to obtain an estimated noise component of the original coil image of each coil;

[0010] A target amplitude image of the scanned target object is acquired based on the estimated noise components of each coil of each slice and an original coil image.

[0011] In a possible implementation, acquiring the amplitude image of the target slice based on the original coil images of each coil includes:

[0012] acquiring an amplitude image of each coil based on an original coil image of each coil;

[0013] After performing a square operation on the amplitude image of each original coil image of the coil, the square operation results are added to obtain a total energy map;

[0014] The square root of the total energy image is taken to generate an amplitude image of the target slice.

[0015] In a possible implementation, acquiring a simulated coil image of each coil based on the amplitude image of the target slice and the sensitivity map and phase map of each coil includes:

[0016] performing smoothing processing on the amplitude image of the target slice and the corresponding amplitude images of each coil respectively;

[0017] acquiring a sensitivity map and a phase map of each coil based on the smoothed amplitude image of the target slice and the corresponding amplitude images of each coil;

[0018] Based on the amplitude image of the target slice and the corresponding sensitivity map and phase map of each coil, a simulated coil image of each coil is acquired.

[0019] In a possible implementation, obtaining a sensitivity map and a phase map of each coil based on the smoothed amplitude image of the target slice and the corresponding amplitude images of each coil includes:

[0020] For any one of the coils, smoothing the amplitude image of the target slice and the amplitude image of the coil respectively is performed using a preset low-pass filter;

[0021] Calculating the ratio of the smoothed amplitude image of the coil to the amplitude image of the target slice, and obtaining the original phase image of the coil using a preset angle acquisition function;

[0022] Based on the smoothed amplitude image of the coil and the amplitude image of the target slice, an original phase image of the coil is calculated using a preset angle acquisition function;

[0023] The original smoothed phase image of the coil after smoothing is normalized to obtain the phase image of the coil.

[0024] In one possible implementation, obtaining an estimated noise component of the original coil image of each coil using a noise estimation model based on the simulated coil image and the original coil image of each coil includes:

[0025] Preprocessing the simulated coil image and the original coil image of each coil to obtain a target simulated coil image and a target original coil image of each coil, wherein the preprocessing includes a standardization process;

[0026] The target simulated coil image and the target original coil image of each coil are respectively input into the noise estimation model to obtain the estimated noise component of each coil output by the noise estimation model.

[0027] In one possible implementation, a method for training a noise estimation model includes:

[0028] Obtaining a plurality of sample original coil images of each sample coil in each sample slice based on the sample coil data of the sample object;

[0029] For each of the sample coils, calculating an arithmetic mean of all sample original coil images of the sample coil to obtain an average coil image of the sample coil;

[0030] generating a sample noisy coil image based on adding a preset sample noise component to the average coil image of each of the sample coils;

[0031] For each of the sample slices, acquiring a sample simulated coil image of each of the sample coils based on the amplitude image of the sample slice and the corresponding sensitivity map and phase map of each of the sample coils;

[0032] A noise estimation model to be trained is trained based on training data until a preset model training condition is reached; wherein the training data includes sample images and labels of the sample images, the sample images include sample noisy coil images and sample simulated coil images, and the labels include sample noise components.

[0033] A second aspect of the present application provides an image noise reduction device, comprising:

[0034] an object scanning unit, configured to obtain coil data of a plurality of coils of a target slice obtained by scanning a target object, wherein the target slice is any slice;

[0035] an original image acquisition unit, configured to obtain an original coil image of each coil based on the coil data of each coil of the target slice;

[0036] an amplitude image acquisition unit, configured to acquire an amplitude image of the target slice based on the original coil images of each of the coils;

[0037] a simulation image acquisition unit, configured to acquire a simulation coil image of each of the coils based on the amplitude image of the target slice and the sensitivity map and phase map of each of the coils;

[0038] a noise estimation unit, configured to obtain an estimated noise component of the original coil image of each coil using a noise estimation model based on the simulated coil image and the original coil image of each coil;

[0039] The noise reduction image reconstruction unit is used to obtain the target amplitude image of the scanned target object based on the estimated noise components of each coil of each slice and the original coil image.

[0040] A third aspect of the present application provides a computer program product, comprising computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements the image denoising method according to the first aspect or any implementation of the first aspect.

[0041] A fourth aspect of the present application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:

[0042] The memory is used to store computer programs;

[0043] The processor is configured to execute the computer program so that the electronic device can implement the image noise reduction method according to the first aspect or any implementation of the first aspect.

[0044] In a fifth aspect, the present application provides a computer storage medium, which carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the image noise reduction method of the first aspect or any implementation of the first aspect.

[0045] By utilizing the above-described technical solution, the image denoising and related apparatus provided by this application acquire coil data for multiple coils of a target slice obtained by scanning a target object, where the target slice is any slice. Based on the coil data of each coil of the target slice, an original coil image of each coil is obtained. Based on the original coil image of each coil, an amplitude image of the target slice is obtained. Based on the amplitude image of the target slice and the sensitivity map and phase map of each coil, a simulated coil image of each coil is obtained. Based on the simulated coil image and the original coil image of each coil, an estimated noise component of the original coil image of each coil is obtained using a noise estimation model. Based on the estimated noise component of each coil of each slice and the original coil image, a target amplitude image of the scanned target object is obtained. This method simultaneously leverages the high signal-to-noise ratio of the amplitude image and the Gaussian distribution characteristics of the coil image. By first estimating and removing noise using the Gaussian distribution characteristics of the coil image, the amplitude image is then generated. This ensures that the DC noise component is removed from the amplitude image generated after noise removal, thereby improving the denoising effect and image quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The above and other features, advantages, and aspects of the various embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that the originals and elements are not necessarily drawn to scale.

[0047] Figure 1 A schematic diagram of the system architecture provided for this application;

[0048] Figure 2 A flowchart of an image noise reduction method provided in an embodiment of the present application;

[0049] Figure 3 A flowchart of a specific implementation of an image noise reduction method provided in an embodiment of the present application;

[0050] Figure 4 A flowchart of a noise estimation model training method provided in an embodiment of the present application;

[0051] Figure 5 A schematic structural diagram of an image noise reduction device provided in an embodiment of the present application;

[0052] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] The following describes the embodiments of the present application in conjunction with the accompanying drawings. The terms used in the implementation methods of the present application are only used to explain the specific embodiments of the present application and are not intended to limit the present application.

[0054] The embodiments of the present application are described below in conjunction with the accompanying drawings. Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0055] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, and this is merely a way of distinguishing the objects of the same attributes when describing them in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0056] An MRI image is constructed based on coil data from multiple coils of at least one slice. The MRI image reconstructed from the coil data of multiple coils of a single slice is a two-dimensional MRI image, and the MRI image reconstructed from the coil data of multiple coils of multiple slices is a three-dimensional MRI image. Specifically, an original coil image is generated based on the coil data collected by each coil, and a two-dimensional MRI image or a three-dimensional MRI image is reconstructed based on the multiple original coil images of at least one slice. Because the coil data is a complex signal, the original coil images generated from the complex signal are all complex images including real and imaginary images, and the noise in the complex image is distributed as a Gaussian distribution. The reconstructed MRI image is an amplitude image, and the noise in the amplitude image is distributed as a Rayleigh distribution.

[0057] The properties of the Gaussian distribution indicate that predicting noise from a noisy image is easy, and preparing a pair of clean and noisy images during machine learning training is straightforward. However, because the signal-to-noise ratio of the original coil image is often extremely low, and the true structure is hidden within the Gaussian noise, predicting the clean image or the noise component is ineffective.

[0058] Furthermore, in the prior art, to overcome the difficulty of noise prediction caused by the low signal-to-noise ratio of coil images, the noise in the amplitude image is directly reduced by using the amplitude image. Although the amplitude image has a higher signal-to-noise ratio and is easier to predict the true fine structure in the noisy image than the coil image, due to the characteristics of the Rayleigh distribution, its shape depends on the noise level and also has a DC noise component. This DC noise component depends on the noise level and the original contrast of the coil image. Therefore, predicting the DC noise component from the amplitude image is very difficult. As can be seen, the complex noise characteristics of the amplitude image also make effective noise estimation difficult.

[0059] In summary, MRI noise exhibits different properties in amplitude images and complex images from the receiving coil. In amplitude images, noise exhibits a Rayleigh distribution, meaning it has both DC and AC components. In coil images, however, noise exhibits a zero-mean Gaussian distribution. Therefore, for amplitude images, both AC and DC noise must be eliminated to effectively denoise the amplitude image and improve its quality. Therefore, existing techniques for processing the complex characteristics of Rayleigh noise in amplitude images, or for estimating the true noise from coil images with extremely low signal-to-noise ratios, leave much to be desired.

[0060] In order to solve the above technical problems, the embodiments of the present application provide an image denoising method and related devices. The present application can be applied to the field of machine learning technology, and specifically can be applied to denoising MRI two-dimensional or three-dimensional images using a machine learning model.

[0061] This application can be applied to, but is not limited to, applications with image noise reduction functions or cloud services provided by cloud-side servers. The following describes each of them:

[0062] See also Figure 1 , Figure 1 A schematic diagram of a system architecture is shown. The system may include a terminal 100 and a server 200. The server 200 may provide the method provided in the embodiment of the present application to one or more terminals.

[0063] Among them, an image noise reduction application can be installed on the terminal 100. The above application and web page can provide an interface. The terminal 100 can receive relevant parameters entered by the user on the image noise reduction interface and send the above parameters to the server 200. The server 200 can obtain processing results based on the received parameters and return the processing results to the terminal 100.

[0064] It should be understood that in some optional implementations, the terminal 100 can also complete the action of obtaining the processing result based on the received parameters by itself without the need for the cooperation of the server, and the embodiments of the present application are not limited to this.

[0065] Next describe Figure 1 The product form of the mid-terminal 100;

[0066] The terminal 100 in the embodiment of the present application can be a mobile phone, a tablet computer, a wearable device, an in-vehicle device, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc., and the embodiment of the present application does not impose any restrictions on this.

[0067] The terminal 100 may include components such as a radio frequency unit, a memory, an input unit, a display unit, a camera (optional), an audio circuit (optional), a speaker (optional), a microphone (optional), a headphone jack (optional), a processor, an external interface, and a power supply. Those skilled in the art will appreciate that the aforementioned components are merely examples and do not limit the terminal or multi-function device, and may include more or fewer components, or a combination of certain components, or different components.

[0068] The input unit can be used to receive input digital or character information and generate key signal input related to user settings and function control of the portable multi-function device. Specifically, the input unit may include a touch screen (optional) and / or other input devices. Specifically, other input devices may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control keys, power keys, etc.), a trackball, a mouse, a joystick, etc.

[0069] Among them, the input device can receive input data and so on.

[0070] The display unit can be used to display information input by the user or information provided to the user, various menus of the terminal, interactive interfaces, file display and / or playback of any multimedia file. In an embodiment of the present application, the display unit can be used to display the interface of the image noise reduction process, processing results, etc.

[0071] Among them, the memory can be used to store software codes related to the image noise reduction method, the processor can execute the steps of the image noise reduction method, and can also schedule other units (such as the above-mentioned input unit and display unit) to achieve corresponding functions.

[0072] The radio frequency unit (optional) can be used to send and receive information or receive and send signals during a call.

[0073] In this embodiment of the present application, the radio frequency unit can send data to the server 200 and receive the processing results sent by the server 200.

[0074] It should be understood that the radio frequency unit is optional and can be replaced by other communication interfaces, such as a network port.

[0075] The terminal 100 further includes a power source (such as a battery) for supplying power to various components.

[0076] The terminal 100 also includes an external interface, which may be a standard Micro USB interface or a multi-pin connector, and may be used to connect the terminal 100 to other devices for communication, or to connect a charger to charge the terminal 100.

[0077] The server 200 includes a bus, a processor, a communication interface, and a memory. The processor, the memory, and the communication interface communicate with each other via the bus.

[0078] Among them, the memory can be used to store software codes related to the image noise reduction method, the processor can execute the steps of the image noise reduction method of the chip, and can also schedule other units to implement corresponding functions.

[0079] In order to solve the above problems, the present invention provides an image noise reduction method. The image noise reduction method of the present invention is described in detail below with reference to the accompanying drawings.

[0080] Reference Figure 2 , Figure 2 The process diagram of an image noise reduction method provided in an embodiment of the present application is as follows: Figure 2 As shown, a data processing method provided in an embodiment of the present application may include steps S201 to S206, and these steps are described in detail below.

[0081] S201 : Acquire coil data of multiple coils of a target slice obtained by scanning a target object.

[0082] In this embodiment, the target slice is any slice.

[0083] In this embodiment, coil data of multiple coils of at least one slice obtained by scanning a target object is acquired. The target object is a preset magnetic resonance imaging target body, such as human tissue, animal model, and non-biological material.

[0084] S202 : Obtaining original coil images of each coil based on the coil data of each coil of the target slice.

[0085] In this embodiment, an original coil image is obtained by performing Fourier transform on the coil data of each coil of the target slice. The original coil image is a complex image including a real image and an imaginary image.

[0086] S203 : Acquire an amplitude image of the target slice based on the original coil images of each coil.

[0087] In this embodiment, specific methods for obtaining the amplitude image of the target slice based on the original coil images of each coil of the target slice include a square sum method, a coil phase combination method based on K space, a SENSE (Sensitivity Encoding) algorithm, and other optional methods.

[0088] S204 : Acquire a simulated coil image of each coil based on the amplitude image of the target slice and the sensitivity map and phase map of each coil.

[0089] In this embodiment, if the sensitivity map and phase map do not exist in the parameter database, the sensitivity map and phase map can be calculated by comparing the amplitude image of the coil and the amplitude image of the slice. The specific calculation method can be found in the following embodiment.

[0090] S205 : Based on the simulated coil image and the original coil image of each coil, obtain the estimated noise component of the original coil image of each coil using a noise estimation model.

[0091] In this embodiment, a noise estimation model is derived from a pre-trained machine learning model and is used to estimate the noise component of the coil. The training data for the noise estimation model includes sample images and labels for the sample images. The sample images include sample noisy coil images and sample simulated coil images, and the labels include the sample noise component. The noise estimation model training method can be found in the following embodiments.

[0092] S206 : Acquire a target amplitude image of the scanned target object based on the estimated noise components of each coil of each slice and the original coil image.

[0093] In this embodiment, an amplitude image of a slice is reconstructed based on clean coil images of each coil, wherein the clean coil image is obtained by removing the estimated noise component from the original coil image.

[0094] As can be seen from the above technical solution, an image denoising method provided by an embodiment of the present application obtains coil data for a target slice obtained by scanning a target object, where the target slice is any slice. Based on the coil data for each coil of the target slice, an original coil image for each coil is obtained. Based on the original coil image for each coil, an amplitude image for the target slice is obtained. Based on the amplitude image for the target slice and the sensitivity map and phase map for each coil, a simulated coil image for each coil is obtained. Based on the simulated coil image and the original coil image for each coil, an estimated noise component for the original coil image for each coil is obtained using a noise estimation model. Based on the estimated noise component for each coil of each slice and the original coil image, a target amplitude image for the scanned target object is obtained. This method simultaneously leverages the high signal-to-noise ratio (SNR) of the amplitude image and the Gaussian distribution of the coil image. Noise is first estimated and removed using the Gaussian distribution of the coil image before generating the amplitude image. This ensures that the DC noise component is removed from the amplitude image generated after noise removal, thereby improving the denoising effect and image quality.

[0095] See also Figure 3 , Figure 3 A specific implementation flow chart of an image noise reduction method provided in an embodiment of the present application is as follows: Figure 3 As shown, it specifically includes S301 to S312, and these steps are described in detail below.

[0096] S301 : Acquire coil data of multiple coils of at least one slice obtained by scanning a target object.

[0097] In this embodiment, coil data of each slice of each coil obtained by scanning the target object is obtained. The coil data is obtained by scanning the target object. The target object is a preset magnetic resonance imaging target body, such as human tissue, animal model, and non-biological material.

[0098] In this embodiment, the number of slices and the number of coils corresponding to each slice are determined based on the actual scenario. The index of any slice is denoted by s. For example, in the context of 2D MRI image noise reduction, the number of slices is 1, where slice s represents the only slice. In the context of 3D MRI image noise reduction, the number of slices is S, and the slices are denoted by slices 1 through S. The multiple coils corresponding to slice s are denoted by coils 1 through J, where coil j is any coil.

[0099] S302 : For a target slice, obtain an original coil image of each coil based on the coil data of each coil of the target slice.

[0100] In this embodiment, an original coil image is obtained by performing Fourier transform on coil data obtained by scanning. The original coil image is a complex image including a real image and an imaginary image.

[0101] In this embodiment, the target slice is any slice, namely slice s, and the target coil is any coil, namely coil j. Taking coil j and slice s as an example, the original coil image X of the target coil obtained based on the coil data of the target slice is s,j Expressed as:

[0102]

[0103] in, Represents X s,j The real part of Represents X s,j The imaginary part of , i represents the imaginary unit.

[0104] It should be noted that the specific method for generating the original coil data based on Fourier transform can be referred to in the prior art.

[0105] S303 . For the target slice, based on the original coil images of each coil, obtain an amplitude image of the target slice.

[0106] In this embodiment, specific methods for obtaining the amplitude image of the target slice based on the original coil images of each coil of the target slice include a square sum method, a coil phase combination method based on K space, a SENSE (Sensitivity Encoding) algorithm, and other optional methods.

[0107] In this embodiment, the method for obtaining the amplitude image of the target slice includes:

[0108] A1. Acquire the amplitude image of each coil based on the original coil image of each coil.

[0109] For example, the amplitude of coil j is |X s,j |.

[0110] A2. Perform a square operation on the amplitude image of each original coil image. After eliminating the influence of phase difference, add the square operation results to obtain a total energy image.

[0111] A3. Take the square root of the total energy image to generate the amplitude image.

[0112] Taking slice s as an example, the calculation method of the amplitude image of slice s is as follows:

[0113]

[0114] Where J represents the total number of coils in slice s.

[0115] S304 : For the target slice, perform smoothing processing on the amplitude image of the target slice and the amplitude images of each coil respectively.

[0116] In this embodiment, each amplitude image is smoothed by a low-pass filter. Optionally, the low-pass filter is used to smooth the amplitude image using any one of the algorithms such as k-space filtering, moving average filtering or image space Gaussian kernel filtering. The amplitude image M of coil j after smoothing is j，smooth and the amplitude image M of slice s s,smooth As shown below:

[0117] M j,smooth =LPF(M j );

[0118] M s,smooth =LPF(Ms);

[0119] Among them, LPF represents low-pass filter, M j is the amplitude of coil j, that is, M j =|X s,j |.

[0120] S305 . For the target slice, based on the smoothed amplitude image of the target slice and the amplitude images of each coil, obtain a sensitivity map and a phase map of each coil.

[0121] In this embodiment, for the target coil, the ratio of the amplitude image of the smoothed target coil to the amplitude image of the target slice is calculated to obtain a sensitivity map. For example, the sensitivity map of coil j in slice s is calculated as follows:

[0122] Ss,j =M j,smooth / M s,smooth ;

[0123] In this embodiment, since the signal-to-noise ratio of the complex image is usually too low, a low-pass filter is required to calculate the phase map. For the target coil, the method for calculating the phase map includes:

[0124] B1. Use the preset angle acquisition function to obtain the original phase image p as follows:

[0125] p=arg(X s,j );

[0126] Wherein, arg represents the preset angle acquisition function.

[0127] B2. Smooth the original phase image p based on low-pass filtering.

[0128] Specifically, the original cosine value cosp and the original sine value sinp of the original phase image p are smoothed to obtain the smoothed cosine value p cos and sine value p sin , calculate the cosine value p cos and sine value p sin The smooth phase diagram is obtained by summing .

[0129] p cos =LPF(cosp);

[0130] p sin =LPF(sinp);

[0131] p complex =p cos +ip sin .

[0132] B3. Normalize the smoothed phase image to obtain the phase image of the target coil.

[0133] Specifically, the phase diagram of the target coil is expressed as:

[0134] p=arg(p complex ).

[0135] It should be noted that the sensitivity map and phase map for each coil can be pre-calculated and stored in a parameter database before scanning the target object. In this step, if the sensitivity map and phase map do not exist in the parameter database, they can be calculated by comparing the amplitude image of the coil and the amplitude image of the slice. Since sensitivity maps are generally smooth, the LPF has little effect on sensitivity map estimation. Due to the non-smooth nature of phase maps, low-pass filtering is required, which has little effect on sensitivity map estimation. Ideal phase maps may have edges, so the low-pass filter needs to be adjusted to avoid phase ambiguity.

[0136] S306 : For the target slice, based on the amplitude image of the target slice and the sensitivity map and phase map of each coil, obtain a simulated coil image of each coil.

[0137] In this embodiment, the real part of the simulated coil image of the target coil is the product of the amplitude image of the target slice, the sensitivity map of the target coil, and the cosine value of the phase map, and the imaginary part is the product of the simulated amplitude image, the sensitivity map of the target coil, and the sine value of the phase map.

[0138] That is, in slice s, the simulated coil image X' of coil j s,j is represented as follows:

[0139] X s,j =Ms·S s,j ·p cos +i×Ms·S s,j ·p sin .

[0140] S307 : For the target slice, pre-process the simulated coil image and the original coil image of each coil to obtain a target simulated coil image and a target original coil image of each coil.

[0141] In this embodiment, taking the target coil as an example, the simulated coil image (denoted as X C ) and the original coil image (denoted as X N ) are the same as those in the previous step, including but not limited to:

[0142] Normalize so that the mean of the real and imaginary parts is zero, and the standard deviation of the real and imaginary parts is 1.

[0143] and / or,

[0144] High-pass filter processing to create a noise-dominated image without contrast.

[0145] Note that the pre-processed simulated coil image X C and the original coil image X N are the target simulated coil images X' C and the target original coil image X'N .

[0146] S308 : For the target slice, the target simulated coil image and the target original coil image of each coil are respectively input into the noise estimation model to obtain the estimated noise component of each coil output by the noise estimation model.

[0147] In this embodiment, a noise estimation model is derived from a pre-trained machine learning model and is used to estimate the noise component of the coil. The training data for the noise estimation model includes sample images and labels for the sample images. The sample images include sample noisy coil images and sample simulated coil images, and the labels include the sample noise component. The noise estimation model training method can be found in the following embodiments.

[0148] In this embodiment, for the target coil, the target original coil image X' of the target coil output by the machine learning model is N The real part Re[N'] and imaginary part Im[N'] of the noise component N' in are expressed as:

[0149] Re[N']=net([Re(X N '),Re(X C ')]);

[0150] Im[N']=net([Im(X N '),Im(X C ')]);

[0151] Among them, net() represents a neural network, and [,] represents the connection between two variables.

[0152] S309 : For the target slice, target processing is performed on the estimated noise components of each coil to obtain the target noise components of each coil.

[0153] In this embodiment, the target processing is determined based on the pre-processing. Specifically, the target processing includes inverse processing for eliminating the influence of each pre-processing.

[0154] For example, for the normalization process in the preprocessing, the target process includes the inverse operation of the normalization process.

[0155] S310 : For a target slice, estimate a clean coil image of each coil based on the original coil image of each coil and a target noise component.

[0156] In this embodiment, the clean coil image X pred It is equal to the original coil image minus the target noise component, that is, the estimation method of the clean coil image is as follows:

[0157] X pred =X N -N

[0158] In summary, this method obtains a clean coil image of each coil of each slice based on S302 to S310. Optionally, a parallel computing system is constructed to execute S302 to S310 in parallel to obtain a clean coil image of each coil of each slice.

[0159] S311 . For a target slice, reconstruct a target amplitude image of the target slice based on clean coil images of all coils.

[0160] In this embodiment, the method for reconstructing the target amplitude image of the target slice based on the clean coil images of all coils is as follows:

[0161]

[0162] Where Mc,s represents the target amplitude image of slice s, X pred,s,j Represents the clean coil image of coil j at slice s.

[0163] S312 : Acquire a target amplitude image of the target object based on the target amplitude images of each slice.

[0164] As can be seen from the above technical solution, the image denoising method provided by the embodiments of the present application utilizes the Gaussian distribution characteristics of the original coil image. Specifically, the noise in the coil image of a complex image has a zero-mean Gaussian distribution. This makes it easy to predict noise from a noisy image, and it also reduces the difficulty of preparing a pair of clean and noisy images during machine learning training. Furthermore, by combining the characteristics of the simulated coil image with the original coil image and the amplitude image, both the simulated coil image and the original coil image are used as input data for the machine learning model, improving the accuracy of the estimated noise component and enabling the safe and efficient removal of DC components. This, combined with the high signal-to-noise ratio characteristics of the amplitude image, enhances the noise reduction effect.

[0165] It should be noted that the image denoising method provided in the embodiment of the present application can be applied to other medical imaging systems besides magnetic resonance imaging. When the images have similar size structures (I, C, D, H, W) or (C, D, H, W), the image with a higher signal-to-noise ratio (SNR) can be calculated by averaging the images in the dimensions of I and C or C. In one possible implementation, the image denoising method provided in the embodiment of the present application can be applied to the denoising scenario of a 3D image with dimensions (I, C, D, H, W), where (D, H, W) is a single volume image.

[0166] See also Figure 4 , Figure 4 The specific implementation process of the training method of a noise estimation model provided in the embodiment of the present application is as follows: Figure 4As shown, the method includes:

[0167] S401 , scanning a sample object multiple times, and obtaining multiple sample original coil images of each sample coil in each sample slice based on the sample coil data of the sample object.

[0168] In this embodiment, the sample coil data includes a collection of coil data of multiple sample coils for multiple sample slices. The sample coil collection data includes coil data obtained from each scan of the sample object. Optionally, for a method of obtaining an original coil image of a target coil for a target slice based on the sample coil data, see S302.

[0169] S402 : For each sample coil, calculate the arithmetic mean of all sample original coil images to obtain an average coil image.

[0170] In this embodiment, the average coil image obtained by calculating the arithmetic mean of multiple sample original coil images of the sample coil can improve the signal-to-noise ratio of the coil image without changing the noise distribution. The average coil image of the sample coil C is denoted as X m , X m It is obtained by calculating the arithmetic mean of the original coil images of each sample of the sample coil C, that is:

[0171]

[0172] Among them, K represents the number of scans, X k It represents the sample original coil image of the sample coil C corresponding to the k-th scan.

[0173] S403 : Generate a sample noisy coil image by adding a preset sample noise component to the average coil image of each sample coil.

[0174] For each target sample slice, an amplitude image of the target sample slice is constructed based on the sample noisy coil images of each sample coil of the target sample slice.

[0175] In this embodiment, the method for acquiring the amplitude image of the target sample slice is described in S303.

[0176] S404 . For the target sample slice, based on the amplitude image of the target sample slice and the sensitivity map and phase map of each sample coil, obtain a sample simulation coil image of each sample coil.

[0177] S405: Train a preset noise estimation model based on the training data until a preset model training condition is met.

[0178] In this embodiment, the training data includes sample images and labels of the sample images. The sample images include sample noisy coil images and sample simulated coil images. The labels include sample noise components.

[0179] It should be noted that complex data is generally not used during the training phase. Therefore, the sample image includes the target complex component of the sample noisy coil image and the target complex component of the sample simulated coil image, and the label includes the target complex component of the noise component. The target complex component includes either the real part or the imaginary part. For example, if the target complex component is the real part, the training data includes the real part of the sample noisy coil image and the real part of the sample simulated coil image, and the label is the real part of the sample noise component. The input data of the noise estimation model is the real part of the sample noisy coil image and the real part of the sample simulated coil image for each sample coil, and the target output is the real part of the sample noise component.

[0180] In this embodiment, the model training conditions include reaching a preset number of training times, or the model loss is less than a preset loss threshold.

[0181] It can be seen from the above technical solution that an image denoising method provided in an embodiment of the present application utilizes the zero-mean feature of the Gaussian distribution of the original coil image to construct a noisy image in the sample data, uses the noisy image and the corresponding simulated coil image as sample data at the same time, and uses the sample noise component as a label to train a machine learning model, thereby improving the noise estimation accuracy of the machine learning model.

[0182] The above describes an image denoising method provided by an embodiment of the present application. The following describes an apparatus for executing the above image denoising method.

[0183] See also Figure 5 , Figure 5 A schematic diagram of the structure of an image noise reduction device provided in an embodiment of the present application.

[0184] like Figure 5 As shown, the image noise reduction device 500 includes:

[0185] An object scanning unit 501 is configured to obtain coil data of a plurality of coils of a target slice obtained by scanning a target object, where the target slice is any slice;

[0186] An original image acquisition unit 502 is configured to obtain an original coil image of each coil based on the coil data of each coil of the target slice;

[0187] an amplitude image acquisition unit 503, configured to acquire an amplitude image of the target slice based on the original coil images of each coil;

[0188] A simulation image acquisition unit 504 is configured to acquire a simulation coil image of each coil based on the amplitude image of the target slice and the sensitivity map and phase map of each coil;

[0189] a noise estimation unit 505 configured to obtain an estimated noise component of the original coil image of each coil using a noise estimation model based on the simulated coil image and the original coil image of each coil;

[0190] The noise reduction image reconstruction unit 506 is configured to obtain a target amplitude image of the scanned target object based on the estimated noise components of each coil of each slice and the original coil image.

[0191] In a possible implementation, the amplitude image acquisition unit is configured to acquire the amplitude image of the target slice based on the original coil images of each coil, specifically to:

[0192] acquiring an amplitude image of each coil based on an original coil image of each coil;

[0193] After performing a square operation on the amplitude image of each original coil image of the coil, the square operation results are added to obtain a total energy map;

[0194] The square root of the total energy image is taken to generate an amplitude image of the target slice.

[0195] In a possible implementation, the simulation image acquisition unit is configured to acquire the simulation coil image of each coil based on the amplitude image of the target slice and the sensitivity map and phase map of each coil, specifically for:

[0196] performing smoothing processing on the amplitude image of the target slice and the corresponding amplitude images of each coil respectively;

[0197] acquiring a sensitivity map and a phase map of each coil based on the smoothed amplitude image of the target slice and the corresponding amplitude images of each coil;

[0198] Based on the amplitude image of the target slice and the corresponding sensitivity map and phase map of each coil, a simulated coil image of each coil is acquired.

[0199] In a possible implementation, the simulation image acquisition unit is configured to acquire the sensitivity map and the phase map of each coil based on the smoothed amplitude image of the target slice and the corresponding amplitude images of each coil, specifically for:

[0200] For any one of the coils, smoothing the amplitude image of the target slice and the amplitude image of the coil respectively is performed using a preset low-pass filter;

[0201] Calculating the ratio of the smoothed amplitude image of the coil to the amplitude image of the target slice, and obtaining the original phase image of the coil using a preset angle acquisition function;

[0202] Calculating the original phase image of the coil using a preset angle acquisition function based on the smoothed amplitude image of the coil and the amplitude image of the target slice;

[0203] The original smoothed phase image of the coil after smoothing is normalized to obtain the phase image of the coil.

[0204] In one possible implementation, the noise estimation unit is configured to obtain the estimated noise component of the original coil image of each coil using a noise estimation model based on the simulated coil image and the original coil image of each coil, specifically for:

[0205] Preprocessing the simulated coil image and the original coil image of each coil to obtain a target simulated coil image and a target original coil image of each coil, wherein the preprocessing includes a standardization process;

[0206] The target simulated coil image and the target original coil image of each coil are respectively input into the noise estimation model to obtain the estimated noise component of each coil output by the noise estimation model.

[0207] In a possible implementation, the image denoising apparatus further includes a model training unit configured to:

[0208] Obtaining a plurality of sample original coil images of each sample coil in each sample slice based on the sample coil data of the sample object;

[0209] For each of the sample coils, calculating an arithmetic mean of all sample original coil images of the sample coil to obtain an average coil image of the sample coil;

[0210] generating a sample noisy coil image based on adding a preset sample noise component to the average coil image of each of the sample coils;

[0211] For each of the sample slices, acquiring a sample simulated coil image of each of the sample coils based on the amplitude image of the sample slice and the corresponding sensitivity map and phase map of each of the sample coils;

[0212] A noise estimation model to be trained is trained based on training data until a preset model training condition is reached; wherein the training data includes sample images and labels of the sample images, the sample images include sample noisy coil images and sample simulated coil images, and the labels include sample noise components.

[0213] An electronic device is also provided in an embodiment of the present application. Figure 6 , which shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present application. The electronic device in the embodiments of the present application may include but is not limited to fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 6 The electronic device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0214] like Figure 6 As shown, the electronic device may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. When the electronic device is powered on, the RAM 603 also stores various programs and data required for the operation of the electronic device. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0215] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a memory card, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although Figure 6 The electronic device is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0216] An embodiment of the present application also provides a computer program product including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements any one of the image noise reduction methods provided in the embodiments of the present application.

[0217] An embodiment of the present application also provides a computer-readable storage medium, which carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any image noise reduction method provided in the embodiment of the present application.

[0218] It should also be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided in this application, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines.

[0219] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course can also be implemented by special hardware including application-specific integrated circuits, special CPUs, special memories, special components, etc. In general, all functions performed by computer programs can be easily implemented with corresponding hardware, and the specific hardware structures used to implement the same function can also be diverse, such as analog circuits, digital circuits or special circuits, etc. However, for the present application, software program implementation is a better implementation method in most cases. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer's floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc., and includes a number of instructions to enable a computer device (which can be a personal computer, training equipment, or network equipment, etc.) to execute the methods described in each embodiment of the present application.

[0220] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0221] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website, a computer, a training device or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website, a computer, a training device or a data center. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium can be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

Claims

1. An image denoising method, characterized in that: include: Acquire coil data of a plurality of coils of a target slice obtained by scanning a target object, wherein the target slice is any slice; obtaining an original coil image of each coil based on the coil data of each coil of the target slice; acquiring an amplitude image of the target slice based on the original coil images of each coil; acquiring a simulated coil image of each coil based on the amplitude image of the target slice and the sensitivity map and phase map of each coil; Based on the simulated coil image and the original coil image of each coil, using a noise estimation model to obtain an estimated noise component of the original coil image of each coil; A target amplitude image of the scanned target object is acquired based on the estimated noise components of each coil of each slice and an original coil image.

2. The image denoising method according to claim 1, wherein: The acquiring of the amplitude image of the target slice based on the original coil images of each coil includes: acquiring an amplitude image of each coil based on an original coil image of each coil; After performing a square operation on the amplitude image of each original coil image of the coil, the square operation results are added to obtain a total energy map; The square root of the total energy image is taken to generate an amplitude image of the target slice.

3. The image denoising method according to claim 1, wherein: The acquiring of a simulated coil image of each coil based on the amplitude image of the target slice and the sensitivity map and phase map of each coil comprises: performing smoothing processing on the amplitude image of the target slice and the corresponding amplitude images of each coil respectively; Acquire a sensitivity map and a phase map of each coil based on the smoothed amplitude image of the target slice and the corresponding amplitude images of each coil; Based on the amplitude image of the target slice and the corresponding sensitivity map and phase map of each coil, a simulated coil image of each coil is acquired.

4. The image denoising method according to claim 3, wherein: The step of obtaining a sensitivity map and a phase map of each coil based on the smoothed amplitude image of the target slice and the corresponding amplitude images of each coil includes: For any one of the coils, smoothing the amplitude image of the target slice and the amplitude image of the coil respectively is performed using a preset low-pass filter; Calculating the ratio of the smoothed amplitude image of the coil to the amplitude image of the target slice, and obtaining the original phase image of the coil using a preset angle acquisition function; Calculating the original phase image of the coil using a preset angle acquisition function based on the smoothed amplitude image of the coil and the amplitude image of the target slice; The original smoothed phase image of the coil after smoothing is normalized to obtain the phase image of the coil.

5. The image denoising method according to claim 1, wherein: The step of obtaining an estimated noise component of the original coil image of each coil using a noise estimation model based on the simulated coil image and the original coil image of each coil comprises: Preprocessing the simulated coil image and the original coil image of each coil to obtain a target simulated coil image and a target original coil image of each coil, wherein the preprocessing includes a standardization process; The target simulated coil image and the target original coil image of each coil are respectively input into the noise estimation model to obtain the estimated noise component of each coil output by the noise estimation model.

6. The image denoising method according to claim 1, wherein: The training method of the noise estimation model comprises: Obtaining a plurality of sample original coil images of each sample coil in each sample slice based on the sample coil data of the sample object; For each of the sample coils, calculating an arithmetic mean of all sample original coil images of the sample coil to obtain an average coil image of the sample coil; generating a sample noisy coil image based on adding a preset sample noise component to the average coil image of each of the sample coils; For each of the sample slices, acquiring a sample simulated coil image of each of the sample coils based on the amplitude image of the sample slice and the corresponding sensitivity map and phase map of each of the sample coils; A noise estimation model to be trained is trained based on training data until a preset model training condition is reached; wherein the training data includes sample images and labels of the sample images, the sample images include sample noisy coil images and sample simulated coil images, and the labels include sample noise components.

7. An image noise reduction device, characterized in that: include: an object scanning unit, configured to obtain coil data of a plurality of coils of a target slice obtained by scanning a target object, wherein the target slice is any slice; an original image acquisition unit, configured to obtain an original coil image of each coil based on the coil data of each coil of the target slice; an amplitude image acquisition unit, configured to acquire an amplitude image of the target slice based on the original coil images of each of the coils; a simulation image acquisition unit, configured to acquire a simulation coil image of each of the coils based on the amplitude image of the target slice and the sensitivity map and phase map of each of the coils; a noise estimation unit, configured to obtain an estimated noise component of the original coil image of each coil using a noise estimation model based on the simulated coil image and the original coil image of each coil; The noise reduction image reconstruction unit is configured to obtain a target amplitude image of the scanned target object based on the estimated noise components of each coil of each slice and the original coil image.

8. A computer program product, characterized in that The method comprises computer-readable instructions, which, when executed on an electronic device, enable the electronic device to implement the image noise reduction method according to any one of claims 1 to 6.

9. An electronic device, characterized in that: comprising at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is configured to execute the computer program so that the electronic device can implement the image noise reduction method according to any one of claims 1 to 6.

10. A computer storage medium, characterized in that The storage medium carries one or more computer programs, and when the one or more computer programs are executed by an electronic device, the electronic device can implement the image noise reduction method according to any one of claims 1 to 6.

Images