Image noise reduction method, device, equipment, storage medium and program product

Abstract

The present application relates to an image denoising method, apparatus, device, storage medium and program product. The method comprises: acquiring a single plain scan image and a single delayed enhanced image; performing denoising on the single delayed enhanced image based on the single plain scan image, the single delayed enhanced image and a preset image denoising network, and determining the denoised images corresponding to the single delayed enhanced image and the single plain scan image; wherein the image denoising network is trained based on a single noisy plain scan image and a corresponding gold standard single plain scan image, a single noisy delayed enhanced image and a corresponding gold standard single delayed enhanced image; and the gold standard single plain scan image and the gold standard single delayed enhanced image are both denoised images. The present method can reduce the scanning radiation of the object to be measured.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to an image noise reduction method, apparatus, device, storage medium, and program product. Background Art

[0002] Computed Tomography Perfusion (CTP) can provide a high-quality image foundation when analyzing quantitative indicators of tissues. In the CTP scanning and analysis process, images from three periods are generally obtained, which are, in order of the time of contrast agent injection, plain scan images during the plain scan period, enhanced sequence images during the enhanced period, and delayed enhanced sequence images. Among them, many clinically significant parameters can be obtained through delayed enhanced sequence images, and delayed enhanced sequences are usually acquired when the tissue is not fully filled with contrast agent. Therefore, the imaging quality of delayed enhanced sequence images is low, and the noise on the images is relatively large.

[0003] In related technologies, when performing noise reduction on delayed enhancement sequence images, most of the methods are to repeatedly scan the object to be tested to obtain multiple scanned delayed enhancement sequence images, and then perform weighted averaging on the multiple scanned delayed enhancement sequence images to finally obtain the noise reduced delayed enhancement sequence images.

[0004] However, the above technology has the problem of large scanning radiation to the object to be measured. Summary of the Invention

[0005] Based on this, it is necessary to provide an image noise reduction method, device, equipment, storage medium and program product that can reduce the scanning radiation of the object to be measured in order to solve the above technical problems.

[0006] In a first aspect, the present application provides an image noise reduction method, the method comprising:

[0007] Acquire a single plain scan image and a single delayed enhanced image;

[0008] performing noise reduction processing on the single delayed enhanced image according to the single plain scan image, the single delayed enhanced image, and a preset image noise reduction network, and determining noise reduction images corresponding to the single delayed enhanced image and the single plain scan image;

[0009] Among them, the above-mentioned image denoising network is trained based on a single noisy plain scan image and its corresponding gold standard single plain scan image, a single noisy delayed enhanced image and its corresponding gold standard single delayed enhanced image; the above-mentioned gold standard single plain scan image and gold standard single delayed enhanced image are both denoised images.

[0010] In one embodiment, the above-mentioned performing noise reduction processing on the single delayed-enhanced image based on the single plain scan image and the single delayed-enhanced image and a preset image noise reduction network to determine the noise reduction images corresponding to the single delayed-enhanced image and the single plain scan image includes:

[0011] Registering a single plain scan image to the spatial coordinate system of a single delayed enhancement image to obtain a single registered plain scan image;

[0012] A single registered plain scan image and a single delayed enhanced image are input into the image denoising network for denoising to determine the denoised image.

[0013] In one embodiment, before inputting the single registered plain scan image and the single delayed enhanced image into the image denoising network for denoising, and determining the denoised image, the method further includes:

[0014] Determining a first segmentation result of the first region of interest and a second segmentation result of the second region of interest based on the single plain scan image and the single delayed enhanced image, wherein both the single plain scan image and the single delayed enhanced image include the first region of interest and the second region of interest;

[0015] An image value variation parameter corresponding to the second region of interest is determined according to the second segmentation result.

[0016] In one embodiment, the image denoising network includes a denoising subnetwork and a parameter calculation subnetwork; the step of inputting a single registered plain scan image and a single delayed enhanced image into the image denoising network for denoising to determine a denoised image includes:

[0017] Inputting the single registered plain scan image and the single delayed enhanced image into the denoising subnetwork for denoising, and determining a first denoised image corresponding to the single delayed enhanced image and a second denoised image corresponding to the single registered plain scan image;

[0018] Determining a target parameter map corresponding to the first region of interest based on the first denoised image, the second denoised image, the image value change parameter, and the parameter calculation subnetwork; the target parameter map including target parameters of each pixel point in the first region of interest;

[0019] The first denoised image, the second denoised image, and the target parameter map are determined as denoised images.

[0020] In one embodiment, determining the target parameter map corresponding to the first region of interest based on the first denoised image, the second denoised image, the image value change parameter, and the parameter calculation subnetwork includes:

[0021] In the parameter calculation subnetwork, an image difference between each pixel point in the first region of interest of the first denoised image and a corresponding pixel point in the first region of interest of the second denoised image is calculated to obtain a first difference corresponding to each pixel point in the first region of interest;

[0022] According to each first difference and the image value change parameter, a target parameter corresponding to each pixel point in the first region of interest is determined to obtain a target parameter map.

[0023] In one embodiment, determining the image value change parameter corresponding to the second region of interest based on the second segmentation result includes:

[0024] Calculating a first mean parameter corresponding to a second region of interest in a second segmentation result of the single plain scan image;

[0025] Calculating a second mean parameter corresponding to a second region of interest in a second segmentation result of the single delayed enhancement image;

[0026] An image value variation parameter corresponding to the second region of interest is determined according to the first mean parameter and the second mean parameter.

[0027] In one embodiment, the training method of the image denoising network includes:

[0028] Acquire a sample plain scan sequence and a sample delayed enhancement sequence; the sample plain scan sequence includes a plurality of sample plain scan images, and the sample delayed enhancement sequence includes a plurality of sample delayed enhancement images;

[0029] Determine a gold standard single plain scan image and a gold standard single delayed enhanced image according to the sample plain scan sequence and the sample delayed enhanced sequence;

[0030] Noise is added to the gold standard single plain scan image and the gold standard single delayed enhanced image respectively to determine a single noise plain scan image and a single noise delayed enhanced image;

[0031] The initial image denoising network is trained according to a single noisy plain scan image, a single noisy delayed enhanced image, a gold standard single plain scan image and a gold standard single delayed enhanced image to determine the image denoising network.

[0032] In one embodiment, the training of the initial image denoising network based on the single noisy plain scan image, the single noisy delayed enhanced image, the gold standard single plain scan image, and the gold standard single delayed enhanced image includes:

[0033] Determining, based on a gold standard single plain scan image and a gold standard single delayed enhanced image, a sample image value change parameter corresponding to a second region of interest of the sample and a gold standard parameter map corresponding to a first region of interest of the sample;

[0034] The initial image denoising network is trained based on the sample image value variation parameters, the gold standard parameter map, a single noisy plain scan image, a single noisy delayed enhanced image, a gold standard single plain scan image, and a gold standard single delayed enhanced image.

[0035] In one embodiment, the step of adding noise to the gold standard single delayed-enhanced image to determine the single noise delayed-enhanced image includes:

[0036] In the process of adding noise to the gold standard single delayed enhanced image, the signal-to-noise ratio of the gold standard single delayed enhanced image after adding noise is calculated;

[0037] When the signal-to-noise ratio reaches a signal-to-noise ratio threshold, a single noise delay enhanced image is determined.

[0038] In one embodiment, determining the gold standard single plain scan image and the gold standard single delayed enhanced image based on the sample plain scan sequence and the sample delayed enhanced sequence includes:

[0039] Perform registration and averaging on multiple sample plain scan images in the sample plain scan sequence to determine the gold standard single plain scan image corresponding to the sample plain scan sequence;

[0040] Multiple sample delayed enhancement images in the sample delayed enhancement sequence are registered and averaged to determine the gold standard single delayed enhancement image corresponding to the sample delayed enhancement sequence.

[0041] In one embodiment, before determining the gold standard single plain scan image and the gold standard single delayed enhancement image based on the sample plain scan sequence and the sample delayed enhancement sequence, the method further includes:

[0042] Performing registration processing on multiple sample plain scan images in a sample plain scan sequence to obtain a sample plain scan sequence after initial registration;

[0043] Performing registration processing on multiple sample delay enhancement images in the sample delay enhancement sequence to obtain a sample delay enhancement sequence after initial registration;

[0044] The sample plain scan sequence after the initial registration is registered to the spatial coordinate system where the sample delayed enhancement sequence after the initial registration is located, so as to obtain a registered sample plain scan sequence.

[0045] In a second aspect, the present application further provides an image noise reduction device, the device comprising:

[0046] An image acquisition module, used for acquiring a single plain scan image and a single delayed enhanced image;

[0047] a noise reduction module for performing noise reduction processing on the single delayed enhanced image based on the single plain scan image, the single delayed enhanced image, and a preset image noise reduction network, and determining noise reduction images corresponding to the single delayed enhanced image and the single plain scan image;

[0048] Among them, the above-mentioned image denoising network is trained based on a single noisy plain scan image and its corresponding gold standard single plain scan image, a single noisy delayed enhanced image and its corresponding gold standard single delayed enhanced image; the above-mentioned gold standard single plain scan image and gold standard single delayed enhanced image are both denoised images.

[0049] In a third aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0050] Acquire a single plain scan image and a single delayed enhanced image;

[0051] performing noise reduction processing on the single delayed enhanced image according to the single plain scan image, the single delayed enhanced image, and a preset image noise reduction network, and determining noise reduction images corresponding to the single delayed enhanced image and the single plain scan image;

[0052] Among them, the above-mentioned image denoising network is trained based on a single noisy plain scan image and its corresponding gold standard single plain scan image, a single noisy delayed enhanced image and its corresponding gold standard single delayed enhanced image; the above-mentioned gold standard single plain scan image and gold standard single delayed enhanced image are both denoised images.

[0053] In a fourth aspect, the present application further provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the following steps are implemented:

[0054] Acquire a single plain scan image and a single delayed enhanced image;

[0055] performing noise reduction processing on the single delayed enhanced image according to the single plain scan image, the single delayed enhanced image, and a preset image noise reduction network, and determining noise reduction images corresponding to the single delayed enhanced image and the single plain scan image;

[0056] Among them, the above-mentioned image denoising network is trained based on a single noisy plain scan image and its corresponding gold standard single plain scan image, a single noisy delayed enhanced image and its corresponding gold standard single delayed enhanced image; the above-mentioned gold standard single plain scan image and gold standard single delayed enhanced image are both denoised images.

[0057] In a fifth aspect, the present application further provides a computer program product, the computer program product comprising a computer program, which, when executed by a processor, implements the following steps:

[0058] Acquire a single plain scan image and a single delayed enhanced image;

[0059] performing noise reduction processing on the single delayed enhanced image according to the single plain scan image, the single delayed enhanced image, and a preset image noise reduction network, and determining noise reduction images corresponding to the single delayed enhanced image and the single plain scan image;

[0060] Among them, the above-mentioned image denoising network is trained based on a single noisy plain scan image and its corresponding gold standard single plain scan image, a single noisy delayed enhanced image and its corresponding gold standard single delayed enhanced image; the above-mentioned gold standard single plain scan image and gold standard single delayed enhanced image are both denoised images.

[0061] The above-described image denoising method, apparatus, device, storage medium, and program product acquire a single plain image and a single delayed-enhanced image, and perform denoising on the single delayed-enhanced image based on the single plain image, the single delayed image, and an image denoising network, thereby determining the denoised images corresponding to the single delayed-enhanced image and the single plain image. The image denoising network is trained based on the single noisy plain image and its gold standard single plain image, the single noisy delayed-enhanced image, and its gold standard single delayed-enhanced image, both of which are denoised images. In this method, since the denoising of the single delayed-enhanced image can be performed using the image denoising network trained using the single plain image, the single delayed-enhanced image, and their respective gold standards, the denoised single delayed-enhanced image has a high image quality. This eliminates the need to repeatedly acquire multiple delayed-enhanced images and average them for denoising the delayed-enhanced image. Therefore, scanning radiation caused by repeated scanning of the object under test during repeated acquisition can be avoided, thereby reducing radiation damage to the object under test. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a diagram of the internal structure of a computer device in one embodiment;

[0063] Figure 2 1 is a flow chart of an image noise reduction method according to an embodiment;

[0064] Figure 3 is a flow chart of an image noise reduction method according to another embodiment;

[0065] Figure 4 is a flow chart of an image noise reduction method according to another embodiment;

[0066] Figure 5 is a flow chart of an image noise reduction method according to another embodiment;

[0067] Figure 6 is a flow chart of an image noise reduction method according to another embodiment;

[0068] Figure 7 is an example diagram of a low-noise image and a high-noise image in another embodiment;

[0069] Figure 8 is a flow chart of an image noise reduction method according to another embodiment;

[0070] Figure 9 FIG. 4 is a structural block diagram of an image noise reduction device in one embodiment. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0072] The image noise reduction method provided in the embodiment of the present application can be applied to a computer device, which can be a terminal or a server. Taking the terminal as an example, its internal structure diagram can be as follows: Figure 1 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, an image noise reduction method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0073] Those skilled in the art will understand that Figure 1The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0074] In one embodiment, Figure 2 As shown, an image noise reduction method is provided, which is applied to Figure 1 Taking the computer device in the example as an example, the method may include the following steps:

[0075] S202, acquiring a single plain scan image and a single delayed enhanced image.

[0076] A plain scan image refers to an image obtained by reconstructing the scanned data after scanning the subject's test area without injecting a contrast agent. An enhanced image refers to an enhanced image obtained by reconstructing the scanned data after injecting a contrast agent into the subject's test area during the enhancement period. A delayed enhanced image refers to a delayed enhanced image obtained by reconstructing the scanned data after scanning the subject's test area during a delay period after the enhancement period.

[0077] In this step, the subject's target area can be scanned and image reconstructed during a plain scan phase without contrast agent injection to obtain a single plain scan image. Similarly, the subject's target area can be scanned and image reconstructed during a delay period after the enhancement phase to obtain a single delayed enhancement image. The single plain scan image and the single delayed enhancement image can be two-dimensional images.

[0078] This article focuses on plain scan images and delayed contrast-enhanced images. Enhanced images obtained during the enhancement phase generally have high quality and do not require noise reduction. Single plain scan images obtained during the plain scan phase generally do not require contrast injection, so image quality is also lower. Single delayed contrast-enhanced images also require noise reduction due to the low contrast agent concentration in the area under examination and excessive image noise.

[0079] In addition, the aforementioned test site may be the chest of the subject to be tested. Of course, the test site may also be other parts of the subject to be tested, such as the brain. Taking the chest as an example, the test site may include a first region of interest and a second region of interest of the chest. The first region of interest and the second region of interest may be, for example, the myocardium, a blood pool within the myocardium, and the like.

[0080] S204: performing noise reduction processing on the single delayed enhanced image according to the single plain scan image, the single delayed enhanced image, and a preset image noise reduction network to determine noise reduction images corresponding to the single delayed enhanced image and the single plain scan image.

[0081] The image denoising network can be a neural network, and its specific network type and architecture are not specifically limited herein. The image denoising network is primarily used to denoise a single delayed-enhanced image to improve its quality. Of course, the image denoising network can also denoise a single plain scan image to obtain a single plain scan image of higher quality. Alternatively, the image denoising network can calculate relevant parameters of the target area on the denoised single delayed-enhanced image to obtain corresponding parameter calculation results.

[0082] Of course, the image denoising network can be trained before use. During the training process, the image denoising network can be trained based on a single noisy plain scan image and the corresponding gold standard single plain scan image, a single noisy delayed enhanced image and the corresponding gold standard single delayed enhanced image; the above-mentioned gold standard single plain scan image and the gold standard single delayed enhanced image are both denoised images; that is, the single noisy plain scan image and the single noise delayed enhanced image can be used as the input of the initial image denoising network, and the gold standard single delayed enhanced image and the gold standard single delayed enhanced image can be used as the reference output of the initial image denoising network. By training the initial image denoising network, a trained image denoising network can be obtained.

[0083] After the image denoising network is trained, the single plain scan image and the single delayed enhanced image obtained above can be directly or after image preprocessing, input into the trained image denoising network for denoising to obtain the corresponding denoised image. The denoised image here can include only the denoised image corresponding to the single delayed enhanced image, or can include the denoised image corresponding to the single delayed enhanced image and the denoised image corresponding to the single plain scan image, or can include the denoised image corresponding to the single delayed enhanced image and the denoised image corresponding to the single plain scan image, as well as relevant parameter maps of the region of interest of the measured part, etc.

[0084] It should be noted that the quality of the image after denoising is higher than that of the image before denoising, and therefore can be used for subsequent parameter calculation, image analysis, etc. In addition, the above-mentioned image preprocessing may include image normalization, resampling, registration, etc.

[0085] In the above-mentioned image denoising method, a single plain image and a single delayed-enhanced image are acquired, and denoising is performed on the single delayed-enhanced image based on the single plain image, the single delayed image, and an image denoising network, thereby determining the denoised images corresponding to the single delayed-enhanced image and the single plain image. The image denoising network is trained based on the single noisy plain image and its gold standard single plain image, the single noisy delayed-enhanced image and its gold standard single delayed-enhanced image, and both the gold standard single plain image and the gold standard single delayed-enhanced image are denoised images. In this method, since the denoising of the single delayed-enhanced image can be performed using the image denoising network trained using the single plain image, the single delayed-enhanced image, and their respective gold standards, the denoised single delayed-enhanced image obtained has high image quality. This eliminates the need to repeatedly acquire multiple delayed-enhanced images and average them for denoising the delayed-enhanced image. Therefore, scanning radiation caused by repeated scanning of the object under test during repeated acquisition can be avoided, thereby reducing radiation damage to the object under test.

[0086] When using an image denoising network to denoise a single delayed enhanced image, in order to improve the denoising effect and facilitate subsequent parameter calculation, preprocessing such as image registration can usually be performed. The following embodiment describes this process in detail.

[0087] In another embodiment, another image noise reduction method is provided. Based on the above embodiment, Figure 3 As shown, the above S202 may include the following steps:

[0088] S302 , registering the single plain scan image to the spatial coordinate system of the single delayed enhancement image to obtain a single registered plain scan image.

[0089] In this step, when a single plain scan image and a single enhanced image are collected, the spatial coordinate systems in which they are located can be obtained. Then, by selecting several feature points on the two images, the transformation relationship between the two spatial coordinate systems can be calculated. After that, each point on the single plain scan image can be transformed according to the transformation relationship to align the single plain scan image to the spatial coordinate system where the single delayed enhanced image is located, thereby obtaining a single registered plain scan image.

[0090] S304: Input the single registered plain scan image and the single delayed enhanced image into an image denoising network for denoising, and determine a denoised image.

[0091] In this step, after obtaining a single registered plain scan image and a single delayed enhanced image, both can be input into the image denoising network for denoising processing to obtain the denoised images corresponding to the single delayed enhanced image and the single registered plain scan image.

[0092] In this embodiment, a single plain scan image is registered to the spatial coordinate system of a single delayed enhanced image and then inputted into the image denoising network for denoising processing, thereby obtaining denoised images corresponding to the single delayed enhanced image and the single registered plain scan image. In this way, through image registration, image denoising can be facilitated while also facilitating subsequent parameter calculation and image analysis of the denoised image, thereby obtaining more accurate calculation and analysis results.

[0093] The above-mentioned image denoising network can also calculate relevant parameters on the image while actually reducing noise. Therefore, the process of calculating the parameters needs to calculate the relevant parameters in advance. The following embodiment illustrates this process.

[0094] In another embodiment, another image noise reduction method is provided. Based on the above embodiment, Figure 4 As shown, before the above S304, the above method may further include the following steps:

[0095] S402, determining a first segmentation result of a first region of interest and a second segmentation result of a second region of interest based on a single plain scan image and a single delayed enhanced image; both the single plain scan image and the single delayed enhanced image include the first region of interest and the second region of interest.

[0096] Taking the chest as an example, the first region of interest may be the myocardium, and the second region of interest may be the blood pool in the myocardium.

[0097] In this step, a segmentation network or a segmentation algorithm can be specifically used to perform segmentation processing on a single plain scan image to obtain a first segmentation result of the first region of interest and a second segmentation result of the second region of interest in the single plain scan image. The first segmentation result of the first region of interest may include a binary mask image of the first region of interest, so that the first region of interest and the background can be quickly distinguished; similarly, the second segmentation result of the second region of interest may include a binary mask image of the second region of interest, so that the second region of interest and the background can be quickly distinguished. Alternatively, the first segmentation result of the first region of interest and the second segmentation result of the second region of interest can be the same multi-valued mask image, which can distinguish the first region of interest, the second region of interest and the background. It should be noted that the single plain scan image here is an image registered with a single delayed enhanced image.

[0098] Similarly, a segmentation network or segmentation algorithm can also be used to segment a single delayed-enhanced image to obtain a first segmentation result for a first region of interest and a second segmentation result for a second region of interest in the single delayed-enhanced image. The first segmentation result and the second segmentation result can be similar to the segmentation results of the single plain scan image described above, and will not be further described here.

[0099] Since the positions of the first region of interest and the second region of interest in the above-mentioned single registered plain scan image and the single delayed enhanced image are corresponding, the segmentation results obtained by the single registered plain scan image and the single delayed enhanced image should also be the same. Therefore, the first segmentation result of the first region of interest and the second segmentation result of the second region of interest here can be selected by segmenting the single delayed enhanced image.

[0100] In addition, the above-mentioned segmentation network can be pre-trained, and the samples used in the image denoising network training can be used during the specific training. For example, a sample plain scan sequence including multiple sample plain scan images and a sample delayed enhanced sequence including multiple sample delayed enhanced images can be obtained, wherein each sample image is pre-marked with the first region of interest and the second region of interest. After that, each sample image can be used to train the initial segmentation network to obtain a trained segmentation network.

[0101] S404: Determine an image value change parameter corresponding to the second region of interest according to the second segmentation result.

[0102] In this step, the image value may include an image value in a CT (Computed Tomography) image or an image value in an MR (Magnetic Resonance) image. The image value in the CT image may be expressed as a CT value or a Hounsfield Unit (HU) value, which are equivalent concepts.

[0103] Correspondingly, the above-mentioned image value change parameters may include image value change parameters in CT images or image value change parameters in MR images, etc.

[0104] After obtaining the first segmentation result and the second segmentation result as described above, the first segmentation result is mainly for the first region of interest. Here, the parameters are calculated mainly for the second region of interest. Then, as an optional embodiment, the first mean parameter corresponding to the second region of interest in the second segmentation result of a single plain scan image can be calculated; the second mean parameter corresponding to the second region of interest in the second segmentation result of a single delayed enhancement image can be calculated; and the image value change parameter corresponding to the second region of interest is determined based on the first mean parameter and the second mean parameter.

[0105] Specifically, for a single plain image, the image values ​​corresponding to each point in the second region of interest in the second segmentation result can be obtained on the single plain image, and then the image values ​​at each point are averaged to obtain a first mean parameter. For a single delayed-enhanced image, the image values ​​corresponding to each point in the second region of interest in the second segmentation result can be obtained on the single delayed-enhanced image, and then the image values ​​at each point are averaged to obtain a second mean parameter. The second mean parameter can then be subtracted from the first mean parameter, and the difference obtained is the image value change parameter.

[0106] For example, taking the second region of interest as the blood pool in the myocardium as an example, the first mean parameter calculated above can be recorded as HU blood,pre , the calculated second mean parameter can be recorded as HU blood,post , then the image value change parameter can be expressed as ΔHU blood , which can be expressed as: ΔHU blood =HU blood,post -HU blood,pre .

[0107] In this embodiment, a single plain scan image and a single delayed-contrast image are segmented into a first region of interest and a second region of interest, and the image value change parameter of the second region of interest is determined based on the segmentation results of the second region of interest. This segmentation and subsequent calculation of the image value change parameter allows calculations to be performed only on the second region of interest, improving the efficiency and accuracy of the image value change parameter calculation. Furthermore, the image value change parameter calculation process is refined by using the segmentation results of the second region of interest for both the single plain scan image and the single delayed-contrast image, resulting in a more accurate image value change parameter.

[0108] In the above embodiment, it is mentioned that the image denoising network can perform image denoising and parameter calculation on the image. The image denoising network may include a denoising subnetwork and a parameter calculation subnetwork to simultaneously perform image denoising and parameter calculation processes. The following embodiment illustrates this process.

[0109] In another embodiment, another image noise reduction method is provided. Based on the above embodiment, Figure 5 As shown, the above S304 may include the following steps:

[0110] S502: Input the single registered plain scan image and the single delayed enhanced image into a denoising subnetwork for denoising, and determine a first denoised image corresponding to the single delayed enhanced image and a second denoised image corresponding to the single registered plain scan image.

[0111] In this step, both the denoising subnetwork and the parameter calculation subnetwork can be neural networks, and the specific network type and network architecture are not specifically limited here.

[0112] As mentioned above, a single registered plain scan image and a single delayed enhanced image can be input into the denoising subnetwork of the image denoising network for denoising, yielding respective denoised images. The denoised image of the delayed enhanced image is referred to as the first denoised image, and the denoised image of the registered plain scan image is referred to as the second denoised image. Both the first and second denoised images have higher image quality than the corresponding images before denoising.

[0113] S504: Determine a target parameter map corresponding to the first region of interest based on the first denoised image, the second denoised image, the image value change parameter, and the parameter calculation subnetwork; the target parameter map includes target parameters of each pixel in the first region of interest.

[0114] Taking the example where the first ROI is the myocardium and the second ROI is the blood pool in the myocardium, the target parameter here may be an ECV (Extracellular Volume) parameter, and of course may also be other parameters.

[0115] When the parameter calculation subnetwork is used to perform parameter calculation for the above-mentioned denoised image, as an optional embodiment, the image difference between each pixel point in the first region of interest of the first denoised image and the corresponding pixel point in the first region of interest of the second denoised image is calculated in the parameter calculation subnetwork to obtain the first difference corresponding to each pixel point in the first region of interest; based on each first difference and the image value change parameter, the target parameter corresponding to each pixel point in the first region of interest is determined to obtain a target parameter map.

[0116] Specifically, in the parameter calculation subnetwork, the first image value of each pixel in the first region of interest can be obtained on the first denoised image of a single delayed-enhanced image based on the first segmentation result corresponding to the single delayed-enhanced image. Similarly, the second image value of each pixel in the first region of interest can be obtained on the second denoised image of a single plain scan image based on the first segmentation result corresponding to the single plain scan image. Subsequently, the image value difference between the pixels at corresponding positions on the two denoised images can be calculated, and the calculated difference for each pixel can be recorded as the first difference. A pre-set constant related to the object to be measured can then be obtained. Then, subtraction, multiplication, and division operations can be performed using the first difference for each pixel, the constant, and the calculated image value change parameter to obtain the target parameter for each pixel in the first region of interest. By combining the target parameters for each point in the first region of interest, a target parameter map can be obtained.

[0117] For example, taking the example that the first region of interest is the myocardium and the second region of interest is the blood pool in the myocardium, the image value change parameter calculated above is ΔHU blood , the following formula can be used to calculate the target parameters of each pixel in the first region of interest. Taking the target parameter as ECV as an example, the formula is as follows:

[0118] ECV=(ΔHU myocardium / ΔHU blood )*(1-HCT blood )

[0119] Among them, HCT blood is a constant related to the object to be measured, which can generally be taken as 0.45; ΔHU myocardium It represents the first difference and can be expressed by the following formula:

[0120] ΔHU myocardium =HU myocardium,post -HU myocardium,pre

[0121] Among them, HU myocardium,post HU represents the image value of a pixel in the first region of interest (i.e., myocardium) in the first denoised image. myocardium,pre Represents the image value of a pixel in the first region of interest (ie, the myocardium) in the second denoised image.

[0122] The target parameters of each pixel point on the first region of interest can be calculated by the above formula, and then the target parameter map including the target parameters of all pixel points can be obtained.

[0123] S506: Determine the first denoised image, the second denoised image, and the target parameter map as denoised images.

[0124] After obtaining the first denoised image corresponding to the single delayed enhanced image, the second denoised image corresponding to the registered single plain scan image, and the target parameter map corresponding to the first region of interest, these two denoised images and the target parameter map can be used as the denoised images corresponding to the single delayed enhanced image and the single plain scan image mentioned above.

[0125] In this embodiment, a denoising subnetwork within an image denoising network is employed to denoise a single registered plain scan image and a single delayed enhanced image, obtaining a corresponding denoised image. The denoised image and image value variation parameters are then used in the parameter calculation subnetwork to calculate a target parameter map for the first region of interest. This two-part subnetwork for denoising and parameter calculation allows for a simple and rapid implementation of the denoising and parameter calculation process for a single delayed enhanced image, improving the overall efficiency of denoising and parameter calculation. Furthermore, during the parameter calculation process, the image difference between the first region of interest in the two denoised images is calculated, and then combined with the image value variation parameters to obtain a corresponding target parameter map within the parameter calculation subnetwork. This refinement of the target parameter map calculation process results in a more accurate target parameter map.

[0126] The above embodiment mentioned a simple training process of the image denoising network. The following embodiment mainly describes the detailed training process.

[0127] In another embodiment, another image noise reduction method is provided. Based on the above embodiment, Figure 6 As shown, the training method of the above image denoising network may include the following steps:

[0128] S602 , obtaining a sample plain scan sequence and a sample delayed enhanced sequence; the sample plain scan sequence includes a plurality of sample plain scan images, and the sample delayed enhanced sequence includes a plurality of sample delayed enhanced images.

[0129] In this step, multiple sample plain scan images can be obtained by scanning the object to be tested multiple times and reconstructing the image during the plain scan period of scanning each object to be tested, thereby forming a sample plain scan sequence; similarly, multiple sample delayed enhancement images can be obtained by scanning the object to be tested multiple times and reconstructing the image within a period of time after the enhancement period of scanning each object to be tested, thereby forming a sample delayed enhancement sequence.

[0130] It should be noted that the size, resolution and other parameters of each sample plain scan image and each sample delayed enhancement image are consistent.

[0131] The aforementioned sample plain scan sequence and delayed contrast-enhanced sample sequence can be a cerebral perfusion CTP (CT Perfusion) image sequence or a myocardial perfusion CTP image sequence, i.e., the site to be measured can be the brain or chest, but can also be an image sequence of other sites. In other words, both the aforementioned sample plain scan sequence and delayed contrast-enhanced sample sequence include a first region of interest and a second region of interest of the sample site.

[0132] S604 , determining a gold standard single plain scan image and a gold standard single delayed enhanced image according to the sample plain scan sequence and the sample delayed enhanced sequence.

[0133] In this step, when obtaining the gold standard image, as an optional embodiment, multiple sample plain scan images in the sample plain scan sequence can be aligned and averaged to determine the gold standard single plain scan image corresponding to the sample plain scan sequence; multiple sample delayed enhanced images in the sample delayed enhancement sequence can be aligned and averaged to determine the gold standard single delayed enhanced image corresponding to the sample delayed enhancement sequence.

[0134] Specifically, after obtaining a sample plain scan sequence comprising multiple sample plain scan images, the multiple sample plain scan images can be registered. After registration, a weighted average of the pixels at the same location on the multiple sample plain scan images can be performed to obtain a pixel value corresponding to each location. The pixel values ​​at each location are then applied to a blank image of the same size as the sample plain scan image to obtain a gold standard single plain scan image. Similarly, the multiple sample delayed-enhanced images can be processed in the same manner as the gold standard single plain scan image to obtain a gold standard single delayed-enhanced image.

[0135] S606 , adding noise to the gold standard single plain scan image and the gold standard single delayed enhanced image respectively, to determine a single noisy plain scan image and a single noisy delayed enhanced image.

[0136] In this step, since the noise of each of the above-mentioned sample plain scan images and each sample delayed enhanced image is uncertain, noise is added to the gold standard single plain scan image and the gold standard single delayed enhanced image respectively to infer the corresponding noise image, and then participate in the subsequent image denoising network training process.

[0137] When adding noise to the gold standard image, taking the gold standard single delayed enhanced image as an example, as an optional embodiment, it can be that in the process of adding noise to the gold standard single delayed enhanced image, the signal-to-noise ratio of the gold standard single delayed enhanced image after the noise is added is counted; when the signal-to-noise ratio reaches the signal-to-noise ratio threshold, the single noise delayed enhanced image is determined.

[0138] When adding noise, a noise model can be used to add noise to the image. The noise of the noise model can include Gaussian noise, white noise, random noise, etc. The specific addition can be made by referring to the following formula:

[0139]

[0140] in, is a hyperparameter of the noise model, usually using linear interpolation from 0.001 to 0.02 (T = 100); X t It represents the image after adding t times of noise to the low-noise image X0; q(X t |Xt-1 ) indicates that it is based on X t-1 As a condition, X t where N(·) represents the true conditional probability distribution of ; N(·) represents the multivariate normal distribution. When the formula is expressed as N(μ,ε), μ is the mean and ε is the variance; when the formula is expressed as N(X,μ,ε), X is the corresponding random variable; I represents the random variable of the standard multivariate normal distribution.

[0141] Taking the gold standard single delayed enhanced image as an example, during the noise addition process, the signal-to-noise ratio between the gold standard single delayed enhanced image after noise addition and the gold standard single delayed enhanced image can be calculated, which is recorded as SNR. The following formula can be used for calculation:

[0142]

[0143] Where X represents an image, the subscript represents the time position (0, t) of the image in the forward process, and the superscript represents the sequence number of the pixel in the image, that is, traversing all pixels.

[0144] Of course, while calculating the SNR, you can also determine whether the SNR exceeds the SNR threshold under the current hyperparameters. If so, the image exceeding the SNR threshold is used as a single noise-delayed enhanced image. Depending on the duration that the SNR exceeds the SNR threshold under the current hyperparameters, you can obtain multiple single noise-delayed enhanced images with different durations.

[0145] Similarly, according to the above method, noise can be added to the gold standard single plain scan image to obtain multiple single noisy plain scan images.

[0146] It should be noted that the gold standard single plain scan image and the gold standard single delayed enhanced image are essentially low-noise and high-quality images, while the single noisy plain scan image and the single noisy delayed enhanced image after adding noise are essentially high-noise and low-quality images. For example, see Figure 7 The schematic diagram shown in the figure shows that the left image is a low-noise image and the right image is a high-noise image. It is obvious that the tissue in the low-noise image is clearer.

[0147] S608 , training an initial image denoising network based on the single noisy plain scan image, the single noisy delayed enhanced image, the gold standard single plain scan image, and the gold standard single delayed enhanced image to determine an image denoising network.

[0148] In this step, after obtaining the gold standard sample image and the noisy sample image as described above, in order to jointly train the image denoising network, samples and gold standards related to parameter calculation may also be added for training. For this process, as an optional embodiment, the sample image value change parameters corresponding to the second region of interest of the sample and the gold standard parameter map corresponding to the first region of interest of the sample may be determined based on the gold standard single plain scan image and the gold standard single delayed enhanced image; and the initial image denoising network is trained based on the sample image value change parameters, the gold standard parameter map, the single noisy plain scan image, the single noisy delayed enhanced image, the gold standard single plain scan image, and the gold standard single delayed enhanced image.

[0149] In this case, the image value change parameter corresponding to the second region of interest between the two gold standard sample images can be calculated according to the above-mentioned method for calculating the target parameters. The difference between each pixel in the first region of interest between the two gold standard images can also be calculated. The gold standard parameter map can then be obtained by calculating the image value change parameter and the difference. Subsequently, the image value change parameter, the gold standard sample image, and the noisy image can be input into the initial image denoising network for image denoising and parameter calculation, thereby obtaining a predicted single sample plain image, a predicted single delayed enhanced image, and a predicted parameter map. Subsequently, the loss between the predicted single sample plain image and the corresponding gold standard single sample plain image, the loss between the predicted single sample delayed enhanced image and the corresponding gold standard single sample delayed enhanced image, and the loss between the predicted parameter map and the gold standard parameter map can be calculated. The initial image denoising network is trained by combining the total loss of these three losses. When the total loss reaches a threshold or stabilizes, the network parameters are fixed to obtain a trained image denoising network. The aforementioned losses can be achieved by calculating the mean, variance, or mean square error.

[0150] In this embodiment, gold standard sample plain scan images and gold standard sample delayed enhanced images are obtained through a sample plain scan sequence and a sample delayed enhanced sequence. After adding noise to the gold standard sample images to obtain noisy sample images, the gold standard sample images and noisy sample images are used to train an image denoising network. By adding noise to the gold standard sample images to obtain noisy sample images for network training, the noise can be controlled, thereby making the network training process controllable and enabling more accurate and rapid network training. Furthermore, the gold standard sample images can be obtained by registering and averaging multiple sample images in the sample sequence. This reduces noise and improves image quality, thereby improving the accuracy of the gold standard sample images. Furthermore, when adding noise to the gold standard sample images, the signal-to-noise ratio is controlled, making the noise addition process more efficient and improving the efficiency of obtaining the noisy sample images. Furthermore, the network can be trained in conjunction with a gold standard parameter map. This combination of multiple parameters improves the accuracy of the trained network, resulting in more accurate results from noise reduction and parameter calculation using the network.

[0151] The following embodiment illustrates the image registration process involved in the image denoising network training process. In another embodiment, another image denoising method is provided. Based on the above embodiment, Figure 8 As shown, before the above S604, the above method may further include the following steps:

[0152] S702 , performing registration processing on a plurality of sample flat scan images in a sample flat scan sequence to obtain a sample flat scan sequence after initial registration.

[0153] After obtaining the sample flat scan sequence including multiple sample flat scan images as mentioned above, the multiple sample flat scan images can be aligned first. The specific alignment method can be to select a sample flat scan image with the optimal phase from the multiple sample flat scan images as the reference phase image, and align the other images to the reference image to obtain the sample flat scan sequence after initial alignment.

[0154] When selecting the sample flat scan image of the optimal phase, the average image of the multiple sample flat scan images in the sample flat scan sequence can be averaged with equal weights first, and then the normalized mutual information (NMI) between each sample flat scan image and the average image can be calculated to obtain the normalized mutual information corresponding to each sample flat scan image; then the sample flat scan image with the highest value of the normalized mutual information can be selected as the sample flat scan image of the optimal phase.

[0155] S704 , performing registration processing on a plurality of sample delay enhancement images in the sample delay enhancement sequence to obtain a sample delay enhancement sequence after initial registration.

[0156] The process of registering multiple sample delayed enhanced images in this step can also refer to the registration process of multiple sample plain scan images in the above S702, which will not be repeated here. In short, the sample delayed enhanced sequence after initial registration can be obtained in the end.

[0157] S706 , registering the initially registered sample plain scan sequence to the spatial coordinate system where the initially registered sample delayed enhancement sequence is located, to obtain a registered sample plain scan sequence.

[0158] In this step, after obtaining the sample plain scan sequence after initial registration and the sample delayed enhancement sequence after initial registration, the sample plain scan sequence after initial registration can be rigidly registered first, so that the respective test parts of the sample plain scan sequence after initial registration and the sample delayed enhancement sequence after initial registration basically overlap in space; thereafter, the sample plain scan sequence after initial registration can continue to be non-rigidly registered with multiple degrees of freedom, so that the spatial positions of various regions of interest in the test part basically overlap; finally, the sample plain scan sequence after initial registration is registered to the spatial coordinate system where the sample delayed enhancement sequence after initial registration is located, and the registered sample plain scan sequence is obtained.

[0159] In this embodiment, the sample plain scan sequence and the sample delayed enhancement sequence are first subjected to internal image registration, and then image registration is performed between the two sequences. Through this two-layer registration process, the final registration result of the two sequences can be made more accurate, thereby making the results obtained by denoising and parameter calculation of the registered sequences more accurate.

[0160] A detailed embodiment is given below to illustrate the technical solution of the present application. The method may include the following steps:

[0161] S1, acquiring a sample plain scan sequence and a sample delayed enhanced sequence; the sample plain scan sequence includes a plurality of sample plain scan images, and the sample delayed enhanced sequence includes a plurality of sample delayed enhanced images;

[0162] S2, performing registration processing on multiple sample plain scan images in the sample plain scan sequence to obtain a sample plain scan sequence after initial registration;

[0163] S3, performing registration processing on multiple sample delay enhancement images in the sample delay enhancement sequence to obtain a sample delay enhancement sequence after initial registration;

[0164] S4, registering the sample plain scan sequence after the initial registration to the spatial coordinate system where the sample delayed enhancement sequence after the initial registration is located, to obtain a registered sample plain scan sequence;

[0165] S5, performing mean processing on multiple sample plain scan images in the sample plain scan sequence to determine a gold standard single plain scan image corresponding to the sample plain scan sequence;

[0166] S6, performing mean processing on multiple sample delayed enhancement images in the sample delayed enhancement sequence to determine a gold standard single delayed enhancement image corresponding to the sample delayed enhancement sequence;

[0167] S7, in the process of adding noise to the gold standard single delayed enhanced image, calculating the signal-to-noise ratio of the gold standard single delayed enhanced image after the noise is added, and determining the single noise delayed enhanced image when the signal-to-noise ratio reaches a signal-to-noise ratio threshold;

[0168] S8, in the process of adding noise to the gold standard single plain scan image, calculating the signal-to-noise ratio of the gold standard single plain scan image after the noise is added, and determining the single noisy plain scan image when the signal-to-noise ratio reaches a signal-to-noise ratio threshold;

[0169] S9, determining a sample image value change parameter corresponding to the second region of interest of the sample and a gold standard parameter map corresponding to the first region of interest of the sample based on the gold standard single plain scan image and the gold standard single delayed enhancement image;

[0170] S10, training the initial image denoising network according to the sample image value variation parameter, the gold standard parameter map, the single noisy plain scan image, the single noisy delayed enhanced image, the gold standard single plain scan image, and the gold standard single delayed enhanced image to obtain a trained image denoising network;

[0171] S11, acquiring a single plain scan image and a single delayed enhanced image;

[0172] S12, registering the single plain scan image to the spatial coordinate system of the single delayed enhancement image to obtain a single registered plain scan image;

[0173] S13, determining a first segmentation result of the first region of interest and a second segmentation result of the second region of interest based on the single plain scan image and the single delayed enhanced image, wherein both the single plain scan image and the single delayed enhanced image include the first region of interest and the second region of interest;

[0174] S14, calculating a first mean parameter corresponding to the second region of interest in the second segmentation result of the single plain scan image;

[0175] S15, calculating a second mean parameter corresponding to a second region of interest in a second segmentation result of the single delayed enhancement image;

[0176] S16, determining an image value change parameter corresponding to the second region of interest based on the first mean parameter and the second mean parameter;

[0177] S17, inputting the single registered plain scan image and the single delayed enhanced image into the denoising subnetwork of the image denoising network for denoising, and determining a first denoised image corresponding to the single delayed enhanced image and a second denoised image corresponding to the single registered plain scan image;

[0178] S18, in the parameter calculation subnetwork, calculating an image difference between each pixel point in the first region of interest of the first denoised image and a corresponding pixel point in the first region of interest of the second denoised image to obtain a first difference corresponding to each pixel point in the first region of interest;

[0179] S19, determining target parameters corresponding to each pixel point in the first region of interest according to each first difference and the image value change parameter, and obtaining a target parameter map.

[0180] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of 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 these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0181] Based on the same inventive concept, embodiments of the present application also provide an image noise reduction device for implementing the aforementioned image noise reduction method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more of the following embodiments of the image noise reduction device can be found in the above-described limitations of the image noise reduction method and are not further elaborated here.

[0182] In one embodiment, Figure 9 As shown, an image noise reduction device is provided, comprising: an image acquisition module and a noise reduction module, wherein:

[0183] An image acquisition module, used for acquiring a single plain scan image and a single delayed enhanced image;

[0184] a noise reduction module for performing noise reduction processing on the single delayed enhanced image based on the single plain scan image, the single delayed enhanced image, and a preset image noise reduction network, and determining noise reduction images corresponding to the single delayed enhanced image and the single plain scan image;

[0185] Among them, the above-mentioned image denoising network is trained based on a single noisy plain scan image and its corresponding gold standard single plain scan image, a single noisy delayed enhanced image and its corresponding gold standard single delayed enhanced image; the above-mentioned gold standard single plain scan image and gold standard single delayed enhanced image are both denoised images.

[0186] In another embodiment, another image noise reduction device is provided. Based on the above embodiment, the noise reduction module may include:

[0187] A registration unit, configured to register a single plain scan image to a spatial coordinate system where a single delayed enhancement image is located, to obtain a single registered plain scan image;

[0188] The denoising unit is used to input a single registered plain scan image and a single delayed enhanced image into an image denoising network for denoising, and determine a denoised image.

[0189] In another embodiment, another image noise reduction device is provided. Based on the above embodiment, the device may further include:

[0190] a segmentation module, configured to determine a first segmentation result of the first region of interest and a second segmentation result of the second region of interest based on the single plain scan image and the single delayed enhanced image, wherein both the single plain scan image and the single delayed enhanced image include the first region of interest and the second region of interest;

[0191] The parameter determination module is used to determine the image value change parameter corresponding to the second region of interest according to the second segmentation result.

[0192] Optionally, the parameter determination module may include:

[0193] A first calculation unit is used to calculate a first mean parameter corresponding to a second region of interest in a second segmentation result of a single plain scan image;

[0194] A second calculation unit is used to calculate a second mean parameter corresponding to a second region of interest in a second segmentation result of a single delayed enhancement image;

[0195] The parameter determination unit is configured to determine an image value variation parameter corresponding to the second region of interest based on the first mean parameter and the second mean parameter.

[0196] In another embodiment, another image denoising device is provided. Based on the above embodiment, the image denoising network includes a denoising subnetwork and a parameter calculation subnetwork; the denoising unit may include:

[0197] a denoising subunit, configured to input the single registered plain scan image and the single delayed enhanced image into the denoising subnetwork for denoising, and determine a first denoised image corresponding to the single delayed enhanced image and a second denoised image corresponding to the single registered plain scan image;

[0198] a parameter map determination subunit, configured to determine a target parameter map corresponding to the first region of interest based on the first denoised image, the second denoised image, the image value variation parameter, and the parameter calculation subnetwork; the target parameter map including target parameters for each pixel in the first region of interest;

[0199] The denoised image determining unit is configured to determine the first denoised image, the second denoised image, and the target parameter map as the denoised image.

[0200] Optionally, the above-mentioned parameter map determination subunit is specifically used to calculate the image difference between each pixel point in the first region of interest of the first denoised image and the corresponding pixel point in the first region of interest of the second denoised image in the parameter calculation subnetwork, and obtain the first difference corresponding to each pixel point in the first region of interest; determine the target parameter corresponding to each pixel point in the first region of interest based on each first difference and the image value change parameter, and obtain the target parameter map.

[0201] In another embodiment, another image denoising device is provided. Based on the above embodiment, the device may further include a training module. The training module may include:

[0202] A sequence acquisition unit, configured to acquire a sample plain scan sequence and a sample delayed enhanced sequence; the sample plain scan sequence includes a plurality of sample plain scan images, and the sample delayed enhanced sequence includes a plurality of sample delayed enhanced images;

[0203] A gold standard determination unit, configured to determine a gold standard single plain scan image and a gold standard single delayed enhanced image according to a sample plain scan sequence and a sample delayed enhanced sequence;

[0204] a noise adding unit, used for adding noise to the gold standard single plain scan image and the gold standard single delayed enhanced image, respectively, to determine a single noisy plain scan image and a single noise delayed enhanced image;

[0205] The training unit is used to train the initial image denoising network according to the single noisy plain scan image, the single noisy delayed enhanced image, the gold standard single plain scan image and the gold standard single delayed enhanced image to determine the image denoising network.

[0206] Optionally, the above training unit may include:

[0207] a standard parameter map determining subunit, configured to determine the sample image value variation parameter corresponding to the second region of interest of the sample and the gold standard parameter map corresponding to the first region of interest of the sample based on the gold standard single plain scan image and the gold standard single delayed enhanced image;

[0208] The training subunit is used to train the initial image denoising network according to the sample image value change parameter, the gold standard parameter map, the single noisy plain scan image, the single noisy delayed enhanced image, the gold standard single plain scan image and the gold standard single delayed enhanced image.

[0209] Optionally, the noise adding unit is specifically used to calculate the signal-to-noise ratio of the gold standard single delayed enhanced image after adding noise to the gold standard single delayed enhanced image; when the signal-to-noise ratio reaches a signal-to-noise ratio threshold, determine the single noise delayed enhanced image.

[0210] Optionally, the above-mentioned gold standard determination unit is specifically used to align and average multiple sample plain scan images in a sample plain scan sequence to determine the gold standard single plain scan image corresponding to the sample plain scan sequence; and to align and average multiple sample delayed enhanced images in a sample delayed enhancement sequence to determine the gold standard single delayed enhanced image corresponding to the sample delayed enhancement sequence.

[0211] In another embodiment, another image noise reduction device is provided. Based on the above embodiment, the device may further include:

[0212] A flat scan registration module is used to perform registration processing on multiple sample flat scan images in a sample flat scan sequence to obtain a sample flat scan sequence after initial registration;

[0213] An enhanced registration module is used to perform registration processing on multiple sample delayed enhancement images in a sample delayed enhancement sequence to obtain a sample delayed enhancement sequence after initial registration;

[0214] The sequence registration module is used to register the sample plain scan sequence after the initial registration to the spatial coordinate system where the sample delayed enhancement sequence after the initial registration is located, so as to obtain a registered sample plain scan sequence.

[0215] Each module in the above-mentioned image noise reduction device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in hardware form, or can be stored in a memory in the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0216] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0217] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0218] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0219] It should be noted that the image data involved in this application (including but not limited to data used for analysis, storage, display, etc.) are all image data that have been fully authorized by all parties.

[0220] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0221] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0222] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. An image denoising method, characterized in that: The method comprises: Acquire a single plain scan image and a single delayed enhanced image; performing noise reduction processing on the single delayed-enhanced image according to the single plain scan image, the single delayed-enhanced image, and a preset image noise reduction network to determine noise reduction images corresponding to the single delayed-enhanced image and the single plain scan image; The image denoising network is trained based on a single noisy plain scan image and its corresponding gold standard single plain scan image, a single noisy delayed enhanced image and its corresponding gold standard single delayed enhanced image; the gold standard single plain scan image and the gold standard single delayed enhanced image are both denoised images.

2. The method according to claim 1, characterized in that The performing noise reduction processing on the single delayed enhanced image according to the single plain scan image, the single delayed enhanced image, and a preset image noise reduction network to determine noise reduction images corresponding to the single delayed enhanced image and the single plain scan image includes: Registering the single plain scan image to the spatial coordinate system of the single delayed enhancement image to obtain a single registered plain scan image; The single registered plain scan image and the single delayed enhanced image are input into the image denoising network for denoising to determine the denoised image.

3. The method according to claim 2, characterized in that Before inputting the single registered plain scan image and the single delayed enhanced image into the image denoising network for denoising, and determining the denoised image, the method further comprises: determining a first segmentation result of a first region of interest and a second segmentation result of a second region of interest based on the single plain scan image and the single delayed enhanced image; wherein both the single plain scan image and the single delayed enhanced image include the first region of interest and the second region of interest; An image value change parameter corresponding to the second region of interest is determined according to the second segmentation result.

4. The method according to claim 3, characterized in that The image denoising network includes a denoising subnetwork and a parameter calculation subnetwork; inputting the single registered plain scan image and the single delayed enhanced image into the image denoising network for denoising, and determining the denoised image, includes: Inputting the single registered plain scan image and the single delayed enhanced image into the denoising subnetwork for denoising, and determining a first denoised image corresponding to the single delayed enhanced image and a second denoised image corresponding to the single registered plain scan image; Determining a target parameter map corresponding to the first region of interest based on the first denoised image, the second denoised image, the image value change parameter, and the parameter calculation subnetwork; the target parameter map including target parameters of each pixel point in the first region of interest; The first denoised image, the second denoised image, and the target parameter map are determined as the denoised image.

5. The method according to claim 4, characterized in that The determining, according to the first denoised image, the second denoised image, the image value change parameter, and the parameter calculation subnetwork, a target parameter map corresponding to the first region of interest includes: In the parameter calculation subnetwork, an image difference between each pixel point in the first region of interest of the first denoised image and a corresponding pixel point in the first region of interest of the second denoised image is calculated to obtain a first difference corresponding to each pixel point in the first region of interest; According to each of the first differences and the image value change parameter, a target parameter corresponding to each pixel point in the first region of interest is determined to obtain the target parameter map.

6. The method according to any one of claims 1 to 5, characterized in that The training method of the image denoising network includes: Acquire a sample plain scan sequence and a sample delayed enhancement sequence; the sample plain scan sequence includes a plurality of sample plain scan images, and the sample delayed enhancement sequence includes a plurality of sample delayed enhancement images; Determining a gold standard single plain scan image and a gold standard single delayed enhanced image according to the sample plain scan sequence and the sample delayed enhanced sequence; adding noise to the gold standard single plain scan image and the gold standard single delayed enhanced image respectively to determine a single noisy plain scan image and a single noise delayed enhanced image; An initial image denoising network is trained according to the single noisy plain scan image, the single noisy delayed enhanced image, the gold standard single plain scan image and the gold standard single delayed enhanced image to determine the image denoising network.

7. An image noise reduction device, characterized in that: The device comprises: An image acquisition module, used for acquiring a single plain scan image and a single delayed enhanced image; a noise reduction module, configured to perform noise reduction processing on the single delayed-enhanced image and the single plain image according to the single plain image, the single delayed-enhanced image, and a preset image noise reduction network, and determine a first noise-reduced image corresponding to the single delayed-enhanced image and a second noise-reduced image corresponding to the single plain image; The image denoising network is trained based on a single noisy plain scan image and its corresponding gold standard single plain scan image, a single noisy delayed enhanced image and its corresponding gold standard single delayed enhanced image; the gold standard single plain scan image and the gold standard single delayed enhanced image are both denoised images.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images