Method for correcting motion artifacts in images, electronic device and storage medium

Abstract

This invention discloses a method, electronic device, and storage medium for correcting motion artifacts in images. The correction method includes: acquiring an image of the current patient to be corrected; acquiring offset parameters corresponding to motion artifacts generated by the current patient's movement in the target examination scene based on the image to be corrected; and performing artifact correction processing on the image to be corrected using a matching preset artifact correction network according to the target examination scene and the offset parameters to obtain the target image of the current patient. In this invention, for any image to be corrected containing motion artifacts, a matching preset artifact correction network is promptly identified to perform artifact removal correction on the current image to be corrected, thereby optimizing the image containing motion artifacts, eliminating the motion artifacts, and ultimately displaying an image free of motion artifacts. This effectively improves image quality and facilitates the accuracy of subsequent assessments of the patient's condition by doctors and other personnel.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method, electronic device, and storage medium for correcting motion artifacts in images. Background Technology

[0002] During an MRI scan, the patient needs to remain still on the examination bed to cooperate with the entire procedure. If the patient moves during the scan, motion artifacts will appear in the reconstructed images. Sometimes, due to the long examination time, the patient inevitably moves, causing motion artifacts. In addition, it is difficult for the elderly and children to remain still on the bed for a certain period of time, making them more susceptible to motion artifacts during the examination.

[0003] Currently, although some methods for processing image artifacts have emerged, they all suffer from problems such as complex processing steps, low processing efficiency, and insufficient processing quality, which do not meet the actual image processing needs. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of existing image artifact removal schemes that are not accurate enough and cannot meet the needs of actual scenarios, and to provide a method, electronic device and storage medium for correcting motion artifacts in images.

[0005] The present invention solves the above-mentioned technical problems through the following technical solution:

[0006] This invention provides a method for correcting motion artifacts in an image, the method comprising:

[0007] Acquire the image to be corrected for the current patient;

[0008] Based on the image to be corrected, obtain the offset parameters corresponding to the motion artifacts generated by the current patient's movement in the target examination scene;

[0009] Based on the target examination scenario and the offset parameters, a matching preset artifact correction network is used to perform artifact correction processing on the image to be corrected, thereby obtaining the target image of the current patient.

[0010] Preferably, before the step of acquiring the image to be corrected for the current patient, the method further includes:

[0011] Acquire first images of different patients at different preset deflection angles under different types of examination scenarios; wherein each first image identifies the examination scenario in which it is located;

[0012] Obtain the first image parameters corresponding to the first image;

[0013] The first pixel offset of the image pixel corresponding to the motion artifact in the first image is calculated based on the first image parameters.

[0014] The preset neural network is trained based on the first pixel offset to construct the corresponding preset artifact correction network.

[0015] Preferably, the preset neural network includes a Gaussian low-pass filtering algorithm;

[0016] And / or,

[0017] The preset neural network includes a convolutional neural network, which corresponds to a number of convolutional kernels that are randomly configured or pre-fixed.

[0018] Preferably, the first image parameters include image size and / or image resolution in a set direction;

[0019] The step of calculating the first pixel offset of the image pixel corresponding to the motion artifact in the first image based on the image parameters includes:

[0020] The first displacement amount generated in the set direction is collected when the patient moves in the corresponding examination scenario;

[0021] Based on the first displacement and the image parameters, the first pixel offset corresponding to the image pixel in the first image in the set direction is calculated;

[0022] And / or,

[0023] The step of performing artifact correction processing on the image to be corrected using a matching preset artifact correction network based on the target inspection scene and the offset parameters includes:

[0024] Obtain the actual amount of movement of the current patient in a set direction when moving in the target examination scenario;

[0025] Obtain the target image parameters of the image to be corrected;

[0026] The target image parameters include the image size and / or image resolution in the set direction;

[0027] Based on the actual movement, the target image parameters, and the target inspection scene, a corresponding preset artifact correction network is obtained, and the preset artifact correction network is used to perform artifact correction processing on the image to be corrected.

[0028] Preferably, the correction method further includes:

[0029] An image database is constructed based on several of the first images;

[0030] The steps for determining the target inspection scene corresponding to the image to be corrected include:

[0031] From a plurality of the first images in the image database, a second image is selected whose matching degree with the image to be corrected is greater than a set threshold;

[0032] The inspection scene of the second image is taken as the target inspection scene corresponding to the image to be corrected.

[0033] Preferably, the examination scenario information corresponding to each of the examination scenarios includes at least one of the following: patient basic information, patient posture information, examination system information, examination protocol information, and examination operation mode information.

[0034] Preferably, before the step of acquiring the image to be corrected for the current patient, the method further includes:

[0035] Obtain the current patient's status change information;

[0036] If the state change information indicates that the current patient has moved, then the step of obtaining the image to be corrected for the current patient is performed.

[0037] Preferably, the state change information includes at least one of the following: positional offset information, pressure change information, vibration change information, and light signal change information generated when the patient's body moves.

[0038] The present invention also provides a system for correcting motion artifacts in an image, the system comprising:

[0039] The image to be corrected acquisition module is used to acquire the image to be corrected for the current patient;

[0040] The offset parameter acquisition module is used to acquire the offset parameters corresponding to the motion artifacts generated by the movement of the current patient in the target examination scene based on the image to be corrected.

[0041] The artifact correction module is used to perform artifact correction processing on the image to be corrected using a matching preset artifact correction network based on the target examination scene and the offset parameters, so as to obtain the target image of the current patient.

[0042] Preferably, the correction system further includes:

[0043] The first image acquisition module is used to acquire first images of different patients at different preset deflection angles under different types of examination scenarios; wherein each first image identifies the examination scenario in which it is located;

[0044] The first image parameter acquisition module is used to acquire the first image parameters corresponding to the first image.

[0045] The first pixel offset calculation module is used to calculate the first pixel offset of the image pixel corresponding to the motion artifact in the first image based on the first image parameters.

[0046] The correction network construction module is used to train a preset neural network based on the first pixel offset to construct the corresponding preset artifact correction network.

[0047] Preferably, the preset neural network includes a Gaussian low-pass filtering algorithm;

[0048] And / or,

[0049] The preset neural network includes a convolutional neural network, which corresponds to a number of convolutional kernels that are randomly configured or pre-fixed.

[0050] Preferably, the first image parameters include image size and / or image resolution in a set direction;

[0051] The first pixel offset calculation module includes:

[0052] The first displacement amount generated in the set direction is collected when the patient moves in the corresponding examination scenario;

[0053] Based on the first displacement and the image parameters, the first pixel offset corresponding to the image pixel in the first image in the set direction is calculated;

[0054] And / or,

[0055] The artifact correction module includes:

[0056] The actual movement acquisition unit is used to acquire the actual movement of the current patient in a set direction when the patient moves in the target examination scenario.

[0057] A target image parameter acquisition unit is used to acquire the target image parameters of the image to be corrected.

[0058] The target image parameters include the image size and / or image resolution in the set direction;

[0059] The artifact correction unit is used to match the corresponding preset artifact correction network based on the actual movement amount, the target image parameters and the target inspection scene, and to use the preset artifact correction network to perform artifact correction processing on the image to be corrected.

[0060] Preferably, the correction system further includes:

[0061] An image database construction module is used to construct an image database based on several of the first images;

[0062] An image filtering module is used to filter out a second image from a plurality of first images in the image database that has a matching degree greater than a set threshold with respect to the image to be corrected;

[0063] The target inspection scene determination module is used to determine the inspection scene of the second image as the target inspection scene corresponding to the image to be corrected.

[0064] Preferably, the examination scenario information corresponding to each of the examination scenarios includes at least one of the following: patient basic information, patient posture information, examination system information, examination protocol information, and examination operation mode information.

[0065] Preferably, the correction system includes:

[0066] The status change information acquisition module is used to acquire the current patient's status change information;

[0067] The judgment module is used to call the image acquisition module to be corrected if the state change information indicates that the current patient has moved.

[0068] Preferably, the state change information includes at least one of the following: positional offset information, pressure change information, vibration change information, and light signal change information generated when the patient's body moves.

[0069] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method for correcting motion artifacts in an image.

[0070] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-described method for correcting motion artifacts in an image.

[0071] Based on common knowledge in the field, the preferred conditions described can be combined arbitrarily to obtain various preferred embodiments of the present invention.

[0072] The positive and progressive effects of this invention are as follows:

[0073] In this invention, for any image to be corrected, which contains motion artifacts caused by the movement of the current patient in the target examination scene, the image pixel offset corresponding to the motion artifact in the target examination scene and the actual movement of the detected patient are calculated. A matching preset artifact correction network is used to perform artifact correction processing on the image to be corrected, resulting in a corrected target image of the current patient. This image is then transmitted to the scanning device for image display, thereby optimizing the image containing motion artifacts, eliminating the motion artifacts, and ultimately displaying an image without motion artifacts. This effectively improves image quality and facilitates the accuracy of subsequent determination of the patient's condition by doctors and other personnel. Attached Figure Description

[0074] Figure 1 This is a flowchart of the method for correcting motion artifacts in images according to Embodiment 1 of the present invention.

[0075] Figure 2 This is a first flowchart of the method for correcting motion artifacts in images according to Embodiment 2 of the present invention.

[0076] Figure 3 This is a second flowchart of the method for correcting motion artifacts in images according to Embodiment 2 of the present invention.

[0077] Figure 4 This is the third flowchart of the method for correcting motion artifacts in images according to Embodiment 2 of the present invention.

[0078] Figure 5 This is a schematic diagram of the module of the motion artifact correction system in the image of Embodiment 3 of the present invention.

[0079] Figure 6 This is a schematic diagram of the module of the motion artifact correction system in the image of Embodiment 4 of the present invention.

[0080] Figure 7 This is a schematic diagram of the electronic device according to Embodiment 5 of the present invention. Detailed Implementation

[0081] The present invention will be further illustrated by way of embodiments below, but the present invention is not limited to the scope of the embodiments described herein.

[0082] Example 1

[0083] like Figure 1 As shown, the method for correcting motion artifacts in images in this embodiment includes:

[0084] S101. Obtain the image to be corrected for the current patient;

[0085] The images to be corrected include, but are not limited to, magnetic resonance (MR) images, computed tomography (CT) images, positron emission tomography (PET) images, and ultrasound images. During the examination, voluntary or involuntary movements of the subject can cause motion artifacts in medical images. Involuntary movements (physiological movements) produce physiological motion artifacts, primarily caused by the subject's breathing, blood flow, and cardiac contraction. Voluntary movements mainly include coughing, swallowing, and other body movements. For example, in CT scans, the movement of the subject can disrupt the consistency and integrity of the projection data. Similarly, in MR scans, the pulsation of blood vessels, the heart, or the flow of cerebrospinal fluid can produce periodic motion artifacts in the phase encoding direction; voluntary patient movements can produce parallel strip-like artifacts in the phase encoding direction.

[0086] S102. Based on the image to be corrected, obtain the offset parameters corresponding to the motion artifacts generated by the current patient's movement in the target examination scene.

[0087] S103. Based on the target examination scene and offset parameters, a matching preset artifact correction network is used to perform artifact correction processing on the image to be corrected, thereby obtaining the target image of the current patient.

[0088] Among them, the target image contains relatively little artifact information or the artifact information is suppressed, or even does not contain any artifact information, compared to the image to be corrected.

[0089] In this embodiment, for any image to be corrected, which contains motion artifacts caused by the movement of the current patient in the target examination scene, the offset parameters corresponding to the motion artifacts in the target examination scene are calculated, and a matching preset artifact correction network is used to perform artifact correction processing on the image to be corrected, so as to obtain the corrected target image of the current patient, which is then transmitted to the scanning device for image display. This achieves the effect of optimizing the image containing motion artifacts, eliminating the motion artifacts, and finally displaying an image without motion artifacts, effectively improving the image quality and facilitating the accuracy of subsequent determination of the patient's condition by doctors and other personnel.

[0090] Example 2

[0091] The method for correcting motion artifacts in the image in this embodiment is a further improvement on Embodiment 1, specifically:

[0092] In a feasible solution, such as Figure 2As shown, the procedure before step S101 also includes:

[0093] S1001. Acquire first images of different patients at different preset deflection angles under different types of examination scenarios; wherein each first image is identified by the examination scenario in which it is located;

[0094] S1002. Obtain the first image parameters corresponding to the first image;

[0095] S1003. Calculate the first pixel offset of the image pixel corresponding to the motion artifact in the first image based on the first image parameters;

[0096] S1004. Train the preset neural network based on the first pixel offset to construct the corresponding preset artifact correction network.

[0097] Among them, a preset artifact correction network corresponding to each different inspection scenario is trained, so that it can be directly matched and called in the actual inspection scenario without the need for real-time model training. This saves the total time of real-time model training and artifact correction, and effectively improves the overall efficiency and rationality of image correction processing.

[0098] Among them, the same preset artifact correction network corresponds to one inspection scene;

[0099] Specifically, the preset neural network includes processing the image using a Gaussian low-pass filtering algorithm;

[0100] The process of training a neural network using Gaussian low-pass filtering is as follows:

[0101] The basic principle of the Gaussian low-pass filtering algorithm is as follows: First, a Gaussian kernel is obtained. Based on a Gaussian distribution, the weights of pixels surrounding the center pixel are obtained and normalized. Then, a Gaussian filter is calculated, and the weighted average of the pixel's neighborhood is used to replace the pixel value. The weight of each neighboring pixel monotonically increases with its distance from the center pixel. Based on the Gaussian kernel, a weighted average is calculated on the current image, i.e., convolution is performed on the center pixel. The resulting image is the Gaussian blurred image, which corrects artifacts by smoothing the details of the image. In other words, high-frequency components in the non-smooth parts of the image are filtered out, while low-frequency components in the smooth parts are retained. After Gaussian blurring, Gaussian normal distribution noise is suppressed, and the image becomes blurred, thus correcting motion artifacts in the image.

[0102] Before image processing, an image database indexed by image pixel offsets needs to be established. In this embodiment, firstly, the pixel offset is calculated based on the image size and resolution; an image with a pixel offset of 0 is an image without motion artifacts. Then, images with motion artifacts corresponding to different pixel offsets at various deflection angles, including 0°, 45°, 90°, 135°, 180°, 225°, 270°, and 315°, are set up to construct various scenes.

[0103] Furthermore, the image pixels are determined based on the image size and image resolution. When the monitoring device detects the displacement caused by the patient's movement, the system obtains the image pixel offset corresponding to different motion offsets based on the motion offset and pixel offset, thereby establishing a motion artifact correction database (i.e., an image pixel offset database).

[0104] In one embodiment, the Gaussian low-pass filter uses a two-dimensional Gaussian filter formula, with the center point coordinates set to (0, 0):

[0105]

[0106] Where G(x,y) is a two-dimensional Gaussian function, σ is the variance, and x and y are the coordinates of the point.

[0107] In another embodiment, the two-dimensional Gaussian filter can be expanded in two independent one-dimensional spaces and calculated separately. That is, Gaussian filtering is first calculated in the x-direction, and then Gaussian filtering is calculated in the y-direction based on this.

[0108]

[0109] A neural network is trained on an established motion artifact correction database using Gaussian filtering. This allows for the generation of a corresponding preset artifact correction network for each inspection scenario. This ensures that for any image to be corrected, a suitable preset artifact correction network can be matched for correction, guaranteeing the timeliness and accuracy of the correction process. The preset neural network includes a convolutional neural network, which corresponds to several randomly or pre-fixed convolutional kernels.

[0110] Of course, other model training algorithms can be used for the preset training network, as long as they can achieve the corresponding image artifact correction function. The specific model training algorithm used can be determined or adjusted according to actual needs.

[0111] In one feasible approach, the first image parameters include the image size and / or image resolution in a given orientation.

[0112] like Figure 3 As shown, step S1003 specifically includes:

[0113] S10031, the step of calculating the first pixel offset of the image pixel corresponding to the motion artifact in the first image based on image parameters includes:

[0114] S10032. Collect the first displacement amount generated in a set direction when the patient moves in the corresponding examination scenario;

[0115] S10033. Based on the first displacement and image parameters, calculate the first pixel offset corresponding to the image pixel in the set direction in the first image;

[0116] Specifically, the calculation formula corresponding to step S10033 is as follows:

[0117]

[0118] Where Δpixel is the pixel offset in the x-direction, x is the image size in the x-direction, and R... x R x Let Δx be the image resolution in the x-direction, and Δx be the displacement corresponding to the patient's movement in the x-direction.

[0119] In addition, in order to better train the images and obtain a more accurate neural network for motion artifact removal, and to process images containing motion artifacts more accurately and efficiently, it is necessary to acquire a large number of motion images in various scanning scenarios. For example, it is necessary to acquire images of examination scenarios with related parameters such as different machine configurations, different patient genders, different patient ages, different patient weights, different patient positions, and different scanning protocols, so as to cover as many clinical scenarios as possible.

[0120] In one feasible solution, step S103 specifically includes:

[0121] S1031. Obtain the actual amount of movement of the current patient in the set direction when moving in the target examination scene;

[0122] S1032. Obtain the target image parameters of the image to be corrected;

[0123] The target image parameters include the image size and / or image resolution in the specified direction;

[0124] S1033. Based on the actual movement, target image parameters and target inspection scenario, a corresponding preset artifact correction network is obtained, and the preset artifact correction network is used to perform artifact correction processing on the image to be corrected.

[0125] In one feasible approach, the correction method also includes:

[0126] An image database is constructed based on several first images;

[0127] The steps to determine the target inspection scene corresponding to the image to be corrected include:

[0128] Select a second image from a set threshold that matches the image to be corrected better than a set threshold from a set number of first images in the image database;

[0129] In this process, the first image with the highest similarity is selected as the second image that matches the image to be corrected. If multiple images have the same similarity, one of the first images is randomly selected as the second image that matches the image to be corrected, or a reminder message is generated and pushed to relevant staff, and the corresponding first image is selected as the second image that matches the image to be corrected based on the staff's feedback. The specific method used to determine the second image that matches the image to be corrected can be selected or set according to the actual needs of the scenario, so as to more flexibly meet the control needs of different scenarios.

[0130] The inspection scene of the second image is used as the target inspection scene corresponding to the image to be corrected.

[0131] The examination scenario information for each examination scenario includes patient basic information, patient posture information, examination system information, examination protocol information, and examination operation mode information.

[0132] Specifically, DICOM (Digital Imaging and Communication in Medicine) images of patients under different examination scenarios are acquired. Reconstructed images of patients of different body types under various scanning scenarios (such as different positioning, different movement distances, etc.) are collected as much as possible to establish an image database for different examination scenarios. Raw data images and DICOM images are stored on a pre-defined server. Simultaneously, a specific identification method is used to distinguish and store images from different examination scenarios on the pre-defined server. The classification schemes corresponding to different examination scenarios are shown in Table 1 below:

[0133] Table 1

[0134]

[0135] In the table above, HFS indicates head-first, supine position; HFP indicates head-first, prone position; FFS indicates feet-first, supine position; FFP indicates feet-first, prone position; HFDL indicates head-first, left lateral decubitus position. GRE indicates gradient echo pulse sequence; EPI indicates planar echo imaging sequence; FSE indicates fast spin echo sequence.

[0136] According to the classification scheme in the table above, reconstructed images of different patients under various protocols in different examination scenarios are collected. Since the main purpose of this scheme is to correct motion artifacts, it is necessary to collect reconstructed images of patients in scenarios with no positional offset, offset of 1cm, offset of 2cm, offset of 3cm, etc., and to consider various deflection angles (e.g., 0°, 45°, 90°, 135°, 180°, 225°, 270°, 315°) to simulate as many possible movements of patients as possible during actual examinations. These images are stored on a specific server, managed by category, and a multi-scenario image database (or motion correction database) for reconstructed images is created.

[0137] Through the refined construction of the aforementioned image database, images of any patient in any movement state can be matched and queried in the constructed image database in actual examination scenarios, thereby ensuring the feasibility of subsequent artifact correction operations and the accuracy of matching correction with high quality.

[0138] Of course, other classification criteria besides those mentioned above can be used to collect reconstructed images of patients in more examination scenarios. The specific criteria can be flexibly determined or adjusted according to the actual situation.

[0139] In a feasible solution, such as Figure 4 As shown, the procedure before step S101 also includes:

[0140] S1005. Obtain information on the current patient's status changes;

[0141] S1006. If the status change information indicates that the current patient has moved, then execute step S101.

[0142] The status change information includes positional shifts, pressure changes, vibration changes, and light signal changes caused by the patient's body movement.

[0143] For example, 3D cameras can be set up in detection equipment (such as MRI machines) to capture patient images, pressure sensors can be set up to detect changes in pressure values ​​within a set time period, vibration sensors can be set up to detect changes in vibration values ​​within a set time period, and light sensors can be set up to detect changes in light signals within a set time period, so as to analyze and determine the corresponding positional movement in a timely manner.

[0144] In addition, two or more types of sensors can be set up simultaneously in the detection device, which can effectively avoid false triggering and achieve more timely and accurate trigger control. This avoids performing the above-mentioned artifact correction process in unnecessary scenarios, which would waste computing resources and cause the image generation process to be slow and fail to meet actual needs. This improves the accuracy, rationality and timeliness of image artifact correction control.

[0145] The following examples and... Figure 4 The implementation principle of the motion artifact correction method in images based on deep learning algorithms in this embodiment is explained in detail below:

[0146] (1) Constructing an image database

[0147] Based on the established classification scheme, reconstructed images of different patients under various protocols in different examination scenarios are collected. Reconstructed images of patients under different offset scenarios are acquired, and various deflection angles are considered to simulate as many possible movement patterns of patients as possible during actual examinations. These images are stored on a specific server, managed by category, and a multi-scenario image database for reconstructed image analysis is created.

[0148] (2) Is artifact correction triggered?

[0149] The system acquires information on changes in the patient's condition, including but not limited to positional shifts, pressure changes, vibration changes, and light signal changes, to determine in a timely and accurate manner whether the patient has moved during the examination. If movement occurs, the scanned image is determined to contain motion artifacts, and the artifact correction function is automatically triggered; otherwise, no triggering is required, and the image can be reconstructed directly.

[0150] (3) Preset artifact correction network

[0151] First images of different patients under different preset deflection angles in different types of examination scenarios are acquired, and each first image is labeled with its own examination scenario; the first image parameters (size, resolution, etc.) corresponding to the first image are obtained and the displacement caused by the detected patient movement is calculated to obtain the pixel displacement of the corresponding image pixels; then, a preset neural network is trained based on the pixel offset using the Gaussian low-pass filtering method to construct the corresponding preset artifact correction network.

[0152] In this approach, model training is performed at the image level, rather than using K-space data for neural network training, thus effectively ensuring the efficiency and accuracy of model training.

[0153] In addition, a preset artifact correction network corresponding to each different inspection scenario can be pre-trained before the actual inspection, so that it can be directly matched and called in the actual inspection scenario without the need for real-time model training.

[0154] Whether to use real-time model training or pre-trained model training can be selected or adjusted according to actual needs.

[0155] (4) Image artifact correction

[0156] Based on the inspection scene corresponding to the image to be corrected, the pixel offset of the image pixels, the image size, and the image resolution, a preset artifact correction network is determined to match it. The image to be corrected is input into the preset artifact correction network, which directly performs artifact correction processing on the image to be corrected. At the image level, the image containing motion artifacts is optimized, eliminating the motion artifacts and displaying an image without motion artifacts, thus obtaining a DICOM scan image after motion artifact removal. In addition, the image artifact correction in this embodiment can support online and offline correction, which can more flexibly meet the requirements of more image artifact correction scenarios.

[0157] The process involves scanning an image containing motion artifacts using a scanning device, calculating the corresponding pixel offset based on the motion offset caused by the patient's movement, image parameters such as the image size and resolution of the image to be corrected, and considering factors such as the current examination setup, patient age, gender, weight, patient positioning, and scanning sequence. Based on this information, a preset artifact correction network is retrieved from a database that best matches the current artifact image in the examination scenario. Finally, the preset artifact correction network outputs the motion artifact-corrected image and transmits it to the scanning device for display.

[0158] In this embodiment, for any image to be corrected, which contains motion artifacts caused by the movement of the current patient in the target examination scene, the image pixel offset corresponding to the motion artifact in the target examination scene and the actual movement of the detected patient are calculated. A matching preset artifact correction network is used to perform artifact correction processing on the image to be corrected, and the corrected target image of the current patient is obtained. Finally, it is transmitted to the scanning device for image display, thereby optimizing the image containing motion artifacts, eliminating the motion artifacts, and finally displaying an image without motion artifacts. This effectively improves the image quality and facilitates the accuracy of subsequent determination of the patient's condition by doctors and other personnel.

[0159] Example 3

[0160] like Figure 5 As shown, the motion artifact correction system in the image of this embodiment includes:

[0161] Image acquisition module 1 is used to acquire the image to be corrected for the current patient.

[0162] The images to be corrected include, but are not limited to, MRI images, computed tomography (CT) images, positron emission tomography (PET) images, and ultrasound images. During the detection process, voluntary or involuntary movements of the subject can cause motion artifacts in medical images. Involuntary movements (physiological movements) produce physiological motion artifacts, primarily caused by the subject's breathing, blood flow, and cardiac contraction. Voluntary movements mainly include coughing, swallowing, and other body movements. For example, in CT scans, the movement of the subject during the scan can disrupt the consistency and integrity of the projection data. Similarly, during MR scans, the pulsation of blood vessels and the heart, or the flow of cerebrospinal fluid, can produce periodic motion artifacts in the phase encoding direction; voluntary patient movements can produce parallel strip-like artifacts in the phase encoding direction.

[0163] Offset parameter acquisition module 2 is used to acquire the offset parameters corresponding to the motion artifacts generated by the movement of the current patient in the target examination scene based on the image to be corrected;

[0164] The artifact correction module 3 is used to perform artifact correction processing on the image to be corrected using a matching preset artifact correction network based on the target examination scene and offset parameters, so as to obtain the target image of the current patient.

[0165] Among them, the target image contains relatively little artifact information or the artifact information is suppressed, or even does not contain any artifact information, compared to the image to be corrected.

[0166] In this embodiment, for any image to be corrected, which contains motion artifacts caused by the movement of the current patient in the target examination scene, the offset parameters corresponding to the motion artifacts in the target examination scene are calculated, and a matching preset artifact correction network is used to perform artifact correction processing on the image to be corrected, so as to obtain the corrected target image of the current patient, which is then transmitted to the scanning device for image display. This achieves the effect of optimizing the image containing motion artifacts, eliminating the motion artifacts, and finally displaying an image without motion artifacts, effectively improving the image quality and facilitating the accuracy of subsequent determination of the patient's condition by doctors and other personnel.

[0167] Example 4

[0168] like Figure 6 As shown, the motion artifact correction system in this embodiment is a further improvement on Embodiment 3, specifically:

[0169] In one feasible solution, the correction system also includes:

[0170] The first image acquisition module 4 is used to acquire first images of different patients at different preset deflection angles under different types of examination scenarios; wherein each first image is identified by the examination scenario in which it is located;

[0171] The first image parameter acquisition module 5 is used to acquire the first image parameters corresponding to the first image;

[0172] The first pixel offset calculation module 6 is used to calculate the first pixel offset of the image pixel corresponding to the motion artifact in the first image based on the first image parameters.

[0173] The correction network construction module 7 is used to train the preset neural network based on the first pixel offset and construct the corresponding preset artifact correction network.

[0174] In one feasible approach, the pre-defined neural network includes processing the image using a Gaussian low-pass filtering algorithm;

[0175] The process of training a neural network using Gaussian low-pass filtering is as follows:

[0176] The basic principle of the Gaussian low-pass filtering algorithm is as follows: First, a Gaussian kernel is obtained. Based on a Gaussian distribution, the weights of pixels surrounding the center pixel are obtained and normalized. Then, a Gaussian filter is calculated, and the weighted average of the pixel's neighborhood is used to replace the pixel value. The weight of each neighboring pixel monotonically increases with its distance from the center pixel. Based on the Gaussian kernel, a weighted average is calculated on the current image, i.e., convolution is performed on the center pixel. The resulting image is the Gaussian blurred image, which corrects artifacts by smoothing the details of the image. In other words, high-frequency components in the non-smooth parts of the image are filtered out, while low-frequency components in the smooth parts are retained. After Gaussian blurring, Gaussian normal distribution noise is suppressed, and the image becomes blurred, thus correcting motion artifacts in the image.

[0177] Before image processing, an image database indexed by image pixel offsets needs to be established. In this embodiment, firstly, the pixel offset is calculated based on the image size and resolution; an image with a pixel offset of 0 is an image without motion artifacts. Then, images with motion artifacts corresponding to different pixel offsets at various deflection angles, including 0°, 45°, 90°, 135°, 180°, 225°, 270°, and 315°, are set up to construct various scenes.

[0178] Furthermore, the image pixels are determined based on the image size and image resolution. When the monitoring device detects the displacement caused by the patient's movement, the system obtains the image pixel offset corresponding to different motion offsets based on the motion offset and pixel offset, thereby establishing a motion artifact correction database (i.e., an image pixel offset database).

[0179] In one embodiment, the Gaussian low-pass filter uses a two-dimensional Gaussian filter formula, with the center point coordinates set to (0, 0):

[0180]

[0181] Where G(x,y) is a two-dimensional Gaussian function, σ is the variance, and x and y are the coordinates of the point.

[0182] In another embodiment, the two-dimensional Gaussian filter can be expanded in two independent one-dimensional spaces and calculated separately. That is, Gaussian filtering is first calculated in the x-direction, and then Gaussian filtering is calculated in the y-direction based on this.

[0183]

[0184] By using Gaussian filtering, a neural network is trained on the established motion artifact correction database to obtain a corresponding preset artifact correction network for each inspection scenario. This ensures that for any image to be corrected, a suitable preset artifact correction network can be matched for correction processing, guaranteeing the timeliness and accuracy of the image to be corrected.

[0185] The preset neural network also includes a convolutional neural network, which corresponds to a number of convolutional kernels that are randomly configured or pre-fixed.

[0186] In one feasible embodiment, the first image parameters include the image size and / or image resolution in a given orientation;

[0187] The first pixel offset calculation module 6 includes:

[0188] The first displacement acquisition unit 8 is used to acquire the first displacement generated in a set direction when the patient moves in the corresponding examination scenario.

[0189] The first pixel offset calculation unit 9 is used to calculate the first pixel offset corresponding to the image pixel in the first image in a set direction based on the first displacement and image parameters.

[0190] Specifically, the calculation formula for the first pixel offset obtained by the first pixel offset calculation unit is as follows:

[0191]

[0192] Where Δpixel is the pixel offset in the x-direction, x is the image size in the x-direction, and R... x R x Let Δx be the image resolution in the x-direction, and Δx be the displacement corresponding to the patient's movement in the x-direction.

[0193] In addition, in order to better train the images and obtain a more accurate neural network for motion artifact removal, and to process images containing motion artifacts more accurately and efficiently, it is necessary to acquire a large number of motion images in various scanning scenarios. For example, it is necessary to acquire images of examination scenarios with related parameters such as different machine configurations, different patient genders, different patient ages, different patient weights, different patient positions, and different scanning protocols, so as to cover as many clinical scenarios as possible.

[0194] In one feasible solution, the artifact correction module 3 includes:

[0195] The actual movement acquisition unit 10 is used to acquire the actual movement of the current patient in a set direction when the patient moves in the target examination scene.

[0196] The target image parameter acquisition unit 11 is used to acquire the target image parameters of the image to be corrected.

[0197] The target image parameters include the image size and / or image resolution in the specified direction;

[0198] The artifact correction unit 12 is used to match the corresponding preset artifact correction network based on the actual movement amount, target image parameters and target inspection scene, and use the preset artifact correction network to perform artifact correction processing on the image to be corrected.

[0199] In one feasible solution, the correction system also includes:

[0200] Image database construction module 13 is used to construct an image database based on several first images;

[0201] Image filtering module 14 is used to filter out a second image from a number of first images in the image database that has a matching degree greater than a set threshold with the image to be corrected.

[0202] In this process, the first image with the highest similarity is selected as the second image that matches the image to be corrected. If multiple images have the same similarity, one of the first images is randomly selected as the second image that matches the image to be corrected, or a reminder message is generated and pushed to relevant staff, and the corresponding first image is selected as the second image that matches the image to be corrected based on the staff's feedback. The specific method used to determine the second image that matches the image to be corrected can be selected or set according to the actual needs of the scenario, so as to more flexibly meet the control needs of different scenarios.

[0203] The target inspection scene determination module 15 is used to determine the inspection scene of the second image as the target inspection scene corresponding to the image to be corrected.

[0204] In a feasible solution, the examination scenario information corresponding to each examination scenario includes at least one of the following: patient basic information, patient posture information, examination system information, examination protocol information, and examination operation mode information.

[0205] Specifically, DICOM (Digital Imaging and Communication in Medicine) images of patients under different examination scenarios are acquired. Reconstructed images of patients of different body types under various scanning scenarios (such as different positioning, different movement distances, etc.) are collected as much as possible to establish an image database for different examination scenarios. Raw data images and DICOM images are stored on a pre-defined server. Simultaneously, a specific identification method is used to distinguish and store images from different examination scenarios on the pre-defined server. The classification schemes corresponding to different examination scenarios are shown in Table 1 below:

[0206] Table 1

[0207]

[0208] In the table above, HFS indicates head-first, supine position; HFP indicates head-first, prone position; FFS indicates feet-first, supine position; FFP indicates feet-first, prone position; HFDL indicates head-first, left lateral decubitus position. GRE indicates gradient echo pulse sequence; EPI indicates planar echo imaging sequence; FSE indicates fast spin echo sequence.

[0209] According to the classification scheme in the table above, reconstructed images of different patients under various protocols in different examination scenarios are collected. Since the main purpose of this scheme is to correct motion artifacts, it is necessary to collect reconstructed images of patients in scenarios with no positional offset, offset of 1cm, offset of 2cm, offset of 3cm, etc., and to consider various deflection angles (e.g., 0°, 45°, 90°, 135°, 180°, 225°, 270°, 315°) to simulate as many possible movements of patients as possible during actual examinations. These images are stored on a specific server, managed by category, and a multi-scenario image database (or motion correction database) for reconstructed images is created.

[0210] Through the refined construction of the aforementioned image database, images of any patient in any movement state can be matched and queried in the constructed image database in actual examination scenarios, thereby ensuring the feasibility of subsequent artifact correction operations and the accuracy of matching correction with high quality.

[0211] Of course, other classification criteria besides those mentioned above can be used to collect reconstructed images of patients in more examination scenarios. The specific criteria can be flexibly determined or adjusted according to the actual situation.

[0212] In one feasible solution, the correction system includes:

[0213] The status change information acquisition module 16 is used to acquire the current patient's status change information;

[0214] The judgment module 17 is used to call the image acquisition module 1 to be corrected if the status change information indicates that the current patient has moved.

[0215] In a feasible solution, the state change information includes positional shifts, pressure changes, vibration changes, and light signal changes caused by the patient's body movement.

[0216] For example, 3D cameras can be set up in detection equipment (such as MRI machines) to capture patient images, pressure sensors can be set up to detect changes in pressure values ​​within a set time period, vibration sensors can be set up to detect changes in vibration values ​​within a set time period, and light sensors can be set up to detect changes in light signals within a set time period, so as to analyze and determine the corresponding positional movement in a timely manner.

[0217] In addition, two or more types of sensors can be set up simultaneously in the detection device, which can effectively avoid false triggering and achieve more timely and accurate trigger control. This avoids performing the above-mentioned artifact correction process in unnecessary scenarios, which would waste computing resources and cause the image generation process to be slow and fail to meet actual needs. This improves the accuracy, rationality and timeliness of image artifact correction control.

[0218] The following examples and... Figure 4 The implementation principle of the motion artifact correction method in images based on deep learning algorithms in this embodiment is explained in detail below:

[0219] (1) Constructing an image database

[0220] Based on the established classification scheme, reconstructed images of different patients under various protocols in different examination scenarios are collected. Reconstructed images of patients under different offset scenarios are acquired, and various deflection angles are considered to simulate as many possible movement patterns of patients as possible during actual examinations. These images are stored on a specific server, managed by category, and a multi-scenario image database for reconstructed image analysis is created.

[0221] (2) Is artifact correction triggered?

[0222] The system acquires information on changes in the patient's condition, including but not limited to positional shifts, pressure changes, vibration changes, and light signal changes, to determine in a timely and accurate manner whether the patient has moved during the examination. If movement occurs, the scanned image is determined to contain motion artifacts, and the artifact correction function is automatically triggered; otherwise, no triggering is required, and the image can be reconstructed directly.

[0223] (3) Preset artifact correction network

[0224] First images of different patients under different preset deflection angles in different types of examination scenarios are acquired, and each first image is labeled with its own examination scenario; the first image parameters (size, resolution, etc.) corresponding to the first image are obtained and the displacement caused by the detected patient movement is calculated to obtain the pixel displacement of the corresponding image pixels; then, a preset neural network is trained based on the pixel offset using the Gaussian low-pass filtering method to construct the corresponding preset artifact correction network.

[0225] In this approach, model training is performed at the image level, rather than using K-space data for neural network training, thus effectively ensuring the efficiency and accuracy of model training.

[0226] In addition, a preset artifact correction network corresponding to each different inspection scenario can be pre-trained before the actual inspection, so that it can be directly matched and called in the actual inspection scenario without the need for real-time model training.

[0227] Whether to use real-time model training or pre-trained model training can be selected or adjusted according to actual needs.

[0228] (4) Image artifact correction

[0229] Based on the inspection scene corresponding to the image to be corrected, the pixel offset of the image pixels, the image size, and the image resolution, a preset artifact correction network is determined to match it. The image to be corrected is input into the preset artifact correction network, which directly performs artifact correction processing on the image to be corrected. At the image level, the image containing motion artifacts is optimized, eliminating the motion artifacts and displaying an image without motion artifacts, thus obtaining a DICOM scan image after motion artifact removal. In addition, the image artifact correction in this embodiment can support online and offline correction, which can more flexibly meet the requirements of more image artifact correction scenarios.

[0230] The process involves scanning an image containing motion artifacts using a scanning device, calculating the corresponding pixel offset based on the motion offset caused by the patient's movement, image parameters such as the image size and resolution of the image to be corrected, and considering factors such as the current examination setup, patient age, gender, weight, patient positioning, and scanning sequence. Based on this information, a preset artifact correction network is retrieved from a database that best matches the current artifact image in the examination scenario. Finally, the preset artifact correction network outputs the motion artifact-corrected image and transmits it to the scanning device for display.

[0231] In this embodiment, for any image to be corrected, which contains motion artifacts caused by the movement of the current patient in the target examination scene, the image pixel offset corresponding to the motion artifact in the target examination scene and the actual movement of the detected patient are calculated. A matching preset artifact correction network is used to perform artifact correction processing on the image to be corrected, and the corrected target image of the current patient is obtained. Finally, it is transmitted to the scanning device for image display, thereby optimizing the image containing motion artifacts, eliminating the motion artifacts, and finally displaying an image without motion artifacts. This effectively improves the image quality and facilitates the accuracy of subsequent determination of the patient's condition by doctors and other personnel.

[0232] Example 5

[0233] Figure 7 This is a schematic diagram of an electronic device according to Embodiment 5 of the present invention. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the method of the above embodiment. Figure 7 The electronic device 30 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0234] like Figure 7As shown, the electronic device 30 can be represented in the form of a general computing device, such as a server device. The components of the electronic device 30 may include, but are not limited to: at least one processor 31, at least one memory 32, and a bus 33 connecting different system components (including memory 32 and processor 31).

[0235] Bus 33 includes a data bus, an address bus, and a control bus.

[0236] The memory 32 may include volatile memory, such as random access memory (RAM) 321 and / or cache memory 322, and may further include read-only memory (ROM) 323.

[0237] The memory 32 may also include a program / utility 325 having a set (at least one) of program modules 324, including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0238] The processor 31 executes various functional applications and data processing, such as the method of the above embodiments of the present invention, by running computer programs stored in the memory 32.

[0239] Electronic device 30 can also communicate with one or more external devices 34 (e.g., keyboard, pointing device, etc.). This communication can be performed via input / output (I / O) interface 35. Furthermore, the model-generating device 30 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 36. Figure 7 As shown, network adapter 36 communicates with other modules of the model-generated device 30 via bus 33. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the model-generated device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.

[0240] It should be noted that although several units / modules or sub-units / modules of the electronic device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0241] Example 6

[0242] This embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the method of the above embodiment.

[0243] The readable storage medium may be more specifically adopted, including but not limited to: portable disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0244] In a possible implementation, the present invention can also be implemented as a program product comprising program code, which, when the program product is run on a terminal device, causes the terminal device to perform the steps of the method implementing the above embodiments.

[0245] The program code for executing the present invention can be written using any combination of one or more programming languages. The program code can be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on a remote device.

[0246] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of the present invention is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but all such changes and modifications fall within the scope of protection of the present invention.

Claims

1. A method for correcting motion artifacts in an image, characterized in that, The correction method includes: Acquire the image to be corrected for the current patient; Based on the image to be corrected, obtain the offset parameters corresponding to the motion artifacts generated by the current patient's movement in the target examination scene; Based on the target examination scenario and the offset parameters, a matching preset artifact correction network is used to perform artifact correction processing on the image to be corrected to obtain the target image of the current patient. Before the step of acquiring the image to be corrected for the current patient, the method further includes: Acquire first images of different patients at different preset deflection angles under different types of examination scenarios; wherein each first image identifies the examination scenario in which it is located; Obtain the first image parameters corresponding to the first image; The first pixel offset of the image pixel corresponding to the motion artifact in the first image is calculated based on the first image parameters; The preset neural network is trained based on the first pixel offset to construct the corresponding preset artifact correction network.

2. The method for correcting motion artifacts in an image as described in claim 1, characterized in that, The preset neural network includes a Gaussian low-pass filtering algorithm; And / or, The preset neural network includes a convolutional neural network, which corresponds to a number of convolutional kernels that are randomly configured or pre-fixed.

3. The method for correcting motion artifacts in an image as described in claim 1, characterized in that, The first image parameters include the image size and / or image resolution in a set direction; The step of calculating the first pixel offset of the image pixel corresponding to the motion artifact in the first image based on the image parameters includes: The first displacement amount generated in the set direction is collected when the patient moves in the corresponding examination scenario; Based on the first displacement and the image parameters, the first pixel offset corresponding to the image pixel in the first image in the set direction is calculated; And / or, The step of performing artifact correction processing on the image to be corrected using a matching preset artifact correction network based on the target inspection scene and the offset parameters includes: Obtain the actual amount of movement of the current patient in the set direction when moving in the target examination scenario; Obtain the target image parameters of the image to be corrected; The target image parameters include the image size and / or image resolution in the set direction; Based on the actual movement, the target image parameters, and the target inspection scene, a corresponding preset artifact correction network is obtained, and the preset artifact correction network is used to perform artifact correction processing on the image to be corrected.

4. The method for correcting motion artifacts in an image as described in any one of claims 1-3, characterized in that, The correction method further includes: An image database is constructed based on several of the first images; The steps for determining the target inspection scene corresponding to the image to be corrected include: From a plurality of the first images in the image database, a second image is selected whose matching degree with the image to be corrected is greater than a set threshold; The inspection scene of the second image is taken as the target inspection scene corresponding to the image to be corrected.

5. The method for correcting motion artifacts in an image as described in claim 4, characterized in that, The examination scenario information corresponding to each of the aforementioned examination scenarios includes at least one of the following: patient basic information, patient posture information, examination system information, examination protocol information, and examination operation mode information.

6. The method for correcting motion artifacts in an image as described in any one of claims 1-3, characterized in that, Before the step of acquiring the image to be corrected for the current patient, the method further includes: Obtain the current patient's status change information; If the state change information indicates that the current patient has moved, then the step of obtaining the image to be corrected for the current patient is performed.

7. The method for correcting motion artifacts in an image as described in claim 6, characterized in that, The state change information includes at least one of the following: positional offset information, pressure change information, vibration change information, and light signal change information generated when the patient's body moves.

8. A system for correcting motion artifacts in an image, characterized in that, The correction system includes: The image to be corrected acquisition module is used to acquire the image to be corrected for the current patient; The offset parameter acquisition module is used to acquire the offset parameters corresponding to the motion artifacts generated by the movement of the current patient in the target examination scene based on the image to be corrected. The artifact correction module is used to perform artifact correction processing on the image to be corrected using a matching preset artifact correction network based on the target examination scene and the offset parameters, so as to obtain the target image of the current patient. The correction system also includes: The first image acquisition module is used to acquire first images of different patients at different preset deflection angles under different types of examination scenarios; wherein each first image identifies the examination scenario in which it is located; The first image parameter acquisition module is used to acquire the first image parameters corresponding to the first image. The first pixel offset calculation module is used to calculate the first pixel offset of the image pixel corresponding to the motion artifact in the first image based on the first image parameters. The correction network construction module is used to train a preset neural network based on the first pixel offset to construct the corresponding preset artifact correction network.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes a computer program, it implements the method for correcting motion artifacts in the image according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for correcting motion artifacts in the image according to any one of claims 1-7.

Images