Auxiliary film reading analysis system in MRI (Magnetic Resonance Imaging) image examination process

Abstract

The invention relates to the technical field of image processing, and provides an auxiliary film reading analysis system in an MRI (Magnetic Resonance Imaging) image examination process, which comprises the following steps: acquiring a plurality of kidney MRI images of a patient; performing edge detection and contour analysis on the kidney MRI image to obtain a plurality of connected domains; acquiring a banding factor of each connected domain; obtaining an offset artifact possibility of each connected domain, and screening to obtain a plurality of suspected chemical shift artifact regions; acquiring a plurality of kidney regions in the kidney MRI image; the kidney offset distance of each suspected chemical shift artifact area is obtained; obtaining the final artifact possibility of each suspected chemical shift artifact area, and screening to obtain a plurality of chemical shift artifact areas; and marking a chemical shift artifact area in each kidney MRI image, and assisting in reading and analyzing. The invention aims to solve the problem that chemical shift artifacts in MRI images affect film reading analysis.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to an auxiliary image reading and analysis system during MRI image examination. Background Technology

[0002] In MRI imaging examinations, the significance of assisted image interpretation lies in improving the accuracy and efficiency of diagnosis. By combining artificial intelligence algorithms and image processing technology, assisted image interpretation systems can quickly identify abnormal structures, lesions, or lesions in images, helping doctors to understand the patient's condition more clearly. This helps reduce the workload of doctors, improves the speed and accuracy of image interpretation, and can also detect some tiny lesions that are difficult to detect with the naked eye, thus enabling earlier intervention and treatment. It plays a vital role in improving the quality of image diagnosis, ensuring patient safety, and optimizing the allocation of medical resources.

[0003] When patients undergo MRI imaging, chemical shift artifacts can occur because hydrogen atoms in water and fat have different frequencies in a magnetic field. Inhomogeneity of the magnetic field and the selection of scanning parameters can exacerbate this phenomenon. The presence of artifacts in MRI images can affect the identification of tissue structures, thus impacting the doctor's interpretation of the images. Currently, the presence of chemical shift artifacts in MRI images relies primarily on the doctor's judgment, increasing the doctor's burden in image analysis. There is a need to analyze the characteristics related to chemical shift artifacts to assist doctors in quickly analyzing and labeling chemical artifact areas, reducing the doctor's image analysis burden and improving the efficiency of image interpretation. Summary of the Invention

[0004] This invention provides an auxiliary image interpretation and analysis system during MRI image examination to solve the problem of chemical shift artifacts affecting image interpretation and analysis in existing MRI images. The specific technical solution adopted is as follows:

[0005] This invention proposes an auxiliary image interpretation and analysis system during MRI image examination, the system comprising:

[0006] The MRI image acquisition module is used to acquire several MRI images of the patient's kidneys.

[0007] The MRI image processing module is used to obtain several connected regions from kidney MRI images through edge detection and contour analysis; analyze the width of each position on the skeleton line of the connected region in the connected region to obtain the band factor of each connected region; combine the fluctuation of the width of each position on the skeleton line of the connected region in the connected region with the gray-scale performance in the connected region to obtain the probability of displacement artifacts in each connected region, and then screen out several suspected chemical shift artifact regions.

[0008] Several renal regions were obtained from the renal MRI images. Based on the positional distribution relationship between the suspected chemical shift artifact regions and the renal regions, the renal offset distance of each suspected chemical shift artifact region was obtained. Combining the width differences between any suspected chemical shift artifact region and other chemical shift artifact regions before and after it, the final artifact probability of each suspected chemical shift artifact region was obtained, and several chemical shift artifact regions were selected.

[0009] The MRI image analysis module is used to annotate the chemical shift artifact regions in each kidney MRI image and assist in image interpretation and analysis.

[0010] Optionally, the method for obtaining several connected components from the renal MRI images through edge detection and contour analysis includes:

[0011] Edge detection is performed on renal MRI images to obtain several edges of the renal MRI images; based on these edges, connected component analysis is used to obtain several connected components in the renal MRI images.

[0012] Optionally, the banding factor of each connected component is obtained by the following method:

[0013] The Zhang-Suen algorithm is used to operate on each connected component in the kidney MRI image to obtain the skeleton line of each connected component, and the pixels on the skeleton line are used as the skeleton points of each connected component.

[0014] Based on the skeleton line and its skeleton points, as well as their distribution within the connected domain, the width representation factor of each skeleton point on the skeleton line of the connected domain is obtained.

[0015] The band factor of a connected component is obtained by taking the number of skeleton points on the skeleton line, the mean of the width performance factor of all skeleton points, and the sum of the width performance factors of all skeleton points. The band factor is positively correlated with the number of skeleton points and the mean of the width performance factor, and negatively correlated with the sum of the width performance factors.

[0016] Optionally, the specific method for obtaining the width representation factor of each skeleton point on the skeleton line of the connected component includes:

[0017] For any connected domain, draw a perpendicular line through any skeleton point on the skeleton line of the connected domain. The perpendicular line intersects the edge of the connected domain at two points. The distance between the two intersection points is taken as the width representation factor of the skeleton point on the skeleton line of the connected domain.

[0018] Optionally, the specific method for obtaining the offset artifact probability of each connected component includes:

[0019] For any adjacent skeleton points on a connected skeleton line, the absolute value of the difference between the width representation factors of the two skeleton points is taken as the width difference between the two skeleton points; the average width difference between all adjacent skeleton points on the connected skeleton line is taken as the adjacent width difference of the connected skeleton line.

[0020] The probability of offset artifacts in the i-th connected component is R. i The calculation method is as follows:

[0021]

[0022] Among them, W i Let σ represent the banding factor of the i-th connected component. i (h) represents the standard deviation of all pixels in the i-th connected component, E i This represents the number of skeleton points on the skeleton line of the i-th connected component. ΔY represents the difference in width between adjacent skeleton lines of the i-th connected component. i (t, t-1) represents the width difference between the i-th skeleton point and the (i-1)-th skeleton point on the skeleton line of the i-th connected component; || represents the absolute value function; exp() represents the exponential function with the natural constant as the base; norm() represents the linear normalization function.

[0023] Optionally, the specific method for obtaining several renal regions in the renal MRI image is as follows:

[0024] The kidney region in a kidney MRI image is segmented using a semantic segmentation network. The kidney MRI image to be segmented is input into the trained semantic segmentation network, and several kidney regions in the kidney MRI image are output.

[0025] Optionally, the specific method for obtaining the kidney offset distance of each suspected chemical shift artifact region includes:

[0026] For any suspected chemical shift artifact region in a kidney MRI image, and for any kidney region, for any skeleton point on the skeleton line of the suspected chemical shift artifact region, obtain the minimum distance between the skeleton point and all edge pixels of the kidney region, and use it as the offset distance between the skeleton point and the kidney region.

[0027] The average offset distance between the suspected chemical shift artifact region and the kidney region is taken as the offset distance between the suspected chemical shift artifact region and the kidney region; the minimum offset distance between the suspected chemical shift artifact region and each kidney region is taken as the kidney offset distance of the suspected chemical shift artifact region.

[0028] Optionally, the specific method for obtaining the final artifact probability of each suspected chemical shift artifact region includes:

[0029] Based on the frequency coding direction during kidney MRI image scanning, all suspected chemical shift artifact regions were sorted.

[0030] For any suspected chemical shift artifact region in a renal MRI image, the mean of the width representation factors of all skeletal points on its skeleton line is used as the width representation of the suspected chemical shift artifact region. The final artifact probability F of the j-th suspected chemical shift artifact region is then calculated. j The calculation method is as follows:

[0031]

[0032] Among them, S j G represents the kidney offset distance of the j-th suspected chemical shift artifact region. j P represents the width of the j-th suspected chemical shift artifact region. j Q represents the number of suspected chemical shift artifact regions preceding the j-th suspected chemical shift artifact region. j G represents the number of suspected chemical shift artifact regions following the j-th suspected chemical shift artifact region. j,p G represents the width of the p-th suspected chemical shift artifact region preceding the j-th suspected chemical shift artifact region. j,q This represents the width of the qth suspected chemical shift artifact region after the jth suspected chemical shift artifact region; exp() represents an exponential function with the natural constant as the base; norm() represents a linear normalization function.

[0033] Optionally, the method for sorting all suspected chemical shift artifact regions based on the frequency encoding direction during renal MRI image scanning includes:

[0034] Obtain the centroid of each suspected chemical shift artifact region in the renal MRI image, and sort all suspected chemical shift artifact regions according to the order in which the centroid of each suspected chemical shift artifact region appears in the frequency coding direction during the renal MRI image scanning process.

[0035] Optionally, the specific method for annotating the chemical shift artifact region in each renal MRI image includes:

[0036] Chemical shift artifact regions were acquired for each kidney MRI image, and the edge pixels of the chemical shift artifact regions were marked in red in the kidney MRI image.

[0037] The beneficial effects of this invention are as follows: This invention extracts connected regions from renal MRI images, analyzes the changes in the width of the connected regions to quantify their band-like appearance, and combines the width fluctuations with the grayscale values ​​of internal pixels to obtain the probability of displacement artifacts. This reflects the band-like appearance of the connected regions and the overall tendency of the internal structure to be white or black due to the superposition or cancellation of signals that conform to the characteristics of chemical shift artifacts, thereby screening out suspected chemical shift artifact regions. Then, by analyzing the distribution relationship between the suspected chemical shift artifact regions and the renal region, the renal offset distance is quantified to reflect whether they are distributed around the renal region. Combined with the factor that the increase in the frequency coding direction gradient during MRI scanning leads to an increase in offset, the width changes of the suspected chemical shift artifact regions are comprehensively considered to obtain the final artifact probability, thereby ultimately screening out the chemical shift artifact regions. This eliminates the interference of possible chemical shift artifact regions in renal MRI images on the overall appearance of renal MRI images, reduces the burden on medical staff in interpreting and analyzing MRI images, and improves the efficiency of auxiliary image interpretation and analysis. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a structural block diagram of an auxiliary image reading and analysis system provided in an embodiment of the present invention during MRI image examination.

[0040] Figure 2 This is a schematic diagram of the connected component skeleton lines and the band analysis process.

[0041] Figure 3 An example image showing the sorting of suspected chemical shift artifact regions based on frequency encoding direction. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Please see Figure 1 The diagram illustrates a structural block diagram of an auxiliary image reading and analysis system for MRI image examination provided by an embodiment of the present invention. The system includes:

[0044] MRI image acquisition module 101: Acquires several kidney MRI images of the patient.

[0045] The purpose of this embodiment is to extract and label the chemical shift artifact regions that may occur in the patient's renal MRI images, so as to reduce the burden on medical staff in the process of interpreting and analyzing MRI images and improve the efficiency of auxiliary interpretation and analysis. First, it is necessary to obtain several renal MRI images of the patient. This embodiment uses renal MRI examination as an example for description.

[0046] Specifically, the patient's kidneys are scanned using an MRI scanner to obtain several kidney images. Each kidney image is then processed using histogram equalization to enhance contrast, resulting in several kidney MRI images. In this embodiment, any kidney MRI image is used as an example for subsequent processing.

[0047] MRI Image Processing Module 102:

[0048] It should be noted that when artifacts appear in renal MRI images, they typically manifest as white and black band-like areas around the kidney tissue. This is due to the difference in signal phase caused by the different resonance frequencies of hydrogen atoms in water and fat in a magnetic field. When the signals from water and fat are superimposed, if the phase difference leads to signal enhancement, it appears as white; if the phase difference leads to signal cancellation, it appears as black. Furthermore, the width of the artifact area is uniform. Because renal tissue is complex, some normal areas also appear as band-like areas, thus interfering with artifact screening. Therefore, it is necessary to utilize other characteristics of chemical shift artifacts for screening. The displacement of chemical shift artifacts increases with increasing field strength, thereby allowing for the screening of chemical shift artifact areas in renal MRI images.

[0049] (1) Several connected regions were obtained from the kidney MRI images by edge detection and contour analysis; the width of each position on the skeleton line of the connected region was analyzed to obtain the band factor of each connected region; the fluctuation of the width of each position on the skeleton line of the connected region and the gray scale in the connected region were combined to obtain the possibility of displacement artifacts in each connected region, and then several suspected chemical displacement artifact regions were screened out.

[0050] Preferably, in one embodiment of the present invention, several connected regions are obtained from renal MRI images through edge detection and contour analysis, including the following specific methods:

[0051] Edge detection is performed on the renal MRI image to obtain several edges of the renal MRI image; the edge detection uses the Canny edge detection algorithm, which is a well-known technology and will not be described in detail in this embodiment; based on the several edges of the renal MRI image, several connected components in the renal MRI image are obtained through connected component analysis.

[0052] It should be further explained that after obtaining the connected regions from the renal MRI images, the connected regions are analyzed based on the band-like appearance of the artifact regions. This is done by obtaining the skeleton line of the connected regions and drawing perpendicular lines to each pixel on the skeleton line. The distance between the intersection of the perpendicular line and the edge of the connected regions reflects the width of each pixel on the skeleton line. Then, based on the width appearance and the length of the skeleton line, the band-like appearance of the connected regions is comprehensively quantified.

[0053] Preferably, in one embodiment of the present invention, the method for analyzing the width performance of each position on the skeleton line of the connected domain and obtaining the banding factor of each connected domain includes:

[0054] The Zhang-Suen algorithm is used to process each connected component in the renal MRI image to obtain the skeleton line of each connected component. Pixels on the skeleton line are then used as skeleton points of each connected component. The Zhang-Suen algorithm is a well-known technique for skeleton acquisition and will not be described in detail in this embodiment. For any connected component, a perpendicular line is drawn through any skeleton point on the skeleton line of that connected component. This perpendicular line intersects the edge of the connected component at two points. The distance between these two intersection points is used as the width representation factor of that skeleton point on the skeleton line of the connected component. The perpendicular line is perpendicular to the tangent of the skeleton line at that skeleton point. Figure 2 As shown; obtain the width representation factor of each skeleton point on the skeleton line of the connected component, then the banding factor W of the i-th connected component is... i The calculation method is as follows:

[0055]

[0056] Among them, E i This represents the number of skeleton points on the skeleton line of the i-th connected component. Y represents the mean width performance factor of all skeleton points on the skeleton line of the i-th connected component. i,t It represents the width representation factor of the t-th skeleton point on the skeleton line of the i-th connected component.

[0057] It should be noted that the product of the number of skeleton points of a connected region and the mean of the width performance factor can reflect the overall area performance of the connected region. The closer the cumulative value of the width performance factor is to the mean, the smaller the difference between the width performance factors of each skeleton point. The entire connected region presents a band-like shape, that is, the width at each position is similar. This is used to quantify the band-like factor of the connected region.

[0058] Furthermore, it should be noted that, based on the banded representation of connected components, the more similar the widths of adjacent skeleton points on the skeleton line, the smaller the overall fluctuation in the width differences between adjacent skeleton points, and the closer the offsets at different positions within the connected component, the more consistent it is with the representation of chemical shift artifact regions. At the same time, due to the superposition of water and fat signals, the phase difference leads to signal enhancement, which makes the gray values ​​of pixels within the artifact approach white, while the phase difference leads to signal cancellation, which makes the gray values ​​of pixels within the artifact approach black. The difference in gray values ​​between pixels in the same artifact region is small, and thus the standard deviation is smaller, thereby quantifying the probability of shift artifacts in connected components.

[0059] Preferably, in one embodiment of the present invention, by combining the fluctuation of the width of each position on the connected domain skeleton line with the grayscale performance within the connected domain, the probability of offset artifacts in each connected domain is obtained, and then several suspected chemical shift artifact regions are screened out. The specific method includes:

[0060] For any two skeleton points on a connected component skeleton line, the absolute value of the difference between the width representation factors of the two skeleton points is taken as the width difference between the two skeleton points; the average width difference between all adjacent skeleton points on the connected component skeleton line is taken as the adjacent width difference of the connected component skeleton line; then the probability R of the offset artifact in the i-th connected component is... i The calculation method is as follows:

[0061]

[0062] Among them, W i Let σ represent the banding factor of the i-th connected component. i (h) represents the standard deviation of all pixels in the i-th connected component, E i This represents the number of skeleton points on the skeleton line of the i-th connected component. ΔY represents the difference in width between adjacent skeleton lines of the i-th connected component. i (t, t-1) represents the width difference between the ith skeleton point and the (i-1)th skeleton point on the skeleton line of the ith connected component; || represents the absolute value function; exp() represents the exponential function with the natural constant as the base. In this embodiment, the exp(-x) model is used to represent the inverse proportional relationship, where x is the input of the model. The implementer can set the inverse proportional function according to the actual situation; norm() represents the linear normalization function, which normalizes the values ​​of all connected components.

[0063] It should be noted that, given a larger banding factor, connected components are more consistent with banding patterns. Furthermore, the smaller the width difference between adjacent skeleton points on the skeleton line and the smaller the overall fluctuation, the more similar the offsets at various positions within the connected component, and the smaller the standard deviation of the gray values ​​of the pixels within the connected component, the more consistent the gray values ​​of the pixels within it become. This is more consistent with the characteristics of chemical shift artifact regions, and the greater the likelihood of shift artifacts in the corresponding connected components.

[0064] Furthermore, the probability of displacement artifacts in each connected region in the renal MRI image is obtained according to the above method. A preset displacement artifact threshold is set. In this embodiment, the displacement artifact threshold is described as 0.6. If the probability of displacement artifacts in any connected region is greater than or equal to the displacement artifact threshold, the connected region is regarded as a suspected chemical displacement artifact region, thus obtaining several suspected chemical displacement artifact regions in the renal MRI image.

[0065] (2) Obtain several kidney regions in the MRI images of the kidney; based on the positional distribution relationship between the suspected chemical displacement artifact regions and the kidney regions, obtain the kidney offset distance of each suspected chemical displacement artifact region; combine the width difference between any suspected chemical displacement artifact region and other chemical displacement artifact regions before and after it to obtain the final artifact probability of each suspected chemical displacement artifact region, and screen out several chemical displacement artifact regions.

[0066] It should be noted that, due to the complexity of kidney tissue, some normal areas may also appear as bands. Therefore, it is necessary to analyze the distribution relationship between the suspected chemical shift artifact area and the kidney area to ensure that the chemical shift artifact area is close to the kidney area, thereby quantifying the kidney offset distance of the suspected chemical shift artifact area.

[0067] Preferably, in one embodiment of the present invention, the specific method for acquiring several renal regions in a renal MRI image includes:

[0068] The semantic segmentation network is used to segment the kidney region in kidney MRI images. In this embodiment, the semantic segmentation network adopts the DeepLabV3 neural network structure and uses a large number of kidney MRI images as training sets. The pixels in each kidney MRI image in the training set are manually labeled, with background pixels labeled as 0 and kidney region pixels labeled as 1. The task of the network is classification, and the cross-entropy loss function is used. The semantic segmentation network is trained by making the loss function converge, and the trained semantic segmentation network is obtained.

[0069] Furthermore, the kidney MRI image to be segmented is input into the trained semantic segmentation network, and the output is several kidney regions in the kidney MRI image.

[0070] It should be further noted that artifacts are usually distributed around the kidney tissue. By analyzing the distribution relationship between the position of each skeleton point on the connected domain skeleton line and its nearest kidney region, the closer the suspected chemical displacement artifact region is to the kidney region, the greater the probability that it is a displacement artifact region around the kidney region. This can be used to determine the kidney displacement distance of each suspected chemical artifact region.

[0071] Preferably, in one embodiment of the present invention, the renal offset distance of each suspected chemical shift artifact region is obtained based on the positional distribution relationship between the suspected chemical shift artifact region and the renal region. The specific method includes:

[0072] For any suspected chemical shift artifact region in a renal MRI image, and any renal region (there may be two renal regions in a renal MRI image), for any skeleton point on the skeleton line of the suspected chemical shift artifact region, the minimum distance between the skeleton point and all edge pixels of the renal region is obtained as the offset distance between the skeleton point and the renal region; the average offset distance between all skeleton points on the skeleton line of the suspected chemical shift artifact region and the renal region is taken as the offset distance between the suspected chemical shift artifact region and the renal region; and the minimum offset distance between the suspected chemical shift artifact region and each renal region is taken as the renal offset distance of the suspected chemical shift artifact region.

[0073] Preferably, in one embodiment of the present invention, by combining the width differences between any suspected chemical shift artifact region and other chemical shift artifact regions before and after it, the final artifact probability of each suspected chemical shift artifact region is obtained, and a number of chemical shift artifact regions are screened out. The specific method includes:

[0074] It should be noted that during scanning, the intensity of the frequency-encoded gradient usually changes gradually along a specific direction. That is, signals emitted from different locations will have different frequencies depending on the intensity of this gradient, so each point can be located by the difference in frequency. However, in renal MRI images, the intensity of the frequency encoding increases sequentially along the frequency encoding direction. The greater the intensity of the frequency encoding, the greater the offset of the chemical shift artifact region. Therefore, it is necessary to sort the suspected chemical shift artifact regions according to the frequency encoding direction, and then, based on the width difference between each suspected chemical shift artifact region and other suspected chemical shift artifact regions before and after it, combined with the renal offset distance, determine the final artifact probability and screen out the chemical shift artifact regions.

[0075] Specifically, the centroid (or centrature) of each suspected chemical shift artifact region in the renal MRI image is obtained using existing techniques. The renal MRI image corresponds to a frequency coding direction during the scanning process. All suspected chemical shift artifact regions are sorted according to the order in which their centroids appear in the frequency coding direction. Figure 3 As shown.

[0076] Furthermore, for any suspected chemical shift artifact region in a renal MRI image, the average width representation factor of all skeletal points on its skeleton line is used as the width representation of the suspected chemical shift artifact region. Then, the final artifact probability F of the j-th suspected chemical shift artifact region is calculated. j The calculation method is as follows:

[0077]

[0078] Among them, S j G represents the kidney offset distance of the j-th suspected chemical shift artifact region. j P represents the width of the j-th suspected chemical shift artifact region. j Q represents the number of suspected chemical shift artifact regions preceding the j-th suspected chemical shift artifact region. j G represents the number of suspected chemical shift artifact regions following the j-th suspected chemical shift artifact region. j,p G represents the width of the p-th suspected chemical shift artifact region preceding the j-th suspected chemical shift artifact region. j,q This represents the width of the q-th suspected chemical shift artifact region after the j-th suspected chemical shift artifact region; exp() represents an exponential function with the natural constant as the base. In this embodiment, the exp(-x) model is used to represent the inverse proportional relationship, where x is the input of the model. The implementer can set the inverse proportional function according to the actual situation; norm() represents a linear normalization function, which normalizes all suspected chemical shift artifact regions.

[0079] It should be noted that, along the frequency coding direction, the greater the frequency coding intensity, the greater the offset of the chemical shift artifact region. Consequently, the width of the suspected chemical shift artifact region is larger than that of the previous suspected chemical shift artifact region, and smaller than that of the subsequent one. At the same time, the smaller its kidney offset distance, the more it matches the offset behavior of the chemical shift artifact region, and it is distributed around the kidney region, thus obtaining the final artifact probability.

[0080] Furthermore, a preset artifact probability threshold is set. In this embodiment, the artifact probability threshold is described as 0.7. If the final artifact probability of any suspected chemical shift artifact region is greater than or equal to the artifact probability threshold, the suspected chemical shift artifact region is taken as the chemical shift artifact region in the renal MRI image, thus obtaining several chemical shift artifact regions.

[0081] Therefore, by extracting connected regions from renal MRI images and analyzing the width changes of these regions to quantify their band-like appearance, the likelihood of offset artifacts is obtained by combining the width fluctuations with the grayscale values ​​of internal pixels. This reflects the band-like appearance of the connected regions and the overall tendency of the internal area to be white or black due to signal superposition or cancellation that conforms to the characteristics of chemical shift artifacts, thus screening out suspected chemical shift artifact regions. Then, by analyzing the distribution relationship between suspected chemical shift artifact regions and the renal region, the renal offset distance is quantified to reflect whether it is distributed around the renal region. Combined with the factor that the increase in the frequency coding direction gradient during MRI scanning leads to an increase in offset, the width changes of suspected chemical shift artifact regions are comprehensively considered to obtain the final artifact likelihood, thereby ultimately screening out chemical shift artifact regions.

[0082] MRI Image Analysis Module 103: Marks the chemical shift artifact areas in each kidney MRI image and assists in image interpretation and analysis.

[0083] Specifically, the chemical shift artifact region is obtained for each renal MRI image according to the above method and marked in the renal MRI image. In this embodiment, the edge pixels of the chemical shift artifact region are marked with red to reduce the interference of the chemical shift artifact region in the image reading and analysis process, and the marked renal MRI image is obtained. Medical staff can use the marked renal MRI image for image reading and analysis.

[0084] This concludes the embodiment.

[0085] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An auxiliary image reading and analysis system for MRI image examination, characterized in that, The system includes: The MRI image acquisition module is used to acquire several MRI images of the patient's kidneys. The MRI image processing module is used to obtain several connected regions from kidney MRI images through edge detection and contour analysis; analyze the width of each position on the skeleton line of the connected region in the connected region to obtain the band factor of each connected region; combine the fluctuation of the width of each position on the skeleton line of the connected region in the connected region with the gray-scale performance in the connected region to obtain the probability of displacement artifacts in each connected region, and then screen out several suspected chemical shift artifact regions. Several renal regions were obtained from the renal MRI images. Based on the positional distribution relationship between the suspected chemical shift artifact regions and the renal regions, the renal offset distance of each suspected chemical shift artifact region was obtained. Combining the width differences between any suspected chemical shift artifact region and other chemical shift artifact regions before and after it, the final artifact probability of each suspected chemical shift artifact region was obtained, and several chemical shift artifact regions were selected. The MRI image analysis module is used to annotate the chemical shift artifact regions in each kidney MRI image and assist in image interpretation and analysis.

2. The auxiliary image reading and analysis system for MRI image examination according to claim 1, characterized in that, The method for obtaining several connected components from renal MRI images through edge detection and contour analysis includes the following specific methods: Edge detection is performed on renal MRI images to obtain several edges of the renal MRI images; based on these edges, connected component analysis is used to obtain several connected components in the renal MRI images.

3. The auxiliary image reading and analysis system for MRI image examination according to claim 1, characterized in that, The specific method for obtaining the banding factor of each connected component is as follows: The Zhang-Suen algorithm is used to operate on each connected component in the kidney MRI image to obtain the skeleton line of each connected component, and the pixels on the skeleton line are used as the skeleton points of each connected component. Based on the skeleton line and its skeleton points, as well as their distribution within the connected domain, the width representation factor of each skeleton point on the skeleton line of the connected domain is obtained. The band factor of a connected component is obtained by taking the number of skeleton points on the skeleton line, the mean of the width performance factor of all skeleton points, and the sum of the width performance factors of all skeleton points. The band factor is positively correlated with the number of skeleton points and the mean of the width performance factor, and negatively correlated with the sum of the width performance factors.

4. The auxiliary image reading and analysis system for MRI image examination according to claim 3, characterized in that, The specific method for obtaining the width representation factor of each skeleton point on the skeleton line of the connected domain is as follows: For any connected domain, draw a perpendicular line through any skeleton point on the skeleton line of the connected domain. The perpendicular line intersects the edge of the connected domain at two points. The distance between the two intersection points is taken as the width representation factor of the skeleton point on the skeleton line of the connected domain.

5. The auxiliary image reading and analysis system for MRI image examination according to claim 4, characterized in that, The specific methods for obtaining the offset artifact probability of each connected component are as follows: For any adjacent skeleton points on a connected skeleton line, the absolute value of the difference between the width representation factors of the two skeleton points is taken as the width difference between the two skeleton points; the average width difference between all adjacent skeleton points on the connected skeleton line is taken as the adjacent width difference of the connected skeleton line. The probability of offset artifacts in the i-th connected component is R. i The calculation method is as follows: Among them, W i Let σ represent the banding factor of the i-th connected component. i (h) represents the standard deviation of all pixels in the i-th connected component, E i This represents the number of skeleton points on the skeleton line of the i-th connected component. ΔY represents the difference in width between adjacent skeleton lines of the i-th connected component. i (t, t-1) represents the width difference between the i-th skeleton point and the (i-1)-th skeleton point on the skeleton line of the i-th connected component; || represents the absolute value function; exp() represents the exponential function with the natural constant as the base; norm() represents the linear normalization function.

6. The auxiliary image reading and analysis system for MRI image examination according to claim 1, characterized in that, The specific method for obtaining several renal regions in the renal MRI image is as follows: The kidney region in a kidney MRI image is segmented using a semantic segmentation network. The kidney MRI image to be segmented is input into the trained semantic segmentation network, and several kidney regions in the kidney MRI image are output.

7. The auxiliary image reading and analysis system for MRI image examination according to claim 3, characterized in that, The specific method for obtaining the kidney offset distance of each suspected chemical shift artifact region is as follows: For any suspected chemical shift artifact region in a kidney MRI image, and for any kidney region, for any skeleton point on the skeleton line of the suspected chemical shift artifact region, obtain the minimum distance between the skeleton point and all edge pixels of the kidney region, and use it as the offset distance between the skeleton point and the kidney region. The average offset distance between the suspected chemical shift artifact region and the kidney region is taken as the offset distance between the suspected chemical shift artifact region and the kidney region; the minimum offset distance between the suspected chemical shift artifact region and each kidney region is taken as the kidney offset distance of the suspected chemical shift artifact region.

8. The auxiliary image reading and analysis system for MRI image examination according to claim 3, characterized in that, The specific methods for obtaining the final artifact probability of each suspected chemical shift artifact region are as follows: Based on the frequency coding direction during kidney MRI image scanning, all suspected chemical shift artifact regions were sorted. For any suspected chemical shift artifact region in a renal MRI image, the mean of the width representation factors of all skeletal points on its skeleton line is used as the width representation of the suspected chemical shift artifact region. The final artifact probability F of the j-th suspected chemical shift artifact region is then calculated. j The calculation method is as follows: Among them, S j G represents the kidney offset distance of the j-th suspected chemical shift artifact region. j P represents the width of the j-th suspected chemical shift artifact region. j Q represents the number of suspected chemical shift artifact regions preceding the j-th suspected chemical shift artifact region. j G represents the number of suspected chemical shift artifact regions following the j-th suspected chemical shift artifact region. j,p G represents the width of the p-th suspected chemical shift artifact region preceding the j-th suspected chemical shift artifact region. j,q This represents the width of the qth suspected chemical shift artifact region after the jth suspected chemical shift artifact region; exp() represents an exponential function with the natural constant as the base; norm() represents a linear normalization function.

9. The auxiliary image reading and analysis system for MRI image examination according to claim 8, characterized in that, The method for sorting all suspected chemical shift artifact regions based on the frequency encoding direction during renal MRI image scanning includes the following specific steps: Obtain the centroid of each suspected chemical shift artifact region in the renal MRI image, and sort all suspected chemical shift artifact regions according to the order in which the centroid of each suspected chemical shift artifact region appears in the frequency coding direction during the renal MRI image scanning process.

10. The auxiliary image reading and analysis system for MRI image examination according to claim 1, characterized in that, The specific method for annotating the chemical shift artifact regions in each renal MRI image includes: Chemical shift artifact regions were acquired for each kidney MRI image, and the edge pixels of the chemical shift artifact regions were marked in red in the kidney MRI image.

Images