Method for removing artifacts of upper respiratory tract image

Abstract

The invention relates to the technical field of image enhancement, in particular to an artifact removal method for an upper respiratory tract image. According to the method, the optimal analysis areas are selected from the local areas of the upper respiratory tract image of the respiratory cycle, and the local areas adjacent to the optimal analysis areas correspond to the lengths, the direction distribution concentration degrees and the detail richness degrees of the edge lines in the optimal analysis areas, and the closeness degrees of the edge lines in the optimal analysis areas. Obtaining an edge detail loss degree, and obtaining a fusion fit factor in combination with a detail richness degree of the optimal analysis area and a complexity degree of a breathing mode of a breathing cycle; and adjusting the filtering intensity of filtering processing on the optimal analysis region by using the image fusion algorithm to obtain an artifact correction region, and fusing the regions to obtain an upper respiratory tract enhanced image. According to the method, targeted filtering adjustment is carried out on the optimal analysis area, and the effect of removing the breathing artifacts in the artifact correction area is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of image enhancement, and particularly relates to a method for removing artifacts from upper respiratory tract images. Background Art

[0002] The upper respiratory tract imaging examination mainly scans the patient's upper respiratory tract through X-ray irradiation to understand the morphological abnormalities of the upper respiratory tract; the upper respiratory tract images are crucial in the clinical diagnosis of bronchial asthma, allergic rhinitis, sinusitis, vocal cord diseases, etc. During the scanning process, the patient's respiratory movement is the main factor causing serious artifacts in the upper respiratory tract images, affecting the quality of the upper respiratory tract images and thus reducing the accuracy of the upper respiratory tract examination.

[0003] In order to remove the artifacts generated by respiration in the upper respiratory tract images, the existing filtering methods usually directly perform filtering processing on the entire upper respiratory tract image and the filtering intensity of the pixel points in the image is the same; however, due to the complex structure and tissue morphological differences of the upper respiratory tract, the artifacts generated by respiratory movement in different parts of the upper respiratory tract are different. For example, the image of the airway shows a clear edge, but the image of the soft tissue lacks an obvious boundary, making the soft tissue more vulnerable to respiratory artifacts, resulting in a poor removal effect of the unified filtering intensity on the respiratory artifacts of the upper respiratory tract images. Summary of the Invention

[0004] In order to solve the technical problem that the artifacts generated by respiratory movement in different parts of the upper respiratory tract are different, resulting in a poor removal effect of the unified filtering intensity on the respiratory artifacts of the upper respiratory tract images, the purpose of the present invention is to provide a method for removing artifacts from upper respiratory tract images, and the specific technical solution adopted is as follows: The present invention proposes a method for removing artifacts from upper respiratory tract images, and the method includes: Obtain the upper respiratory tract images at each moment within the respiratory cycle of the patient, and the upper respiratory images include different local regions; Match the local regions of the upper respiratory tract images within the respiratory cycle, and select the optimal analysis region for each local region from the matched local regions based on the richness of details of the local regions; Perform edge detection on the optimal analysis region to obtain an edge line, and obtain the edge detail loss degree of each optimal analysis region according to the length, direction distribution concentration degree and richness of details of the edge line in the optimal analysis region corresponding to the local region adjacent to each optimal analysis region, as well as the tightness of the edge line in each optimal analysis region; According to the complexity of the breathing pattern of the patient in the breathing cycle, the richness of details of each optimal analysis region and the edge detail loss degree, obtain the fusion adaptation factor of each optimal analysis region; use the fusion adaptation factor to adjust the filtering intensity for filtering the optimal analysis region to obtain an artifact correction region, and fuse all the artifact correction regions to obtain an upper respiratory tract enhanced image.

[0005] Further, the selection of the optimal analysis region for each local region includes: Optionally, record the upper respiratory tract image at a certain moment as the reference image, perform region matching on each local region of the reference image and the local regions of the upper respiratory tract images other than the reference image in the breathing cycle, and form a region set by each local region of the reference image and its corresponding local regions matched in the other upper respiratory tract images; Optionally, record a local region as the example region. After obtaining the average gray value of the pixel points in the example region, calculate the average of the absolute values of the differences between the average values of the example region and its adjacent local regions as the adjacent gray difference of the example region; perform corner detection on the example region, and obtain the richness of details of the example region according to the total number of corner points in the example region and the adjacent gray difference; Select the local region corresponding to the maximum richness of details from each region set as the optimal analysis region for all local regions within each region set.

[0006] Further, the obtaining of the edge detail loss degree of each optimal analysis region includes: Obtain the edge blurriness according to the length and direction distribution concentration degree of the edge lines in each optimal analysis region; Calculate the product of the edge blurriness and the richness of details of the corresponding optimal analysis region of each local region adjacent to each optimal analysis region, and take the average of all the products as the edge detail annoyance degree of each optimal analysis region; obtain the edge compactness according to the position distribution of the pixel points on the edge lines in each optimal analysis region; Obtain the edge detail loss degree of each optimal analysis region according to the edge detail annoyance degree and the edge compactness. The edge detail annoyance degree and the edge detail loss degree are in a positive correlation relationship, and the edge compactness and the edge detail loss degree are in a negative correlation relationship.

[0007] Further, the obtaining of the edge blurriness includes: Perform linear fitting on the pixel points on each edge line in the optimal analysis region to obtain a fitting line; according to the included angle between each edge line in the optimal analysis region and the fitting line corresponding to each other edge line, and the difference in the average slope of the pixel points on the corresponding two edge lines, obtain the local concentration degree of the corresponding two edge lines; For each edge line within the optimal analysis region, record the mean of the local concentration degrees with respect to the remaining edge lines as the direction concentration degree of each edge line; calculate the product of the number of pixel points on each edge line within the optimal analysis region and the direction concentration degree, and take the mean of all products as the edge blur degree of the optimal analysis region.

[0008] Furthermore, the edge tightness is equal to the mean distance between pairwise pixel points on all edge lines within each optimal analysis region.

[0009] Furthermore, the obtaining of the fusion adaptation factor for each optimal analysis region includes: Obtain the respiration signal of the patient during the respiratory cycle, and based on the respiration symmetry and the fluctuation of the respiration signal of the patient during the respiratory cycle, obtain the pattern complexity; Perform a negative correlation mapping on the product of the detail richness and the edge detail loss degree, and perform a normalization process on the product of the mapping result and the pattern complexity to obtain the fusion adaptation factor for each optimal analysis region.

[0010] Furthermore, the obtaining of the pattern complexity includes: Obtain the upper and lower envelope lines of the respiration signal; calculate the cumulative sum of the absolute values of the differences in the slopes of the corresponding data points of all data points of the respiration signal on the two envelope lines, and record it as the signal fluctuation degree; convert each envelope line into frequency domain data, and calculate the phase difference between the corresponding frequency domain data of the two envelope lines of the respiration signal, and record it as the respiration asynchrony degree; Take the product of the signal fluctuation degree and the respiration asynchrony degree as the pattern complexity.

[0011] Furthermore, the using of the fusion adaptation factor to adjust the filtering intensity for filtering the optimal analysis region to obtain the artifact correction region includes: Use the sum of the fusion adaptation factor and the constant 1 to weight the preset filtering standard deviation to obtain the adjusted standard deviation for each optimal analysis region; based on the adjusted standard deviation, perform a filtering process on each optimal analysis region using Gaussian filtering to obtain the artifact correction region corresponding to each optimal analysis region.

[0012] Furthermore, the edge line is a curve obtained by curve fitting of the edge pixel points detected for the optimal analysis region.

[0013] Furthermore, the number of local regions of the upper respiratory tract images at different moments within the respiratory cycle is equal.

[0014] The present invention has the following beneficial effects: First aspect: By matching the upper respiratory tract images within the respiratory cycle to align the images, a local area with a relatively rich level of details for each part of the upper respiratory tract is selected as the optimal analysis area for the corresponding part, thereby merging the same structural information and reducing the artifacts caused by breathing. Fusing the artifact correction area after filtering the optimal analysis area can better remove the respiratory artifacts caused by the respiratory cycle changes and avoid introducing excessive noise.

[0015] Second aspect: The length and direction concentration degree of the edge line can measure the edge blur degree of the optimal analysis area. The local area adjacent to each optimal analysis area corresponds to the detail richness and edge blur degree of the optimal analysis area, which directly reflects the edge information loss situation of each optimal analysis area during the area fusion process, and combines the tightness degree of the edge lines within the optimal analysis area that may indicate the presence of joints to obtain the edge detail loss degree. Considering that the breathing patterns of patients may vary due to individual differences, and the artifacts generated by the breathing movement in different parts of the upper respiratory tract are different, different optimal analysis areas are adjusted for targeted filtering according to the influence of the complexity of the breathing pattern, the richness of details, and the edge detail loss degree. When accurately removing artifacts, it can better retain the important details of different parts of the upper respiratory tract in the image, enhance the removal effect of respiratory artifacts in the optimal analysis area, and thus improve the quality of the enhanced upper respiratory tract image. Brief Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 It is a flowchart of the steps of a method for removing artifacts from upper respiratory tract images provided by an embodiment of the present invention; Figure 2 It is a flowchart of a method for obtaining the edge detail loss degree provided by an embodiment of the present invention; Figure 3 It is a flowchart of a method for obtaining the fusion adaptation factor provided by an embodiment of the present invention; Figure 4 It is a system structure diagram of a system for removing artifacts from upper respiratory tract images provided by an embodiment of the present invention; Figure 5 It is a schematic diagram of a computer device of a device for removing artifacts from upper respiratory tract images provided by an embodiment of the present invention. Detailed Embodiments

[0018] To further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following specifically describes, in conjunction with the accompanying drawings and preferred embodiments, a method for removing artifacts from upper respiratory tract images proposed according to the present invention, including its specific implementation manner, structure, features, and effects. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0020] The following specifically describes, in conjunction with the accompanying drawings, the specific solution of a method for removing artifacts from upper respiratory tract images provided by the present invention.

[0021] Embodiment 1: The present invention proposes a method for removing artifacts from upper respiratory tract images. Please refer to Figure 1 , which shows a flowchart of the steps of a method for removing artifacts from upper respiratory tract images provided by an embodiment of the present invention. The method includes: Step S1: Obtain upper respiratory tract images at each moment during the patient's respiratory cycle. The upper respiratory tract images include different local regions.

[0022] Scan the patient's upper respiratory tract using a computed tomography scanner to obtain upper respiratory tract images at each moment during a single respiratory cycle. To more accurately analyze the detailed information of the upper respiratory tract, each moment's upper respiratory tract image is segmented into multiple local regions. During the image acquisition process, the position and shape of the upper respiratory tract change due to the patient's breathing, causing the image to be blurred or distorted, resulting in artifacts in the upper respiratory tract image.

[0023] It should be noted that, to avoid subsequent matching errors, the number of local regions in the upper respiratory tract images at all moments during the respiratory cycle is equal.

[0024] In one implementation manner of the embodiment of the present invention, the data acquisition frequency of the computed tomography scanner is set to 30 frames per second.

[0025] In one implementation manner of the embodiment of the present invention, the superpixel segmentation algorithm is selected to divide the chest image into local regions. In other embodiments, a trained neural network can also be used to divide the chest image into multiple local regions.

[0026] Step S2: Match the local regions of the upper respiratory tract images during the respiratory cycle, and select the optimal analysis region for each local region from the matched local regions based on the richness of details in the local regions. The upper respiratory tract of a patient, such as the nasal cavity, larynx, etc., undergoes periodic displacement during breathing. Due to the complex structure and tissue morphological differences of the upper respiratory tract, different parts may show varying degrees of deformation during the breathing process, such as the expansion and contraction of soft tissues, the narrowing and dilation of the airway, etc., resulting in some parts of the upper respiratory tract image being blurred or distorted, forming artifacts. Traditional image processing methods cannot effectively eliminate the artifacts in the upper respiratory tract image, leading to the loss of details or incorrect representation in some areas. By matching the upper respiratory tract images within the breathing cycle to align the images, local regions with a relatively rich degree of detail in each part of the upper respiratory tract are selected as the optimal analysis regions for the corresponding parts, thereby merging the same structural information and reducing the artifacts caused by breathing.

[0027] Step S3: Perform edge detection on the optimal analysis regions to obtain edge lines. According to the length, direction distribution concentration degree, detail richness degree of the edge lines within the optimal analysis regions corresponding to the local regions adjacent to each optimal analysis region, and the tightness degree of the edge lines within each optimal analysis region, obtain the edge detail loss degree of each optimal analysis region.

[0028] After selecting the optimal analysis regions for all parts of the upper respiratory tract, perform filtering processing on them to remove breathing artifacts, and perform region fusion on the optimal analysis regions after artifact removal to obtain the complete upper respiratory tract image after artifact removal. Since the upper respiratory tract presents different detail information at different breathing phases, the edges of the optimal analysis regions may become blurred or lost during the region fusion process, affecting the clarity and accuracy of the final image. It is necessary to obtain the edge lines within the optimal analysis regions to analyze the degree of edge blurring. The upper respiratory tract undergoes regular deformation during breathing, making the artifacts generated during its breathing movement show a parallel arrangement characteristic, that is, the direction distribution is relatively concentrated, and the artifacts are relatively blurred; some structures of the upper respiratory tract show relatively detailed characteristics in the image, the edges of the airway region are clear and continuous, but the soft tissue region is vulnerable to breathing artifacts and noise due to strong homogeneity and lack of obvious boundaries, which will exacerbate edge blurring and make the edges of the soft tissue structure shorter. Therefore, the degree of edge blurring of the optimal analysis region can be measured by combining the length and direction concentration degree of the edge lines.

[0029] Taking an arbitrarily selected optimal analysis region as the target region as an example, during the fusion process of the optimal analysis regions, when the optimal analysis regions adjacent to the target region have rich details but blurred boundaries, these optimal analysis regions may affect the recognition of the boundary details of the target region, resulting in an unclear transition between the boundary of the target region and its adjacent optimal analysis regions, and causing greater loss of edge information in the target region during the region fusion process, thereby reducing the accuracy of the fusion result. The structure of the upper respiratory tract is complex, and the region may contain small joints. When the joint is located at the edge of the soft tissue, the edge effect of the soft tissue may cause the loss or blurring of the boundary details of the joint, affecting the accurate recognition of the regional structure; due to the small and complex structure of the airway joint, its edge usually presents more bends and changes, making the edge of the joint closer than that of the normal region, thus resulting in greater loss of edge information in the joint region during the region fusion process. Therefore, by comprehensively considering the degree of edge blurring and detail richness of the optimal analysis regions corresponding to the local regions adjacent to the target region, as well as the tightness of the edge lines within the target region, the degree of edge information loss in the target region during the region fusion process is analyzed to obtain the edge detail loss degree; and the greater the above factors, the greater the loss of edge information in the target region during the region fusion process.

[0030] It should be noted that the edge line is a curve obtained by curve fitting of the edge pixel points detected on the optimal analysis region. The least squares method is selected for curve fitting. Part of the edge of the optimal analysis region is the edge of its adjacent local region.

[0031] In an implementation manner of the embodiment of the present invention, the Canny operator is selected for edge detection, and the Sobel operator and the Laplacian operator can also be selected, etc.

[0032] Step S4: According to the complexity of the breathing pattern of the patient during the breathing cycle, the detail richness of each optimal analysis region, and the edge detail loss degree, obtain the fusion adaptation factor for each optimal analysis region; use the fusion adaptation factor to adjust the filtering intensity for filtering the optimal analysis region to obtain the artifact correction region, and fuse all the artifact correction regions to obtain the enhanced image of the upper respiratory tract.

[0033] Due to the differences in breathing patterns such as frequency, depth, and rhythm among different patients, the impact of patient breathing on the upper respiratory tract is different, that is, there are significant differences in artifacts in the upper respiratory tract images of different patients. Complex breathing patterns will exacerbate the movement of the patient's upper respiratory tract during image acquisition, generating obvious respiratory artifacts. To suppress high-frequency artifacts, a relatively large filtering intensity needs to be set; when the edge information loss of the optimal analysis region is greater and it has more abundant detailed information during the region fusion process, the high-detail region usually contains more valuable information. Strong filtering will cause detail loss, especially at the edge positions of the region. To retain the local detail features of the upper respiratory tract structure, a relatively small filtering intensity needs to be set so that when suppressing noise during the region fusion process of the optimal analysis region, the transmission of effective details is maximized, thereby ensuring the accuracy of region fusion. Therefore, considering the above three factors, the filtering intensity for filtering the optimal analysis region is adjusted to obtain a fusion adaptation factor. The larger the fusion adaptation factor of the optimal analysis region, the greater the required filtering intensity. Filtering using the fusion adaptation factor can better retain the important details of different parts of the upper respiratory tract in the image when accurately removing artifacts, enhancing the removal effect of respiratory artifacts in the artifact correction region, thereby realizing the enhancement processing of upper respiratory tract images. Each breathing cycle corresponds to an enhanced upper respiratory tract image.

[0034] In an embodiment of the present invention, an optical flow algorithm such as FlowNet is used to calculate the displacement field of the upper respiratory tract image during the breathing cycle, a continuous deformation field is generated by interpolating the timestamps of the optimal analysis regions, and each optimal analysis region is mapped to a reference coordinate system through the deformation field, and then Poisson fusion is used to fuse the optimal analysis regions to obtain an enhanced upper respiratory tract image. Among them, optical flow registration, deformation field interpolation, image fusion, etc. are all well-known technologies to those skilled in the art and will not be elaborated here.

[0035] In another embodiment of the present invention, a graph model is constructed with the optimal analysis regions as nodes and the adjacency relationship between regions as edges. The sum of the error term of optical flow feature alignment and the deformation smoothing term determined by the Demons algorithm is used as the optimization objective function, and graph cut or the Gauss-Newton method is used to solve the optimal stitching. The optimal analysis regions are stitched and fused according to the optimal stitching to obtain an enhanced upper respiratory tract image.

[0036] Preferably, in some possible implementation manners of the embodiments of the present invention, the method for selecting the optimal analysis region includes: taking the upper respiratory tract image at any moment as the reference image, performing region matching on each local region of the reference image with the local regions of the remaining upper respiratory tract images in the respiratory cycle except the reference image, and forming a region set by each local region of the reference image and its corresponding matched local regions in the remaining upper respiratory tract images; arbitrarily selecting a local region as the example region, after obtaining the average gray value of the pixel points in the example region, calculating the average value of the absolute values of the differences between the average values of the example region and its adjacent local regions as the adjacent gray difference of the example region; performing corner detection on the example region, and obtaining the detail richness of the example region according to the total number of corner points in the example region and the adjacent gray difference; selecting the local region corresponding to the maximum detail richness from each region set as the optimal analysis region for all local regions in each region set.

[0037] It should be noted that the gray difference between the example region and its adjacent local regions reflects the possibility of the existence of high-frequency information in the example region, and the total number of corner points in the example region presents the degree of tortuosity of the edge structure in the example region. The larger both are, the richer the details in the example region are. Therefore, both the total number of corner points in the example region and the adjacent gray difference are positively correlated with the detail richness. In the embodiments of the present invention, the product of the total number of corner points in the example region and the adjacent gray difference is used as the detail richness. To ensure the matching effect of the local regions of the reference image and the remaining upper respiratory tract images, the reference image should be selected as the upper respiratory tract image without respiratory artifacts. The method for obtaining the detail richness of all local regions and the example region is the same.

[0038] In one implementation manner of the embodiments of the present invention, a motion compensation algorithm is used to perform region matching on each local region of the reference image with the local regions of the remaining upper respiratory tract images in the respiratory cycle except the reference image.

[0039] In one implementation method of the embodiments of the present invention, the Shi-Tomasi corner detection algorithm is used for corner detection.

[0040] Preferably, in some possible implementation manners of the embodiments of the present invention, for the method for obtaining the edge detail loss degree, please refer to Figure 2 , which shows a flowchart of a method for obtaining the edge detail loss degree provided by an embodiment of the present invention. The method includes: Step S310: Obtain the edge blurriness according to the length and direction distribution concentration degree of the edge lines in each optimal analysis region.

[0041] Preferably, in some possible implementation manners of the embodiments of the present invention, the method for obtaining the edge blur degree includes: performing linear fitting on the pixel points on each edge line in the optimal analysis region to obtain a fitting line; obtaining the local concentration degree of two corresponding edge lines according to the included angle between the fitting lines corresponding to each edge line in the optimal analysis region and the remaining edge lines, and the difference between the mean values of the slopes of the pixel points on the two corresponding edge lines; taking the mean value of the local concentration degrees of each edge line in the optimal analysis region and the remaining edge lines as the direction concentration degree of each edge line; calculating the product of the number of pixel points on each edge line in the optimal analysis region and the direction concentration degree, and taking the mean value of all the products as the edge blur degree of the optimal analysis region.

[0042] It should be noted that the greater the possibility that the optimal analysis region with the known edge lines having more parallel directions and shorter lengths is a region with blurred boundaries or a soft tissue region, the greater the edge blur degree. When the included angle between each edge line and the fitting line corresponding to each of the remaining edge lines, and the difference between the mean values of the slopes of the pixel points on the edge line are both smaller, it indicates that the overall directions of the two corresponding edge lines are more parallel, the edge contours are closer, the two edge lines are more parallel, and the local concentration degree is greater. Therefore, both the included angle between the fitting lines corresponding to the two edge lines and the difference between the mean values of the slopes of the pixel points on the edge line are negatively correlated with the local concentration degree. In the embodiments of the present invention, the product of the included angle between each edge line and the fitting line corresponding to each of the remaining edge lines and the absolute value of the difference between the mean values of the slopes of the pixel points on the two corresponding edge lines is subjected to a negative correlation mapping to obtain the local concentration degree. By analyzing the parallelism of each edge line and the remaining edge points through the mean value of the local concentration degrees of each edge line and the remaining edge lines, the possibility of each edge line being affected by respiratory artifacts is measured. The edge line with a greater possibility of artifacts is more blurred. The length of the edge line is measured by the number of pixel points on the edge line. The shorter the length of the edge line, the greater the possibility that it is the edge of the soft tissue region, and the more blurred the edge line with a greater possibility of soft tissue. Considering the blur degrees of all the edge lines in the optimal analysis region, the edge blur degree of the optimal analysis region is obtained.

[0043] In an implementation method of the embodiments of the present invention, the least squares method is used to perform linear fitting on the pixel points on the edge line. The polynomial fitting method and the support vector machine and other methods can also be used for linear fitting.

[0044] In an implementation method of the embodiments of the present invention, the opposite number of the above product is used as the exponent of the exponential function with the natural constant as the base to implement the negative correlation mapping process of the product.

[0045] Step S320: Calculate the product of the edge blur degree and the detail richness of the optimal analysis region corresponding to each local region adjacent to each optimal analysis region, and take the mean of all products as the edge detail annoyance degree of each optimal analysis region; according to the position distribution of the pixel points on the edge line within each optimal analysis region, obtain the edge compactness.

[0046] Preferably, in some possible implementation manners of the embodiments of the present invention, the edge compactness is equal to the mean distance between every two pixel points on all the edge lines within each optimal analysis region.

[0047] It should be noted that the position distribution of the edge line within the optimal analysis region with a smaller mean distance is more compact; the edge detail annoyance degree represents the degree of interference of the edge information of each optimal analysis region by the local region corresponding to the optimal analysis region during the region fusion process.

[0048] Step S330: Obtain the edge detail loss degree of each optimal analysis region according to the edge detail annoyance degree and the edge compactness.

[0049] It should be noted that it is known that the edge information loss is relatively large in the region corresponding to the airway joint structure during the region fusion process. The greater the possibility that the optimal analysis region with a smaller edge compactness has an airway joint structure, the greater the edge information loss during the region fusion process. The greater the edge detail annoyance degree of the optimal analysis region, the greater the edge information loss caused by the influence of its surrounding local regions during the region fusion process. Therefore, there is a positive correlation between the edge detail annoyance degree and the edge detail loss degree, and a negative correlation between the edge compactness and the edge detail loss degree. In the embodiments of the present invention, the ratio of the edge detail annoyance degree to the edge compactness of each optimal analysis region is used as the edge detail loss degree.

[0050] Preferably, in some possible implementation manners of the embodiments of the present invention, for the method of obtaining the fusion adaptation factor, please refer to Figure 3 , which shows a flowchart of a method for obtaining a fusion adaptation factor provided by an embodiment of the present invention. The method includes: Step S410: Obtain the respiratory signal of the patient during the respiratory cycle, and obtain the pattern complexity according to the respiratory symmetry situation and the fluctuation situation of the respiratory signal of the patient during the respiratory cycle.

[0051] Use an abdominal belt pressure sensor to collect the respiratory signal of the patient during the respiratory cycle. The horizontal axis represents time, and the vertical axis represents the pressure generated by the abdomen during breathing. In an implementation manner of the embodiments of the present invention, the data acquisition frequency of the pressure sensor is set to 20 Hz.

[0052] Preferably, in some possible implementation manners of the embodiments of the present invention, the method for obtaining the pattern complexity includes: obtaining the upper and lower envelope lines of the respiratory signal; calculating the sum of the absolute values of the differences in the slopes of the corresponding data points of all the data points of the respiratory signal on the two envelope lines, which is denoted as the signal fluctuation degree; converting each envelope line into frequency-domain data, and calculating the phase difference between the frequency-domain data corresponding to the two envelope lines of the respiratory signal, which is denoted as the respiratory asynchrony degree; taking the product of the signal fluctuation degree and the respiratory asynchrony degree as the pattern complexity.

[0053] It should be noted that in a normal respiratory cycle, the rhythm of inhalation and exhalation is relatively consistent, indicating that the acceleration and deceleration processes of inhalation and exhalation are coordinated, keeping the depth fluctuation of the respiratory signal within a relatively constant range. The greater the depth fluctuation of the respiratory signal, the more diverse the respiratory changes and the more complex the respiratory pattern. The respiratory cycle includes the inhalation and exhalation phases, and the two phases are relatively balanced under normal breathing conditions, that is, the ratio of inhalation to exhalation is approximately equal; in the case where the durations of inhalation and exhalation are asymmetric, the respiratory changes will be more diverse, and more irregular rhythms will appear, making the respiratory pattern more complex. The signal fluctuation degree and the respiratory asynchrony degree reflect the fluctuation situation of the respiratory signal and the respiratory symmetry of the patient during the respiratory cycle in turn. When the signal fluctuation degree and the respiratory asynchrony degree are greater, the depth fluctuation of the respiratory signal is more obvious, the durations of inhalation and exhalation are more asymmetric, the respiratory pattern of the patient during the respiratory cycle is more complex, and the pattern complexity is greater.

[0054] It should be noted that the phase difference between the frequency-domain data corresponding to the two envelope lines of the respiratory signal is equal to the average of the phase differences of the components corresponding to all equal frequencies of the two envelope lines. The time of each data point of the respiratory signal is the same as the time of the corresponding data point on the envelope line.

[0055] In one implementation method of the embodiments of the present invention, the moving extreme value method is selected to extract the envelope line of the respiratory signal, or the Hilbert transform method can also be selected to ensure that the number of data points of the respiratory signal is equal to the number of data points of each of its envelope lines.

[0056] In one implementation method of the embodiments of the present invention, the Fourier transform is selected for frequency-domain conversion.

[0057] Step S420: Perform a negative correlation mapping on the product of the detail richness and the edge detail loss degree, and perform a normalization process on the product of the mapping result and the pattern complexity to obtain the fusion matching factor of each optimal analysis region.

[0058] In a specific implementation manner of the embodiments of the present invention, the fusion matching factor is expressed by the formula: In the formula, is the fusion matching factor of the a-th optimal analysis region; is the pattern complexity; is the richness of details of the a-th optimal analysis region; is the edge detail loss degree of the a-th optimal analysis region; Norm is the normalization function; exp is the exponential function with the natural constant as the base. In this implementation, the negative correlation mapping is achieved through the exp function. It should be noted that the more complex the breathing pattern, i.e., the larger W is, the more obvious the breathing artifacts will be, and the greater the filtering intensity is required to suppress the artifacts. Then is larger; the edge detail loss degree is larger and the richness of details is larger for the optimal analysis region containing more valuable information. In order to retain the local detail features of the upper respiratory tract structure, a smaller filtering intensity is required. Then is smaller.

[0059] Preferably, in some possible implementation manners of the embodiments of the present invention, the method for obtaining the artifact correction region includes: weighting the preset filtering standard deviation by using the sum value of the fusion adaptation factor and the constant 1 to obtain the adjusted standard deviation of each optimal analysis region; based on the adjusted standard deviation, performing filtering processing on each optimal analysis region by using Gaussian filtering to obtain the artifact correction region corresponding to each optimal analysis region. Among them, Gaussian filtering is a well-known technology to those skilled in the art and will not be elaborated here. It should be noted that the larger the fusion adaptation factor is, the greater the required filtering intensity is, and the filtering intensity is enhanced by increasing the filtering standard deviation.

[0060] In an implementation method of the embodiments of the present invention, the preset filtering standard deviation is set to 2.

[0061] In other embodiments of the present invention, the preset size of the filtering window is weighted by using the sum value of the fusion adaptation factor and the constant 1 to obtain the adjusted filtering window of each optimal analysis region; based on the adjusted filtering window, performing filtering processing on each optimal analysis region by using Gaussian filtering to obtain the artifact correction region corresponding to each optimal analysis region.

[0062] In an implementation method of the embodiments of the present invention, the preset size is set to 13.

[0063] So far, the present invention is completed.

[0064] Embodiment 2: The present invention proposes an artifact removal system for upper respiratory tract images. Please refer to Figure 4 , which shows the system structure diagram of an artifact removal system for upper respiratory tract images provided by an embodiment of the present invention. The system includes: A data acquisition module 510, configured to acquire upper respiratory tract images at each moment during the breathing cycle of a patient. The upper respiratory tract images include different local regions; The region screening module 520 is used to match local regions of the upper respiratory tract image within the respiratory cycle, and select the optimal analysis region for each local region from the matched local regions based on the richness of details in the local regions; The detail loss analysis module 530 is used to perform edge detection on the optimal analysis region to obtain an edge line, and obtain the edge detail loss degree of each optimal analysis region according to the length, direction distribution concentration degree and detail richness of the edge line in the optimal analysis region corresponding to the local region adjacent to each optimal analysis region, and the tightness of the edge line in each optimal analysis region; The image enhancement module 540 is used to obtain the fusion matching factor for each optimal analysis region according to the complexity of the breathing pattern of the patient in the respiratory cycle, the detail richness and the edge detail loss degree of each optimal analysis region; use the fusion matching factor to adjust the filtering intensity for filtering the optimal analysis region to obtain an artifact correction region, and fuse all the artifact correction regions to obtain an enhanced upper respiratory tract image.

[0065] It should be noted that: the device provided in the above embodiment is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules according to needs, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the functions described above. In addition, an artifact removal system for an upper respiratory tract image and an embodiment of an artifact removal method for an upper respiratory tract image provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0066] Embodiment 3: Figure 5 FIG. is a schematic diagram of a computer device of an artifact removal device for an upper respiratory tract image provided by an embodiment of the present invention. Exemplarily, as Figure 5 shown, the computer device includes: a memory 601, a processor 602, and a computer program 603 stored in the memory 601 and running on the processor 602. When the processor 602 executes the computer program 603, the computer device can execute any one of the artifact removal methods for an upper respiratory tract image introduced above.

[0067] In addition, an embodiment of the present application also protects a device, which may include a memory and a processor. The memory stores executable program code, and the processor is used to call and execute the executable program code to execute an artifact removal method for an upper respiratory tract image provided by an embodiment of the present application.

[0068] In this embodiment, the device can be divided into functional modules according to the above method examples. For example, each functional module can be corresponded, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware. It should be noted that the division of modules in this embodiment is illustrative, only a logical function division, and there can be other division methods in actual implementation.

[0069] It should be understood that the device provided in this embodiment is used to execute the above method for removing artifacts in upper respiratory tract images, so the same effects as the above implementation method can be achieved.

[0070] In the case of adopting an integrated unit, the device can include a processing module and a storage module. Among them, when the device is applied to a device, the processing module can be used to control and manage the actions of the device. The storage module can be used to support the device to execute mutual program codes, etc.

[0071] Among them, the processing module can be a processor or a controller, which can implement or execute various exemplary logical blocks, modules, and circuits included in the disclosure of this application. The processor can also be a combination that realizes computing functions, such as including a combination of one or more microprocessors, a combination of digital signal processing (DSP) and a microprocessor, etc. The storage module can be a memory.

[0072] It should be noted that the above sequence of embodiments of the present invention is only for description and does not represent the advantages or disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0073] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the key points of each embodiment are the differences from other embodiments.

[0074] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for removing artifacts from upper respiratory tract images, characterized in that: The method includes: Acquire an upper respiratory tract image at each moment in the patient's respiratory cycle, wherein the upper respiratory image includes different local areas; Matching local areas of the upper respiratory tract image during the respiratory cycle, and selecting the optimal analysis area of ​​each local area from the matched local areas based on the detail richness of the local areas; Perform edge detection on the optimal analysis area to obtain edge lines, and obtain the edge detail loss degree of each optimal analysis area according to the length, direction distribution concentration and detail richness of the edge lines in the optimal analysis area corresponding to the local area adjacent to each optimal analysis area, as well as the compactness of the edge lines in each optimal analysis area; According to the complexity of the patient's breathing pattern in the respiratory cycle, the detail richness of each optimal analysis area and the edge detail loss, a fusion adaptation factor of each optimal analysis area is obtained; the filter intensity of the optimal analysis area is adjusted by using the fusion adaptation factor to obtain an artifact correction area, and all artifact correction areas are fused to obtain an upper respiratory tract enhanced image; Methods for obtaining the detail richness of a local area include: Any local area is recorded as the sample area. After obtaining the grayscale value mean of the pixels in the sample area, the average of the absolute values ​​of the differences between the sample area and the local areas adjacent to it is calculated as the adjacent grayscale difference of the sample area. Corner point detection is performed on the sample area, and the detail richness of the sample area is obtained according to the total number of corner points in the sample area and the adjacent grayscale difference. Get the edge detail loss of each optimal analysis area, including: Obtain edge fuzziness according to the length and direction distribution concentration of edge lines in each optimal analysis area; Calculate the product of the edge fuzziness and detail richness of each local area adjacent to each optimal analysis area corresponding to the optimal analysis area, and take the average of all products as the edge detail annoyance of each optimal analysis area; obtain the edge density according to the position distribution of pixel points on the edge line in each optimal analysis area; Obtaining the edge detail loss degree of each optimal analysis area according to the edge detail annoyance and the edge compactness; Obtain the fusion fitness factor for each optimal analysis area, including: Acquire the patient's respiratory signal during the respiratory cycle, and acquire the pattern complexity according to the patient's respiratory symmetry and the fluctuation of the respiratory signal during the respiratory cycle; A negative correlation mapping is performed on the product of the detail richness and the edge detail loss, and the product of the mapping result and the pattern complexity is normalized to obtain a fusion adaptation factor for each optimal analysis area.

2. The method for removing artifacts from upper respiratory tract images according to claim 1, characterized in that: The selecting the optimal analysis area for each local area comprises: An upper respiratory tract image at any time is recorded as a reference image, and each local area of ​​the reference image is matched with the local areas of the remaining upper respiratory tract images except the reference image in the respiratory cycle, and each local area of ​​the reference image and its matching local areas in the remaining upper respiratory tract images form a region set; A local region corresponding to the largest detail richness is selected from each region set as the optimal analysis region for all local regions in each region set.

3. The method for removing artifacts from upper respiratory tract images according to claim 1, characterized in that: The step of obtaining edge blur comprises: Perform straight line fitting on the pixel points on each edge line in the optimal analysis area to obtain a fitting straight line; obtain the local concentration of the corresponding two edge lines according to the angle between each edge line in the optimal analysis area and the corresponding fitting straight line of each other edge line, and the difference in the mean slope of the pixel points on the two edge lines; The average of the local concentrations of each edge line in the optimal analysis area and the remaining edge lines is recorded as the directional concentration of each edge line; the product of the number of pixel points on each edge line in the optimal analysis area and the directional concentration is calculated, and the average of all products is taken as the edge blur of the optimal analysis area.

4. The method for removing artifacts from upper respiratory tract images according to claim 1, characterized in that: The edge density is equal to the average distance between two pixels on all edge lines in each optimal analysis area.

5. The method for removing artifacts from upper respiratory tract images according to claim 1, characterized in that: The acquisition mode complexity includes: Obtain the upper and lower envelopes of the respiratory signal; calculate the cumulative sum of the absolute values ​​of the slope differences of all data points of the respiratory signal on the two envelopes corresponding to the data points, which is recorded as the signal fluctuation degree; convert each envelope into frequency domain data, calculate the phase difference of the frequency domain data corresponding to the two envelopes of the respiratory signal, which is recorded as the respiratory asynchrony degree; The product of the signal fluctuation and the respiratory asynchrony is taken as the pattern complexity.

6. The method for removing artifacts from upper respiratory tract images according to claim 1, characterized in that: The step of using the fusion adaptation factor to adjust the filtering intensity of the optimal analysis area to obtain the artifact correction area includes: The preset filtering standard deviation is weighted by using the sum of the fusion adaptation factor and the constant 1 to obtain the adjusted standard deviation of each optimal analysis area; based on the adjusted standard deviation, each optimal analysis area is filtered using Gaussian filtering to obtain the artifact correction area corresponding to each optimal analysis area.

7. The method for removing artifacts from upper respiratory tract images according to claim 1, characterized in that: The edge line is a curve obtained by performing curve fitting on edge pixel points obtained by edge detection in the optimal analysis area.

8. The method for removing artifacts from upper respiratory tract images according to claim 1, characterized in that: The number of local areas of the upper respiratory tract image at different moments in the respiratory cycle is equal.

Images