A fusion processing method for multiple review images of a COVID-19 patient

Abstract

The application discloses a fusion processing method for multiple review images of COVID-19 patients, image processing is performed on each first examination CT image to screen and save the first examination CT image with a lesion, and a patient point coordinate threshold of the corresponding lesion is marked on the first examination CT image; pixel points of each review CT image and the patient point coordinate threshold of the corresponding first examination CT image in the first examination are sequentially subjected to pixel comparison to calculate the change of pixel values in the patient point coordinate threshold; the recovery condition of the original lesion is determined; pixel values of the pixel points in all patient point coordinate thresholds of the first examination CT image and the review CT image are adjusted and assigned, the multiple review CT images after the adjustment and assignment are compared with the matched first examination CT image to review whether new lesions appear and output new review pathological results; the application improves the review accuracy, obtains the specific recovery range of each lesion point, and can also determine whether new lesion points appear in the lung of the patient.

Description

Technical Field

[0001] This invention relates to the field of computer image processing technology, specifically to a method for fusing images from multiple follow-up examinations of COVID-19 patients. Background Technology

[0002] The main symptoms of COVID-19 are fever, dry cough, and fatigue. A small number of patients may also experience upper respiratory and digestive symptoms such as nasal congestion, runny nose, and diarrhea. Severe cases often develop respiratory distress after one week, and in severe cases, the condition rapidly progresses to acute respiratory distress syndrome, septic shock, uncorrectable metabolic acidosis, coagulation dysfunction, and multiple organ failure. It is noteworthy that severe and critically ill patients may experience low-grade fever or even no obvious fever during the course of the disease. Mild cases may only present with low-grade fever and mild fatigue, without signs of pneumonia.

[0003] Therefore, the most reliable way to detect COVID-19 is through lung CT imaging showing viral pneumonia, which allows for the detection of COVID-19 lesions in patients. Currently, the method for judging the results of follow-up examinations for COVID-19 patients is mostly to have doctors directly face the patient and compare the lesion data in multiple lung CT images to judge the development of lesions and determine the patient's degree of recovery. However, this method is slow. There are now computer image processing methods that automatically process lung CT images, fusion and comparing each follow-up lung CT image with historical lung CT images to determine the patient's lung recovery.

[0004] The existing computer image processing methods for determining recovery status do not have the function of determining whether new lesions have appeared in the CT images during follow-up examinations. Summary of the Invention

[0005] The purpose of this invention is to provide a fusion processing method for multiple follow-up images of COVID-19 patients, in order to solve the technical problem that existing computer image processing does not have the function of determining whether new lesions have appeared in CT images during follow-up examinations.

[0006] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution:

[0007] A method for fusing images from multiple follow-up examinations of COVID-19 patients, comprising the following steps:

[0008] Step 100: Obtain multiple initial CT images of each patient's first examination, perform image processing on each initial CT image to filter and save the initial CT images with lesions, and mark the lesion point coordinate threshold on the initial CT image.

[0009] Step 200: Obtain multiple follow-up CT images of the follow-up patient and filter out the follow-up CT images with lesion point coordinate thresholds. Compare the pixels of each follow-up CT image with the corresponding first CT image in the first examination pixel by pixel to calculate the change of pixel value within the lesion point coordinate threshold.

[0010] Step 300: Determine the recovery status of the original lesion during the follow-up examination based on the changes in pixel values ​​within the lesion point coordinate threshold of each of the follow-up CT images;

[0011] Step 400: Adjust and assign pixel values ​​of all lesion point coordinates within the threshold of the initial CT image and the follow-up CT image so that the pixel values ​​within the range of the pixel point are restored to the normal state. Compare the adjusted follow-up CT images with the matching initial CT image to check whether new lesions have appeared and output new follow-up pathology results.

[0012] As a preferred embodiment of the present invention, in step 100, the method for marking and delineating the coordinate threshold of lesion points on each initial CT image is as follows:

[0013] Step 101: Image uniform preprocessing. Denoise the two lung lobes in each of the first CT images and correct the image gray level non-uniformity to obtain a noise-free preprocessed image with uniform gray level distribution.

[0014] Step 102: Calculate the average grayscale value of the preprocessed images of the two lung lobes in each of the initial CT images, based on the difference in the average grayscale value of the preprocessed images of the two lung lobes.

[0015] Step 103: Image filling. Establish a coordinate system on the preprocessed images of the two lung lobes, calculate the size of the two lung lobes respectively, and fill the missing position of the smaller lung image with the larger lung image through mirror filling to obtain two lung images of the same size.

[0016] Step 104: Image segmentation. The two lung images of the same size are sequentially segmented into multiple lung slice layers. The total pixel value of the lung slice layers with the same sequence number in the two lung images is compared to identify areas with large-area nodular pneumonia.

[0017] Step 105: Image feature extraction. Based on the calculated size of the two lung halves, a standard image without lesions is obtained. The standard image of the two lung halves is then divided into lung slice layers. The lung slice layer of each lung halves is compared with the lung slice layer of the same sequence number in the standard image to identify the feature regions and number of small nodules that produce differences and are randomly distributed.

[0018] In a preferred embodiment of the present invention, in step 103, when the difference in the average grayscale value of the preprocessed images of the two lung lobes is greater than a set threshold, a coordinate system is established on the preprocessed images of the two lung lobes. By mirroring the excess portion of one of the larger lung images to fill the missing position of the smaller lung image, the lung images of the two lung lobes are made to be the same size, which facilitates the identification of areas with large nodular pneumonia. The specific implementation steps are as follows:

[0019] A two-dimensional plane coordinate system is established with the midpoint of the preprocessed images of the two lungs as the origin. The two-dimensional plane coordinate systems of the preprocessed images of the two lungs share a common Y-axis, and the X-axis directions of the two two-dimensional plane coordinate systems are opposite.

[0020] Divide the same region of the two lungs into a single longitudinal order and a single transverse order, and determine the coordinate values ​​corresponding to the edges of the same region.

[0021] Using the Y-axis, which is shared by the two-dimensional plane coordinate system, as the axis of symmetry, the excess parts of the two lungs are mirrored and filled in at the corresponding coordinate positions;

[0022] The missing areas of the two lung lobes were filled sequentially along a uniform vertical and horizontal direction to obtain images of two lung lobes of the same size.

[0023] In a preferred embodiment of the present invention, in step 100, each patient has a code for identification, and after image processing of each of the initial CT images, the initial CT images containing the patient's location are selected and retained, and the capture sequence number of the initial CT image is determined.

[0024] As a preferred embodiment of the present invention, in step 200, multiple follow-up CT images are acquired sequentially at the same angle as the initial CT image, and the multiple follow-up CT images are sequentially selected according to the captured sequence number of the retained initial CT image to obtain multiple follow-up CT images with the same size and the same angle as the initial CT image.

[0025] As a preferred embodiment of the present invention, in step 200, a two-dimensional coordinate system identical to that of the matching initial CT image is established within each of the follow-up CT images, and the same lesion point coordinate threshold is selected within the follow-up CT images. The method for comparing the follow-up CT images with the initial CT images with matching imaging sequences is as follows:

[0026] Determine the lesion point coordinate threshold that is the same in the follow-up CT image and the initial CT image with the matching shooting sequence number, obtain the pixel value of each pixel of the lesion point coordinate threshold in sequence, and compare the pixel values ​​of each pixel of the lesion point coordinate threshold.

[0027] The number of pixels whose pixel values ​​in the follow-up CT image are less than those in the initial CT image is calculated, and this number of pixels is compared with a judgment error value. If the number of pixels is greater than the error value, it is considered that the lesion area in the follow-up CT image is decreasing; otherwise, it is considered that the lesion area in the follow-up CT image remains unchanged.

[0028] In a preferred embodiment of the present invention, in step 300, the pixel values ​​of all lesion point coordinates within the threshold of each of the follow-up CT images are compared with those of the matching initial CT images. The number of recovered lesion areas in which the pixel values ​​of the lesion point coordinates within the threshold of the follow-up CT images are less than those in the initial CT images is counted. The number of such recovered lesion areas is then compared with the number of all lesion point coordinates corresponding to the threshold, so as to roughly determine the recovery status of COVID-19 patients.

[0029] As a preferred embodiment of the present invention, the number of pixels in each recovered lesion region of each follow-up CT image whose pixel value is less than that of the first CT image is counted, and the number of pixels in the recovered lesion region whose pixel value is less than that of the first CT image is compared with the total number of pixels at the lesion point coordinate threshold, so as to accurately determine the specific degree of recovery of the COVID-19 patient.

[0030] As a preferred embodiment of the present invention, in step 300, the pixel value of the pixel points within the lesion point coordinate threshold of the initial CT image and the follow-up CT image is 255. Adjusting the pixel value of each pixel point within the lesion point coordinate threshold on the initial CT image to 0 restores the pixel value within that pixel point range to a normal state, thereby obtaining a healthy state image after the original lesion has been eliminated.

[0031] In a preferred embodiment of the present invention, the initial CT image after the removal of the original lesion is used as the reference image. The follow-up CT images of each subsequent examination are compared with the initial CT image of the corresponding imaging sequence number to determine whether new lesions have appeared. The steps for determining new lesions are as follows:

[0032] The lung slice layer of each lung lobe in the follow-up CT image is compared with the lung slice layer of the same sequence number in the corresponding first CT image. The lung slice layer of each lung lobe is used as a comparison area. The lung slice layer of each lung lobe is compared with the lung slice layer of the same sequence number in the corresponding standard image in the same traversal direction to see if the pixel value of the same pixel point is the same.

[0033] Mark the pixels in the lung slice layer that show differences for each lung lobe;

[0034] The two adjacent lung slice layers in each lung lobe are stitched together to determine whether the pixels that differ between the two adjacent lung slice layers can be stitched together.

[0035] The size of the stitched pixels is calculated, and a comparison is made with the size and number of small nodule features within the error range to identify new lesions.

[0036] Compared with the prior art, the present invention has the following advantages:

[0037] This invention can not only determine the number of lesions that are tending to recover, but also specifically calculate the recovery range of each lesion that is tending to recover, thereby improving the accuracy of follow-up examinations, obtaining the specific recovery range of each lesion, and also determining whether new lesions have appeared in the patient's lungs. Attached Figure Description

[0038] To more clearly illustrate the embodiments of the present invention or the technical solutions in 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 merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0039] Figure 1 This is a flowchart illustrating the image fusion processing method provided in an embodiment of the present invention. Detailed Implementation

[0040] 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.

[0041] like Figure 1 As shown, this invention provides a fusion processing method for multiple follow-up images of COVID-19 patients. This embodiment uses image fusion processing to determine the state changes of the original lesions corresponding to the first examination during multiple follow-up examinations, and can also determine whether new lesions have appeared during the follow-up examinations. The specific implementation method includes the following steps:

[0042] Step 100: Obtain multiple initial CT images of each patient's first examination, establish a two-dimensional coordinate system for each initial CT image, perform image processing on each initial CT image to filter out initial CT images with lesions, and mark the lesion point coordinate threshold on the initial CT image.

[0043] In step 100, the method for marking and defining the coordinate threshold of lesion points on each initial CT image is as follows:

[0044] Step 101: Image uniform preprocessing. Denoise the two lung lobe in the CT image and correct the image gray level non-uniformity to obtain a noise-free preprocessed image with uniform gray level distribution.

[0045] Noise reduction and grayscale uniformity can be achieved using existing image processing methods. A special instruction is required: the grayscale value is taken as the median value for the CT image film grayscale. Since most lung CT images are binary grayscale images, the infiltrated areas of the lung are black areas with a pixel value of 0, while the blood vessels and trachea in the lung are white areas with a pixel value of 255. Therefore, this embodiment sets the grayscale of the CT image film to 123, which allows for relatively accurate identification of both black and white areas, facilitating the identification of nodule areas and the calculation of nodule size.

[0046] Step 102: Calculate the average grayscale value of the preprocessed images of the two lung halves. Based on the difference in the average grayscale value of the preprocessed images of the two lung halves, perform image filling or image feature extraction operations on the preprocessed images of the two lung halves respectively.

[0047] In step 102, the specific implementation method for selecting image inpainting or image feature extraction operations for pneumonia identification is as follows:

[0048] If the difference in the average grayscale value of the preprocessed images of the two lungs is greater than a set threshold, then pneumonia abnormality is found in one of the lung flaps. The location of the pneumonia abnormality and the size of the pneumonia lesion are determined by image filling.

[0049] If the difference in the average grayscale value of the preprocessed images of the two lungs is less than or equal to a set threshold, then both lungs show pneumonia abnormalities or there are no lesions in either lung. In this case, the preprocessed images of the two lungs are compared with the standard images by means of image feature extraction.

[0050] It should be noted that hepatitis is generally classified into large nodules (>1cm) and randomly distributed small nodules (<2mm in diameter). Large nodules (>1cm) and cavities often indicate infectious causes, most likely septic embolism, bacterial lung abscess, fungal and mycobacterial infections. Small nodules (<2mm in diameter) are highly suggestive of miliary tuberculosis, nontuberculous mycobacterial or fungal infections (histomalaciae, coccidioidomycosis, cryptococcosis).

[0051] Therefore, this implementation method, based on existing hepatitis classifications, identifies large and small nodules in pneumonia through image processing of CT images. It is well known that small nodular pneumonia is generally distributed in both lungs, while large nodular pneumonia is generally distributed in one lung. Therefore, comparing small nodular pneumonia cannot be done by simply comparing the grayscale of CT images from both lungs. Instead, it is necessary to compare the nodules with standard images of roughly the same size, and determine whether pneumonia is present by the number of nodules detected, within a certain margin of error.

[0052] The specific identification process for large nodules is as follows: 103.

[0053] Step 103: Image filling. Calculate the size of the two lung lobes separately. Establish a coordinate system on the preprocessed images of the two lung lobes based on the size difference. Fill the missing position of the smaller lung image with the larger lung image through mirror filling to obtain two lung images of the same size.

[0054] The rationale for image filling is as follows: since lung CT images may contain multiple sets, the sizes of the two lung lobes in these sets are not entirely identical. For CT images containing nodules, if the size difference between the two lung lobes exceeds a set value, image filling is required on the smaller lung lobe until the two lung CT images are the same size, facilitating the determination of the size and location of the larger nodule. For CT images where the two lungs are roughly the same size, step 103 can be omitted, and the process can proceed directly to step 105 for subsequent image processing.

[0055] In processing two CT images with significant differences in lung size, in step 103, when the difference in the average grayscale value of the preprocessed images of the two lung lobes exceeds a set threshold, a coordinate system is established on the preprocessed images of the two lung lobes. The excess portion of one of the larger lung images is mirrored and filled into the missing area of ​​the smaller lung image, thus making the lung images of the two lobes the same size. This facilitates the identification of areas with large nodular pneumonia. The specific implementation steps are as follows:

[0056] 1. Establish a two-dimensional plane coordinate system with the midpoint of the preprocessed images of the two lung lobes as the origin. The two-dimensional plane coordinate systems of the preprocessed images of the two lung lobes share a common Y-axis, and the X-axis directions of the two two-dimensional plane coordinate systems are opposite.

[0057] The method for determining the position of the origin of a two-dimensional plane coordinate system is as follows:

[0058] (1) Obtain a line of image units with a fixed width of the preprocessed image of the two lungs, divide the film area between two image units at the same location of the two lungs, and divide the film area with the same width as the image unit.

[0059] (2) Calculate the number of pixels in each film area and set the origin area as the film area with the smallest number of pixels;

[0060] (3) Using the center position of the origin-defined area as the reference point, establish the Y-axis of two two-dimensional plane coordinate systems;

[0061] (4) Traverse the preprocessed images of the two lungs from bottom to top, determine the pixels representing the edge of the lung in the preprocessed image of the single lung, and establish the X-axis of two two-dimensional plane coordinate systems in different directions, with the first pixel representing the edge of the lung as the reference point.

[0062] (5) The intersection of the Y-axis and the X-axis is taken as the origin of the two two-dimensional plane coordinate systems, and the lung images of the two lungs are located in the first quadrant of the corresponding two-dimensional plane coordinate system.

[0063] 2. Divide the same region of the two lungs into a single longitudinal sequence and a single transverse sequence, and determine the coordinate values ​​corresponding to the edges of the same region.

[0064] Third, using the Y-axis, which is shared by the two-dimensional plane coordinate system, as the axis of symmetry, the excess parts of the two lungs are mirrored and filled in at the corresponding coordinate positions.

[0065] Fourth, fill the missing areas of the two lung lobes in a unified vertical and horizontal direction to obtain images of two lung lobes of the same size.

[0066] In this embodiment, the image filling is achieved by copying the protruding part of the CT image of the larger lung relative to the CT image of the smaller lung, and mirroring and pasting it onto the CT image of the smaller lung using the Y-axis shared by the two-dimensional plane coordinate system as the axis of symmetry, thereby obtaining two CT images of roughly the same lung size.

[0067] In this way, the CT images are divided into different lung slice layers from top to bottom. By comparing the pixel differences of the same layer in two lung CT images, the size and distribution of nodules can be determined. The specific operation for comparing and identifying the location of pneumonia nodules is as follows: step 104.

[0068] Step 104: Image segmentation. Two lung images of the same size are segmented into multiple lung slice layers in sequence. The total pixel value of the lung slice layers with the same sequence number in the two lung images is compared to identify areas with large-area nodular pneumonia.

[0069] In step 104, two lung images of the same size are divided into multiple lung slice layers in order from top to bottom or from bottom to top. The total pixel value of the lung slice layers with the same number in the two lung images is compared to determine the difference in the total pixel value of the two lung slice layers with the same number. If the difference in the total pixel value is less than or equal to a set threshold, the two lung slice layers are indistinguishable; if the difference in the total pixel value is greater than the set threshold, a lesion point appears in one of the lung slice layers.

[0070] In this implementation, the filled image grayscale values ​​are omitted, and only the pixels that are different from the pixels in the two lung slice layers with the same sequence number and the corresponding pixel range are obtained.

[0071] Therefore, the pixel difference between two lung slice layers with the same serial number that is greater than a set threshold is further compared to determine the pixel points with different pixel values. After comparing all lung slice layers of the two lungs, the pixel coordinate values ​​of two adjacent lung slice layers are counted to determine the lesion points connected in the two adjacent lung slice layers. The nodule type and location of pneumonia are calculated by calculating the size of the lesion points.

[0072] Based on the above statement, for two CT images with similar lung sizes, step 103 can be omitted, and the process can proceed directly to step 105 for subsequent image processing. That is, step 105 is the identification method step for small nodular pneumonia.

[0073] Step 105: Image feature extraction. Based on the calculated size of the two lung halves, a standard image without lesions is obtained. The standard image of the two lung halves is then divided into lung slice layers. The lung slice layer of each lung halves is compared with the lung slice layer of the same sequence number in the standard image to identify the feature regions and number of small nodules that produce differences and are randomly distributed.

[0074] In step 105, the steps for obtaining standard images without lesions based on the calculated size of the two lung flaps are as follows: obtain standard images without lesions corresponding to each lung size based on the calculated size of the two lung flaps; and select unique standard images based on the patient's gender and age as the second search criteria.

[0075] The steps to perform feature comparison between lung slice layers of each lung lobe and lung slice layers of the same sequence number in the corresponding standard image are as follows:

[0076] Using the lung slice layer of each lung lobe as a comparison area, the lung slice layer of each lung lobe and the lung slice layer of the corresponding standard image with the same sequence number are compared in the same traversal direction to see if the pixel value of the same pixel is the same.

[0077] Mark the pixels in the lung slice layer that show differences for each lung lobe;

[0078] By stitching together two adjacent lung slice layers in each lung lobe, it is determined whether the pixels that differ between the two adjacent lung slice layers can be stitched together.

[0079] The size of the stitched pixels is calculated, and the size and number are compared with the features of the small nodules within the error range to identify small nodular pneumonia.

[0080] As an innovation of this embodiment, this embodiment uses two different methods to identify small nodular pneumonia and large nodular pneumonia respectively. The method for identifying large nodular pneumonia can be summarized as: directly comparing the two lungs of an individual. Because the image filling method can more easily and accurately determine the nodular area and size between the two lungs, it further reduces the identification error.

[0081] Regarding the identification of pneumonia in small nodules, since small nodules are usually distributed in both lungs, it is difficult to determine whether an individual has pneumonia and the location of the pneumonia by comparing the two lungs. Therefore, for the identification of this nodule, this embodiment uses the captured CT image to compare with the existing standardized CT image of the same size. Since there are generally many small nodules, the number of small nodules identified is compared with the number of identified nodules within a set error range to determine whether the individual has pneumonia.

[0082] This implementation method is based on two types of pneumonia: large nodules and small nodules. By calculating the average difference between the CT images of the two lungs, different methods are selected for image processing and pneumonia identification. These methods include direct comparison of CT images of the two lungs and indirect comparison of CT images of a single lung with standard images, which can improve the identification accuracy of the two types of nodular pneumonia.

[0083] Step 200: Obtain multiple follow-up CT images of the patient, filter out the follow-up CT images and lesion point coordinate thresholds that are the same as each initial CT image of the patient's first examination, and compare the pixels of each follow-up CT image with the corresponding initial CT image in the first examination pixel by pixel to calculate the change of pixel value within the lesion point coordinate threshold.

[0084] It should be noted that in step 100, each patient has a code for identification, and after image processing of each initial CT image, the initial CT images containing the patient's location are selected and retained, and the capture sequence number of the initial CT image is determined.

[0085] In step 200, multiple follow-up CT images are acquired sequentially at the same angle as the initial CT image, and the multiple follow-up CT images are sequentially selected according to the capture sequence number of the retained initial CT image to obtain multiple follow-up CT images with the same size and angle as the initial CT image.

[0086] Therefore, comparing the follow-up CT images with the initial CT images taken at the same sequence number for each patient is necessary to ensure that the lung size, lesion status, and lesion pixel values ​​are the same in the two CT images, thereby improving the accuracy of subsequent judgments on changes in the original lesions and the presence of new lesions.

[0087] The method for establishing a two-dimensional coordinate system identical to the matching initial CT image within each follow-up CT image, and selecting the same lesion point coordinate threshold within the follow-up CT images, and comparing the follow-up CT images with the matching initial CT images is as follows:

[0088] (1) Determine the same lesion point coordinate threshold in the follow-up CT image and the first CT image with the matching shooting sequence number, obtain the pixel value of each pixel of the lesion point coordinate threshold in sequence, and compare the pixel value of each pixel of the lesion point coordinate threshold.

[0089] (2) Calculate the number of pixels whose pixel value of the re-examination CT image is less than that of the first examination CT image, and compare the number of pixels with the judgment error value. If the number of pixels is greater than the error value, it is considered that the lesion area of ​​the re-examination CT image is decreasing; otherwise, it is considered that the lesion area of ​​the re-examination CT image has not changed.

[0090] If the number of pixels in the follow-up CT image that are less than the pixel value of the initial CT image is 0, or the number of pixels is less than the judgment error value, it means that the lesion point corresponding to the coordinate threshold of the lesion point has not changed, or may even be getting bigger. As for how to judge the increase, it will be further judged in step 400.

[0091] If the number of pixels in the follow-up CT image with a pixel value lower than that in the initial CT image is greater than the judgment error value, it means that the lesion area corresponding to the coordinate threshold of that lesion point is decreasing. The specific reduction in the lesion area is further calculated in step 300.

[0092] Step 300: Determine the recovery status of the original lesion during the follow-up examination based on the change of pixel values ​​within the coordinate threshold of the lesion point in each of the follow-up CT images.

[0093] By comparing the pixel values ​​of all lesion point coordinates within the threshold of each follow-up CT image with the matching initial CT image, the recovered lesion areas where the pixel values ​​of the lesion point coordinates within the threshold of the follow-up CT image are less than those of the lesion point coordinates within the threshold of the initial CT image are counted. The number of recovered lesion areas corresponding to the threshold of all lesion point coordinates is compared to roughly determine the recovery status of COVID-19 patients.

[0094] The number of pixels in each recovered lesion region of each follow-up CT image whose pixel value is less than that of the initial CT image is counted. The number of pixels in the recovered lesion region whose pixel value is less than that of the initial CT image is compared with the total number of pixels at the lesion point coordinate threshold to accurately determine the specific degree of recovery of COVID-19 patients.

[0095] In step 300, the specific criteria for determining the lesion points that have recovered to normal are as follows: the pixel value of the pixel points within the coordinate threshold of the lesion points in the initial CT image and the follow-up CT image is 255. Adjusting the pixel value of each pixel point within the coordinate threshold of the lesion points in the initial CT image to 0 will restore the pixel value within that pixel point range to a normal state, thus obtaining a healthy image after the original lesion has been eliminated.

[0096] Similarly, for COVID-19, there are multiple lesions in the lungs. Therefore, this implementation method first calculates the number of lesion coordinate thresholds in the lesion area that has shrunk compared to the initial CT image to roughly determine the recovery status of COVID-19 patients. For example, if there are 5 lesions in the initial CT image and 3 lesions in the follow-up CT image that have shrunk, it can be determined that the patient is gradually recovering.

[0097] This embodiment further determines the recovery status by calculating the reduction size of the lesion point coordinate threshold of the lesion area that has shrunk compared with the initial CT image. When the reduction size of the lesion point in three lesion areas is 30%, 20%, and 50%, the recovery status of the lesion point in each lesion area can be determined in turn.

[0098] Therefore, the image processing method of this embodiment can not only determine the number of lesions that tend to recover, but also specifically calculate the recovery range of each lesion that tends to recover, thereby improving the accuracy of the follow-up examination and obtaining the specific recovery range of each lesion.

[0099] Step 400: Adjust and assign pixel values ​​to all lesion points within the coordinate threshold of the initial CT image and the follow-up CT image so that the pixel values ​​within the range of the pixel point are restored to the normal state. Compare the multiple follow-up CT images after adjustment and assignment with the matching initial CT image to check whether new lesions have appeared and output new follow-up pathology results.

[0100] The specific procedure for comparing follow-up CT images with the initial CT images of the corresponding sequence number and with the same lung size and pixel count is as follows: using the initial CT image after the original lesion has been eliminated as the baseline image, the follow-up CT images of each follow-up examination are compared with the initial CT images of the corresponding sequence number to determine whether new lesions have appeared.

[0101] As an advantage of this embodiment, the image processing method for follow-up CT images is not only used to determine the recovery status of the original lesion, but also to determine whether new lesions have appeared in the patient's lungs. When determining whether new lesions have appeared in the follow-up CT images, this embodiment first adjusts the pixel values ​​of all pixels within the coordinate threshold of the lesion points in the initial CT image and the follow-up CT image to 0, that is, first eliminates the original lesion. At this time, the initial CT image can be used as the reference image for determining new lesions. There are no lesions in this image, but if there are pixels in the follow-up CT image that are different from the reference image, then new lesions will appear.

[0102] The steps to identify new lesions are as follows:

[0103] The lung slice layer of each lung lobe in the follow-up CT image is compared with the lung slice layer of the same sequence number in the corresponding first CT image. The lung slice layer of each lung lobe is used as a comparison area. The lung slice layer of each lung lobe is compared with the lung slice layer of the same sequence number in the corresponding standard image in the same traversal direction to see if the pixel value of the same pixel point is the same.

[0104] Mark the pixels in the lung slice layer that show differences for each lung lobe;

[0105] By stitching together two adjacent lung slice layers in each lung lobe, it is determined whether the pixels that differ between the two adjacent lung slice layers can be stitched together.

[0106] The size of the stitched pixels is calculated, and new lesions are identified by comparing the size and number with the features of small nodules within the error range.

[0107] The lung slices of each lung lobe in the upcoming follow-up CT images are compared with the lung slices of the same number in the baseline image to identify new lesions that produce differences and are randomly distributed.

[0108] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. A method for fusing images from multiple follow-up examinations of COVID-19 patients, characterized in that, Includes the following steps: Step 100: Obtain multiple initial CT images of each patient's first examination, perform image processing on each initial CT image to filter and save the initial CT images with lesions, and mark the lesion point coordinate threshold on the initial CT image. Step 200: Obtain multiple follow-up CT images of the follow-up patient and filter out the follow-up CT images with lesion point coordinate thresholds. Compare the pixels of each follow-up CT image with the corresponding first CT image in the first examination pixel by pixel to calculate the change of pixel value within the lesion point coordinate threshold. Step 300: Determine the recovery status of the original lesion during the follow-up examination based on the changes in pixel values ​​within the lesion point coordinate threshold of each of the follow-up CT images; Step 400: Adjust and assign pixel values ​​of all lesion point coordinates within the threshold of the initial CT image and the follow-up CT image so that the pixel values ​​within the range of the pixel point are restored to the normal state. Compare the multiple follow-up CT images after adjustment and assignment with the matching initial CT image to check whether new lesions have appeared and output new follow-up pathology results. In step 100, the method for marking and defining the coordinate threshold of lesion points on each initial CT image is as follows: Step 101: Image uniform preprocessing. Denoise the two lung lobes in each of the first CT images and correct the image gray level non-uniformity to obtain a noise-free preprocessed image with uniform gray level distribution. Step 102: Calculate the average grayscale value of the preprocessed images of the two lung lobes in each of the initial CT images, based on the difference in the average grayscale value of the preprocessed images of the two lung lobes. In step 102, the specific implementation method for selecting image inpainting or image feature extraction operations for pneumonia identification is as follows: If the difference in the average grayscale value of the preprocessed images of the two lungs is greater than a set threshold, the location of the pneumonia abnormality and the size of the pneumonia lesion are determined by the image filling method in step 103. If the difference in the average grayscale value of the preprocessed images of the two lungs is less than or equal to a set threshold, the preprocessed images of the two lungs are compared with the standard image by the image feature extraction method in step 105. Step 103: Image filling. Establish a coordinate system on the preprocessed images of the two lung lobes, calculate the size of the two lung lobes respectively, and fill the missing position of the smaller lung image with the larger lung image through mirror filling to obtain two lung images of the same size. Step 104: Image segmentation. The two lung images of the same size are sequentially segmented into multiple lung slice layers. The total pixel value of the lung slice layers with the same sequence number in the two lung images is compared to identify areas with large-area nodular pneumonia. Step 105: Image feature extraction. Based on the calculated size of the two lung halves, a standard image without lesions is obtained. The standard image of the two lung halves is then divided into lung slice layers. The lung slice layer of each lung halves is compared with the lung slice layer of the same sequence number in the standard image to identify the feature regions and number of small nodules that produce differences and are randomly distributed.

2. The method for fusion processing of multiple follow-up images of COVID-19 patients according to claim 1, characterized in that: In step 103, when the difference in the average grayscale value of the preprocessed images of the two lung lobes exceeds a set threshold, a coordinate system is established on the preprocessed images of the two lung lobes. By mirroring the excess portion of one of the larger lung images to fill the missing position of the smaller lung image, the lung images of the two lung lobes are made to be the same size, which facilitates the identification of areas with large nodular pneumonia. The specific implementation steps are as follows: A two-dimensional plane coordinate system is established with the midpoint of the preprocessed images of the two lungs as the origin. The two-dimensional plane coordinate systems of the preprocessed images of the two lungs share a common Y-axis, and the X-axis directions of the two two-dimensional plane coordinate systems are opposite. Divide the same region of the two lungs into a single longitudinal order and a single transverse order, and determine the coordinate values ​​corresponding to the edges of the same region. Using the Y-axis, which is shared by the two-dimensional plane coordinate system, as the axis of symmetry, the excess parts of the two lungs are mirrored and filled in at the corresponding coordinate positions; The missing areas of the two lung lobes were filled sequentially along a uniform vertical and horizontal direction to obtain images of two lung lobes of the same size.

3. The method for fusion processing of multiple follow-up images of COVID-19 patients according to claim 1, characterized in that: In step 100, each patient has a code for identification, and after image processing of each of the initial CT images, the initial CT images containing the patient's location are selected and retained, and the capture sequence number of the initial CT image is determined.

4. The fusion processing method for multiple follow-up images of COVID-19 patients according to claim 3, characterized in that: In step 200, multiple follow-up CT images are acquired sequentially at the same angle as the initial CT image, and the multiple follow-up CT images are sequentially filtered according to the captured sequence number of the retained initial CT image to obtain multiple follow-up CT images with the same size and angle as the initial CT image.

5. The method for fusion processing of multiple follow-up images of COVID-19 patients according to claim 4, characterized in that: In step 200, a two-dimensional coordinate system identical to that of the matching initial CT image is established within each follow-up CT image, and the same lesion point coordinate threshold is selected within the follow-up CT images. The method for comparing the follow-up CT images with the initial CT images of the matching imaging sequence is as follows: Determine the lesion point coordinate threshold that is the same in the follow-up CT image and the initial CT image with the matching shooting sequence number, obtain the pixel value of each pixel of the lesion point coordinate threshold in sequence, and compare the pixel values ​​of each pixel of the lesion point coordinate threshold. The number of pixels whose pixel values ​​in the follow-up CT image are less than those in the initial CT image is calculated, and this number of pixels is compared with a judgment error value. If the number of pixels is greater than the error value, it is considered that the lesion area in the follow-up CT image is decreasing; otherwise, it is considered that the lesion area in the follow-up CT image remains unchanged.

6. The method for fusion processing of multiple follow-up images of COVID-19 patients according to claim 5, characterized in that: In step 300, the pixel values ​​of all lesion point coordinates within the threshold of each of the follow-up CT images are compared with those of the matching initial CT images. The number of recovered lesion areas where the pixel values ​​of the lesion point coordinates within the threshold of the follow-up CT images are less than those of the lesion point coordinates within the threshold of the initial CT images is counted. The number of such recovered lesion areas is then compared with the number of all lesion point coordinates within the threshold to roughly determine the recovery status of COVID-19 patients.

7. The method for fusion processing of multiple follow-up images of COVID-19 patients according to claim 6, characterized in that: The number of pixels in each recovered lesion region of each follow-up CT image whose pixel value is less than that of the initial CT image is counted, and the number of pixels in the recovered lesion region whose pixel value is less than that of the initial CT image is compared with the total number of pixels at the lesion point coordinate threshold to accurately determine the specific degree of recovery of COVID-19 patients.

8. The method for fusion processing of multiple follow-up images of COVID-19 patients according to claim 1, characterized in that, In step 300, the pixel value of the pixels within the lesion point coordinate threshold of the initial CT image and the follow-up CT image is 255. Adjusting the pixel value of each pixel within the lesion point coordinate threshold on the initial CT image to 0 restores the pixel value within that pixel range to a normal state, resulting in a healthy image after the original lesion has been eliminated.

9. The method for fusion processing of multiple re-examination images of COVID-19 patients according to claim 8, characterized in that, Using the initial CT image after the removal of the original lesion as the baseline image, the follow-up CT images of each subsequent examination are compared with the initial CT image of the corresponding imaging sequence number to determine whether new lesions have appeared. The steps for determining new lesions are as follows: The lung slice layer of each lung lobe in the follow-up CT image is compared with the lung slice layer of the same sequence number in the corresponding first CT image. The lung slice layer of each lung lobe is used as a comparison area. The lung slice layer of each lung lobe is compared with the lung slice layer of the same sequence number in the corresponding standard image in the same traversal direction to see if the pixel value of the same pixel point is the same. Mark the pixels in the lung slice layer that show differences for each lung lobe; The two adjacent lung slice layers in each lung lobe are stitched together to determine whether the pixels that differ between the two adjacent lung slice layers can be stitched together. The size of the stitched pixels is calculated, and a comparison is made with the size and number of small nodule features within the error range to identify new lesions.

Images