An auxiliary reading system for patients with pulmonary cryptococcosis

Abstract

The present application relates to the technical field of lung image marking auxiliary reading, in particular to an auxiliary reading system suitable for lung cryptococcosis patients. The present application firstly acquires continuous CT slice images of the lung, and extracts the foreground connected domain of the lung parenchyma region as the connected domain to be analyzed; further, according to the monotonicity of the width change of the target connected domain, the blood vessel possibility of the target connected domain is acquired; further, according to the monotonicity of the area change of the connected domain to be analyzed at the same position, the possibility of the target connected domain penetrating through the blood vessel is acquired; the degree of the target connected domain being similar to a circle is acquired; finally, according to the degree of being similar to a circle, the blood vessel possibility and the possibility of penetrating through the blood vessel are fused, and the target connected domain is marked. The present application measures the possibility of the target connected domain belonging to the blood vessel from the width change and the area change, and fuses the blood vessel possibility and the possibility of penetrating through the blood vessel by means of the degree of being similar to a circle, so as to accurately mark the connected domain and assist the relevant personnel in reading.

Description

Technical Field

[0001] This invention relates to the field of lung image annotation-assisted image reading technology, specifically to an auxiliary image reading system suitable for patients with pulmonary cryptococcosis. Background Technology

[0002] In cases of cryptococcal lung infection, when the patient's immune system is normal, the lesions usually present as nodular masses. While these features are easy to identify, they can easily be misdiagnosed as tumors or tuberculosis. However, when the patient's immune function is suppressed, various "satellite" lesions may appear. These lesions have varied morphology and commonly exhibit diffuse, disseminated infiltration, and consolidation. Because the density of the diseased tissue and vascular tissue is generally high, they produce a similar whitish appearance, making it easy for the grayscale of the lesion pixels to overlap with that of the pulmonary vessels, causing mutual interference and making it difficult to distinguish small lesions. Summary of the Invention

[0003] To address the technical problem of distinguishing between pulmonary cryptococcal disease lesions and vascular tissues, the present invention aims to provide an auxiliary image reading system suitable for patients with pulmonary cryptococcal disease. The specific technical solution adopted is as follows:

[0004] Image acquisition module: acquires continuous CT slice images of the current patient's lungs; extracts the lung parenchyma region from each CT slice image through image segmentation; extracts the foreground connected component of the lung parenchyma region as the connected component to be analyzed; and matches the connected components to be analyzed at the same position in all CT slice images.

[0005] Image analysis module: Selects any of the connected components to be analyzed as the target connected component; obtains the probability of blood vessels in the target connected component based on the monotonicity of the width change of the target connected component; obtains the probability of blood vessels penetrating the target connected component based on the monotonicity of the area change of the connected component to be analyzed matched with the target connected component; obtains the degree of circularity of the target connected component.

[0006] Image annotation module: Based on the degree of circularity, the probability of blood vessels and the probability of penetrating blood vessels are fused together to label the target connected region.

[0007] Furthermore, the method for obtaining the probability of the blood vessels includes:

[0008] Using the line connecting the two farthest pixels in the target connected region as a reference line, move from one end of the reference line to the other with a step size of a single pixel; after each move, obtain the pixel width of the nearest target connected region in the vertical direction of the reference line, and construct the width sequence of the target connected region.

[0009] Based on the monotonicity of the width sequence, the probability of blood vessels in the target connected region is obtained.

[0010] Furthermore, the method for obtaining the probability of blood vessels in the target connected region based on the monotonicity of the width sequence includes:

[0011] The probability of blood vessels in the target connected region is obtained based on the correlation between the width sequence and the sequence number formed by the indices of the elements within the width sequence; the correlation between the width sequence and the sequence number is positively correlated with the probability of blood vessels.

[0012] Furthermore, the method for obtaining the probability of penetrating a blood vessel includes:

[0013] The area of ​​the connected component to be analyzed that matches the target connected component is obtained, and the area is sorted according to the acquisition order of the CT slice images to obtain an area sequence;

[0014] Based on the monotonicity of the area sequence, the probability of the target connected region penetrating a blood vessel is obtained.

[0015] Furthermore, the method for obtaining the probability of a penetrating blood vessel in the target connected region based on the monotonicity of the area sequence includes:

[0016] The probability of a penetrating blood vessel in the target connected region is obtained based on the number of extreme points at non-endpoints in the area sequence; the number of extreme points at non-endpoints is negatively correlated with the probability of a penetrating blood vessel.

[0017] Furthermore, the method for marking the target connected component includes:

[0018] A first normal probability is obtained by combining the degree of roundness and the probability of blood vessels; the degree of roundness is negatively correlated with the first normal probability; the probability of blood vessels is positively correlated with the first normal probability.

[0019] A second normal probability is obtained by combining the degree of circularity and the probability of penetrating a blood vessel; both the degree of circularity and the probability of penetrating a blood vessel are positively correlated with the second normal probability.

[0020] The normal organization probability of the target connected component is obtained by fusing the first normal probability and the second normal probability; both the first normal probability and the second normal probability are positively correlated with the normal organization probability.

[0021] When the probability of a normal organization is lower than a preset normal threshold, the target connected component is marked.

[0022] Furthermore, the degree of circularity refers to the roundness of the outermost pixel outline of the target connected region.

[0023] Furthermore, the method for matching the connected components to be analyzed at the same location in all the CT slice images includes:

[0024] In acquiring two adjacent frames of the CT slice images, when the overlap ratio between the pixels in the connected component to be analyzed in one frame and the pixels in the connected component to be analyzed in another frame is greater than a preset threshold, the two connected components to be analyzed are determined to be matched.

[0025] Furthermore, the lung parenchyma region of the CT slice image is extracted using a pre-trained neural network.

[0026] Furthermore, the method for obtaining the foreground connected component includes:

[0027] A region growing algorithm is selected to randomly select seed points from pixels whose gray values ​​are greater than the average gray value of the lung parenchyma region to obtain the foreground connected region.

[0028] The present invention has the following beneficial effects:

[0029] This invention first acquires continuous CT slice images of the lungs to establish an analytical foundation. It then extracts the lung parenchyma region and its foreground connected components as the connected components to be analyzed, gradually narrowing the analysis scope and eliminating interference from irrelevant regions. Furthermore, based on the monotonicity of the width variation of the target connected component and the monotonicity of the area variation of connected components at the same location, it obtains the probability of the target connected component being a blood vessel and the probability of it penetrating a blood vessel. Morphological changes are analyzed from the perspectives of width and area variations to measure the probability that the target connected component belongs to a blood vessel, providing a basis for subsequent image labeling. Next, the circularity of the target connected component is obtained to prepare for subsequent adjustment and fusion of the probability of blood vessel penetration, comprehensively analyzing the probability that the target connected component belongs to vascular tissue, thus preparing for labeling. Finally, based on the circularity, the probability of blood vessel fusion, and the probability of penetrating a blood vessel, the target connected component is labeled. This invention measures the probability of a target connected component being a blood vessel from the perspectives of width and area variations, and uses the circularity to fuse the probability of blood vessel penetration, accurately labeling the connected component and assisting relevant personnel in interpreting images. Attached Figure Description

[0030] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a system block diagram of an auxiliary image reading system for patients with pulmonary cryptococcosis, provided in one embodiment of the present invention;

[0032] Figure 2 A CT slice image provided in one embodiment of the present invention;

[0033] Figure 3 An image of a lung parenchyma region provided in one embodiment of the present invention;

[0034] Figure 4 This is a schematic diagram illustrating the distribution of connected components to be analyzed, provided as an embodiment of the present invention.

[0035] Figure 5 This is a schematic diagram showing the appearance of pulmonary blood vessels in a series of sections, according to an embodiment of the present invention.

[0036] Figure 6 This is a schematic diagram illustrating the appearance of pulmonary blood vessels in a single slice, as provided in one embodiment of the present invention.

[0037] Figure 7 This is a schematic diagram illustrating the calculation of pixel width according to an embodiment of the present invention;

[0038] Figure 8 This is a schematic diagram illustrating the process of shrinking the circumcircle of a connected domain, as provided in an embodiment of the present invention. Detailed Implementation

[0039] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an auxiliary image reading system for patients with pulmonary cryptococcosis proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0041] The following description, in conjunction with the accompanying drawings, details a specific scheme for an auxiliary image reading system for patients with pulmonary cryptococcosis provided by the present invention.

[0042] Please see Figure 1 The diagram illustrates a system block diagram of an auxiliary image reading system for patients with pulmonary cryptococcosis, provided by an embodiment of the present invention. The system includes: an image acquisition module 101, an image analysis module 102, and an image annotation module 103.

[0043] Image acquisition module 101: acquires continuous CT slice images of the current patient's lungs; extracts the lung parenchyma region from each CT slice image through image segmentation; extracts the foreground connected component of the lung parenchyma region as the connected component to be analyzed; and matches the connected components to be analyzed at the same location in all CT slice images.

[0044] In one embodiment of the present invention, a high-resolution thin-slice CT scanner is used to perform a plain chest scan, with the slice thickness limited to no more than 2 mm; the scan time should be completed within 30 seconds if possible. Simultaneously, the completed chest CT images are stored in real time, according to the scanning sequence, to obtain continuous CT slice images of the patient's lungs and establish a basis for analysis.

[0045] Please see Figure 2 It illustrates a CT slice image provided by an embodiment of the present invention; from Figure 2 As can be seen, the obtained CT slice images include irrelevant areas such as the patient's chest and spine. Therefore, the lung parenchyma region in each CT slice image is extracted first through image segmentation to narrow down the analysis scope.

[0046] In one embodiment of the present invention, the lung parenchyma region of a CT slice image is extracted using a pre-trained neural network. The neural network architecture is a CNN network; the input is a scanned chest CT image, and the output is an image containing only the lung parenchyma region; the training set consists of manually annotated lung parenchyma region contours; the loss function is the cross-entropy loss function. Please refer to [link to relevant documentation]. Figure 3 This image illustrates a lung parenchyma region provided by an embodiment of the present invention.

[0047] It should be noted that neural networks and their training methods are well-known techniques to those skilled in the art. In other embodiments of the present invention, implementers may also use existing graph cut algorithms and other segmentation methods, which will not be elaborated here.

[0048] During CT imaging, organs with lower tissue density generally appear darker, resulting in soft tissues appearing grayish-white. Therefore, most parenchymal areas of the lungs appear darker because they contain air, while the blood in the lung tissue has a higher density, making the blood vessels appear lighter in gray.

[0049] Furthermore, since pulmonary cryptococcal infection can cause nodules, exudates, and ground-glass opacities in local tissues, leading to increased local density, the whitish pixel areas in the lungs could be normal blood vessels or diseased tissue. Therefore, further extraction of the foreground connected domain of the lung parenchyma as the connected domain to be analyzed further narrows the scope of the analysis.

[0050] Preferably, in one embodiment of the present invention, a region growing algorithm is selected. Considering that the gray values ​​of the blood vessel region and the lesion tissue region are relatively high, seed points are randomly selected from the pixels whose gray values ​​are greater than the average gray value of the lung parenchyma region to obtain the foreground connected region.

[0051] Specifically, seed points are randomly selected evenly, with a minimum of 20 seed points, set to 25. The growth criterion is set to a relative threshold of 30, and the neighborhood is set to eight neighborhoods. Growth stops when the absolute value of the difference between the gray values ​​of the ungrown pixels in the eight neighborhoods of all newly grown pixels and the average gray value of the grown area is greater than 30.

[0052] Please see Figure 4 It shows a schematic diagram of the distribution of connected components to be analyzed according to an embodiment of the present invention. Figure 4 The connected components with a gray value of 0 are considered as background connected components, and the connected components with a gray value of 255 are considered as foreground connected components, which are used as the connected components to be analyzed.

[0053] It should be noted that in other embodiments of the present invention, the Otsu threshold segmentation algorithm can also be used to extract foreground pixels. Both the Otsu threshold segmentation algorithm and the region growing algorithm are well known to those skilled in the art, and will not be described in detail here.

[0054] Considering that the same vascular tissue or lesion tissue may appear differently in CT slices at different depths, the connected regions to be analyzed at the same location in all CT slices are matched to facilitate the analysis of the variation characteristics of connected regions of the same tissue at different depths.

[0055] Preferably, in one embodiment of the present invention, when two adjacent CT slice images are acquired, if the overlap ratio between the pixels in the connected component to be analyzed in one frame image and the pixels in the connected component to be analyzed in another frame image is greater than a preset threshold, the two connected components to be analyzed are determined to be matched.

[0056] As an example, the overlapping area of ​​the connected components to be analyzed in two frames of images is used as the numerator, the minimum area of ​​the two connected components to be analyzed is used as the denominator, and the ratio of the fractions is used as the overlap ratio. The preset threshold is 0.5. When the overlap ratio is greater than 0.5, the corresponding two connected components to be analyzed are determined to be matched.

[0057] In another embodiment of the present invention, a matching constraint condition of proximity of centroids can be added, such as calculating the Euclidean distance between the centroid coordinates of the connected domains to be analyzed in adjacent CT slices, setting a matching threshold such as 5, and determining that the two connected domains to be analyzed are matched when the overlap ratio is higher than 0.5 and the Euclidean distance is less than 5.

[0058] It should be noted that obtaining the centroid of a connected component, establishing a coordinate system, and calculating the Euclidean distance are all techniques well-known to those skilled in the art, and will not be elaborated upon here.

[0059] Image analysis module 102: Selects any connected component to be analyzed as the target connected component; obtains the probability of blood vessels in the target connected component based on the monotonicity of the width change of the target connected component; obtains the probability of blood vessels penetrating the target connected component based on the monotonicity of the area change of the connected component to be analyzed matched with the target connected component; obtains the degree of circularity of the target connected component.

[0060] Considering that vascular tissue and diseased tissue have similar gray levels, making them difficult to distinguish from a gray level perspective, we analyze morphological changes to determine the likelihood that the connected region to be analyzed belongs to vascular tissue. This allows us to ultimately label the lesion area that is very likely to be caused by Cryptococcus pneumoniae, assisting relevant personnel in interpreting the images. First, any connected region to be analyzed is selected as the target connected region for easy analysis. The analysis method for each connected region to be analyzed is consistent, and only the target connected region is described here as an example, without repeating the description.

[0061] Please see Figure 5 It shows a schematic diagram of the appearance of pulmonary blood vessels in a series of slices according to an embodiment of the present invention; Figure 5 The image shows three consecutive CT slices and illustrates the spatial relationship between the pulmonary vessels and the slices.

[0062] Please see Figure 6 It shows a schematic diagram of the appearance of pulmonary blood vessels in a single slice according to an embodiment of the present invention; Figure 6 The image shows a slice that passes horizontally through pulmonary vessels, appearing tubular; it also shows a slice that passes vertically through pulmonary vessels, appearing roughly circular.

[0063] Combination Figure 5 and Figure 6 It can be seen that in three-dimensional space, normal blood vessels spread outward from the main pulmonary vessels, and the diameter of the blood vessels from the main vessels to the ends shows a characteristic of decreasing from large to small, thus the width of normal blood vessels shows a continuous monotonic change.

[0064] Meanwhile, some blood vessels may appear as a hole in a single CT image, indicating that the blood vessel runs through the current CT slice. Therefore, the size of the through hole at the same location in consecutive slices also has strong monotonicity. Thus, the probability of the target connected region being a blood vessel is obtained based on the monotonicity of the width change of the target connected region; the probability of the target connected region being a through blood vessel is obtained based on the monotonicity of the area change of the connected region to be analyzed matched with the target connected region; the probability of the target connected region belonging to a blood vessel is measured from the perspectives of width change and area change, respectively, providing a basis for subsequent image labeling.

[0065] Preferably, in one embodiment of the present invention: considering that blood vessels usually appear as slender structures on CT slices, the line connecting the two farthest pixels in the target connected region is used as a reference line, which can make the reference line roughly aligned with the main axis direction of the blood vessel, thereby presenting the main shape of the target connected region and making the width measurement as close as possible to the cross-sectional width of the real blood vessel.

[0066] The reference line is moved from one end to the other with a step size of a single pixel. After each movement, the pixel width of the nearest target connected region in the vertical direction of the reference line is obtained, and a sequence of target connected region widths is constructed. In order to avoid inaccurate analysis of width changes due to the measurement of the width of multiple blood vessel branches, only the pixel width of the nearest target connected region in the vertical direction of the reference line is obtained.

[0067] Please see Figure 7 It shows a schematic diagram of a method for calculating pixel width provided by the present invention; Figure 7 It contains a target connected component. The dashed line in the figure is the reference line. The line perpendicular to the reference line is used to count the pixel width and shows only the pixel width of the nearest target connected component in the direction perpendicular to the reference line. The pixel width is the number of pixels.

[0068] The width sequence represents the width variation characteristics of the target connected region, so the possibility of blood vessels in the target connected region can be obtained further based on the monotonicity of the width sequence.

[0069] As an example: Considering that the width variation exhibits strong monotonicity, the width sequence also exhibits monotonic variation with the position number. Based on the correlation between the width sequence and the sequence number formed by the element numbers within the width sequence, the probability of blood vessels in the target connected region can be obtained. Since the sequence number is monotonic, the stronger the correlation between the width sequence and the sequence number, the more it matches the width variation characteristics of blood vessel structures. Therefore, the correlation between the width sequence and the sequence number is positively correlated with the probability of blood vessels.

[0070] Specifically, considering that the Spearman rank correlation coefficient is a statistical method used to measure the strength of the monotonic relationship between two variables, the larger the absolute value of the Spearman rank correlation coefficient, the stronger the correlation and the stronger the monotonicity of the width sequence, the absolute value of the Spearman rank correlation coefficient of the width sequence and the sequence number sequence is used as the probability of blood vessels in the target connected region.

[0071] In another embodiment of the present invention, the skeleton of the target connected domain can also be extracted, its topological structure and key morphological features are preserved, the skeleton line where the two farthest pixels are located is selected as the reference line, the pixel width of the original connected domain corresponding to the reference line is obtained, and the width sequence is obtained.

[0072] In other embodiments of the present invention, the correlation between the width sequence and the sequence number can also be analyzed using the Pearson correlation coefficient. For example, the absolute value of the Pearson correlation coefficient can be used as the probability of blood vessels. The skeleton extraction method, the Pearson correlation coefficient and the Spearman rank correlation coefficient are all existing technologies and will not be described in detail here.

[0073] Preferably, in one embodiment of the present invention, considering that some blood vessels run through continuous CT slices, the area change of the connected regions of such blood vessels in the continuous slices reflects the change of the blood vessel diameter. Therefore, the area of ​​the connected region to be analyzed matched by the target connected region is obtained, and sorted according to the acquisition order of the CT slice images to obtain the area sequence.

[0074] Based on the monotonicity of the area sequence, the probability of penetrating blood vessels in the target connected region is obtained.

[0075] As an example: considering that the stronger the monotonicity of the data in the area sequence, the fewer the number of extreme points, and that the extreme points at both ends of the sequence cannot reflect the monotonicity of the data change, the probability of the target connected region penetrating the blood vessel is obtained based on the number of extreme points at the non-ends in the area sequence; the number of extreme points at the non-ends is negatively correlated with the probability of penetrating the blood vessel.

[0076] Specifically, the sum of the number of extreme points at the non-end points of the area sequence and the preset positive parameter 1 (divided by zero) is taken as the reciprocal, and the reciprocal is linearly normalized. The normalized result is then used as the probability of penetrating the blood vessel.

[0077] It should be noted that the area of ​​a connected region is the number of pixels in the connected region, and the method for obtaining the extreme points is a well-known technique in the art, so it will not be described in detail here.

[0078] In CT slice images, the cross-sectional shape of blood vessels may vary depending on the slice orientation and the shape of the blood vessel itself. This leads to different reference values ​​for the probability of a blood vessel and the probability of it penetrating a blood vessel in different situations. Therefore, it is also necessary to obtain the degree of circularity of the target connected region as an additional judgment factor, which can improve accuracy.

[0079] Preferably, in one embodiment of the present invention, the degree of circularity is the roundness of the outermost pixel outline of the target connected region.

[0080] It should be noted that the method for calculating roundness is already existing technology and will not be elaborated further.

[0081] In another embodiment of the present invention, considering that as the outermost connecting circle of a near-circular connected region gradually shrinks inward, the smooth pits in the continuous space appear as stepped edges in the discrete pixels, causing the number of different pixels to increase dramatically when the outer circle first comes into contact. The number of pixels that change is larger when the near-circular boundary is closest. This characteristic is used to obtain the degree of near-circularity.

[0082] Please see Figure 8 It shows a schematic diagram of a process for reducing the circumcircle of a connected domain according to an embodiment of the present invention; Figure 8 It contains a connected component to be analyzed. The largest circle is the smallest outer circle of the connected component to be analyzed. The circle is gradually reduced from the center with individual pixels. The arrows represent the direction of reduction. The smallest circle is the circle after the first reduction. After each reduction, the number of pixels of the connected component to be analyzed that the circle passes through is obtained. The absolute value of the difference between the number of pixels and the average radius of the corresponding two reductions are recorded.

[0083] The absolute value of the difference in the number of pixels corresponding to each pair of adjacent shrinking circles is used as the numerator, and the average radius of the two shrinking circles is used as the denominator. The ratio of the fractions is linearly normalized and used as the circularity of each shrinking operation. The maximum value of the circularity of all shrinking operations is used as the degree of circularity.

[0084] Image annotation module 103: Based on the degree of circularity, the probability of blood vessels and the probability of penetrating blood vessels are fused to label the target connected regions.

[0085] The probability of a target connected region belonging to a blood vessel is measured from the perspectives of width change and area change, respectively. The degree of circularity is also used to adjust the reference weights of the two probabilities. Therefore, based on the degree of circularity, the probability of a blood vessel and the probability of a blood vessel are fused to improve the recognition accuracy of blood vessels with different morphological manifestations. The target connected region is then labeled to ensure labeling accuracy and improve the auxiliary image reading effect.

[0086] Preferably, in one embodiment of the invention, the degree of circularity and the likelihood of blood vessels are combined to obtain a first normal probability;

[0087] Considering that the smaller the degree of roundness, the more likely the target connected region is to be a cross-section of a blood vessel, and the greater the reference value of the probability of blood vessel, the greater the probability of blood vessel, the stronger the monotonicity of the width change of the target connected region, the more it matches the characteristics of blood vessel width change, and the more likely it is to be normal tissue; therefore, the degree of roundness is negatively correlated with the first normal probability, and the probability of blood vessel is positively correlated with the first normal probability.

[0088] Further integration of the roundness and the likelihood of vascular penetration yields a second normal possibility;

[0089] Considering that the greater the degree of circularity, the more likely the target connected region is to be the surface through which a blood vessel passes, and the greater the reference value of the probability of penetrating a blood vessel, the greater the probability of penetrating a blood vessel, the stronger the monotonicity of the area change of the target connected region in continuous slices, the more it conforms to the characteristics of blood vessel width change, and the more likely it is to be normal tissue; therefore, both the degree of circularity and the probability of penetrating a blood vessel are positively correlated with the second probability of normality.

[0090] Further integrate the first and second normal possibilities to obtain the normal organization possibility of the target connected domain;

[0091] Considering that the higher the probability of the first and second normal possibilities, the more likely it is to be normal tissue, both the probability of the first and second normal possibilities are positively correlated with the probability of normal tissue.

[0092] Finally, when the probability of normal organization is lower than the preset normal threshold, the target connected component is marked.

[0093] As an example, for a target connected region, the product of the difference between the constant 1 and the degree of circularity and the probability of blood vessels is taken as the first normal probability, the product of the degree of circularity and the probability of penetrating blood vessels is taken as the second probability, and the sum of the first normal probability and the second normal probability is taken as the normal tissue probability; the preset normal threshold is 0.6, and when the normal tissue probability is lower than 0.6, the target connected region is marked as a pulmonary cryptococcal infection area.

[0094] In calculating the difference between the constant 1 and the degree of roundness, the degree of roundness is the minuend. By subtracting the degree of roundness from 1, a negative correlation mapping is applied to the degree of roundness, thus adjusting the calculation logic.

[0095] It should be noted that marking the target connected components is only for assisting relevant personnel in reading the images and is not for the purpose of diagnosing the disease.

[0096] In another embodiment of the present invention, the patient's medical records are also obtained, and relevant keywords for symptoms of pulmonary cryptococcal infection, such as hemoptysis, dyspnea, and shortness of breath, are set. By comparing and extracting these keywords with the patient's relevant medical records, if keyword overlap occurs, an early warning of pulmonary cryptococcal infection is provided to the diagnosing doctor.

[0097] In summary, addressing the technical challenge of distinguishing between pulmonary cryptococcal lesions and vascular tissue, this invention proposes an auxiliary image reading system suitable for patients with pulmonary cryptococcal disease. This invention first acquires continuous CT slice images of the lungs, including the lung parenchyma region, and extracts the foreground connected regions of the lung parenchyma region as the connected regions to be analyzed. Next, any connected region to be analyzed is selected as the target connected region. Based on the monotonicity of the width change of the target connected region, the probability of blood vessels in the target connected region is obtained. Further, based on the monotonicity of the area change of the connected regions to be analyzed at the same location in the target connected region, the probability of blood vessels penetrating the target connected region is obtained. The degree of circularity of the target connected region is also obtained. Finally, based on the degree of circularity, the probability of blood vessels and the probability of blood vessels penetrating the target connected region are fused, and the target connected region is labeled.

[0098] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0099] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. An auxiliary image reading system for patients with pulmonary cryptococcosis, characterized in that, The system includes: Image acquisition module: acquires continuous CT slice images of the current patient's lungs; extracts the lung parenchyma region from each CT slice image through image segmentation; extracts the foreground connected component of the lung parenchyma region as the connected component to be analyzed; and matches the connected components to be analyzed at the same position in all CT slice images. Image analysis module: Selects any of the connected components to be analyzed as the target connected component; obtains the probability of blood vessels in the target connected component based on the monotonicity of the width change of the target connected component; obtains the probability of blood vessels penetrating the target connected component based on the monotonicity of the area change of the connected component to be analyzed matched with the target connected component; obtains the degree of circularity of the target connected component. Image annotation module: Based on the degree of circularity, the probability of blood vessels and the probability of penetrating blood vessels are fused together to label the target connected region; The method for obtaining the probability of the blood vessel includes: Using the line connecting the two farthest pixels in the target connected region as a reference line, move from one end of the reference line to the other with a step size of a single pixel; after each move, obtain the pixel width of the nearest target connected region in the vertical direction of the reference line, and construct the width sequence of the target connected region. Based on the monotonicity of the width sequence, the probability of blood vessels in the target connected region is obtained; The method for obtaining the probability of penetrating a blood vessel includes: The area of ​​the connected component to be analyzed that matches the target connected component is obtained, and the area is sorted according to the acquisition order of the CT slice images to obtain an area sequence; Based on the monotonicity of the area sequence, the probability of the target connected region penetrating a blood vessel is obtained.

2. The auxiliary image reading system for patients with pulmonary cryptococcosis according to claim 1, characterized in that, The method for obtaining the probability of blood vessels in the target connected region based on the monotonicity of the width sequence includes: The probability of blood vessels in the target connected region is obtained based on the correlation between the width sequence and the sequence number formed by the indices of the elements within the width sequence; the correlation between the width sequence and the sequence number is positively correlated with the probability of blood vessels.

3. The auxiliary image reading system for patients with pulmonary cryptococcosis according to claim 1, characterized in that, The method for obtaining the probability of a penetrating blood vessel in the target connected region based on the monotonicity of the area sequence includes: The probability of a penetrating blood vessel in the target connected region is obtained based on the number of extreme points at non-endpoints in the area sequence; the number of extreme points at non-endpoints is negatively correlated with the probability of a penetrating blood vessel.

4. The auxiliary image reading system for patients with pulmonary cryptococcosis according to claim 1, characterized in that, The method for marking the target connected component includes: A first normal probability is obtained by combining the degree of roundness and the probability of blood vessels; the degree of roundness is negatively correlated with the first normal probability; the probability of blood vessels is positively correlated with the first normal probability. A second normal probability is obtained by combining the degree of circularity and the probability of penetrating a blood vessel; both the degree of circularity and the probability of penetrating a blood vessel are positively correlated with the second normal probability. The normal organization probability of the target connected component is obtained by fusing the first normal probability and the second normal probability; both the first normal probability and the second normal probability are positively correlated with the normal organization probability. When the probability of a normal organization is lower than a preset normal threshold, the target connected component is marked.

5. The auxiliary image reading system for patients with pulmonary cryptococcosis according to claim 1, characterized in that, The degree of circularity refers to the roundness of the outermost pixel outline of the target connected region.

6. The auxiliary image reading system for patients with pulmonary cryptococcosis according to claim 1, characterized in that, The method for matching the connected components to be analyzed at the same location in all the CT slice images includes: In acquiring two adjacent frames of the CT slice images, when the overlap ratio between the pixels in the connected component to be analyzed in one frame and the pixels in the connected component to be analyzed in another frame is greater than a preset threshold, the two connected components to be analyzed are determined to be matched.

7. The auxiliary image reading system for patients with pulmonary cryptococcosis according to claim 1, characterized in that, The lung parenchyma region of the CT slice image is extracted using a pre-trained neural network.

8. The auxiliary image reading system for patients with pulmonary cryptococcosis according to claim 1, characterized in that, The method for obtaining the foreground connected component includes: A region growing algorithm is selected to randomly select seed points from pixels whose gray values ​​are greater than the average gray value of the lung parenchyma region to obtain the foreground connected region.

Images