Tumor recurrence and metastasis monitoring method and system

Abstract

The present invention provides a method and system for monitoring tumor recurrence and metastasis, the method comprising: obtaining multiple two-dimensional CT image slices of lung CT images of patients with a history of rectal cancer; determining the number, area information and centroid position information of the first selection boxes in the area where each lesion is located through a lesion detection model; determining the lesion diffusion score and the lesion isolation score; determining the first contour of the pleura area in each two-dimensional CT image slice through a contour detection model, and determining the pleural invasion score; determining the rectal cancer lung metastasis risk score and obtaining a detection report. According to the present invention, when monitoring the risk of rectal cancer recurrence and lung metastasis, multiple evaluations can be performed from the perspectives of lesion diffusion, lesion isolation and pleural invasion, and a detection report can be generated, without relying on personnel experience, thereby improving detection accuracy and the efficiency of rectal cancer lung metastasis risk monitoring.

Description

Technical Field

[0001] The present invention relates to the field of computer computing, and in particular to a method and system for monitoring tumor recurrence and metastasis. Background Art

[0002] When monitoring the risk of recurrent lung metastasis of rectal cancer, medical staff need to spend a lot of time observing the lung CT images of patients with a history of rectal cancer. The examination efficiency is low and it relies on personnel experience. Therefore, there is a possibility of misidentification and low examination efficiency. It is difficult to accurately judge the risk of lung metastasis of rectal cancer and it is also difficult to generate targeted test reports, resulting in low quality and efficiency of monitoring the risk of lung metastasis of rectal cancer.

[0003] The information disclosed in the background technology section of this application is only intended to deepen the understanding of the general background technology of this application, and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art already known to those skilled in the art. Summary of the invention

[0004] The present invention provides a method and system for monitoring tumor recurrence and metastasis, which can solve the technical problems of low inspection efficiency and possibility of misidentification in related technologies.

[0005] According to a first aspect of the present invention, a method for monitoring tumor recurrence and metastasis is provided, comprising:

[0006] Acquire a lung CT image of a patient with a history of rectal cancer, wherein the lung CT image is a three-dimensional image, and the lung CT image includes a plurality of two-dimensional CT image slices;

[0007] Determine a first selection box of the area where the lesion is located in each two-dimensional CT image slice by using the lesion detection model, wherein the first selection box is a minimum rectangular box that selects the area where the lesion is located;

[0008] Determine the number of first selection boxes, area information of each first selection box, and centroid position information of each first selection box;

[0009] Determining a lesion diffusion score according to the number of the first selection boxes, the area information, and the centroid position information;

[0010] Determine the lesion isolation score according to the position information of the first selection box and the centroid position information of the first selection box;

[0011] Determine a first contour of a region where the pleura is located in each two-dimensional CT image slice by using a contour detection model;

[0012] Determining a pleural invasion score based on the first outline;

[0013] According to the lesion diffusion score, the lesion isolation score and the pleural invasion score, the risk score of lung metastasis of rectal cancer is determined, and a test report is obtained.

[0014] According to a second aspect of the present invention, a tumor recurrence and metastasis monitoring system is provided, comprising:

[0015] A two-dimensional CT image slice acquisition module, which acquires a lung CT image of a patient with a history of rectal cancer, wherein the lung CT image is a three-dimensional image, and the lung CT image includes a plurality of two-dimensional CT image slices;

[0016] A first selection box determination module, which determines a first selection box of a lesion area in each two-dimensional CT image slice through a lesion detection model, wherein the first selection box is a minimum rectangular box for selecting the lesion area;

[0017] A module for determining the number of first selection boxes, the area information of each first selection box, and the centroid position information of each first selection box, to determine the number of first selection boxes, the area information of each first selection box, and the centroid position information of each first selection box;

[0018] a lesion diffusion score determination module, which determines a lesion diffusion score according to the number of the first selection boxes, the area information and the centroid position information;

[0019] A lesion isolation score determination module determines the lesion isolation score according to the position information of the first selection box and the centroid position information of the first selection box;

[0020] A first contour determination module, which determines a first contour of a region where the pleura is located in each two-dimensional CT image slice by using a contour detection model;

[0021] a pleural invasion score determination module, which determines a pleural invasion score according to the first contour;

[0022] The rectal cancer lung metastasis risk score determination module determines the rectal cancer lung metastasis risk score according to the lesion diffusion score, the lesion isolation score and the pleural invasion score, and obtains a detection report.

[0023] Technical effect: According to the present invention, multiple two-dimensional CT image slices of lung CT images can be obtained, the first selection box can be determined by the lesion detection model, and the lesion diffusion score and lesion isolation score can be determined based on the number of first selection boxes, the area information of each first selection box, and the centroid position information of each first selection box. The pleural invasion score is determined based on the first contour determined by the contour detection model, and then the rectal cancer lung metastasis risk score is determined and a detection report is generated. The rectal cancer lung metastasis risk can be evaluated from multiple aspects such as lesion diffusion, lesion isolation, and pleural invasion degree, which is helpful to accurately judge the rectal cancer lung metastasis risk, improve the efficiency of rectal cancer lung metastasis risk monitoring, and do not rely on personnel experience, improve detection accuracy, and reduce the probability of misidentification. Further, the lung two-dimensional CT image slices of patients with a history of rectal cancer can be obtained, the first selection box can be determined based on the lesion detection model, and then the number of first selection boxes and the area information of each first selection box are determined, and the centroid position information of each first selection box is determined based on the horizontal and vertical coordinates. Provide basic data for the evaluation of rectal cancer lung metastasis risk. When determining the lesion diffusion score, the second contour of the lung in the target two-dimensional CT image slice can be obtained based on the contour detection model, and the distribution, quantity and independence of the tumor can be obtained according to the first number, the second contour, the second selection box, the number of the first selection box and the area information, so as to evaluate the diffusion of the lesion, and avoid the situation where the number of tumors is small but the distribution is too dispersed, resulting in overestimation of the lesion diffusion, and then the lesion diffusion score of each target two-dimensional CT image slice can be determined, so as to obtain the lesion diffusion score by averaging. The lesion diffusion score can accurately and objectively represent the lesion diffusion of the patient's lung. When determining the lesion isolation score, the first target selection box whose intersection with each first selection box is not an empty set can be determined, and the union can be binarized, so as to determine the connectivity relationship between the centroid position information of each first selection box and the centroid position information of the first target selection box, so as to determine the total number of lesions that are connected to each other, and remove the repeated counts therein, so as to obtain the actual number of isolated lesions in each target two-dimensional CT image slice, so as to make the counting more accurate. Then, an accurate lesion isolation score is determined, so that the lesion isolation score can accurately and objectively represent the isolation of the lesion. When determining the pleural invasion score, multiple nodes can be evenly set on the first contour of the target two-dimensional CT image slice and the coordinate data of each node can be determined. Based on the preset angle and the angle between the vectors of each node, the abnormal protrusions on the pleura are counted, and the situation where the number of tumors on the pleura is overestimated due to the abnormal protrusions formed by the irregular shape of the pleura itself is avoided. The actual number of abnormal protrusions on the pleura caused by the tumor in each target two-dimensional CT image slice can be obtained, so that the counting is more accurate, and then the accurate pleural invasion score is determined, so that the pleural invasion score can accurately and objectively represent the degree of pleural invasion by the tumor.When obtaining the test report, the lesion diffusion score, lesion isolation score and pleural invasion score can be weighted and summed to determine the risk score of rectal cancer lung metastasis, and generate the corresponding test report message. The rectal cancer lung metastasis risk score can accurately describe the results of tumor recurrence and metastasis monitoring, and according to the corresponding test report, it can prompt doctors or patients the risk of rectal cancer lung metastasis.

[0024] It should be understood that the above general description and the following detailed description are exemplary and explanatory only and do not limit the present invention. Other features and aspects of the present invention will become more apparent from the following detailed description of exemplary embodiments with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other embodiments can be obtained based on these drawings without creative work.

[0026] Figure 1 The following is a schematic diagram showing a flow chart of a method for monitoring tumor recurrence and metastasis according to an embodiment of the present invention;

[0027] Figure 2 A schematic diagram of a tumor recurrence and metastasis monitoring system according to an embodiment of the present invention is exemplarily shown. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0030] Figure 1 The following is a schematic flow chart of a method for monitoring tumor recurrence and metastasis according to an embodiment of the present invention. The method includes:

[0031] Step S101, obtaining a lung CT image of a patient with a history of rectal cancer, wherein the lung CT image is a three-dimensional image, and the lung CT image includes a plurality of two-dimensional CT image slices;

[0032] Step S102, determining a first selection box of the area where the lesion is located in each two-dimensional CT image slice by using a lesion detection model, wherein the first selection box is a minimum rectangular box that selects the area where the lesion is located;

[0033] Step S103, determining the number of first selection boxes, area information of each first selection box, and centroid position information of each first selection box;

[0034] Step S104, determining a lesion diffusion score according to the number of the first selection boxes, the area information and the centroid position information;

[0035] Step S105, determining a lesion isolation score according to the position information of the first selection box and the centroid position information of the first selection box;

[0036] Step S106, determining the first contour of the region where the pleura is located in each two-dimensional CT image slice by using the contour detection model;

[0037] Step S107, determining a pleural invasion score according to the first contour;

[0038] Step S108, determining the risk score of lung metastasis of rectal cancer according to the lesion diffusion score, the lesion isolation score and the pleural invasion score, and obtaining a test report.

[0039] According to the tumor recurrence and metastasis monitoring method of the embodiment of the present invention, multiple two-dimensional CT image slices of the lung CT image can be obtained, the first selection box can be determined by the lesion detection model, and the lesion diffusion score and the lesion isolation score can be determined based on the number of first selection boxes, the area information of each first selection box, and the centroid position information of each first selection box. The pleural invasion score is determined based on the first contour determined by the contour detection model, and then the rectal cancer lung metastasis risk score is determined and a detection report is generated. The rectal cancer lung metastasis risk can be evaluated from multiple aspects such as lesion diffusion, lesion isolation and pleural invasion degree, which helps to accurately judge the rectal cancer lung metastasis risk, improve the efficiency of rectal cancer lung metastasis risk monitoring, and does not rely on personnel experience, improve detection accuracy, and reduce the probability of misidentification.

[0040] According to an embodiment of the present invention, in step S101, a lung CT image of a patient with a history of rectal cancer is obtained, wherein the lung CT image is a three-dimensional image, and the lung CT image includes a plurality of two-dimensional CT image slices. For example, the chest of a patient with a history of rectal cancer may be scanned every several millimeters to obtain a two-dimensional CT image slice, and a plurality of two-dimensional CT image slices may constitute the three-dimensional lung CT image.

[0041] According to an embodiment of the present invention, in step S102, a first selection box of the area where the lesion is located in each two-dimensional CT image slice is determined by a lesion detection model, wherein the first selection box is the minimum rectangular box for selecting the area where the lesion is located. The lesion detection model is a deep learning neural network model, for example, a convolutional neural network model, which can be used to detect the area where the lesion is located in each two-dimensional CT image slice, and use the first selection box to select the area to obtain the first selection box of the area where the lesion is located in each two-dimensional CT image slice. In the example, the first selection box is the minimum circumscribed rectangle of the area where the lesion is located.

[0042] According to an embodiment of the present invention, in step S103, the number of first selection boxes, the area information of each first selection box, and the centroid position information of each first selection box are determined. The number and area information of the first selection boxes in each two-dimensional CT image slice can be determined. For example, the area information of the first selection box can be represented by the number of pixels included in the first selection box. The centroid position information can also be represented by the horizontal and vertical coordinates of the centroid of each first selection box in the two-dimensional CT image slice.

[0043] In this way, the lung two-dimensional CT image slices of patients with a history of rectal cancer can be obtained, and the first selection box can be determined based on the lesion detection model, and then the number of first selection boxes and the area information of each first selection box can be determined, and the centroid position information of each first selection box can be determined based on the horizontal and vertical coordinates. Provide basic data for the evaluation of the risk of lung metastasis of rectal cancer.

[0044] According to an embodiment of the present invention, in step S104, determining a lesion diffusion score according to the number of the first selection boxes, the area information, and the centroid position information includes: determining a first number of target two-dimensional CT image slices in which the first selection boxes exist in multiple two-dimensional CT image slices; obtaining a second contour of the lungs in the target two-dimensional CT image slice through a contour detection model; determining a second selection box that can include the centroids of all first selection boxes in the target two-dimensional CT image slice according to the centroid position information of the first selection boxes in the target two-dimensional CT image slice, wherein the second selection box is a minimum rectangular box that selects the centroids of all first selection boxes in the target two-dimensional CT image slice; determining a lesion diffusion score according to the first number, the second contour, the second selection box, the number of the first selection boxes, and the area information.

[0045] According to an embodiment of the present invention, among multiple lung two-dimensional CT image slices, there may be two-dimensional CT images without the first selection frame, that is, there is no tumor in the two-dimensional CT image slice, and these slices do not need to be evaluated. Therefore, the first number of target two-dimensional CT image slices with the first selection frame can be determined, and the target two-dimensional CT image slices can be evaluated. The contour line at the junction of the lung and the pulmonary membrane in the target two-dimensional CT image slice can be determined by contour detection models such as Canny edge detection and Sobel operator, that is, the second contour of the lung in the target two-dimensional CT image slice. The present invention does not limit the specific method for determining the second contour. Since the positions of the first selection frames in each target two-dimensional CT image slice may be different, that is, the distribution of tumors in each target two-dimensional CT image slice may be different, therefore, multiple minimum rectangular frames that can include the centroids of all first selection frames in each target two-dimensional CT image slice can be determined, that is, the second selection frame. The larger the second selection frame, the more dispersed the distribution of the first selection frame in the target two-dimensional CT image slice, the more dispersed the distribution of the tumor, and the larger the distribution range.

[0046] According to an embodiment of the present invention, determining a lesion diffusion score according to the first number, the second outline, the second selection box, the number of the first selection box and the area information includes: determining a lesion diffusion score S according to formula (1): D ,

[0047]

[0048] Among them, Ω S,2,i is the area selected by the second selection box in the i-th target 2D CT image slice, Ω O,2,i is the area within the second contour of the lung in the i-th target 2D CT image slice, S(Ω S,2,i ∩Ω O,2,i ) is Ω S,2,i With Ω O,2,i The area information of the intersection, S(Ω S,2,i ) is Ω S,2,i Area information, Ω S,1,i,j is the area selected by the jth first selection box in the i-th target two-dimensional CT image slice, is the nth target 2D CT image slice i,1 The area information of the union of the areas selected by the first selection boxes, n i,1 is the number of first selection boxes in the i-th target 2D CT image slice, S(Ω S,1,i,j ) is Ω S,1,i,j area information, max is the maximum value function, min is the minimum value function, α1, α2 and α3 are preset weights, N1 is the first number, j≤ni,1 , i≤N1, and i, j, n i,1 and N1 are both positive integers.

[0049] According to an embodiment of the present invention, in formula (1), α1, α2 and α3 are preset weights. In formula (1), S(Ω S,2,i ∩Ω O,2,i ) can be expressed as Ω S,2,i With Ω O,2,i The area information of the intersection of , that is, the area information of the intersection of the area selected by the second selection box in the i-th target two-dimensional CT image slice and the area within the second contour of the lung. Therefore, It can represent the area information of the intersection of the area selected by the second selection box in the i-th target two-dimensional CT image slice and the area within the second contour of the lung, and the ratio of the area information of the area within the second contour of the lung, that is, the ratio of the area information of the lung tumor distribution area in the i-th target two-dimensional CT image slice to the total area information of the lung. The larger the above ratio information is, the more dispersed the distribution of the tumor in the i-th target two-dimensional CT image slice in the lung is, and the larger the distribution range is. Furthermore, the above ratio information is large, which may also be caused by a small number of tumors being far apart. For example, the area selected by the second selection box in the i-th target two-dimensional CT image slice only includes the centroids of the three first selection boxes, but the distance between the three centroids is far. This situation will still cause the above ratio information to be large. Therefore, in addition to the above ratio information, it is also necessary to further evaluate the diffusion of the lesion. In formula (1), It can represent the area information of the union of the areas selected by all the first selection boxes in the i-th target two-dimensional CT image slice. Therefore, It can represent the area information of the union of the areas selected by all the first selection boxes in the i-th target two-dimensional CT image slice, and the ratio of the area information of the area within the second contour of the lung, that is, the ratio of the total area information of the lesion area in the i-th target two-dimensional CT image slice to the total area information of the lung. The larger the ratio, the more tumors there are in the i-th target two-dimensional CT image slice, and the larger the total area of ​​the lesion area where the tumor is located relative to the total area of ​​the lung. Therefore, the diffusivity of the lesion can be evaluated based on the above two ratios. If both ratios are high, the number of lesions in the target two-dimensional CT image slice is large and the distribution range is wide, that is, the lesions are widely distributed in the lungs and the diffusivity is high.

[0050] According to an embodiment of the present invention, in formula (1), It can represent the difference between the maximum and minimum values ​​of the area information of the region selected by the jth first selection box in the i-th target two-dimensional CT image slice. Therefore,

[0051] It can represent the difference between the maximum and minimum values ​​of the area information of the area selected by the jth first selection box in the i-th target two-dimensional CT image slice, and the ratio of the maximum value. The larger the ratio, the larger the area difference between the area where the largest tumor is located and the area where the smallest tumor is located, that is, the more uneven the size of the tumor, the stronger the independence between the tumors, and the stronger the diffusion. Therefore, It can be used as the diffusion score of the i-th target two-dimensional CT image slice. The diffusion scores of each target two-dimensional image slice are averaged to obtain the lesion diffusion score. The higher the lesion diffusion score, the higher the lesion diffusion of the patient, the wider the lesion distribution range, the greater the number, and the stronger the independence of the lesions, the higher the risk of rectal cancer metastasis to the lungs. Conversely, the lower the lesion diffusion score, the lower the lesion diffusion of the patient, and the lower the risk of rectal cancer metastasis to the lungs.

[0052] In this way, the second contour of the lung in the target two-dimensional CT image slice can be obtained based on the contour detection model, and the distribution, quantity and independence of the tumor can be obtained according to the first number, the second contour, the second selection box, the number of the first selection box and the area information, so as to evaluate the diffusion of the lesion, and avoid the situation where the number of tumors is small but the distribution is too dispersed, resulting in overestimation of the diffusion of the lesion, and then the lesion diffusion score of each target two-dimensional CT image slice can be determined, so as to obtain the lesion diffusion score through averaging processing. The lesion diffusion score can accurately and objectively represent the lesion diffusion of the patient's lungs.

[0053] According to an embodiment of the present invention, in step S104, determining a lesion isolation score according to the position information of the first selection box and the centroid position information of the first selection box includes: determining a first number of target two-dimensional CT image slices having the first selection box in multiple two-dimensional CT image slices; determining a first target selection box whose intersection with the j-th first selection box is not an empty set; binarizing an image block in an area where the union of the j-th first selection box and the first target selection box is located to obtain a first binarized image block; determining, according to the centroid position information of the first selection box, in the first binarized image block, a connectivity relationship between the centroid position information of the j-th first selection box and the centroid position information of each first target selection box; and determining a lesion isolation score according to the first number and the connectivity relationship.

[0054] According to an embodiment of the present invention, among multiple lung two-dimensional CT image slices, there may be two-dimensional CT images without the first selection box, that is, there is no tumor in the two-dimensional CT image slice, which does not need to be evaluated. Therefore, the first number of target two-dimensional CT image slices with the first selection box can be determined, and the target two-dimensional CT image slices can be evaluated. Determine the first target selection box whose intersection with the j-th first selection box is not an empty set, that is, the first target selection box and the lesion selected by the j-th first selection box may have a suspected connection relationship with the lesion selected by the first target selection box, and further evaluation of the isolation of the lesion is required. Binarize the image block in the area where the union between the j-th first selection box and the first target selection box is located, for example, the pixel value of the pixel point at the location of the lesion is determined to be 1, and the pixel value of the pixel point at the background is determined to be 0, to obtain the first binary image block. Determine the connectivity relationship between the centroid position information of the jth first selection box and the centroid position information of each first target selection box, that is, whether the centroid of the jth first selection box and the centroid of each first target selection box can be connected by a pixel point with a pixel value of 1 to form a path. If the above path exists, the lesion selected by the jth first selection box and the lesion selected by the first target selection box have a connectivity relationship, otherwise, if the above path does not exist, the lesion selected by the jth first selection box and the lesion selected by the first target selection box do not have a connectivity relationship.

[0055] According to an embodiment of the present invention, determining the lesion isolation score according to the first quantity and the connectivity relationship includes: determining the lesion isolation score S according to formula (2): S ,

[0056]

[0057] Among them, n i,1 is the number of first selection boxes in the i-th target 2D CT image slice, L i,j,k is the connectivity relationship between the kth first target selection box and the jth first selection box whose intersection with the jth first selection box in the i-th target 2D CT image slice is not an empty set, where there is a connectivity relationship between the kth first target selection box and the jth first selection box, then L i,j,k =1, otherwise, L i,j,k =0,n i,j is the number of first target selection boxes in the i-th target 2D CT image slice whose intersection with the j-th first selection box is not an empty set, N1 is the first number, j≤n i,1 , i≤N1, k≤n i,j , and i, j, k, n i,1 、n i,j and N1 are both positive integers.

[0058] According to an embodiment of the present invention, in formula (2), L i,j,k is the connectivity relationship between the kth first target selection box and the jth first selection box in the i-th target 2D CT image slice whose intersection with the j-th first selection box is not an empty set. When there is a connectivity relationship between the kth first target selection box and the j-th first selection box, L i,j,k The function value of L is 1, that is, the number of non-isolated tumors needs to be counted. On the contrary, when there is no connectivity between the kth first target selection box and the jth first selection box, L i,j,k The function value of is 0, that is, there is no need to count the number of non-isolated tumors. By counting all the first target selection boxes that are connected to the jth first selection box, the total number of tumors that are connected to the tumor selected by the jth first selection box can be obtained. By summing all the first target selection boxes that are connected to the first selection box in the i-th target two-dimensional CT image slice, the total number of lesions that are connected to each other in the i-th target two-dimensional CT image slice can be obtained. However, there are repeated counts in this total number. For example, when counting the connectivity between the jth tumor and other tumors, the connectivity between the jth tumor and the kth tumor is counted. When counting the connectivity between the kth tumor and other tumors, the connectivity between the kth tumor and the jth tumor is counted again. Therefore, the total number of connectivity between the above tumors and other tumors is repeated, and the actual number of lesions that are connected to each other in the i-th target two-dimensional CT image slice is half of the above total number. Therefore, It can represent the actual number of lesions that are connected to each other in the i-th target 2D CT image slice. In formula (2), It can be expressed as the actual number of lesions in the i-th target 2D CT image slice if the connected lesions are regarded as an isolated lesion. The ratio between the actual number of isolated lesions in the i-th target two-dimensional CT image slice and the total number of lesions can be used as the lesion isolation score of the i-th target two-dimensional CT image slice. The lesion isolation score of all target two-dimensional CT image slices is averaged to obtain the lesion isolation score. The higher the lesion isolation score, the fewer lesions with connectivity, the higher the isolation of the lesions, and the higher the risk of rectal cancer metastasis to the lungs. Conversely, the lower the lesion isolation score, the lower the risk of rectal cancer metastasis to the lungs.

[0059] In this way, the first target selection frame whose intersection with each first selection frame is not an empty set can be determined, and the union is binarized, and then the connectivity between the centroid position information of each first selection frame and the centroid position information of the first target selection frame is determined, so as to determine the total number of lesions that are connected to each other, and remove the repeated counts, and obtain the actual number of isolated lesions in each target two-dimensional CT image slice, so that the count is more accurate. Then, an accurate lesion isolation score is determined, so that the lesion isolation score can accurately and objectively represent the isolation of the lesion.

[0060] According to an embodiment of the present invention, in step S106, similar to the second contour, the first contour of the region where the pleura is located in each two-dimensional CT image slice is determined by a contour detection model. The contour line at the junction of the lung and the pulmonary membrane in the target two-dimensional CT image slice can be determined by contour detection models such as Canny edge detection and Sobel operator, that is, the first contour of the lung in the target two-dimensional CT image slice. The present invention does not limit the specific method of contour detection.

[0061] According to an embodiment of the present invention, in step S107, a pleural invasion score is determined according to the first contour, including: determining a first number of target two-dimensional CT image slices having a first selection box among multiple two-dimensional CT image slices; evenly setting a plurality of nodes on the first contour of the target two-dimensional CT image slice; acquiring coordinate data of the plurality of nodes in the target two-dimensional CT image slice; and determining the pleural invasion score according to the coordinate data and the first number.

[0062] According to an embodiment of the present invention, among multiple lung two-dimensional CT image slices, there may be two-dimensional CT images in which the first selection box does not exist, that is, there is no tumor in the two-dimensional CT image slice, which does not need to be evaluated. Therefore, the first number of target two-dimensional CT image slices in which the first selection box exists can be determined, and the target two-dimensional CT image slices can be evaluated. On the first contour of the target two-dimensional CT image slice, a plurality of nodes are evenly set. For example, on the first contour of the target two-dimensional CT image slice, a node is set every 2 mm, and a plurality of evenly distributed nodes can be obtained for judging the contour change of the pleura. The coordinate data of the plurality of nodes in the target two-dimensional CT image slice are obtained.

[0063] According to an embodiment of the present invention, determining the pleural invasion score according to the coordinate data and the first quantity includes: determining the pleural invasion score S according to formula (3): P ,

[0064]

[0065] in, is the vector between the sth node and the s+1th node in the i-th target 2D CT image slice, P i,s is the coordinate data of the sth node in the i-th target 2D CT image slice, P i,s+1 is the coordinate data of the s+1th node in the i-th target 2D CT image slice, P i,s+2 is the coordinate data of the s+2th node in the i-th target two-dimensional CT image slice, for and The cosine similarity of P is the preset angle, n i,p is the preset number corresponding to the i-th target two-dimensional CT image slice, if is the conditional function, m i is the number of nodes in the i-th target 2D CT image slice, N1 is the first number, s≤m i -2, i≤N1, and i, s, m i and N1 are both positive integers.

[0066] According to an embodiment of the present invention, in formula (3), since for and The cosine similarity of , therefore, Can be expressed and The angle formed is the angle between the vector from the sth node to the s+1th node in the i-th target 2D CT image slice and the vector from the s+1th node to the s+2th node. In formula (3), θ P It can represent a preset angle, and the present invention does not limit the size of the preset angle. In the conditional function, when the angle formed by the vector between the sth node to the s+1th node in the i-th target two-dimensional CT image slice and the vector between the s+1th node to the s+2th node is greater than the preset angle, it means that the angle formed by the vector between the sth node to the s+1th node and the vector between the s+1th node to the s+2th node is too large, and there are abnormal protrusions on the pleura, and there may be a tumor on the pleura. Therefore, the number of abnormal protrusions needs to be counted, and the value of the conditional function is 1. Therefore, It indicates the total number of abnormal protrusions on the pleura.

[0067] According to an embodiment of the present invention, in formula (3), n i,pis the preset number corresponding to the i-th target two-dimensional CT image slice. Since the pleural contour itself is an irregular shape, there are parts with large angle changes. Therefore, when counting the number of abnormal protrusions formed by tumors on the pleura, it is necessary to subtract the count caused by the large angle caused by the irregular shape of the pleural contour itself, that is, it is necessary to subtract the preset number corresponding to the i-th target two-dimensional CT image slice. Therefore, It indicates the actual number of abnormal protrusions on the pleura due to tumor formation. It can represent the ratio of the actual number of abnormal protrusions formed on the pleura due to the tumor to the preset number, and can be used as the pleural invasion score of the i-th target two-dimensional CT image slice. The pleural invasion score of each target two-dimensional CT image slice is averaged to obtain the pleural invasion score. The higher the pleural invasion score, the more tumors there are on the pleura, and the higher the degree of pleural invasion. Conversely, the lower the pleural invasion score, the fewer tumors there are on the pleura, and the lower the degree of pleural invasion.

[0068] In this way, multiple nodes can be evenly set on the first contour of the target two-dimensional CT image slice and the coordinate data of each node can be determined. The abnormal protrusions on the pleura are counted based on the preset angle and the angle between the vectors of each node, and the situation where the number of tumors on the pleura is overestimated due to the abnormal protrusions formed by the irregular shape of the pleura itself is avoided. The actual number of abnormal protrusions on the pleura caused by the tumor in each target two-dimensional CT image slice can be obtained, making the counting more accurate, and then determining the accurate pleural invasion score, so that the pleural invasion score can accurately and objectively represent the degree of pleural invasion by the tumor.

[0069] According to an embodiment of the present invention, in step S108, the risk score of lung metastasis of rectal cancer is determined according to the lesion diffusion score, the lesion isolation score and the pleural invasion score, and a test report is obtained. The risk score of lung metastasis of rectal cancer can be obtained by weighted summing the lesion diffusion score, the lesion isolation score and the pleural invasion score. The higher the risk score of lung metastasis of rectal cancer, the higher the risk of lung metastasis of rectal cancer, and a test report can be generated to prompt doctors or patients of the risk of lung metastasis of rectal cancer. Therefore, a test report can be obtained based on the above-mentioned risk of lung metastasis of rectal cancer, and the test report can describe the high and low risk of lung metastasis of rectal cancer.

[0070] In this way, the lesion diffusion score, lesion isolation score and pleural invasion score can be weighted and summed to determine the risk score of rectal cancer lung metastasis, and generate a corresponding test report message. The rectal cancer lung metastasis risk score can accurately describe the results of tumor recurrence and metastasis monitoring, and according to the corresponding test report, it can prompt doctors or patients of the risk of rectal cancer lung metastasis.

[0071] According to a method for monitoring tumor recurrence and metastasis of an embodiment of the present invention, multiple two-dimensional CT image slices of a lung CT image can be obtained, a first selection box can be determined through a lesion detection model, and a lesion diffusion score and a lesion isolation score can be determined based on the number of first selection boxes, the area information of each first selection box, and the centroid position information of each first selection box. The pleural invasion score is determined based on the first contour determined by the contour detection model, and then the rectal cancer lung metastasis risk score is determined and a detection report is generated. The rectal cancer lung metastasis risk can be evaluated from multiple aspects such as lesion diffusion, lesion isolation, and pleural invasion degree, which is helpful to accurately judge the rectal cancer lung metastasis risk, improve the efficiency of rectal cancer lung metastasis risk monitoring, and do not rely on personnel experience, improve detection accuracy, and reduce the probability of misidentification. Further, a lung two-dimensional CT image slice of a patient with a history of rectal cancer can be obtained, and a first selection box can be determined based on a lesion detection model, and then the number of first selection boxes and the area information of each first selection box are determined, and the centroid position information of each first selection box is determined based on the horizontal and vertical coordinates. Provide basic data for the assessment of the risk of lung metastasis of rectal cancer. When determining the lesion diffusion score, the second contour of the lung in the target two-dimensional CT image slice can be obtained based on the contour detection model, and the distribution, quantity and independence of the tumor can be obtained according to the first quantity, the second contour, the second selection box, the number of the first selection box and the area information, and then the diffusion of the lesion can be evaluated, and the situation where the tumor quantity is small but the distribution is too dispersed, resulting in an overestimation of the lesion diffusion, can be avoided, and then the lesion diffusion score of each target two-dimensional CT image slice can be determined, and the lesion diffusion score can be obtained by averaging. The lesion diffusion score can accurately and objectively represent the lesion diffusion of the patient's lungs. When determining the lesion isolation score, the first target selection box whose intersection with each first selection box is not an empty set can be determined, and the union can be binarized to determine the connectivity between the centroid position information of each first selection box and the centroid position information of the first target selection box, thereby determining the total number of lesions that are connected to each other, and removing the repeated counts, and obtaining the actual number of isolated lesions in each target two-dimensional CT image slice, making the count more accurate. Then, an accurate lesion isolation score is determined, so that the lesion isolation score can accurately and objectively represent the isolation of the lesion. When determining the pleural invasion score, multiple nodes can be evenly set on the first contour of the target two-dimensional CT image slice and the coordinate data of each node can be determined. The abnormal protrusions on the pleura are counted based on the preset angle and the angle between the vectors of each node, and the overestimation of the number of tumors on the pleura due to the abnormal protrusions formed by the irregular shape of the pleura itself is avoided. The actual number of abnormal protrusions on the pleura caused by the tumor in each target two-dimensional CT image slice can be obtained, making the counting more accurate, and then determining the accurate pleural invasion score, so that the pleural invasion score can accurately and objectively represent the degree of pleural invasion by the tumor.When obtaining the test report, the lesion diffusion score, lesion isolation score and pleural invasion score can be weighted and summed to determine the risk score of rectal cancer lung metastasis, and generate the corresponding test report message. The rectal cancer lung metastasis risk score can accurately describe the results of tumor recurrence and metastasis monitoring, and according to the corresponding test report, it can prompt doctors or patients the risk of rectal cancer lung metastasis.

[0072] Figure 2 A schematic diagram of a tumor recurrence and metastasis monitoring system according to an embodiment of the present invention is exemplarily shown, wherein the system comprises:

[0073] A two-dimensional CT image slice acquisition module is used to acquire a lung CT image of a patient with a history of rectal cancer, wherein the lung CT image is a three-dimensional image and includes a plurality of two-dimensional CT image slices;

[0074] A first selection box determination module, which determines a first selection box of a lesion area in each two-dimensional CT image slice through a lesion detection model, wherein the first selection box is a minimum rectangular box for selecting the lesion area;

[0075] A module for determining the number of first selection boxes, the area information of each first selection box, and the centroid position information of each first selection box, to determine the number of first selection boxes, the area information of each first selection box, and the centroid position information of each first selection box;

[0076] a lesion diffusion score determination module, which determines a lesion diffusion score according to the number of the first selection boxes, the area information and the centroid position information;

[0077] A lesion isolation score determination module determines the lesion isolation score according to the position information of the first selection box and the centroid position information of the first selection box;

[0078] A first contour determination module, which determines a first contour of a region where the pleura is located in each two-dimensional CT image slice by using a contour detection model;

[0079] a pleural invasion score determination module, which determines a pleural invasion score according to the first contour;

[0080] The rectal cancer lung metastasis risk score determination module determines the rectal cancer lung metastasis risk score according to the lesion diffusion score, the lesion isolation score and the pleural invasion score, and obtains a detection report.

[0081] It should be understood by those skilled in the art that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functional and structural principles of the present invention have been demonstrated and explained in the embodiments, and the embodiments of the present invention may be deformed or modified in any way without departing from the principles.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring tumor recurrence and metastasis, characterized in that: include: Acquire a lung CT image of a patient with a history of rectal cancer, wherein the lung CT image is a three-dimensional image, and the lung CT image includes a plurality of two-dimensional CT image slices; Determine a first selection box of the area where the lesion is located in each two-dimensional CT image slice by using the lesion detection model, wherein the first selection box is a minimum rectangular box that selects the area where the lesion is located; Determine the number of first selection boxes, area information of each first selection box, and centroid position information of each first selection box; Determining a lesion diffusion score according to the number of the first selection boxes, the area information, and the centroid position information; Determine the lesion isolation score according to the position information of the first selection box and the centroid position information of the first selection box; Determine a first contour of a region where the pleura is located in each two-dimensional CT image slice by using a contour detection model; Determining a pleural invasion score based on the first outline; Determine the risk score of lung metastasis of rectal cancer according to the lesion diffusion score, the lesion isolation score and the pleural invasion score, and obtain a test report; Based on the first outline, a pleural invasion score is determined, including: Determining, among the plurality of two-dimensional CT image slices, a first number of target two-dimensional CT image slices in which a first selection frame exists; Evenly setting a plurality of nodes on a first contour of the target two-dimensional CT image slice; Acquire coordinate data of a plurality of nodes in the target two-dimensional CT image slice; determining a pleural invasion score based on the coordinate data and the first quantity; Determining a pleural invasion score according to the coordinate data and the first quantity includes: According to the formula Determine the pleural invasion score S P ,in, is the vector between the sth node and the s+1th node in the i-th target 2D CT image slice, P i,s is the coordinate data of the sth node in the i-th target 2D CT image slice, P i,s+1 is the coordinate data of the s+1th node in the i-th target 2D CT image slice, P i,s+2 is the coordinate data of the s+2th node in the i-th target two-dimensional CT image slice, and The cosine similarity of P is the preset angle, n i,p is the preset number corresponding to the i-th target two-dimensional CT image slice, if is the conditional function, m i is the number of nodes in the i-th target 2D CT image slice, N1 is the first number, s≤m i -2, i≤N1, and i, s, m i and N1 are both positive integers.

2. The method for monitoring tumor recurrence and metastasis according to claim 1, characterized in that: Determining a lesion diffusion score according to the number of the first selection boxes, the area information, and the centroid position information includes: Determining, among the plurality of two-dimensional CT image slices, a first number of target two-dimensional CT image slices in which a first selection frame exists; Obtaining a second contour of the lung in the target two-dimensional CT image slice by using a contour detection model; Determining, according to the centroid position information of the first selection box in the target two-dimensional CT image slice, a second selection box that can include the centroids of all the first selection boxes in the target two-dimensional CT image slice, wherein the second selection box is a minimum rectangular box that selects the centroids of all the first selection boxes in the target two-dimensional CT image slice; A lesion diffusion score is determined according to the first number, the second outline, the second selection box, the number of the first selection boxes, and the area information.

3. The method for monitoring tumor recurrence and metastasis according to claim 2, characterized in that: Determining a lesion diffusion score according to the first number, the second outline, the second selection box, the number of the first selection boxes, and the area information includes: According to the formula Determine the diffuseness score S D , where Ω S,2,i is the area selected by the second selection box in the i-th target 2D CT image slice, Ω O,2,i is the area within the second contour of the lung in the i-th target 2D CT image slice, S(Ω S,2,i ∩Ω O,2,i ) is Ω S,2,i With Ω O,2,i The area information of the intersection, S(Ω S,2,i ) is Ω S,2,i Area information, Ω S,1,i,j is the area selected by the jth first selection box in the i-th target two-dimensional CT image slice, is the nth target 2D CT image slice i,1 The area information of the union of the areas selected by the first selection boxes, n i,1 is the number of first selection boxes in the i-th target 2D CT image slice, S(Ω S,1,i,j ) is Ω S,1,i,j area information, max is the maximum value function, min is the minimum value function, α1, α2 and α3 are preset weights, N1 is the first number, j≤n i,1 , i≤N1, and i, j, n i,1 and N1 are both positive integers.

4. The method for monitoring tumor recurrence and metastasis according to claim 1, characterized in that: According to the position information of the first selection box and the centroid position information of the first selection box, the lesion isolation score is determined, including: Determining, among the plurality of two-dimensional CT image slices, a first number of target two-dimensional CT image slices in which a first selection frame exists; Determine a first target selection box corresponding to the j-th first selection box, wherein an intersection between the first target selection box corresponding to the j-th first selection box and the j-th first selection box is not an empty set; Binarize the image block in the area where the union of the j-th first selection box and the first target selection box is located to obtain a first binary image block; According to the centroid position information of the first selection box, determining, in the first binarized image block, a connectivity relationship between the centroid position information of the j-th first selection box and the centroid position information of each first target selection box; A lesion isolation score is determined according to the first quantity and the connectivity relationship.

5. The method for monitoring tumor recurrence and metastasis according to claim 4, characterized in that: Determining a lesion isolation score according to the first quantity and the connectivity relationship includes: According to the formula Determine the lesion isolation score S S , where n i,1 is the number of first selection boxes in the i-th target 2D CT image slice, L i,j,k is the connectivity relationship between the kth first target selection box and the jth first selection box in the i-th target 2D CT image slice, where there is a connectivity relationship between the kth first target selection box and the jth first selection box, then L i,j,k =1, otherwise, L i,j,k =0,n i,j is the number of first target selection boxes in the i-th target 2D CT image slice, N1 is the first number, j≤n i,1 , i≤N1, k≤n i,j , and i, j, k, n i,1 、n i,j and N1 are both positive integers.

6. A tumor recurrence and metastasis monitoring system for executing the method according to any one of claims 1 to 5, characterized in that: include: A two-dimensional CT image slice acquisition module is used to acquire a lung CT image of a patient with a history of rectal cancer, wherein the lung CT image is a three-dimensional image and includes a plurality of two-dimensional CT image slices; A first selection box determination module, which determines a first selection box of a lesion area in each two-dimensional CT image slice through a lesion detection model, wherein the first selection box is a minimum rectangular box for selecting the lesion area; A module for determining the number of first selection boxes, the area information of each first selection box, and the centroid position information of each first selection box, to determine the number of first selection boxes, the area information of each first selection box, and the centroid position information of each first selection box; a lesion diffusion score determination module, which determines a lesion diffusion score according to the number of the first selection boxes, the area information and the centroid position information; A lesion isolation score determination module determines the lesion isolation score according to the position information of the first selection box and the centroid position information of the first selection box; A first contour determination module, which determines a first contour of a region where the pleura is located in each two-dimensional CT image slice by using a contour detection model; a pleural invasion score determination module, which determines a pleural invasion score according to the first contour; The rectal cancer lung metastasis risk score determination module determines the rectal cancer lung metastasis risk score according to the lesion diffusion score, the lesion isolation score and the pleural invasion score, and obtains a detection report.

Images