A method and system for analyzing lesion changes in CT images based on time series

Abstract

The present application provides a method and system for analyzing lesion changes in CT images based on time series, which relates to the field of lesion analysis technology. The method includes: obtaining a CT image of the patient's current examination, and obtaining multiple historical CT images of the patient; obtaining the lesion area where the lesion is located in the CT image through a lesion detection model; obtaining the historical lesion area where the lesion is located in the historical CT image through a lesion detection model; determining the lesion state coefficient of the lesion area, and the historical lesion state coefficient of the historical lesion area; determining the lesion development status; and generating a lesion change analysis report. According to the present application, when determining to analyze the lesion change situation, the lesion area in the CT image and the historical lesion area in the historical CT image can be analyzed, and the lesion state coefficient and the historical lesion state coefficient can be determined to determine the lesion development status. The accuracy of the judgment of the lesion development status is improved, and the efficiency of the lesion change analysis is improved.

Description

Technical Field

[0001] The present application relates to the technical field of lesion analysis, and in particular to a method for analyzing lesion changes in CT images based on time series. Background Art

[0002] When designing further treatment plans for cancer patients, it is necessary to analyze the changes in the patient's lesions. At this time, doctors can only spend a lot of time observing CT images and analyzing the changes in the patient's lesions based on the doctor's experience. The analysis speed is slow and depends on the doctor's experience. Therefore, there is a possibility of inaccurate analysis of lesion changes and low analysis efficiency. It is difficult to accurately judge the changes in the patient's lesions, resulting in low accuracy and efficiency in the analysis of the patient's lesion changes, which affects the patient's further treatment. Summary of the Invention

[0003] The present application provides a time-series-based CT image lesion change analysis method, which can solve the technical problem that related technologies are difficult to accurately judge the lesion changes of patients.

[0004] According to a first aspect of the present application, a time-series-based CT image lesion change analysis method is provided, comprising:

[0005] After obtaining the CT image of the patient's current examination, multiple historical CT images of the patient are obtained from the database;

[0006] The CT image is detected by the lesion detection model to obtain the lesion area where the lesion is located in the CT image;

[0007] Through the lesion detection model, the historical CT images are detected to obtain the historical lesion area where the lesions are located in the historical CT images;

[0008] Determine the lesion state coefficient of the lesion area and the historical lesion state coefficient of the historical lesion area;

[0009] Determine the lesion development status based on the lesion status coefficient, the historical lesion status coefficient, the examination time of the CT image, and the examination time of the historical CT image;

[0010] Generate a lesion change analysis report based on the lesion development status.

[0011] According to this application, determining the lesion status coefficient of the lesion area includes:

[0012] determining a lesion size state coefficient of the lesion region according to the area of ​​the lesion region and a preset lesion area threshold;

[0013] Obtain the outline of the lesion area;

[0014] Multiple sampling points are set on the contour of the lesion area;

[0015] Determine the sampling vector between adjacent sampling points;

[0016] determining a lesion morphological state coefficient of the lesion area according to the sampling vector;

[0017] Determine the lesion shape state coefficient according to the lesion size state coefficient and the lesion morphology state coefficient;

[0018] Determining a lesion density state coefficient of the lesion area according to the CT value of the centroid of the lesion area and a preset CT threshold;

[0019] Connecting the centroid of the lesion area with the sampling point to obtain a detection line;

[0020] According to the CT value of the pixel point on the detection line, the lesion uniformity state coefficient of the lesion area is determined;

[0021] Determine the lesion CT value state coefficient according to the lesion density state coefficient and the lesion uniformity state coefficient;

[0022] The lesion state coefficient is determined according to the lesion appearance state coefficient and the lesion CT value state coefficient.

[0023] According to the present application, determining the lesion morphological state coefficient of the lesion area according to the sampling vector includes:

[0024] According to the formula

[0025]

[0026] Determine the lesion morphological state coefficient of the lesion area ,in, is the coordinate of the i+1th sampling point, is the coordinate of the i-th sampling point, is the sampling vector between the i+1th sampling point and the i-th sampling point, is the coordinate of the i-1th sampling point, is the sampling vector between the i-th sampling point and the i-1-th sampling point, for The transposed vector of , n is the number of sampling points, min is the minimum function, and both i and n are positive integers.

[0027] According to the present application, the lesion uniformity state coefficient of the lesion area is determined based on the CT value of the pixel point on the detection line, including:

[0028] Set multiple detection points evenly on the detection line and determine the CT value of the detection points;

[0029] Divide the detection points with the same serial number on each detection line into a group to obtain multiple detection point groups;

[0030] A lesion uniformity state coefficient of the lesion area is determined according to the CT value of each detection point in the plurality of detection point groups.

[0031] According to the present application, the lesion uniformity state coefficient of the lesion area is determined based on the CT value of each detection point in the plurality of detection point groups, including:

[0032] Determine the CT average value of each detection point in the detection point group;

[0033] Determine the standard deviation of the CT value of each detection point in the detection point group according to the CT average value of each detection point in the detection point group;

[0034] According to the CT value standard deviation and the CT average value, the dispersion coefficient of the CT value of each detection point in the detection point group is determined;

[0035] The maximum value of the discrete coefficients corresponding to the plurality of detection point groups is determined as the lesion uniformity state coefficient of the lesion area.

[0036] According to the present application, the lesion state coefficient includes the lesion appearance state coefficient and the lesion CT value state coefficient, and the historical lesion state coefficient includes the historical lesion appearance state coefficient and the historical lesion CT value state coefficient;

[0037] The lesion development status is determined based on the lesion status coefficient, the historical lesion status coefficient, the examination time of the CT image, and the examination time of the historical CT image, including:

[0038] According to the formula

[0039]

[0040] Determine the lesion condition score of the CT image of this examination ,in, is the lesion shape coefficient, is the CT value status coefficient of the lesion, and is the preset weight;

[0041] According to the formula

[0042]

[0043] Determine the historical lesion condition score of the jth historical CT image ,in, is the j-th historical lesion shape state coefficient, is the CT value status coefficient of the jth historical lesion;

[0044] Establish a coordinate system and determine the coordinates of the lesion state coefficient and each historical lesion state coefficient in the coordinate system, wherein the lesion shape state coefficient is the X value in the coordinate of the lesion state coefficient, the lesion CT value state coefficient is the Y value in the coordinate of the lesion state coefficient, the historical lesion shape state coefficient is the X value in the coordinate of the historical lesion state coefficient, and the historical lesion CT value state coefficient is the Y value in the coordinate of the historical lesion state coefficient;

[0045] Determine the area to be healed in the coordinate system according to the lesion condition score;

[0046] Determine the historical area to be recovered in the coordinate system according to the historical lesion condition score;

[0047] The development status of the lesion is determined based on the area to be recovered and the historical area to be recovered.

[0048] According to this application, the area to be rehabilitated in the coordinate system is determined based on the lesion condition score, including:

[0049] The coordinates of the lesion state coefficient in the coordinate system are used to make the slope The current lesion condition is equivalent to the straight line;

[0050] The area enclosed by the current lesion condition equivalent straight line, X axis and Y axis is determined as the area to be recovered.

[0051] According to this application, the development status of the lesion is determined based on the area to be rehabilitated and the historical area to be rehabilitated, including:

[0052] According to the formula

[0053]

[0054] Determine the lesion development coefficient D, where is the area of ​​the area to be rehabilitated, is the area of ​​the historical area to be recovered corresponding to the m-1th historical CT image, is the inspection time of the CT image, is the inspection time of the m-1th historical CT image, m is the number of inspections, is the area of ​​the historical area to be recovered corresponding to the j+1th historical CT image, is the area of ​​the historical area to be recovered corresponding to the j-th historical CT image, is the inspection time of the j+1th historical CT image, is the inspection time of the jth historical CT image, max is the maximum value function, and both j and m are positive integers;

[0055] The lesion development status is determined based on the lesion development coefficient.

[0056] According to a second aspect of the present application, a time-series-based CT image lesion change analysis system is provided, comprising:

[0057] An image acquisition module, after acquiring the CT image of the patient's current examination, acquires multiple historical CT images of the patient from the database;

[0058] The lesion region module detects the CT image through the lesion detection model and obtains the lesion region where the lesion is located in the CT image;

[0059] The historical lesion region module detects historical CT images through the lesion detection model to obtain the historical lesion region where the lesion is located in the historical CT image;

[0060] A lesion state coefficient module is used to determine the lesion state coefficient of the lesion area and the historical lesion state coefficient of the historical lesion area;

[0061] A lesion development status module determines the lesion development status based on the lesion status coefficient, the historical lesion status coefficient, the examination time of the CT image, and the examination time of the historical CT image;

[0062] The lesion change analysis report module generates a lesion change analysis report based on the lesion development status.

[0063] By adopting the above technical solution, this application can achieve the following technical effects:

[0064] According to the present application, a CT image of the patient's current examination and multiple historical CT images of the patient can be obtained. The CT image and the historical CT images are detected using a lesion detection model to obtain the lesion area where the lesion is located in the CT image and the historical lesion area where the lesion is located in the historical CT images, thereby determining the lesion state coefficient of the lesion area and the historical lesion state coefficient of the historical lesion area. Then, based on the changes in the historical lesion state coefficient and the lesion state coefficient, the lesion development status is determined and a lesion change analysis report is generated. This reduces dependence on manpower, improves the accuracy of lesion development status judgment, and improves the efficiency of lesion change analysis. When detecting the lesion area, the CT image of the patient's current examination and multiple historical CT images can be obtained. Based on the lesion detection model, the CT image is detected to obtain the outline of the lesion area, thereby obtaining the lesion area and the historical lesion area, providing accurate basic data for determining the lesion development status. When determining the lesion state coefficient, the lesion size state coefficient of the lesion area can be determined based on the ratio of the area of ​​the lesion area to a preset lesion area threshold. The lesion detection model can be used to determine the contour of the lesion area. Multiple sampling points are set on the contour of the lesion area at uniform angles to determine the sampling vectors between adjacent sampling points. Based on the sampling vectors, the cosine value of the angle between two adjacent sampling vectors at the most irregular position of the lesion shape is obtained. Based on the number of sampling points, the cosine value of the angle between adjacent sampling vectors is determined when the lesion is completely regular. This allows the lesion morphological state coefficient of the lesion area to be accurately described. The lesion shape state coefficient is then determined through weighted summation. Furthermore, the lesion density state coefficient of the lesion area is determined based on the ratio of the CT value of the centroid of the lesion area to a preset CT threshold. This takes into account the situation where lesions are located in different human organs and have different preset CT thresholds for lesions, thereby improving the accuracy of the lesion density state coefficient. Based on the detection points evenly arranged on the detection line, multiple detection point groups are obtained, thereby obtaining the discrete coefficients of the CT values ​​of each detection point in the detection point group. Based on the relationship between the uniformity of the lesion area and the severity of the disease, the lesion uniformity state coefficient is obtained, thereby improving the accuracy of the lesion uniformity state coefficient. The lesion CT value state coefficient is obtained through weighted summation, and then the lesion state coefficient is obtained, which improves the accuracy, objectivity, and comprehensiveness of the lesion state coefficient, allowing the lesion state coefficient to accurately describe the current state of the lesion. When determining the lesion development coefficient, the lesion state score and the historical lesion state score can be determined based on the lesion state coefficient, the historical lesion state coefficient, the examination time of the CT image, and the examination time of the historical CT image. A coordinate system can be established, with the lesion appearance state coefficient as the horizontal axis and the lesion CT value state coefficient as the vertical axis, to determine the coordinates of the lesion state score in the coordinate system, thereby determining the equivalent line of the current lesion state, and then the area enclosed by the equivalent line and the coordinate axis is determined as the area to be recovered.By accounting for equivalent changes in the lesion's external shape coefficient and CT value coefficient, changes in both lesion's external shape and CT value can be simultaneously observed, enhancing comprehensiveness. Furthermore, the rate of change in the area to be healed and the maximum historical rate of change in the area to be healed can be determined, thereby determining the lesion's development coefficient and, consequently, the lesion's development status. This improves the accuracy, comprehensiveness, and objectivity of lesion development status and enhances the efficiency of lesion change analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 The following is a flow chart showing, by way of example, a method for analyzing lesion changes in CT images based on time series according to an embodiment of the present application;

[0066] Figure 2 A flowchart of determining a lesion state coefficient of a lesion area according to an embodiment of the present application is exemplarily shown;

[0067] Figure 3 The following is a block diagram of a time-series-based CT image lesion change analysis system according to an embodiment of the present application. DETAILED DESCRIPTION

[0068] The following specific embodiments are used to describe the technical solution of the present application in detail. 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.

[0069] Figure 1 The following is a flow chart of a method for analyzing lesion changes in CT images based on time series according to an embodiment of the present application, wherein the method includes:

[0070] Step S1, after obtaining the CT image of the patient's current examination, obtain multiple historical CT images of the patient from the database;

[0071] Step S2, detecting the CT image using a lesion detection model to obtain a lesion region where the lesion is located in the CT image;

[0072] Step S3, detecting the historical CT images using a lesion detection model to obtain the historical lesion region where the lesion is located in the historical CT images;

[0073] Step S4, determining the lesion state coefficient of the lesion area and the historical lesion state coefficient of the historical lesion area;

[0074] Step S5, determining the lesion development status based on the lesion status coefficient, the historical lesion status coefficient, the examination time of the CT image, and the examination time of the historical CT image;

[0075] Step S6: Generate a lesion change analysis report based on the lesion development status.

[0076] According to the time-series-based CT image lesion change analysis method of the embodiment of the present application, the CT image of the patient's current examination and multiple historical CT images of the patient can be obtained. Through the lesion detection model, the CT image and the historical CT images are detected to obtain the lesion area where the lesion is located in the CT image and the historical lesion area where the lesion is located in the historical CT image, thereby determining the lesion state coefficient of the lesion area and the historical lesion state coefficient of the historical lesion area, and then determining the lesion development status based on the historical lesion state coefficient and the change of the lesion state coefficient, and generating a lesion change analysis report. This reduces dependence on manpower, improves the accuracy of judging the lesion development status, and improves the efficiency of lesion change analysis.

[0077] According to an embodiment of the present application, in step S1, after obtaining the CT image of the patient's current examination, multiple historical CT images of the patient are obtained from the database. The patient may undergo an examination every six months or a year, and a CT image will be taken for each examination. After the patient's current examination and the CT image is obtained, all CT images of the patient taken before the current examination can be obtained from the hospital database, i.e., the patient's multiple historical CT images.

[0078] According to an embodiment of the present application, in step S2, the CT image is detected using a lesion detection model to obtain a lesion region in the CT image where the lesion is located. The CT image is detected using a lesion detection model (e.g., Mask R-CNN, U-Net, etc.) to obtain the contour region of the lesion in the CT image, which is the lesion region where the lesion is located.

[0079] According to an embodiment of the present application, in step S3, the historical CT images are detected using a lesion detection model to obtain the historical lesion region where the lesion was located in the historical CT images. The historical CT images are detected using a lesion detection model (e.g., Mask R-CNN, U-Net, etc.) to obtain the contour region of the lesion in the historical CT images, which is the historical lesion region where the lesion was located. Alternatively, if the historical lesion region in the historical CT images has already been obtained using the lesion detection model during a previous examination and has been stored in a hospital database, the relevant information about the historical lesion region can be directly retrieved from the hospital database during this use.

[0080] In this way, the CT image of the patient's current examination and multiple historical CT images can be obtained, and based on the lesion detection model, the CT image is detected to obtain the outline of the lesion area, thereby obtaining the lesion area and the historical lesion area, providing accurate basic data for determining the development status of the lesion.

[0081] Figure 2 A flow chart for determining a lesion state coefficient of a lesion area according to an embodiment of the present application is exemplarily shown.

[0082] According to an embodiment of the present application, in step S4, the lesion state coefficient of the lesion area and the historical lesion state coefficient of the historical lesion area are determined, including: step S41, determining the lesion size state coefficient of the lesion area according to the area of ​​the lesion area and the preset lesion area threshold; step S42, obtaining the outline of the lesion area; step S43, setting a plurality of sampling points on the outline of the lesion area; step S44, determining the sampling vector between adjacent sampling points; step S45, determining the lesion morphology state coefficient of the lesion area according to the sampling vector; step S46, determining the lesion size state coefficient according to the lesion morphology state coefficient state coefficient, determine the lesion shape state coefficient; step S47, determine the lesion density state coefficient of the lesion area according to the CT value of the centroid of the lesion area and the preset CT threshold; step S48, connect the centroid of the lesion area with the sampling point to obtain a detection line; step S49, determine the lesion uniformity state coefficient of the lesion area according to the CT value of the pixel point on the detection line; step S410, determine the lesion CT value state coefficient according to the lesion density state coefficient and the lesion uniformity state coefficient; step S411, determine the lesion state coefficient according to the lesion shape state coefficient and the lesion CT value state coefficient.

[0083] According to an embodiment of the present application, in step S41, a lesion size status coefficient for the lesion region is determined based on the area of ​​the lesion region and a preset lesion area threshold. The area of ​​the lesion region can be determined based on the number of pixels within the lesion region. Because some lesions are small and do not affect the patient's lesion progression, attention is only required when the area of ​​the lesion region exceeds the preset lesion area threshold (for example, 0.5 square centimeters), and the impact of the current lesion on the patient's lesion progression is further considered. The ratio of the lesion area to the preset lesion area threshold is the lesion size status coefficient for the lesion region. The larger the lesion size status coefficient, the larger the lesion area, the greater the likelihood that the lesion is a malignant tumor, and the more attention is needed.

[0084] According to an embodiment of the present application, in step S42, the contour of the lesion area is obtained. The contour of the lesion area can be directly determined by detecting the CT image using a lesion detection model (e.g., Mask R-CNN, U-Net, etc.).

[0085] According to an embodiment of the present application, in step S43, a plurality of sampling points are set on the contour of the lesion area. A ray can be emitted from the centroid of the contour of the lesion area in any direction, and the ray can be rotated with the centroid of the contour of the lesion area as the center. For each rotation of a certain angle, the intersection of the ray and the contour of the lesion area is the sampling point. For example, the lesion area is rotated 360° with the centroid of the contour of the lesion area as the center, and the intersection of the ray and the contour of the lesion area is set as a sampling point every 15° rotation, so that 24 sampling points can be set on the contour of the lesion area. The present application does not limit the specific number of sampling points.

[0086] According to an embodiment of the present application, in step S44, a sampling vector is determined between adjacent sampling points. A vector, i.e., a sampling vector, can be determined for every two sampling points. For example, when 24 sampling points are set on the contour of the lesion area, a sampling vector can be determined that points from the first sampling point to the second sampling point, and a vector that points from the second sampling point to the third sampling point, and so on.

[0087] According to an embodiment of the present application, in step S45, the lesion morphological state coefficient of the lesion area is determined according to the sampling vector, including: determining the lesion morphological state coefficient of the lesion area according to formula (1): ,

[0088] (1)

[0089] in, is the coordinate of the i+1th sampling point, is the coordinate of the i-th sampling point, is the sampling vector between the i+1th sampling point and the i-th sampling point, is the coordinate of the i-1th sampling point, is the sampling vector between the i-th sampling point and the i-1-th sampling point, for The transposed vector of , n is the number of sampling points, min is the minimum function, and i and n are both positive integers.

[0090] According to an embodiment of the present application, in formula (1), It represents the cosine similarity between the sampling vector between the i+1th sampling point and the i-th sampling point and the sampling vector between the i-th sampling point and the i-1th sampling point, that is, the cosine value of the angle between the sampling vector between the i+1th sampling point and the i-th sampling point and the sampling vector between the i-th sampling point and the i-1th sampling point. Represents the minimum value of the above cosine value. Since the smaller the cosine value, the larger the corresponding angle, the minimum value of the above cosine value can also be used as the cosine value corresponding to the maximum value of the angle between each adjacent sampling vector. Since the larger the angle between two adjacent sampling vectors, the sharper and more irregular the shape of the lesion, the minimum value of the above cosine value can be used as the cosine value of the angle between two adjacent sampling vectors at the most irregular position of the lesion shape. Since the shape of the lesion is circular when it is completely regular, It can represent the angle between adjacent sampling vectors when the lesion is completely regular, so, It represents the cosine value of the angle between adjacent sampling vectors when the lesion is completely regular. It can represent the ratio of the cosine value of the angle between two adjacent sampling vectors at the most irregular position of the lesion shape to the cosine value of the angle between two adjacent sampling vectors when the lesion is completely regular. The smaller the ratio, the more irregular the shape of the lesion, and the greater the possibility that the lesion will deteriorate into a malignant tumor. As the lesion morphology state coefficient, the larger the lesion morphology state coefficient is, the more irregular the shape of the lesion is, the greater the possibility that the lesion will deteriorate into a malignant tumor, and the more attention it needs.

[0091] According to an embodiment of the present application, in step S46, a lesion shape state coefficient is determined based on the lesion size state coefficient and the lesion shape state coefficient. The lesion size coefficient and the lesion shape state coefficient are weighted averaged to obtain the lesion shape state coefficient. The larger the lesion shape state coefficient, the worse the lesion's shape state is, and the more attention it requires.

[0092] According to an embodiment of the present application, in step S47, a lesion density state coefficient of the lesion area is determined based on the CT value of the centroid of the lesion area and a preset CT threshold. The CT value of the centroid of the lesion area can represent the density at the center of the lesion. When the lesion is located in different human organs, the preset CT threshold may also vary. For example, when the lesion is in the liver, the preset CT threshold may be 30HU, and when the lesion is in the brain, the preset CT threshold may be 40HU. When the CT value of the lesion centroid is greater than the preset CT threshold, the density of the lesion is relatively high and may be a malignant tumor. The lesion density state coefficient of the lesion area is the ratio of the CT value of the centroid of the lesion area to the preset CT threshold. The larger the lesion density state coefficient of the lesion area, the greater the density of the lesion, the greater the possibility that the lesion is a malignant tumor, and the more attention it requires. This application does not impose any restrictions on this.

[0093] According to an embodiment of the present application, in step S48, the centroid of the lesion area is connected to the plurality of sampling points to obtain a plurality of detection lines, wherein the detection lines are composed of the pixel points of the lesion area.

[0094] According to an embodiment of the present application, in step S49, the lesion uniformity state coefficient of the lesion area is determined based on the CT values ​​of the pixel points on the detection line, including: evenly setting multiple detection points on the detection line and determining the CT values ​​of the detection points; dividing the detection points with the same serial number on each detection line into a group to obtain multiple detection point groups; and determining the lesion uniformity state coefficient of the lesion area based on the CT value of each detection point in the multiple detection point groups.

[0095] According to an embodiment of the present application, since the outline of the lesion area may be an irregular figure, the length of each detection line obtained is different. Therefore, the same number of detection points can be evenly set on each detection line, and the CT value of each detection point can be obtained. For example, by evenly setting 5 detection points on each detection line, each detection line can be divided into 6 equal parts. The sequence of the detection points on each detection line can be determined according to the order from the centroid of the lesion area to the outline of the lesion area along the detection line, and the detection points with the same sequence number on each detection line are divided into a group, namely a detection point group. The shape of the figure obtained after connecting the detection points of each detection point group is similar to the outline of the lesion area. This application does not impose any restrictions on this.

[0096] According to an embodiment of the present application, a lesion uniformity state coefficient of a lesion area is determined based on the CT value of each detection point in a plurality of detection point groups, including: determining the average CT value of each detection point in the detection point group; determining the standard deviation of the CT value of each detection point in the detection point group based on the average CT value of each detection point in the detection point group; determining the dispersion coefficient of the CT value of each detection point in the detection point group based on the standard deviation of the CT value and the average CT value; and determining the maximum value of the dispersion coefficients corresponding to the plurality of detection point groups as the lesion uniformity state coefficient of the lesion area.

[0097] According to an embodiment of the present application, an average CT value of each detection point in a detection point group is obtained through averaging. Based on the average CT value of each detection point in the detection point group, the standard deviation of the CT value of each detection point in the detection point group is determined. The ratio of the CT value standard deviation to the CT value average is the coefficient of variation of the CT value of each detection point group. The larger the coefficient of variation of the CT value of a detection point group, the greater the fluctuation of the CT value of the detection point group. The greater the fluctuation of the CT value on the closed curve formed by the detection points in the group, the more uneven the lesion. For example, malignant tumors may develop lesions such as cysts. The CT values ​​of the area where the cysts are located differ significantly from those of the area where the tumor cells are located. If the closed curve formed by a group of detection points passes through the area where the cysts are located, the CT value fluctuation of the detection points in the group is large, and the coefficient of variation is also large. The maximum value of the coefficient of variation corresponding to multiple detection point groups can be determined as the lesion uniformity state coefficient of the lesion area, thereby describing the maximum degree of fluctuation of the pixels within the lesion area. The larger the lesion uniformity state coefficient, the more uneven the lesion, the greater the possibility that the lesion is a malignant tumor, and the more attention it requires.

[0098] According to an embodiment of the present application, in step S410, a lesion CT value state coefficient is determined based on the lesion density state coefficient and the lesion homogeneity state coefficient. The lesion CT value state coefficient is obtained by weighted summing the lesion density state coefficient and the lesion homogeneity state coefficient. The larger the lesion CT value state coefficient, the greater the possibility that the lesion is a malignant tumor, and the worse the lesion's CT value state, the more attention it requires.

[0099] According to an embodiment of the present application, in step S411, a lesion status coefficient is determined based on the lesion appearance status coefficient and the lesion CT value status coefficient. A weighted sum of the lesion appearance status coefficient and the lesion CT value status coefficient is performed to obtain a lesion status score. The higher the lesion status score, the worse the lesion status, the greater the likelihood that the lesion is a malignant tumor, and the more attention it requires. Using the same method as above, the historical lesion status coefficients for each historical CT image can be obtained.

[0100] In this way, the lesion size state coefficient of the lesion region can be determined based on the ratio of the lesion region's area to a preset lesion area threshold. The lesion region's contour can be determined based on the lesion detection model, and multiple sampling points can be set on the lesion region's contour at uniform angles to determine sampling vectors between adjacent sampling points. Based on the sampling vectors, the cosine value of the angle between two adjacent sampling vectors at the most irregular location of the lesion shape is obtained. Based on the number of sampling points, the cosine value of the angle between adjacent sampling vectors is determined when the lesion is completely regular, thereby obtaining the lesion morphology state coefficient of the lesion region. This lesion morphology state coefficient accurately describes the degree of regularity of the lesion shape, and the lesion shape state coefficient is then determined through weighted summation. Furthermore, the lesion density state coefficient of the lesion region is determined based on the ratio of the CT value of the lesion region's centroid to a preset CT threshold. This takes into account the fact that lesions are located in different human organs and have different preset CT thresholds for lesions, thereby improving the accuracy of the lesion density state coefficient. Based on the uniformly arranged detection points on the detection line, multiple detection point groups are obtained, thereby obtaining the dispersion coefficient of the CT value of each detection point in the detection point group. Based on the relationship between the uniformity of the lesion area and the severity of the disease, the lesion uniformity state coefficient is obtained, which improves the accuracy of the lesion uniformity state coefficient. The lesion CT value state coefficient is obtained through weighted summation, and then the lesion state coefficient is obtained. This improves the accuracy, objectivity and comprehensiveness of the lesion state coefficient, allowing the lesion state coefficient to accurately describe the current state of the lesion.

[0101] According to an embodiment of the present application, in step S5, the lesion state coefficient includes a lesion shape state coefficient and a lesion CT value state coefficient, and the historical lesion state coefficient includes a historical lesion shape state coefficient and a historical lesion CT value state coefficient; according to the lesion state coefficient, the historical lesion state coefficient, the examination time of the CT image and the examination time of the historical CT image, the lesion development status is determined, including: determining the lesion status score of the CT image of the current examination according to formula (2) ,

[0102] (2)

[0103] in, is the lesion shape coefficient, is the CT value status coefficient of the lesion, and is the preset weight; the historical lesion status score of the jth historical CT image is determined according to formula (3) ,

[0104] (3)

[0105] in, is the j-th historical lesion shape state coefficient, is the jth historical lesion CT value state coefficient; establish a coordinate system, and determine the lesion state coefficient and the coordinates of each historical lesion state coefficient in the coordinate system, wherein the lesion shape state coefficient is the X value in the coordinate of the lesion state coefficient, the lesion CT value state coefficient is the Y value in the coordinate of the lesion state coefficient, the historical lesion shape state coefficient is the X value in the coordinate of the historical lesion state coefficient, and the historical lesion CT value state coefficient is the Y value in the coordinate of the historical lesion state coefficient; determine the area to be recovered in the coordinate system according to the lesion status score; determine the historical area to be recovered in the coordinate system according to the historical lesion status score; determine the lesion development status according to the area to be recovered and the historical area to be recovered.

[0106] According to an embodiment of the present application, in formula (2), It represents the weighted sum of the lesion appearance coefficient and the lesion CT value coefficient, which can be used as the lesion condition score of the CT image of this examination. Similarly, in formula (3), The weighted summation of the jth historical lesion appearance coefficient and the jth historical lesion CT value coefficient yields the historical lesion condition score for the jth historical CT image. A higher lesion condition score indicates a worsening lesion condition, a greater likelihood of malignancy, and a greater need for attention. Comparing the lesion condition score with the historical lesion condition scores can be used to determine the lesion's progression, for example, whether the lesion is worsening or improving.

[0107] According to an embodiment of the present application, a plane rectangular coordinate system is established, with the lesion shape state coefficient as the X value in the coordinate system and the lesion CT value state coefficient as the Y value in the coordinate system, to determine the coordinates of each lesion state coefficient. Similarly, the historical lesion shape state coefficient is used as the X value in the coordinate system and the historical lesion CT value state coefficient is used as the Y value in the coordinate system to determine the coordinates of each historical lesion state coefficient.

[0108] According to an embodiment of the present application, the area to be recovered in the coordinate system is determined according to the lesion status score, including: the coordinates of the lesion status coefficient in the coordinate system, the slope of which is The current lesion condition equivalent straight line is used; the area enclosed by the current lesion condition equivalent straight line, the X-axis, and the Y-axis is determined as the area to be recovered. In the lesion condition scoring, there may be a situation where the lesion shape state coefficient and the lesion CT value state coefficient change, but the lesion condition score obtained is the same as the historical lesion condition score of the previous examination, that is, the lesion condition score remains unchanged, but the lesion shape state coefficient and the lesion CT value state coefficient are different from the previous examination. Therefore, the coordinates of the same lesion condition score can be expressed as A straight line is connected to obtain a current lesion condition equivalent line. On the lesion condition equivalent line, the lesion condition score of each point is the same, but the lesion appearance state coefficient and the lesion CT value state coefficient are different. The triangular area enclosed by the current lesion condition equivalent line, the X-axis and the Y-axis is the area to be recovered. When the current lesion condition equivalent line passes through the coordinate origin, it means that the lesion appearance state coefficient and the lesion CT value state coefficient are both 0, the patient is completely recovered, and there is no area to be recovered. Similarly, the historical area to be recovered in the coordinate system can be determined based on the historical lesion condition score. The larger the area of ​​the area to be recovered, the more serious the patient's lesion condition is, and the more attention needs to be paid. Similarly, the historical area to be recovered corresponding to the historical lesion condition score can also be determined.

[0109] According to an embodiment of the present application, the lesion development status is determined based on the area to be rehabilitated and the historical area to be rehabilitated, including: determining the lesion development coefficient D according to formula (4),

[0110] (4)

[0111] in, is the area of ​​the area to be rehabilitated, is the area of ​​the historical area to be recovered corresponding to the m-1th historical CT image, is the inspection time of the CT image, is the inspection time of the m-1th historical CT image, m is the number of inspections, is the area of ​​the historical area to be recovered corresponding to the j+1th historical CT image, is the area of ​​the historical area to be recovered corresponding to the j-th historical CT image, is the inspection time of the j+1th historical CT image, is the inspection time of the jth historical CT image, max is the maximum value function, j and m are both positive integers; the lesion development status is determined according to the lesion development coefficient.

[0112] According to an embodiment of the present application, in formula (4), It represents the difference between the area of ​​the region to be recovered and the area of ​​the historical region to be recovered corresponding to the m-1th historical CT image, where the area of ​​the region to be recovered is the area of ​​the region to be recovered of the mth CT image. represents the difference between the inspection time of the CT image and the inspection time of the m-1th historical CT image, where the inspection time of the CT image is the inspection time of the mth CT image. Therefore, It can represent the rate of change of the area of ​​the area to be healed. The greater the rate of change of the area of ​​the area to be healed, the faster the lesion develops and the more attention it needs. It can represent the change rate of the area of ​​the historical region to be recovered corresponding to the j-th historical CT image and the j+1-th historical CT image. Therefore, The maximum value of the rate of change of the area of ​​the region to be recovered corresponding to each historical CT image can be represented. The ratio of the rate of change of the area of ​​the region to be recovered to the maximum value of the rate of change of the area of ​​the region to be recovered corresponding to each historical CT image It can be used as a lesion development coefficient. Based on the lesion development coefficient, it can be used to determine whether the lesion is worsening, recovering, or unchanged. For example, if the lesion development coefficient is large (for example, greater than 1), it indicates that the rate of change of the current area to be recovered has significantly changed relative to the rate of change of the area of ​​the historical area to be recovered. The lesion is deteriorating rapidly and requires attention and timely treatment. If the lesion development coefficient is greater than 0 and less than or equal to 1, it indicates that the rate of change of the current area to be recovered is small relative to the rate of change of the area of ​​the historical area to be recovered. The lesion is still developing towards a serious condition, but at a slower rate. If the lesion development coefficient is equal to 0, it indicates that the lesion has stopped developing and the condition is stable. If the lesion development coefficient is less than 0, it indicates that the lesion is improving.

[0113] In this way, the lesion status score and historical lesion status score can be determined based on the lesion status coefficient, the historical lesion status coefficient, the examination time of the CT image, and the examination time of the historical CT image. A coordinate system can be established, with the lesion appearance status coefficient as the horizontal coordinate and the lesion CT value status coefficient as the vertical coordinate, to determine the coordinates of the lesion status score in the coordinate system, thereby determining the equivalent straight line of the current lesion status, and then the area enclosed by the equivalent straight line and the coordinate axis as the area to be recovered. Taking into account the situation where the lesion appearance status coefficient and the lesion CT value status coefficient produce equivalent changes, the changes in the lesion appearance status and the lesion CT value status can be observed simultaneously, improving the comprehensiveness of the observation. In addition, the change rate of the area to be recovered and the maximum value of the change rate of the historical area to be recovered can be determined to determine the lesion development coefficient, and then determine the lesion development status. This improves the accuracy, comprehensiveness, and objectivity of the lesion development status and improves the efficiency of lesion change analysis.

[0114] According to an embodiment of the present application, in step S6, a lesion change analysis report is generated based on the lesion development status. The lesion change analysis report may include information such as whether the condition is improving and the review frequency can be reduced, or whether the condition is worsening and the review frequency needs to be increased and treatment measures need to be taken.

[0115] According to the time-series CT image lesion change analysis method of the embodiment of the present application, a CT image of the patient's current examination and multiple historical CT images of the patient can be obtained. The CT image and historical CT images are detected using a lesion detection model to obtain the lesion area where the lesion is located in the CT image and the historical lesion area where the lesion is located in the historical CT images, thereby determining the lesion state coefficient of the lesion area and the historical lesion state coefficient of the historical lesion area. The lesion development status is then determined based on the changes in the historical lesion state coefficient and the lesion state coefficient, and a lesion change analysis report is generated. This reduces reliance on manpower, improves the accuracy of lesion development status judgment, and improves the efficiency of lesion change analysis. When detecting the lesion area, the CT image of the patient's current examination and multiple historical CT images can be obtained. Based on the lesion detection model, the CT images are detected to obtain the outline of the lesion area, thereby obtaining the lesion area and the historical lesion area, providing accurate basic data for determining the lesion development status. When determining the lesion state coefficient, the lesion size state coefficient of the lesion area can be determined based on the ratio of the area of ​​the lesion area to a preset lesion area threshold. The lesion detection model can be used to determine the contour of the lesion area. Multiple sampling points are set on the contour of the lesion area at uniform angles to determine the sampling vectors between adjacent sampling points. Based on the sampling vectors, the cosine value of the angle between two adjacent sampling vectors at the most irregular position of the lesion shape is obtained. Based on the number of sampling points, the cosine value of the angle between adjacent sampling vectors is determined when the lesion is completely regular. This allows the lesion morphological state coefficient of the lesion area to be accurately described. The lesion shape state coefficient is then determined through weighted summation. Furthermore, the lesion density state coefficient of the lesion area is determined based on the ratio of the CT value of the centroid of the lesion area to a preset CT threshold. This takes into account the situation where lesions are located in different human organs and have different preset CT thresholds for lesions, thereby improving the accuracy of the lesion density state coefficient. Based on the uniformly arranged detection points on the detection line, multiple detection point groups are obtained, thereby obtaining the dispersion coefficient of the CT value of each detection point in the detection point group. Based on the relationship between the uniformity of the lesion area and the severity of the disease, the lesion uniformity state coefficient is obtained, which improves the accuracy of the lesion uniformity state coefficient. The lesion CT value state coefficient is obtained through weighted summation, and then the lesion state coefficient is obtained. This improves the accuracy, objectivity, and comprehensiveness of the lesion state coefficient, allowing the lesion state coefficient to accurately describe the current state of the lesion. When determining the lesion development coefficient, the lesion state score and historical lesion state score can be determined based on the lesion state coefficient, the historical lesion state coefficient, the examination time of the CT image, and the examination time of the historical CT image.A coordinate system can be established, with the lesion shape state coefficient as the horizontal coordinate and the lesion CT value state coefficient as the vertical coordinate, to determine the coordinates of the lesion condition score in the coordinate system, thereby determining the equivalent straight line of the current lesion condition, and then the area enclosed by the equivalent straight line and the coordinate axis is determined as the area to be recovered. Taking into account the situation where the lesion shape state coefficient and the lesion CT value state coefficient produce equivalent changes, changes in the lesion shape state and the lesion CT value state can be observed simultaneously, improving the comprehensiveness of the observation. In addition, the rate of change of the area to be recovered and the maximum value of the change rate of the historical area to be recovered can be determined, thereby determining the lesion development coefficient and then determining the lesion development status. This improves the accuracy, comprehensiveness, and objectivity of the lesion development status and improves the efficiency of lesion change analysis.

[0116] Figure 3 A block diagram of a time-series-based CT image lesion change analysis system according to an embodiment of the present application is exemplarily shown, wherein the system includes:

[0117] The image acquisition module obtains multiple historical CT images of the patient from the database after obtaining the CT image of the patient's current examination;

[0118] The lesion region module detects the CT image through the lesion detection model and obtains the lesion region where the lesion is located in the CT image;

[0119] The historical lesion region module detects historical CT images through the lesion detection model to obtain the historical lesion region where the lesions are located in the historical CT images;

[0120] A lesion state coefficient module is used to determine the lesion state coefficient of the lesion area and the historical lesion state coefficient of the historical lesion area;

[0121] A lesion development status module determines the lesion development status based on the lesion status coefficient, the historical lesion status coefficient, the examination time of the CT image, and the examination time of the historical CT image;

[0122] The lesion change analysis report module generates a lesion change analysis report based on the lesion development status.

[0123] The present application may be a method, apparatus, system and / or computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present application.

[0124] Those skilled in the art will understand that the embodiments of the present application described above and shown in the accompanying drawings are intended only as examples and do not limit the present application. The objectives of the present application have been fully and effectively achieved. The functional and structural principles of the present application have been demonstrated and explained in the embodiments. The embodiments of the present application may be modified or altered in any manner without departing from the principles described.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application 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 make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A time-series-based CT image lesion change analysis method, characterized in that: include: After obtaining the CT image of the patient's current examination, multiple historical CT images of the patient are obtained from the database; The CT image is detected by the lesion detection model to obtain the lesion area where the lesion is located in the CT image; Through the lesion detection model, the historical CT images are detected to obtain the historical lesion area where the lesions are located in the historical CT images; Determine the lesion state coefficient of the lesion area and the historical lesion state coefficient of the historical lesion area; Determine the lesion development status based on the lesion status coefficient, the historical lesion status coefficient, the examination time of the CT image, and the examination time of the historical CT image; Generate lesion change analysis report based on lesion development status; Determine the lesion status coefficient of the lesion area, including: determining a lesion size state coefficient of the lesion region according to the area of ​​the lesion region and a preset lesion area threshold; Obtain the outline of the lesion area; Multiple sampling points are set on the contour of the lesion area; Determine the sampling vector between adjacent sampling points; determining a lesion morphological state coefficient of the lesion area according to the sampling vector; Determine the lesion shape state coefficient according to the lesion size state coefficient and the lesion morphology state coefficient; Determining a lesion density state coefficient of the lesion area according to the CT value of the centroid of the lesion area and a preset CT threshold; Connecting the centroid of the lesion area with the sampling point to obtain a detection line; According to the CT value of the pixel point on the detection line, the lesion uniformity state coefficient of the lesion area is determined; Determine the lesion CT value state coefficient according to the lesion density state coefficient and the lesion uniformity state coefficient; Determining a lesion state coefficient according to the lesion appearance state coefficient and the lesion CT value state coefficient; Determining the lesion morphological state coefficient of the lesion area according to the sampling vector includes: According to the formula ； Determine the lesion morphological state coefficient of the lesion area ,in, is the coordinate of the i+1th sampling point, is the coordinate of the i-th sampling point, is the sampling vector between the i+1th sampling point and the i-th sampling point, is the coordinate of the i-1th sampling point, is the sampling vector between the i-th sampling point and the i-1-th sampling point, for The transposed vector of , n is the number of sampling points, min is the minimum function, i and n are both positive integers; The lesion state coefficient includes a lesion appearance state coefficient and a lesion CT value state coefficient, and the historical lesion state coefficient includes a historical lesion appearance state coefficient and a historical lesion CT value state coefficient; The lesion development status is determined based on the lesion status coefficient, the historical lesion status coefficient, the examination time of the CT image, and the examination time of the historical CT image, including: According to the formula ； Determine the lesion condition score of the CT image of this examination ,in, is the lesion shape coefficient, is the CT value status coefficient of the lesion, and is the preset weight; According to the formula ； Determine the historical lesion condition score of the jth historical CT image ,in, is the j-th historical lesion shape state coefficient, is the CT value status coefficient of the jth historical lesion; Establish a coordinate system and determine the coordinates of the lesion state coefficient and each historical lesion state coefficient in the coordinate system, wherein the lesion shape state coefficient is the X value in the coordinate of the lesion state coefficient, the lesion CT value state coefficient is the Y value in the coordinate of the lesion state coefficient, the historical lesion shape state coefficient is the X value in the coordinate of the historical lesion state coefficient, and the historical lesion CT value state coefficient is the Y value in the coordinate of the historical lesion state coefficient; Determine the area to be healed in the coordinate system according to the lesion condition score; Determine the historical area to be recovered in the coordinate system according to the historical lesion condition score; Determine the development status of the lesion based on the area to be rehabilitated and the historical area to be rehabilitated; Determine the development of the lesion based on the area to be rehabilitated and the historical area to be rehabilitated, including: According to the formula ； Determine the lesion development coefficient D, where is the area of ​​the area to be rehabilitated, is the area of ​​the historical area to be recovered corresponding to the m-1th historical CT image, is the inspection time of the CT image, is the inspection time of the m-1th historical CT image, m is the number of inspections, is the area of ​​the historical area to be recovered corresponding to the j+1th historical CT image, is the area of ​​the historical area to be recovered corresponding to the j-th historical CT image, is the inspection time of the j+1th historical CT image, is the inspection time of the jth historical CT image, max is the maximum value function, and both j and m are positive integers; The lesion development status is determined based on the lesion development coefficient.

2. The time-series-based CT image lesion change analysis method according to claim 1, characterized in that: According to the CT value of the pixel point on the detection line, the lesion uniformity state coefficient of the lesion area is determined, including: Set multiple detection points evenly on the detection line and determine the CT value of the detection points; Divide the detection points with the same serial number on each detection line into a group to obtain multiple detection point groups; A lesion uniformity state coefficient of the lesion area is determined according to the CT value of each detection point in the plurality of detection point groups.

3. The time-series-based CT image lesion change analysis method according to claim 2, characterized in that: According to the CT value of each detection point in the plurality of detection point groups, the lesion uniformity state coefficient of the lesion area is determined, including: Determine the CT average value of each detection point in the detection point group; Determine the standard deviation of the CT value of each detection point in the detection point group according to the CT average value of each detection point in the detection point group; According to the CT value standard deviation and the CT average value, the dispersion coefficient of the CT value of each detection point in the detection point group is determined; The maximum value of the discrete coefficients corresponding to the plurality of detection point groups is determined as the lesion uniformity state coefficient of the lesion area.

4. The method for analyzing lesion changes in CT images based on time series according to claim 1, characterized in that: The area to be rehabilitated in the coordinate system is determined based on the lesion condition score, including: The coordinates of the lesion state coefficient in the coordinate system are used to make the slope The current lesion condition is equivalent to the straight line; The area enclosed by the current lesion condition equivalent straight line, X axis and Y axis is determined as the area to be recovered.

5. A time-series-based CT image lesion change analysis system, used to execute the method according to any one of claims 1 to 4, characterized in that: include: The image acquisition module obtains multiple historical CT images of the patient from the database after obtaining the CT image of the patient's current examination; The lesion region module detects the CT image through the lesion detection model and obtains the lesion region where the lesion is located in the CT image; The historical lesion region module detects historical CT images through the lesion detection model to obtain the historical lesion region where the lesions are located in the historical CT images; A lesion state coefficient module is used to determine the lesion state coefficient of the lesion area and the historical lesion state coefficient of the historical lesion area; A lesion development status module determines the lesion development status based on the lesion status coefficient, the historical lesion status coefficient, the examination time of the CT image, and the examination time of the historical CT image; The lesion change analysis report module generates a lesion change analysis report based on the lesion development status.

Images