An imaging-assisted analysis method for endometriosis

Abstract

The present invention relates to the technical field of image data processing, and particularly relates to an imaging-assisted analysis method for endometriosis. This method performs edge detection on endometriosis ultrasound images, determines the lesion indicators at the endometriosis sites, accurately extracts the lesion features and quantifies the size and trend of the lesion areas; in order to provide a basis for the annotation type, based on the lesion indicators at the endometriosis sites, determines the lesion type corresponding to each first image; then, according to the lesion indicators, lesion types and analysis models, determines the deterioration degree determination coefficient corresponding to the first image, thereby quantifying the severity of the lesions; finally, based on the deterioration degree determination coefficient and the lesion types, annotates the endometriosis ultrasound images to obtain ultrasound images including classification labels of the lesion types and deterioration degrees, which can provide classification labels of the lesion types and deterioration degrees obtained through accurate extraction and quantitative analysis, and improve the accuracy of the classification labels.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and particularly relates to an imaging-assisted analysis method for endometriosis. Background Art

[0002] Endometriosis is a common gynecological disease. Under normal circumstances, the endometrial tissue only exists in the uterine cavity. However, under abnormal circumstances, the endometrial tissue appears outside the uterus, such as in the ovaries, fallopian tubes, pelvic cavity and other parts, thus forming pelvic endometriotic foci, cervical endometriotic cysts and abdominal wall scar endometriotic cysts, etc. Such diseases may cause problems such as pain, irregular menstruation and infertility, and the condition may deteriorate over time. Non-invasive ultrasound technology can collect color Doppler ultrasound images of the endometrial tissue, process and analyze the images, and finally obtain a classification result of the degree of endometriosis deterioration through classification. This classification result can be used to assist doctors in subsequent diagnosis. Doctors conduct comprehensive analysis by combining other physical parameters and physical conditions of the patient to obtain the final diagnosis result.

[0003] Related technologies usually use deep learning methods to first process and label several ultrasound images of the endometrium, and then classify and label the images to be labeled according to the labeled ultrasound images of the endometrium and the neural network, so as to assign labels to all ultrasound images of the endometrium.

[0004] However, the processing and labeling in related technologies often only focus on the surface information of endometriosis, lacking precise extraction and quantitative analysis of the lesion characteristics of endometriosis, resulting in the labeling results not being labeled by grade and accurately classified, and further leading to deviations in the classification and labeling results of the neural network. Summary of the Invention

[0005] In view of this, the present invention provides an imaging-assisted analysis method for endometriosis, which can provide classification labels of lesion types and degrees of deterioration obtained through precise extraction and quantitative analysis, and improve the accuracy of the classification labels.

[0006] The object of the present invention is to provide an imaging-assisted analysis method for endometriosis, and the specific technical solution adopted is as follows:

[0007] Obtain sample endometriosis ultrasound images, where the sample endometriosis ultrasound images contain lesion information of sample pelvic endometriotic foci, sample cervical endometriotic cysts and abdominal wall scar endometriotic cysts;

[0008] Preprocess the sample endometriosis ultrasound images to obtain a first image;

[0009] Perform edge detection on the first image to determine the lesion index of the endometriosis site corresponding to each first image;

[0010] Based on the lesion index of the endometriosis site, determine the lesion type corresponding to each first image;

[0011] According to the lesion index of the endometriosis site, lesion type, and analysis model corresponding to the first image, determine the deterioration degree determination coefficient corresponding to the first image;

[0012] Based on the deterioration degree determination coefficient and lesion type corresponding to the first image, annotate the sample endometriosis ultrasound image to obtain a sample annotated endometriosis ultrasound image, where the sample annotated endometriosis ultrasound image includes classification labels of the lesion type and deterioration degree.

[0013] In some embodiments, the preprocessing of the sample endometriosis ultrasound image includes: performing histogram equalization processing on the sample endometriosis ultrasound image.

[0014] In some embodiments, the performing edge detection on the first image to determine the lesion index of the endometriosis site corresponding to each first image includes:

[0015] Perform edge detection on the first image to obtain an edge detection result, where the edge detection result includes a connected lesion region, lesion location, and ultrasonic feature information;

[0016] According to the area and contour shape features of the connected lesion region, determine a first index and a second index, where the first index is the lesion index of cervical endometriosis cyst, and the second index is the lesion index of abdominal wall scar endometriosis cyst and other types of cysts;

[0017] Based on the first index and the second index, determine the lesion index of the endometriosis site.

[0018] In some embodiments, according to the area and contour shape features of the connected lesion region, determine the first index:

[0019]

[0020] Wherein, represents the lesion index of cervical endometriosis cyst, represents the area of the connected lesion region, i represents the intersection point serial number, represents the Euclidean distance from the geometric center of the connected lesion region to the i-th intersection point, represents the Euclidean distance from the geometric center of the connected lesion area to the (i + 1)-th intersection point, where the intersection point and the intersection point serial number i satisfy: a ray is made from the geometric center of the connected lesion area along the horizontal direction towards the contour of the connected lesion area, the ray intersects the contour at the first intersection point, the intersection point serial number i = 1 is recorded, at this time the angle between the ray and the horizontal line is 0°, the ray is rotated counterclockwise, and an intersection point of the ray and the contour is obtained every 30° of counterclockwise rotation. After the ray completes one full rotation, a total of twelve intersection points are obtained, and they are numbered in the order of acquisition to obtain the serial number i.

[0021] In some embodiments, according to the area and contour shape characteristics of the connected lesion area, a second index is determined:

[0022]

[0023] In the formula, represents the lesion index of abdominal wall scar endometriosis cyst and other types of cysts, represents the area of the connected lesion area, j represents the intersection point serial number, represents the difference between the maximum Euclidean distance and the minimum Euclidean distance from the geometric center of the connected lesion area to the intersection point, represents the Euclidean distance from the geometric center of the connected lesion area to the j-th intersection point, represents the Euclidean distance from the geometric center of the connected lesion area to the (j + 6)-th intersection point, where the intersection point and the intersection point serial number j satisfy: a ray is made from the geometric center of the connected lesion area along the horizontal direction towards the contour of the connected lesion area, the ray intersects the contour at the first intersection point, the intersection point serial number j = 1 is recorded, at this time the angle between the ray and the horizontal line is 0°, the ray is rotated counterclockwise, and an intersection point of the ray and the contour is obtained every 30° of counterclockwise rotation. After the ray completes one full rotation, a total of twelve intersection points are obtained, and they are numbered in the order of acquisition to obtain the serial number j.

[0024] In some embodiments, the determining the deterioration degree determination coefficient corresponding to the first image based on the endometriosis site lesion index, lesion type, and analysis model corresponding to the first image includes:

[0025] Obtain the analysis model and determine the deterioration degree determination coefficient corresponding to the first image;

[0026] Determine the image with the largest deterioration degree determination coefficient among the first images with the same lesion type, obtain the image with the largest deterioration degree determination coefficient corresponding to each lesion type, and use the image with the largest deterioration degree determination coefficient corresponding to each lesion type as the reference image for each lesion type;

[0027] Determine that the degree of deterioration of the reference image for each type of lesion is the fifth level;

[0028] Input the reference image, the other images to be determined except the reference image in the first image, and the degree of deterioration of the reference image into the analysis model, and obtain the lesion type corresponding to each of the other images to be determined, and the correlation coefficient between the reference image corresponding to the lesion type of each of the other images to be determined and the other images to be determined.

[0029] Based on the correlation coefficient and the lesion index, determine the deterioration degree determination coefficient corresponding to the first image.

[0030] In some embodiments, based on the first index and the second index, determine the lesion index of the endometriosis site:

[0031]

[0032] where represents the lesion index of the endometriosis site, represents the sign function, represents the lesion index of the endometriotic cyst of the cervix, represents the lesion index of the endometriotic cyst of the abdominal wall scar and other types of cysts.

[0033] In some embodiments, based on the deterioration degree determination coefficient corresponding to the first image, determine the deterioration degree of the sample-labeled endometriosis ultrasound image, including:

[0034] In response to the deterioration degree determination coefficient corresponding to the first image being within the first threshold range, the deterioration degree of the sample-labeled endometriosis ultrasound image is the first level;

[0035] In response to the deterioration degree determination coefficient corresponding to the first image being within the second threshold range, the deterioration degree of the sample-labeled endometriosis ultrasound image is the second level;

[0036] In response to the deterioration degree determination coefficient corresponding to the first image being within the third threshold range, the deterioration degree of the sample-labeled endometriosis ultrasound image is the third level;

[0037] In response to the deterioration degree determination coefficient corresponding to the first image being within the fourth threshold range, the deterioration degree of the sample-labeled endometriosis ultrasound image is the fourth level;

[0038] In response to the deterioration degree determination coefficient corresponding to the first image being within the fifth threshold range, the deterioration degree of the sample-labeled endometriosis ultrasound image is the fifth level.

[0039] In some embodiments, based on the correlation coefficient and the lesion index, a deterioration degree determination coefficient corresponding to the first image is determined:

[0040]

[0041] In the formula, represents the deterioration degree determination coefficient, represents the exponential function with the natural constant as the base, represents the lesion index of cervical endometriotic cyst, represents the lesion index of abdominal wall scar endometriotic cyst and other types of cysts, represents the correlation coefficient calculation symbol, a represents the reference image corresponding to the lesion type of each of the other images to be determined, b represents the other image to be determined, represents the correlation coefficient between the reference image corresponding to the lesion type of each of the other images to be determined and the other image to be determined.

[0042] In some embodiments, the analysis model is a template matching model.

[0043] The present invention has the following beneficial effects:

[0044] An imaging-assisted analysis method for endometriosis provided by the present invention first performs edge detection on the preprocessed sample endometriosis ultrasound images to determine the lesion index of the endometriosis site, so as to accurately extract the lesion features and quantify the size and trend of the lesion area; furthermore, based on the lesion index of the endometriosis site, the lesion type corresponding to each first image is accurately determined, providing a type basis for annotation; then, according to the lesion index, lesion type and analysis model corresponding to the first image, the deterioration degree determination coefficient corresponding to the first image is determined, thereby quantifying the severity of the lesion; finally, based on the deterioration degree determination coefficient and lesion type corresponding to the first image, the sample endometriosis ultrasound images are annotated to obtain the sample annotated endometriosis ultrasound images including the classification labels of the lesion type and deterioration degree, improving the accuracy of the classification labels. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0046] Figure 1The flowchart of a method for imaging-assisted analysis of endometriosis provided by an embodiment of the present invention. Detailed implementation manners

[0047] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following combines the accompanying drawings and preferred embodiments to detail the specific implementation manners, structures, features and effects of a method for imaging-assisted analysis of endometriosis proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0049] The following specifically describes the specific solution of a method for imaging-assisted analysis of endometriosis provided by the present invention with reference to the accompanying drawings.

[0050] Please refer to Figure 1 , which shows the flowchart of a method for imaging-assisted analysis of endometriosis provided by an embodiment of the present invention. The method includes:

[0051] Step 101: Obtain sample endometriosis ultrasound images, where the sample endometriosis ultrasound images contain lesion information of sample pelvic endometriosis foci, sample cervical endometriosis cysts, and abdominal wall scar endometriosis cysts.

[0052] In some embodiments, the sample endometriosis ultrasound images include, but are not limited to, lesion information of sample pelvic endometriosis foci, sample cervical endometriosis cysts, and abdominal wall scar endometriosis cysts, and may also include ultrasound features and bleeding conditions, etc.

[0053] Step 102: Preprocess the sample endometriosis ultrasound images to obtain the first images.

[0054] In some embodiments, the preprocessing in step 102 can be achieved in the following way: perform histogram equalization processing on the sample endometriosis ultrasound images to ensure the observability and contrast of the images.

[0055] Step 103: Perform edge detection on the first images to determine the lesion indicators of the endometriosis sites corresponding to each first image.

[0056] Edge detection can accurately extract lesion features, obtain the lesion ultrasound features and bleeding conditions in the first image, determine the lesion indicators at the endometriosis site, and thus quantify the size and trend of the lesion area.

[0057] It should be noted that endometriosis lesions with a diameter of more than 2 cm can be detected by ultrasound. They have various manifestations according to different growth sites, and different growth sites correspond to different lesion positions, so the specific types are also different. Cervical endometriosis cyst: It appears as a round or oval cloud-like hypoechoic area within the cervical tissue, with a relatively clear boundary and a rough inner wall. Its cyst wall is composed of cervical tissue; Abdominal wall scar endometriosis cyst: It can occur in all layers of the abdominal wall scar. The local abdominal wall thickens, and the lesion is fusiform or oval with a relatively blurred boundary. Therefore, based on its specific distribution characteristics, the corresponding lesion position, lesion type, lesion ultrasound features, bleeding conditions, etc. can be analyzed and determined. The closer the edge detection result is to the lesion features at a certain position above, the greater the probability of it being this ectopia, and the closer the lesion indicators at this ectopic site are to the indicators of this type of lesion.

[0058] In some implementations, the edge detection in step 3 can be Canny edge detection. The pixel values of the detection result are either 0 or 1, which can intuitively reflect the contour shape features and distribution range of the connected lesion area. The distribution range affects the area of the connected area, that is, the size of the lesion.

[0059] Since the contours of some connected lesion areas in the edge detection result may not be closed and have gaps, in some embodiments, the edge detection result can be dilated by a morphological dilation algorithm so that the originally unconnected edge contours can be closed, thereby dividing the connected lesion area.

[0060] In some embodiments, step 103 includes:

[0061] Step 1031, perform edge detection on the first image to obtain the edge detection result. The edge detection result includes the connected lesion area, lesion position, and ultrasound feature information, thereby quantifying the size and trend of the lesion area.

[0062] Step 1032, determine the first index and the second index according to the area and contour shape features of the connected lesion area. The first index is the lesion index of the cervical endometriosis cyst, and the second index is the lesion index of the abdominal wall scar endometriosis cyst and other types of cysts.

[0063] The first index and the second index are the lesion indicators of different lesion types, thereby obtaining the possibility degrees of different lesion types within the connected lesion area.

[0064] In some embodiments, according to the area and contour shape features of the connected lesion region, a first index is determined:

[0065]

[0066] Wherein, represents the lesion index of cervical endometriosis cyst, which is also the first index, represents the area of the connected lesion region, i represents the intersection point serial number, represents the Euclidean distance from the geometric center of the connected lesion region to the i-th intersection point, represents the Euclidean distance from the geometric center of the connected lesion region to the (i + 1)-th intersection point. Among them, the intersection point and the intersection point serial number i satisfy: a ray is made from the geometric center of the connected lesion region along the horizontal direction to the contour of the connected lesion region, the ray intersects the contour at the first intersection point, and the intersection point serial number i = 1 is recorded. At this time, the included angle between the ray and the horizontal line is 0°. Rotate the ray counterclockwise. Each time the ray is rotated counterclockwise by 30°, an intersection point of the ray and the contour is obtained. After the ray completes one circle of rotation, a total of twelve intersection points are obtained, and they are numbered in sequence according to the obtained order to get the serial number i.

[0067] represents the lesion index of cervical endometriosis cyst, aiming to quantify the size and tendency of the lesion corresponding to cervical endometriosis cyst, that is, whether it is endometriosis of this lesion type. Generally, the characteristics of this type of lesion are: a round or oval cloud-like hypoechoic area in the cervical tissue, with a relatively clear boundary and a rough inner wall. Then, the contour shape features of the connected lesion region obtained by edge detection will also show such features.

[0068] The closer the contour shape feature of the connected lesion region is to a circle or an oval, and the larger the area of the connected lesion region, the larger the lesion index. The Euclidean distance from the geometric center of the connected lesion region to the i-th intersection point can be used as an important parameter to quantify whether the contour shape of the connected lesion region is a circle or an oval. When the contour shape is closer to a circle, the Euclidean distances from the geometric center of the connected lesion region to each intersection point are closer, and the distances from the geometric center of the connected lesion region to two adjacent intersection points are also closer. Therefore, represents the degree of proximity of the distance from the geometric center of the connected lesion region to the intersection points; The smaller it is, the closer the distances from the geometric center of the connected lesion region to two adjacent intersection points are, that is, the closer the contour shape feature is to a circle or an oval, The larger it is, the larger the lesion index, so the summation result is larger; multiplying the summation result by the area of the connected lesion region (the number of pixel points in the connected lesion region) to obtain the lesion index of cervical endometriosis cyst , the larger this value is, the larger the lesion area is, and the higher the corresponding lesion index is, that is, the greater the possibility that the lesion type is cervical endometriosis cyst.

[0069] In some embodiments, according to the area and contour shape characteristics of the connected lesion area, a second index is determined:

[0070]

[0071] In the formula, represents the lesion index of abdominal wall scar endometriosis cyst and other types of cysts, that is, the second index, represents the area of the connected lesion area, j represents the intersection serial number, represents the difference between the maximum Euclidean distance and the minimum Euclidean distance from the geometric center of the connected lesion area to the intersection, represents the Euclidean distance from the geometric center of the connected lesion area to the j-th intersection, represents the Euclidean distance from the geometric center of the connected lesion area to the (j + 6)-th intersection. Among them, the intersection and the intersection serial number j satisfy: a ray is made from the geometric center of the connected lesion area along the horizontal direction to the contour of the connected lesion area, and the ray intersects the contour at the first intersection, and the intersection serial number j = 1 is recorded. At this time, the included angle between the ray and the horizontal line is 0°. Rotate the ray counterclockwise, and a ray intersects the contour at an intersection every 30° counterclockwise rotation. After the ray completes a full rotation, a total of twelve intersections are obtained, and they are numbered in the order of acquisition to obtain the serial number j.

[0072] represents the lesion index of abdominal wall scar endometriosis cyst and other types of cysts, aiming to quantify the size and trend of the lesions corresponding to abdominal wall scars and other types of cysts, that is, whether they are abdominal wall scars and other types of cysts of this lesion type. The lesion characteristics formed by abdominal wall scars and other types of cysts do not have a relatively regular geometric shape compared with cervical endometriosis cysts. Its specific characteristics are: the lesions are fusiform or oval, so the contour shape characteristics of the connected lesion area obtained through edge detection will also show a fusiform or oval shape.

[0073] The closer the contour shape characteristics of the connected lesion area are to a fusiform or oval shape, and the larger the area of the connected lesion area is, the larger the lesion index is. The Euclidean distance from the geometric center of the connected lesion area to the j-th intersection can be used as an important parameter to quantify whether the contour shape of the connected lesion area is fusiform or oval. When the contour shape is closer to a fusiform or oval shape, the Euclidean distance from the geometric center of the connected lesion area to the j-th intersection is closer to the Euclidean distance from the geometric center of the connected lesion area to the (j + 6)-th intersection. Therefore, use Characterize the degree of proximity between the Euclidean distance from the geometric center of the connected lesion area to the j-th intersection point and the Euclidean distance from the geometric center of the connected lesion area to the (j + 6)-th intersection point; The closer it is to 1, that is, the more the contour shape feature resembles a spindle or an ellipse, The closer it is to 1, the larger the lesion index, so the summation result is larger; the difference between the maximum Euclidean distance and the minimum Euclidean distance from the geometric center of the connected lesion area to the intersection point represents the flatness degree of the spindle or ellipse. The larger this value is, the flatter the lesion is in the shape of a spindle or an ellipse; Multiply the summation result by the area of the connected lesion area (the number of pixel points in the connected lesion area), to obtain the lesion index of endometriosis cyst of abdominal wall scar and other types of cysts , the larger this value is, the larger the lesion area represents, and the higher the corresponding lesion index is, that is, the greater the possibility that the lesion type is endometriosis cyst of abdominal wall scar and other types of cysts.

[0074] Step 1033, based on the first index and the second index, determine the lesion index of the endometriosis site.

[0075] The obtained lesion index of the endometriosis site can reflect the size and trend of the entire connected lesion area, and determine whether the connected lesion area tends to cervical endometriosis cyst or endometriosis cyst of abdominal wall scar and other types of cysts.

[0076] In some embodiments, according to the following formula, based on the first index and the second index, determine the lesion index of the endometriosis site:

[0077]

[0078] In the formula, represents the lesion index of the endometriosis site, represents the sign function, represents the lesion index of cervical endometriosis cyst, represents the lesion index of endometriosis cyst of abdominal wall scar and other types of cysts.

[0079] Compare the first index and the second index corresponding to the connected lesion area, and judge through the sign function the positive and negative nature of, if the value of the first index is large, then is greater than 0, the classification result of is 1, then the lesion type of this lesion area is cervical endometriosis cyst. If the value of the second index is large, then is less than 0, The classification result is -1, abdominal wall scar endometriosis cyst and other types of cysts. Through the sign function judge the positivity or negativity of, so as to obtain the classification result of the lesion index and determine the lesion index of the endometriosis site .

[0080] Step 104: Based on the lesion index of the endometriosis site, determine the lesion type corresponding to each first image.

[0081] The lesion index of the endometriosis site can quantify the characteristics such as the size and trend of the lesion area, and thus determine the lesion type. By whether the lesion index is 1 or -1, it is determined whether the lesion type corresponding to each first image is a cervical endometriosis cyst or an abdominal wall scar endometriosis cyst and other types of cysts, providing an accurate guidance for the label classification of the lesion type. The determination of the lesion type here is to screen out the corresponding lesion type from the ultrasound image, and then facilitate the subsequent annotation of the sample endometriosis ultrasound images of different lesion types.

[0082] Step 105: According to the lesion index, lesion type and analysis model corresponding to the first image, determine the deterioration degree determination coefficient corresponding to the first image.

[0083] The deterioration degree determination coefficient is used to quantify the severity of the lesion, so as to provide an accurate classification basis for the label classification of the deterioration degree.

[0084] In some embodiments, step 105 includes:

[0085] Step 1051: Obtain the analysis model and determine the deterioration degree determination coefficient corresponding to the first image.

[0086] The analysis model can perform template matching and quantify the deterioration degree corresponding to the first image as the deterioration degree determination coefficient.

[0087] Step 1052: Determine the image with the largest deterioration degree determination coefficient among the first images with the same lesion type, obtain the image with the largest deterioration degree determination coefficient corresponding to each lesion type, and use the image with the largest deterioration degree determination coefficient corresponding to each lesion type as the reference image for each lesion type.

[0088] The reference image can also provide a reference basis for the determination of the deterioration degree of the first image. Since the deterioration degree determination coefficient of the reference image is the largest, it can make the deterioration degree corresponding to each lesion type have a unified judgment standard, and the deterioration degree is quantified by the deterioration degree determination coefficient, so as to more accurately determine the deterioration degree of each first image subsequently.

[0089] Step 1053: Determine that the degree of deterioration of the reference image for each type of lesion is the fifth grade.

[0090] Determine that the degree of deterioration of the reference image is the most severe degree of the lesion, and use the fifth grade to characterize the degree of deterioration of the reference image, so as to set a unified standard for the analysis model, which is convenient for the quantification and classification annotation of the degree of deterioration of other images.

[0091] Step 1054: Input the reference image, other images to be determined except the reference image in the first image, and the degree of deterioration of the reference image into the analysis model, and obtain the corresponding lesion type of each other image to be determined, and the correlation coefficient between the reference image corresponding to the lesion type of each other image to be determined and the other image to be determined.

[0092] In some embodiments, the analysis model is a template matching model to find an image similar to the reference image in other images to be determined according to the matching degree.

[0093] Other images to be determined except the reference image in the first image need to be determined by the analysis model based on the deterioration degree determination coefficient and the degree of deterioration in the reference image, so as to obtain the correlation coefficient between the image and the reference image of the same lesion type. This correlation coefficient represents the degree of closeness between the degree of deterioration of the image and the reference image of the same lesion type (the degree of deterioration is the fifth grade). The correlation coefficient is used to analyze the difference between the reference image and the first image. The larger the correlation coefficient, the closer the lesion type, lesion size and degree of deterioration of the first image are to the reference image, that is, its cyst area, the size corresponding to the cyst, etc. are highly similar.

[0094] Step 1055: Based on the correlation coefficient and the lesion index, determine the deterioration degree determination coefficient corresponding to the first image.

[0095] The larger the correlation coefficient, the relatively larger the deterioration degree determination coefficient; the smaller the correlation coefficient, the relatively smaller the deterioration degree determination coefficient. Adjust the correlation coefficient through the lesion index to obtain the deterioration degree determination coefficient, so as to classify the degree of deterioration more accurately.

[0096] In some embodiments, according to the following formula, step 1055 is implemented to determine the deterioration degree determination coefficient corresponding to the first image based on the correlation coefficient and the lesion index:

[0097]

[0098] In the formula, represents the deterioration degree determination coefficient, represents the exponential function with the natural constant as the base, and exp[-] is used for negative correlation mapping and normalization processing, represents the lesion index of cervical endometriosis cyst, Indicates the lesion index of abdominal wall scar endometriosis cyst and other types of cysts Represents the correlation coefficient calculation symbol. a represents the reference image corresponding to the lesion type of each other image to be determined, and b represents the other image to be determined Represents the correlation coefficient between the reference image corresponding to the lesion type of each other image to be determined and the other image to be determined

[0099] Since the lesion index affects the deterioration degree coefficient, that is, the deterioration degree at different positions of the lesion is different. Specifically, the ratio of the first index and the second index is used to adjust the size of the correlation coefficient, so that the deterioration degree determination coefficient integrates the lesion type information to obtain the deterioration degree determination coefficient, so as to classify the deterioration degree more accurately. The index ratio result before the correlation coefficient Quantify the proximity of the first index and the second index. The closer the two indexes are, the closer the lesion types are. Then, when participating in the calculation of the deterioration degree determination coefficient, its weight is greater

[0100] The correlation coefficient aims to analyze the difference between the reference image and the first image. When the deterioration degrees of the reference image and the first image are highly consistent, the normalized value of the correlation coefficient Is closer to 1. After subtracting 1 from it and performing inverse normalization on the result Is closer to 1, and the corresponding deterioration degree determination coefficient Is higher; conversely, when there is a large difference in their deterioration degrees, for example, there are large differences in the lesion size (that is, the area of the connected lesion area) and the contour shape of the connected lesion area, that is, there are large differences in the lesion characteristics of endometriosis, then there are large differences in their deterioration degrees. At this time, the normalized correlation coefficient is closer to 0, and the inverse normalization result is less than 1, and the corresponding deterioration degree determination coefficient Is lower

[0101] In some embodiments, determining the deterioration degree of the endometriosis ultrasound image with sample annotation based on the deterioration degree determination coefficient corresponding to the first image includes: (1) in response to the deterioration degree determination coefficient corresponding to the first image being within the first threshold range, the deterioration degree of the endometriosis ultrasound image with sample annotation is the first level; (2) in response to the deterioration degree determination coefficient corresponding to the first image being within the second threshold range, the deterioration degree of the endometriosis ultrasound image with sample annotation is the second level; (3) in response to the deterioration degree determination coefficient corresponding to the first image being within the third threshold range, the deterioration degree of the endometriosis ultrasound image with sample annotation is the third level; (4) in response to the deterioration degree determination coefficient corresponding to the first image being within the fourth threshold range, the deterioration degree of the endometriosis ultrasound image with sample annotation is the fourth level; (5) in response to the deterioration degree determination coefficient corresponding to the first image being within the fifth threshold range, the deterioration degree of the endometriosis ultrasound image with sample annotation is the fifth level. Among them, the levels from the first level to the fifth level increase in sequence, and the deterioration degree determination coefficients also increase in sequence.

[0102] The deterioration degree is divided into five levels as the grading standard for quantifying the deterioration degree, and the level of the deterioration degree corresponding to the image is determined by judging the threshold range where the deterioration degree determination coefficient is located.

[0103] In some embodiments, the first threshold range can be , and the second threshold range can be , and the third threshold range can be , and the fourth threshold range can be , and the fifth threshold range can be .

[0104] Exemplarily, when the deterioration degree determination coefficient corresponding to the first image is within , the deterioration degree of the endometriosis ultrasound image with sample annotation is the first level; when the deterioration degree determination coefficient corresponding to the first image is within , the deterioration degree of the endometriosis ultrasound image with sample annotation is the second level; when the deterioration degree determination coefficient corresponding to the first image is within , the deterioration degree of the endometriosis ultrasound image with sample annotation is the third level; when the deterioration degree determination coefficient corresponding to the first image is within , the deterioration degree of the endometriosis ultrasound image with sample annotation is the fourth level; when the deterioration degree determination coefficient corresponding to the first image is within , the deterioration degree of the endometriosis ultrasound image with sample annotation is the fifth level.

[0105] Step 106: Based on the deterioration degree determination coefficient and lesion type corresponding to the first image, annotate the sample endometriosis ultrasound image to obtain the sample annotated endometriosis ultrasound image, where the sample annotated endometriosis ultrasound image includes classification labels for the lesion type and the deterioration degree.

[0106] The sample annotated endometriosis ultrasound image includes classification labels for the lesion type and the deterioration degree. These classification labels can assign recommended classification results for the lesion type and the deterioration degree to each endometriosis ultrasound image, thus providing an effective auxiliary reference basis for doctors to evaluate and determine the lesion type and the deterioration degree, and helping to improve doctors' trust and acceptance of ultrasound technology.

[0107] It should be noted that the classification result of the deterioration degree is only a recommended result, and specifically, professional doctors need to evaluate and determine it according to the actual situation and medical professional knowledge.

[0108] In some embodiments, the sample annotated endometriosis ultrasound image with classification labels can be obtained by training a CNN classification network. The loss function of the CNN classification network uses the cross-entropy function, and the cross-entropy function can measure the amount of information required to eliminate uncertainty and improve the parameter update speed.

[0109] In summary, for the imaging-assisted analysis method for endometriosis provided by the present invention, edge detection is performed on the preprocessed sample endometriosis ultrasound image to determine the lesion indicators at the endometriosis site, so as to accurately extract the lesion features and quantify the size and trend of the lesion area; in order to provide a type basis for annotation, further based on the lesion indicators at the endometriosis site, accurately determine the lesion type corresponding to each first image; then, according to the lesion indicators at the endometriosis site, the lesion type, and the analysis model corresponding to the first image, determine the deterioration degree determination coefficient corresponding to the first image, so as to quantify the severity of the lesion; finally, based on the deterioration degree determination coefficient and the lesion type corresponding to the first image, annotate the sample endometriosis ultrasound image to obtain the sample annotated endometriosis ultrasound image including classification labels for the lesion type and the deterioration degree. This method can provide classification labels for the lesion type and the deterioration degree obtained through accurate extraction and quantitative analysis, and improve the accuracy of the classification labels.

[0110] It should be noted that: the above sequence of embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0111] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. An imaging-assisted analysis method for endometriosis, characterized in that, The method includes: Obtaining a sample endometriosis ultrasound image, where the sample endometriosis ultrasound image contains lesion information of a sample pelvic endometriosis lesion, a sample cervical endometriosis cyst, and an abdominal wall scar endometriosis cyst; Preprocessing the sample endometriosis ultrasound image to obtain a first image; Performing edge detection on the first image to determine the lesion index of the endometriosis site corresponding to each first image; Among them, the method for obtaining the lesion index of the endometriosis site is: performing edge detection on the first image to obtain an edge detection result, where the edge detection result includes a connected lesion area, a lesion position, and ultrasonic feature information; determining a first index and a second index according to the area and contour shape features of the connected lesion area, where the first index is the lesion index of the cervical endometriosis cyst, and the second index is the lesion index of the abdominal wall scar endometriosis cyst and other types of cysts; determining the lesion index of the endometriosis site based on the first index and the second index; Among them, the calculation formula of the first index is: In the formula, represents the lesion index of cervical endometriotic cyst, represents the area of the connected lesion region, i represents the intersection point serial number, represents the Euclidean distance from the geometric center of the connected lesion region to the i-th intersection point, represents the Euclidean distance from the geometric center of the connected lesion region to the (i + 1)-th intersection point, where the intersection point and the intersection point serial number i satisfy: a ray is made from the geometric center of the connected lesion region along the horizontal direction to the contour of the connected lesion region, the ray intersects the contour at the first intersection point, the intersection point serial number i = 1 is recorded, at this time the included angle between the ray and the horizontal line is 0°, the ray is rotated counterclockwise, and each time it is rotated counterclockwise by 30°, an intersection point of the ray and the contour is obtained. After the ray completes one full rotation, a total of twelve intersection points are obtained, and they are numbered in sequence according to the obtained order to get the serial number i; Based on the lesion index of the endometriosis site, determining the lesion type corresponding to each first image; According to the lesion index of the endometriosis site, the lesion type, and an analysis model corresponding to the first image, determining the deterioration degree determination coefficient corresponding to the first image; Based on the deterioration degree determination coefficient and the lesion type corresponding to the first image, annotating the sample endometriosis ultrasound image to obtain a sample annotated endometriosis ultrasound image, where the sample annotated endometriosis ultrasound image includes classification labels of the lesion type and the deterioration degree.

2. The imaging-assisted analysis method for endometriosis according to claim 1, wherein The preprocessing of the sample endometriosis ultrasound image includes: Performing histogram equalization processing on the sample endometriosis ultrasound image.

3. The imaging-assisted analysis method for endometriosis according to claim 1, characterized in that Determining the second index according to the area and contour shape features of the connected lesion area: In the formula, represents the lesion index of abdominal wall scar endometriosis cyst and other types of cysts, represents the area of the connected lesion region, j represents the intersection point serial number, represents the difference between the maximum Euclidean distance and the minimum Euclidean distance from the geometric center of the connected lesion region to the intersection point, represents the Euclidean distance from the geometric center of the connected lesion region to the j-th intersection point, represents the Euclidean distance from the geometric center of the connected lesion region to the (j + 6)-th intersection point, where the intersection point and the intersection point serial number j satisfy: a ray is made from the geometric center of the connected lesion region along the horizontal direction to the contour of the connected lesion region, the ray intersects the contour at the first intersection point, the intersection point serial number j = 1 is recorded, at this time the angle between the ray and the horizontal line is 0°, the ray is rotated counterclockwise, and an intersection point of the ray and the contour is obtained every 30° of counterclockwise rotation. When the ray completes one full rotation, a total of twelve intersection points are obtained, and they are numbered in the order of acquisition to obtain the serial number j.

4. An imaging-assisted analysis method for endometriosis according to claim 1, characterized in that, The determining the deterioration degree determination coefficient corresponding to the first image according to the lesion index of the endometriosis site, the lesion type, and the analysis model corresponding to the first image includes: Obtaining an analysis model and determining the deterioration degree determination coefficient corresponding to the first image; Determining the image with the largest deterioration degree determination coefficient among the first images with the same lesion type, obtaining the image with the largest deterioration degree determination coefficient corresponding to each lesion type, and using the image with the largest deterioration degree determination coefficient corresponding to each lesion type as the reference image for each lesion type; Determining that the deterioration degree of the reference image of each lesion type is the fifth level; Inputting the reference image, the other images to be determined except the reference image in the first image, and the deterioration degree of the reference image into the analysis model to obtain the lesion type corresponding to each of the other images to be determined, and the correlation coefficient between the reference image corresponding to the lesion type of each of the other images to be determined and the other images to be determined; Based on the correlation coefficient and the lesion index, determining the deterioration degree determination coefficient corresponding to the first image.

5. An imaging-assisted analysis method for endometriosis according to claim 1, characterized in that, Determining the lesion index of the endometriosis site based on the first index and the second index: In the formula, represents the lesion index of the endometriosis site, represents the sign function, represents the lesion index of the endometriotic cyst of the cervix, represents the lesion index of the endometriotic cyst of the abdominal wall scar and other types of cysts.

6. The imaging-assisted analysis method for endometriosis according to claim 4, wherein Determine the deterioration degree of the sample-labeled endometriosis ultrasound image based on the deterioration degree determination coefficient corresponding to the first image, including: In response to the deterioration degree determination coefficient corresponding to the first image being within the first threshold range, the deterioration degree of the sample-labeled endometriosis ultrasound image is the first level; In response to the deterioration degree determination coefficient corresponding to the first image being within the second threshold range, the deterioration degree of the sample-labeled endometriosis ultrasound image is the second level; In response to the deterioration degree determination coefficient corresponding to the first image being within the third threshold range, the deterioration degree of the sample-labeled endometriosis ultrasound image is the third level; In response to the deterioration degree determination coefficient corresponding to the first image being within the fourth threshold range, the deterioration degree of the sample-labeled endometriosis ultrasound image is the fourth level; In response to the deterioration degree determination coefficient corresponding to the first image being within the fifth threshold range, the deterioration degree of the sample-labeled endometriosis ultrasound image is the fifth level.

7. The imaging-assisted analysis method for endometriosis according to claim 4, wherein, Determine the deterioration degree determination coefficient corresponding to the first image based on the correlation coefficient and the lesion index: In the formula, represents the deterioration degree determination coefficient, represents the exponential function with the natural constant as the base, represents the lesion index of cervical endometriotic cyst, represents the lesion indexes of abdominal wall scar endometriotic cyst and other types of cysts, represents the correlation coefficient calculation symbol, a represents the reference image corresponding to the lesion type of each of the other images to be determined, and b represents the other image to be determined, represents the correlation coefficient between the reference image corresponding to the lesion type of each of the other images to be determined and the other image to be determined.

8. An imaging-assisted analysis method for endometriosis according to claim 4, characterized in that The analysis model is a template matching model.

Images