Ultrasonic image intelligent processing method for obstetrics and gynecology department

Abstract

The invention relates to the technical field of image filtering enhancement, in particular to an ultrasonic image intelligent processing method for gynaecology and obstetrics. The method comprises the steps of obtaining an abdominal wall ultrasound image and performing region division to obtain a target region; further adjusting a preset filtering window parameter according to the gray level fluctuation degree of the pixel points in the target area; further dividing the connected domain into an organization connected domain and a noise connected domain according to the gray uniformity of the pixel points in each connected domain and the gradient direction change of the edge pixel points; further correcting the initial window parameters according to the relative number of noise connected domains and tissue connected domains in the target area to obtain final window parameters; and finally, filtering the abdominal wall ultrasound image according to the final window parameter of each image area. According to the method, the size of the filtering window is adaptively determined by analyzing the gray fluctuation in the ultrasonic image area and the noise expression of the connected domain, speckle noise is effectively removed, and the ovarian vesicle recognition accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image filtering and enhancement, and in particular to an intelligent processing method of ultrasonic images for obstetrics and gynecology. Background Art

[0002] Ultrasound examination is an important examination method for clinical gynecological and obstetric diseases. With the continuous improvement of ultrasound probe structure technology, the way of obstetric and gynecological ultrasound examination has also developed from simple transabdominal ultrasound examination to transvaginal ultrasound examination, transuterine ultrasound examination and other examination methods.

[0003] Usually, unprocessed gynecological ultrasound images contain speckle noise, which can seriously interfere with the normal diagnosis and analysis of gynecological ultrasound images. For example, polycystic ovary syndrome is a complex endocrine disorder, and its main diagnosis method is to count the number of small cysts in the ovaries in the ultrasound image. Since speckle noise is randomly distributed in ultrasound images, setting a smaller window may not effectively remove large-size speckle noise, while a larger window may cause over-smoothing and blurring, affecting the accuracy of the analysis results. Summary of the invention

[0004] In order to solve the technical problem that the unreasonable setting of the filter window affects the filtering effect of gynecological ultrasound images, the purpose of the present invention is to provide an intelligent processing method for ultrasound images in obstetrics and gynecology. The technical solution adopted is as follows: Acquire an abdominal wall ultrasound image and perform region division to obtain various image regions of the abdominal wall ultrasound image; and select the image regions as target regions one by one; Adjust the preset filter window parameters according to the grayscale fluctuation degree of the pixel points in the target area to obtain the initial window parameters of the target area; obtain the connected domain in the target area; obtain the noise possibility of each connected domain according to the grayscale uniformity of the pixel points in each connected domain and the gradient direction change of the edge pixel points; divide the connected domain into a tissue connected domain and a noise connected domain according to the distribution of the noise possibility; according to the relative number of the noise connected domain and the tissue connected domain in the target area, correct the initial window parameters to obtain the final window parameters of the target area; The abdominal wall ultrasound image is filtered according to the final window parameters of each of the image regions.

[0005] Furthermore, the method for obtaining the initial window parameters includes: According to the grayscale range of the pixels in the target area and the difference between the grayscale value of each pixel and the overall grayscale, a grayscale fluctuation drastic parameter of the target area is obtained; The initial window parameters of the target area are obtained according to the grayscale fluctuation drastic parameter and the preset filter window parameter; the grayscale fluctuation drastic parameter and the preset filter window parameter are both positively correlated with the initial window parameter.

[0006] Furthermore, the method for obtaining the noise possibility includes: According to the difference between the median and mean grayscale values ​​of pixels in each connected domain, combined with the variation range of the gradient direction of all adjacent edge pixels, the noise possibility of each connected domain is obtained; the difference between the median and mean grayscale values, and the variation range of the gradient direction are positively correlated with the noise possibility.

[0007] Furthermore, the method of dividing into tissue connected domains and noise connected domains includes: The noise possibilities of all the connected domains in the target area are sorted from small to large to form a sequence to be analyzed; according to the change of the noise possibility in the sequence to be analyzed, the sequence to be analyzed is divided into two parts, the connected domain corresponding to the one where the maximum noise possibility is located is the noise connected domain, and the connected domain corresponding to the one where the minimum noise possibility is located is the tissue connected domain.

[0008] Furthermore, the method of dividing the sequence to be analyzed into two parts according to the change of the noise possibility in the sequence to be analyzed includes: Sorting the connected domains in the order of the noise possibilities in the sequence to be analyzed to obtain a connected domain sequence; taking the connected domains at non-two ends of the connected domain sequence as connected domains to be analyzed; According to the difference in the noise probability of the connected domain before and after each connected domain to be analyzed, the demarcation probability of each connected domain to be analyzed is obtained; the difference in the noise probability of the connected domain before and after the connected domain to be analyzed is positively correlated with the demarcation probability; the connected domain to be analyzed with the largest demarcation probability is selected as the demarcation point, and the connected domain at and before the demarcation point is divided into one part; and the connected domain after the demarcation point is divided into another part.

[0009] Furthermore, the method for obtaining the final window parameters includes: In the target area, the final window parameters of the target area are obtained according to the difference between the number of the noise connected domains and the number of the tissue connected domains in combination with the initial window parameters; the difference between the number of the noise connected domains and the number of the tissue connected domains and the initial window parameters are positively correlated with the final window parameters.

[0010] Furthermore, the connected domains in the target area are extracted by using the Otsu threshold segmentation algorithm.

[0011] Furthermore, the preset filtering window parameter is 5.

[0012] Furthermore, the size of the image area is .

[0013] Furthermore, the abdominal wall ultrasound image is filtered using a median filter.

[0014] The present invention has the following beneficial effects: The present invention first obtains an abdominal wall ultrasound image and performs regional division to obtain a target area, so as to more carefully analyze the noise performance of different image areas, set more appropriate filtering parameters, and improve the filtering effect; further adjust the preset filtering window parameters according to the grayscale fluctuation degree of the pixel points in the target area, obtain the initial window parameters of the target area, preliminarily adapt to the grayscale fluctuation in the target area, and prepare for the subsequent further adjustment of the filtering window size; further evaluate the possibility that the connected domain is speckle noise from the grayscale uniformity angle and the edge gradient direction change angle, and provide a basis for the subsequent classification of the connected domain and the final correction of the filtering window; further divide the connected domain into a tissue connected domain and a noise connected domain according to the distribution of noise possibility, so as to further determine the noise situation in the target area and correct the initial window parameters; further correct the initial window parameters according to the relative number of the noise connected domain and the tissue connected domain in the target area to obtain the final window parameters of the target area, improve the adaptability of the final window parameters, protect the edge of the image structure, and improve the reliability of the analysis of gynecological and obstetric images such as polycystic ovary; finally, filter the abdominal wall ultrasound image according to the final window parameters of each image area. The present invention analyzes the grayscale fluctuations in the ultrasound image area and the noise performance of the connected domain, adaptively determines the filter window size, effectively removes speckle noise and protects the image detail texture, thereby improving the accuracy of ovarian cyst recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A flowchart of an intelligent processing method for ultrasound images in obstetrics and gynecology provided by an embodiment of the present invention; Figure 2 A flowchart for obtaining a tissue connected domain and a noise connected domain is provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation, structure, features and effects of an intelligent processing method for ultrasound images for obstetrics and gynecology proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0019] The specific scheme of the intelligent processing method of ultrasound images for obstetrics and gynecology provided by the present invention is described in detail below with reference to the accompanying drawings.

[0020] See also Figure 1 , which shows a flow chart of an intelligent processing method for ultrasound images for obstetrics and gynecology provided by an embodiment of the present invention, specifically comprising: Step S1: Acquire an abdominal wall ultrasound image and perform region division to obtain various image regions of the abdominal wall ultrasound image; select image regions as target regions one by one.

[0021] In the embodiment of the present invention, an ultrasonic medical device is first used to obtain an abdominal wall ultrasonic image of a gynecological patient. If the abdominal wall ultrasonic image is not a grayscale image, grayscale preprocessing is required.

[0022] When using ultrasonic medical equipment to collect images of the abdominal wall, when sound waves interact with tissues or objects and return to the sensor, the interference and scattering effects of these sound waves will cause bright and dark spots to appear on the image, presenting random particles or spots, which is the speckle noise of the ultrasound image. Since the noise content of different image areas of the collected abdominal wall is different, the abdominal wall ultrasound image is divided into regions to obtain each image area of ​​the abdominal wall ultrasound image, so as to analyze the noise performance of different image areas more carefully, set more appropriate filtering parameters, and improve the filtering effect.

[0023] As an example, the abdominal wall ultrasound image is evenly divided into image regions, and the size of each image region is .

[0024] It should be noted that the analysis method for each image region is consistent, and image regions are selected one by one as target regions. Only one target region is used as an example for explanation here, and no repetition is given. The implementer may also set other sizes for region division.

[0025] Step S2: Adjust the preset filter window parameters according to the grayscale fluctuation degree of the pixels in the target area to obtain the initial window parameters of the target area; obtain the connected domain in the target area; obtain the noise possibility of each connected domain according to the grayscale uniformity of the pixels in each connected domain and the gradient direction change of the edge pixels; divide the connected domain into tissue connected domain and noise connected domain according to the distribution of noise possibility; correct the initial window parameters according to the relative number of noise connected domain and tissue connected domain in the target area to obtain the final window parameters of the target area.

[0026] Considering that the grayscale fluctuation degree of pixels in the target area reflects the richness of the organizational content or the quantity characteristics of noise points in the target area, the preset filter window parameters are adjusted according to the grayscale fluctuation degree of pixels in the target area, and the initial window parameters of the target area are obtained to preliminarily adapt to the grayscale fluctuation in the target area, so as to prepare for the subsequent further adjustment of the filter window size.

[0027] Preferably, in one embodiment of the present invention, considering that speckle noise appears as randomly distributed bright and dark particles in an image, such as dark particles on a bright background (noise reduces local grayscale) in a high-echo area (such as ovarian stroma), and bright particles on a dark background (noise increases local grayscale) in a low-echo area (such as vesicular fluid), resulting in drastic grayscale fluctuations in the target area, requiring a larger filter window for smoothing and denoising; and the grayscale range of pixels in the area represents the grayscale fluctuation range, the larger the grayscale range, the more drastic the grayscale fluctuation; the greater the difference between the grayscale value of each pixel and the overall grayscale, the greater the grayscale discreteness and the more drastic the grayscale fluctuation; Based on this, the grayscale fluctuation drastic parameter of the target area is obtained according to the grayscale range of the pixels in the target area and the difference between the grayscale value of each pixel and the overall grayscale; The initial window parameters of the target area are obtained according to the grayscale fluctuation drastic parameter and the preset filter window parameter; the grayscale fluctuation drastic parameter and the preset filter window parameter are both positively correlated with the initial window parameter.

[0028] As an example, the product of the sum of the absolute values ​​of the difference between the grayscale value of each pixel in the target area and the mean of the grayscale values ​​of all pixels and the range of the grayscale values ​​is taken as the grayscale fluctuation coefficient of the target area; the ratio of the grayscale fluctuation coefficient of the target area to the sum of the grayscale fluctuation coefficients of all image areas is taken as the grayscale fluctuation intensity parameter of the target area, representing the grayscale fluctuation degree of the pixels in the target area; The preset filter window parameter is 5, the sum of the grayscale fluctuation parameter and the constant 1 is used as the initial adjustment coefficient, and the product of the initial adjustment coefficient and the preset filter window parameter is used as the initial window parameter of the target area.

[0029] As another example, considering that the variance of the grayscale values ​​of pixels in the target area can also reflect the degree of grayscale fluctuation, the larger the variance, the more drastic the grayscale fluctuation, the sum of the absolute values ​​of the difference between the grayscale value of each pixel in the target area and the mean of the grayscale values ​​of all pixels, the variance of the grayscale value, and the range of the grayscale value are multiplied, and the product is taken as the grayscale fluctuation coefficient of the target area.

[0030] In other embodiments of the present invention, the sum of the absolute values ​​of the differences between the grayscale value of each pixel in the target area and the mean of the grayscale values ​​of all pixels, the variance of the grayscale values, and the range of the grayscale values ​​can be fused by addition or weighted summation; the grayscale fluctuation coefficient can also be normalized by linear normalization to obtain a grayscale fluctuation drastic parameter.

[0031] Considering the high richness of tissue content in the target area, such as multiple adjacent vesicles, blood vessels, tissue interfaces, etc., it will also cause pixel grayscale fluctuations. In order to avoid confusion between rich tissue texture and dense speckle noise, resulting in unreasonable initial window parameter settings, the initial window parameters need to be further adjusted.

[0032] Considering that the small cysts in the ovary appear as multiple circular dark areas with clear edges and low echoes in the image, with uniform internal grayscale, while speckle noise shows the characteristics of uneven grayscale, blurred boundaries, and random gradient directions, the connected domains formed by the two are quite different, so the connected domains in the target area are obtained; according to the grayscale uniformity of the pixels in each connected domain and the gradient direction change of the edge pixels, the noise possibility of each connected domain is obtained, and the possibility of the connected domain being speckle noise is evaluated from the perspective of grayscale uniformity and edge gradient direction change, which provides a basis for the subsequent classification of the connected domain and the final correction of the filter window; Considering that the noise probability of the connected domain of normal tissues such as vesicles is quite different from that of the connected domain of speckle noise, the connected domain is divided into tissue connected domain and noise connected domain according to the distribution of noise probability, in preparation for the subsequent filter window correction.

[0033] In one embodiment of the present invention, the connected regions in the target region are extracted using the Otsu threshold segmentation algorithm.

[0034] In other embodiments of the present invention, the implementer may also extract the connected domain by using the Sobel operator, morphological opening operation, etc. The methods for obtaining the connected domain are well known to those skilled in the art and will not be described in detail.

[0035] Preferably, in one embodiment of the present invention, considering that speckle noise is manifested as isolated extremely high or extremely low grayscale pixels, a few extreme values ​​will significantly affect the grayscale mean but the median is almost unchanged, and the difference between the median and the mean is large; while the grayscale in the connected domain of normal tissue is uniform, and the median is close to the mean; at the same time, the gradient direction of the edge of speckle noise is random, without directional trend, and the gradient direction change amplitude is larger than that of normal tissue; Based on this, the noise possibility of each connected domain is obtained according to the difference between the median and mean grayscale values ​​of pixels in each connected domain and the variation amplitude of the gradient direction of all adjacent edge pixels; the difference between the median and mean grayscale values ​​and the variation amplitude of the gradient direction are positively correlated with the noise possibility.

[0036] As an example, starting from the edge pixel point of the connected domain closest to the lower left corner of the target area, the absolute value of the difference in the gradient direction of any two adjacent edge pixels is calculated in a clockwise direction, and the absolute value of the difference between the median and the mean of the grayscale values ​​of the pixels in the connected domain is multiplied by the average of the absolute values ​​of the difference in the gradient direction of all adjacent edge pixels. The product is linearly normalized and used as the noise possibility of the corresponding connected domain.

[0037] The difference between the median and the mean is reflected by the absolute value of the difference, which shows the grayscale uniformity of the pixels in the connected domain. The smaller the absolute value of the difference between the median and the mean, the stronger the grayscale uniformity. Similarly, the absolute value of the difference is used to show the change amplitude of the gradient direction of adjacent pixels, and then the change of the gradient direction of all adjacent edge pixels is reflected by averaging, and the noise possibility of the connected domain is obtained by multiplication.

[0038] In other embodiments of the present invention, the implementer may also fuse the absolute value of the difference between the median and mean grayscale values ​​of pixels in a connected domain with the average of the absolute values ​​of the difference in the gradient direction of all adjacent edge pixels by addition or weighted summation; the implementer may also add the mean absolute deviation, variance, etc. of the grayscale value to measure grayscale discrete features and analyze grayscale uniformity, which will not be elaborated herein.

[0039] Preferably, in one embodiment of the present invention, please refer to Figure 2 , which shows a flowchart of obtaining a tissue connectivity domain and a noise connectivity domain provided by an embodiment of the present invention, specifically including: Step S201: sort the noise possibilities of all connected domains in the target area from small to large to form a sequence to be analyzed.

[0040] Considering that it is easier to analyze the changes in noise possibilities after sorting the noise possibilities, we first construct the sequence to be analyzed and use it to analyze the distribution of noise possibilities.

[0041] Step S202: according to the change of noise possibility in the sequence to be analyzed, the sequence to be analyzed is divided into two parts, the connected domain corresponding to the part where the noise possibility is the maximum is the noise connected domain, and the connected domain corresponding to the part where the noise possibility is the minimum is the tissue connected domain.

[0042] Considering that the noise probability of the connected domain of normal tissue is smaller, and the noise probability of the connected domain of speckle noise is larger, and they show obvious boundary characteristics in the sequence to be analyzed, the sequence to be analyzed is divided into two parts according to the change of noise probability in the sequence to be analyzed. At the same time, in the two sequence segments, the maximum noise probability is most likely to correspond to speckle noise, so the connected domain corresponding to the one with the maximum noise probability is the noise connected domain, and the connected domain corresponding to the one with the minimum noise probability is the tissue connected domain.

[0043] In one embodiment of the present invention, in order to analyze the boundary characteristics of the sequence to be analyzed, the sequence to be analyzed is divided into two parts, and the connected domains are sorted in the order of the noise possibility in the sequence to be analyzed to obtain a connected domain sequence; the connected domains other than the two ends of the connected domain sequence are used as the connected domains to be analyzed; Considering that the difference in noise probability between the connected domains before and after the connected domain to be analyzed is the largest, it indicates that the boundary feature is the most obvious. Therefore, the boundary probability of each connected domain to be analyzed is obtained according to the difference in noise probability between the connected domains before and after each connected domain to be analyzed. The difference in noise probability between the connected domains before and after the connected domain to be analyzed is positively correlated with the boundary probability. As an example, the connected domain to be analyzed and the previous connected domain are regarded as the connected domain before the connected domain to be analyzed, and the connected domain after the connected domain to be analyzed is regarded as the connected domain before the connected domain to be analyzed; The absolute value of the difference between the mean noise probability of the connected domain before the connected domain to be analyzed and the mean noise probability of the connected domain after the connected domain to be analyzed is taken as the first difference parameter; the absolute value of the difference between the noise probability of the connected domain to be analyzed and the adjacent next connected domain is taken as the second difference parameter; the product of the first difference parameter and the second difference parameter is taken as the boundary probability of the connected domain to be analyzed.

[0044] Among them, in the form of mean, the mean of the noise possibility of the connected domain is used to express the overall characteristics of the noise possibility of the connected domain before and after the connected domain to be analyzed. In the form of absolute value of difference, the first difference parameter is used to express the difference in noise possibility before and after the connected domain to be analyzed from an overall perspective; the second difference parameter is used to express the sudden increase in the noise possibility before and after the connected domain to be analyzed, and the difference in noise possibility before and after the connected domain to be analyzed is expressed from a local perspective; finally, they are fused by multiplication to obtain the boundary possibility.

[0045] The connected domain to be analyzed with the greatest demarcation possibility is selected as the demarcation point, and the connected domain before and at the demarcation point is divided into one part (tissue connected domain); the connected domain after the demarcation point is divided into another part (noise connected domain).

[0046] In another embodiment of the present invention, considering that the maximum value in the first-order difference sequence of the sequence to be analyzed corresponds to the most obvious sudden increase in noise possibility, and thus corresponds to the boundary of the noise possibility of two types of connected domains, the sequence to be analyzed is segmented from the maximum value of the elements of the first-order difference sequence.

[0047] In other embodiments of the present invention, the implementer may also adopt a threshold screening method to set a noise possibility threshold, and use a connected domain less than the noise possibility threshold as a tissue connected domain, and use a connected domain greater than or equal to the noise possibility threshold as a noise connected domain; wherein, a number of (e.g., 100) ultrasound images may be selected in advance, and the tissue connected domain and the noise connected domain may be manually labeled to calculate the noise possibility, and the position where the two types of distributions overlap the least may be selected as the threshold, such as 95% of the tissue connected domains have a noise possibility less than the noise possibility threshold, and 80% of the noise connected domains have a noise possibility greater than or equal to the noise possibility threshold.

[0048] You can also use the K-means clustering algorithm to set K=2 to classify the noise possibilities and thus classify the connected domains.

[0049] Furthermore, considering that the relative number of noise connected domains and tissue connected domains in the target area reflects the dominance of the two connected domains, it is judged whether the target area is dominated by normal tissue or noise, so as to dynamically adjust the initial window parameters. Therefore, according to the relative number of noise connected domains and tissue connected domains in the target area, the initial window parameters are corrected to obtain the final window parameters of the target area, thereby improving the adaptability of the final window parameters, protecting the edges of image structures, and improving the reliability of gynecological and obstetric image analysis such as polycystic ovary.

[0050] Preferably, in one embodiment of the present invention, considering that the greater the number of noise connected domains compared to the number of tissue connected domains, the stronger the dominance of noise, the filter window should be enlarged, smoothing should be enhanced, and random speckle noise should be suppressed as much as possible; on the contrary, the smaller the number of noise connected domains compared to the number of tissue connected domains, the more tissue structures there are in the region, and excessive smoothing should not be allowed, and the filter window should be reduced to avoid damaging important structures such as edges and blurring vesicles; Based on this, in the target area, the final window parameters of the target area are obtained according to the difference between the number of noise connected domains and the number of tissue connected domains combined with the initial window parameters; the difference between the number of noise connected domains and the number of tissue connected domains and the initial window parameters are positively correlated with the final window parameters.

[0051] As an example, the difference between the number of noise connected domains and the number of tissue connected domains is used as the numerator, the sum of the number of noise connected domains and the number of tissue connected domains is used as the denominator, and the sum of the fractional ratio and the constant 1 is used as the final correction coefficient; the product of the final correction coefficient and the initial window parameter is rounded up to an odd integer, and the rounded result is used as the final window parameter.

[0052] It should be noted that the minimum value of the final window parameter is 3. When it is less than 3, it is set to 3, and the corresponding filter window size is .

[0053] In another embodiment of the present invention, the initial window parameters are corrected by integrating the area angle. As an example, the difference between the number of noise connected domains and the number of tissue connected domains is used as the numerator, the sum of the number of noise connected domains and the number of tissue connected domains is used as the denominator, and the fractional ratio is used as the first correction factor. The difference between the total area of ​​the noise connected domain and the total area of ​​the tissue connected domain is used as the numerator, the total area of ​​the noise connected domain and the total area of ​​the tissue connected domain are used as the denominator, and the fraction ratio is used as the second correction factor; The first correction factor and the second correction factor are weighted and summed with weights of 0.8 and 0.2, and the sum of the weighted summation result and the constant 1 is used as the final correction coefficient; among them, since the area of ​​normal tissue is usually larger than the noise connectivity domain, it will not show too much noise dominance from the area perspective, and is more sensitive from the quantity perspective, so the weight assigned to the second correction factor is smaller.

[0054] Step S3: filtering the abdominal wall ultrasound image according to the final window parameters of each image region.

[0055] By analyzing the grayscale fluctuation and noise evaluation of abdominal wall ultrasound images in different regions, the filter window is dynamically corrected to obtain the most suitable final window parameters for each image area, balancing the needs of protecting image details and filtering out noise interference. Finally, the abdominal wall ultrasound images are filtered according to the final window parameters of each image area.

[0056] Preferably, in one embodiment of the present invention, considering that the speckle noise in the ultrasound image is random point noise, median filtering is naturally suitable for eliminating sharp isolated points while retaining edge information, which just matches the characteristics of the ultrasound image. Therefore, median filtering is used to filter the abdominal wall ultrasound image to provide reliable image data support for relevant personnel.

[0057] It should be noted that in other embodiments of the present invention, the implementer may also adopt other window filtering methods such as weighted median filtering, which and median filtering are technical means well known to those skilled in the art and will not be described in detail here.

[0058] In summary, in order to solve the technical problem that the setting of the filter window is unreasonable and affects the filtering effect of gynecological ultrasound images, the present invention proposes an intelligent processing method for ultrasound images in obstetrics and gynecology. The present invention obtains an ultrasound image of the abdominal wall and performs regional division to obtain a target area; further adjusts the preset filter window parameters according to the grayscale fluctuation degree of the pixels in the target area; further divides the connected domain into a tissue connected domain and a noise connected domain according to the grayscale uniformity of the pixels in each connected domain and the gradient direction change of the edge pixels; further corrects the initial window parameters according to the relative number of noise connected domains and tissue connected domains in the target area to obtain the final window parameters; finally, the abdominal wall ultrasound image is filtered according to the final window parameters of each image area. The present invention adaptively determines the filter window size by analyzing the grayscale fluctuation and the noise performance of the connected domain in the ultrasound image area, effectively removes speckle noise, and improves the accuracy of ovarian cyst recognition.

[0059] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying 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.

[0060] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. An intelligent processing method for ultrasound images in obstetrics and gynecology, characterized in that: The method comprises: Acquire an abdominal wall ultrasound image and perform region division to obtain various image regions of the abdominal wall ultrasound image; and select the image regions as target regions one by one; Adjust the preset filter window parameters according to the grayscale fluctuation degree of the pixel points in the target area to obtain the initial window parameters of the target area; obtain the connected domain in the target area; obtain the noise possibility of each connected domain according to the grayscale uniformity of the pixel points in each connected domain and the gradient direction change of the edge pixel points; divide the connected domain into a tissue connected domain and a noise connected domain according to the distribution of the noise possibility; according to the relative number of the noise connected domain and the tissue connected domain in the target area, correct the initial window parameters to obtain the final window parameters of the target area; filtering the abdominal wall ultrasound image according to the final window parameters of each of the image regions; The method of dividing into tissue connected domain and noise connected domain comprises: The noise possibilities of all the connected domains in the target area are sorted from small to large to form a sequence to be analyzed; according to the change of the noise possibility in the sequence to be analyzed, the sequence to be analyzed is divided into two parts, the connected domain corresponding to the one where the maximum noise possibility is located is the noise connected domain, and the connected domain corresponding to the one where the minimum noise possibility is located is the tissue connected domain.

2. The method for intelligent processing of ultrasound images for obstetrics and gynecology according to claim 1, characterized in that: The method for obtaining the initial window parameters includes: According to the grayscale range of the pixels in the target area and the difference between the grayscale value of each pixel and the overall grayscale, a grayscale fluctuation drastic parameter of the target area is obtained; The initial window parameters of the target area are obtained according to the grayscale fluctuation drastic parameter and the preset filter window parameter; the grayscale fluctuation drastic parameter and the preset filter window parameter are both positively correlated with the initial window parameter.

3. The method for intelligent processing of ultrasound images for obstetrics and gynecology according to claim 1, characterized in that: The method for obtaining the noise possibility includes: According to the difference between the median and mean grayscale values ​​of pixels in each connected domain, combined with the variation range of the gradient direction of all adjacent edge pixels, the noise possibility of each connected domain is obtained; the difference between the median and mean grayscale values, and the variation range of the gradient direction are positively correlated with the noise possibility.

4. The method for intelligent processing of ultrasound images for obstetrics and gynecology according to claim 1, characterized in that: The method of dividing the sequence to be analyzed into two parts according to the change of the noise possibility in the sequence to be analyzed includes: Sorting the connected domains in the order of the noise possibilities in the sequence to be analyzed to obtain a connected domain sequence; taking the connected domains at non-two ends of the connected domain sequence as connected domains to be analyzed; According to the difference in the noise probability of the connected domain before and after each connected domain to be analyzed, the demarcation probability of each connected domain to be analyzed is obtained; the difference in the noise probability of the connected domain before and after the connected domain to be analyzed is positively correlated with the demarcation probability; the connected domain to be analyzed with the largest demarcation probability is selected as the demarcation point, and the connected domain at and before the demarcation point is divided into one part; and the connected domain after the demarcation point is divided into another part.

5. The method for intelligent processing of ultrasound images for obstetrics and gynecology according to claim 1, characterized in that: The method for obtaining the final window parameters includes: In the target area, the final window parameters of the target area are obtained according to the difference between the number of the noise connected domains and the number of the tissue connected domains in combination with the initial window parameters; the difference between the number of the noise connected domains and the number of the tissue connected domains and the initial window parameters are positively correlated with the final window parameters.

6. The method for intelligent processing of ultrasound images for obstetrics and gynecology according to claim 1, characterized in that: The connected regions in the target area are extracted by Otsu threshold segmentation algorithm.

7. The method for intelligent processing of ultrasound images for obstetrics and gynecology according to claim 1, characterized in that: The preset filter window parameter is 5.

8. The method for intelligent processing of ultrasound images for obstetrics and gynecology according to claim 1, characterized in that: The size of the image area is .

9. The method for intelligent processing of ultrasound images for obstetrics and gynecology according to claim 1, characterized in that: The abdominal wall ultrasound image is filtered using a median filter.

Images