An Adaptive Segmentation Method for Ultrasound Images

Abstract

This invention relates to the field of image segmentation technology, specifically to an adaptive segmentation method for ultrasound images. The method includes: processing neck ultrasound images of patients with thyroid nodules using multiple high-pass filters of different scales to obtain filtered images at each scale; filtering target directions for each region based on the degree of change in high-frequency information content in different directions within each region of the filtered image at each scale; determining the high-frequency information change direction of the filtered image by combining the proportion of high-frequency information content in the target direction across all regions; obtaining a direction change consistency factor for each region based on the difference between the target direction and the high-frequency information change direction; and obtaining a target image by combining the differences in direction change consistency factor, target direction, and high-frequency information content of the same region in filtered images at adjacent scales, and then segmenting the target image. This invention improves the accuracy of neck ultrasound image segmentation results for patients with thyroid nodules.

Description

Technical Field

[0001] This invention relates to the field of image segmentation technology, and more specifically to an adaptive segmentation method for ultrasound images. Background Technology

[0002] Medical ultrasound imaging is a practical medical imaging technology that achieves non-invasive visualization of soft tissues through the propagation and reflection of sound waves. With the continuous development and advancement of medical technology, ultrasound medical image processing technology has become an indispensable part of the medical field. Clinically, ultrasound imaging is a common and widely used medical diagnostic method, and ultrasound medical image processing technology provides doctors with more reliable and accurate diagnostic information, playing a vital role in clinical diagnosis.

[0003] Despite significant advancements in ultrasound medical image processing technology, various challenges remain in segmenting neck images of patients with thyroid nodules. These challenges include the presence of substantial stray noise, image blurring, and distortion in the original ultrasound images. These issues make it difficult to effectively segment target lesions or organs in neck ultrasound images using traditional thresholding methods, resulting in poor segmentation of the target region in the patient's neck image. Summary of the Invention

[0004] To address the issue of low accuracy in segmenting target regions in patient neck images using existing methods, this invention aims to provide an adaptive segmentation method for ultrasound images. The specific technical solution employed is as follows:

[0005] This invention provides an adaptive segmentation method for ultrasound images, the method comprising the following steps:

[0006] Neck ultrasound images of patients with thyroid nodules were acquired, and the ultrasound images were processed using multiple high-pass filters of different scales to obtain filtered images of each scale.

[0007] For each scale of the filtered image, the target direction of each region is selected based on the degree of change of high-frequency information content in different preset directions within each region of the filtered image; the high-frequency information change direction of the filtered image is determined by combining the proportion of high-frequency information content in the target direction of all regions; and the direction change consistency factor of each region is obtained based on the difference between the target direction and the high-frequency information change direction of a single region.

[0008] By combining the differences in the direction change consistency factor, target direction, and high-frequency information content of the same region in filtered images of adjacent scales, the segmentation parameters of each region are obtained;

[0009] The target image is obtained by adjusting the gray values ​​of pixels in the ultrasound image using segmentation parameters, and then adaptive segmentation is performed on the target image.

[0010] Preferably, the acquisition of the degree of change in high-frequency information content in different preset directions within each region includes:

[0011] Based on the standard deviation and average value of the high-frequency information content in each preset direction within the candidate region of the image to be analyzed, the degree of change of the high-frequency information content in each preset direction of the candidate region is obtained, and both the standard deviation and the average value are positively correlated with the degree of change.

[0012] The image to be analyzed is a filtered image of any scale, and the candidate region is any region in the image to be analyzed.

[0013] Preferably, the target direction for each region is filtered, including:

[0014] The direction corresponding to the maximum value of the change in high-frequency information content in all preset directions of each region is taken as the target direction of each region.

[0015] Preferably, determining the direction of high-frequency information change in the filtered image by combining the proportion of high-frequency information content in the target direction of all regions includes:

[0016] By utilizing the proportion of high-frequency information content in the target direction of each region in the image to be analyzed, a weighted average is calculated on the target direction to obtain the direction of high-frequency information change in the image to be analyzed.

[0017] The proportion of high-frequency information content in the target direction of each region is the ratio between the high-frequency information content in the target direction of each region and the total high-frequency information content in the image to be analyzed.

[0018] Preferably, obtaining the direction change consistency factor for each region based on the difference between the target direction and the high-frequency information change direction of a single region includes:

[0019] Calculate the first difference between the target orientation of the candidate region and the high-frequency information change direction of the image to be analyzed;

[0020] Based on the first difference, a direction change consistency factor for the candidate region is obtained, and the first difference is negatively correlated with the direction change consistency factor.

[0021] Preferably, the segmentation parameters for each region are obtained by combining the differences in the direction change consistency factor, target direction, and high-frequency information content of the same region in the filtered images of adjacent scales, including:

[0022] For any region in an ultrasound image:

[0023] Based on the change consistency factor of the direction change and the change of the target direction in any region of the filtered images of adjacent scales, the change index of any region is obtained.

[0024] The segmentation parameters of any region are obtained based on the change index of any region and the difference in high-frequency information content within any region in all filtered images of adjacent scales.

[0025] Preferably, obtaining the change index of any region based on the change consistency factor of the direction change and the change of the target direction in the filtered images of adjacent scales includes:

[0026] The product of the direction change consistency factor of any region in the filtered image at each scale and the target direction is denoted as the first feature value of any region in the filtered image at each scale.

[0027] The filtered image at the previous scale among the filtered images at adjacent scales is designated as the first image, and the filtered image at the next scale among the filtered images at adjacent scales is designated as the second image.

[0028] Calculate the second difference between the first feature value of any region in the first image and any region in the second image; record the ratio between the second difference and the first feature value of any region in the first image as the first ratio.

[0029] The negative correlation mapping result of the first ratio is used as a change index for any region in the first image.

[0030] Preferably, obtaining the segmentation parameters of any region based on the change index of any region and the difference in high-frequency information content within any region in all adjacent scale filtered images includes:

[0031] The difference in high-frequency information content within any region of the filtered images at adjacent scales, summed with a preset adjustment parameter, is denoted as the second feature value of any region of the filtered images at adjacent scales; wherein the preset adjustment parameter is greater than 0.

[0032] The normalized result of the sum of the changes in any region and the ratios between the second feature values ​​of any region in the filtered images of adjacent scales is determined as the segmentation parameter of any region.

[0033] Preferably, the step of adjusting the grayscale values ​​of pixels in the ultrasound image using segmentation parameters to obtain the target image includes:

[0034] For any pixel in any region of an ultrasound image: calculate the first product of the gray value of the pixel in the ultrasound image and the segmentation parameters of the region where the pixel is located; and use the sum of the reference gray value corresponding to the pixel and the first product as the gray feature value of the pixel.

[0035] The feature gray values ​​of all pixels in the ultrasound image are mapped to the interval [0, 255] to obtain the target image.

[0036] Preferably, the reference gray value corresponding to any pixel is the minimum gray value of all pixels in the region where the pixel is located in the ultrasound image.

[0037] The present invention has at least the following beneficial effects:

[0038] This invention first uses multiple high-pass filters of different scales to filter neck ultrasound images of patients with thyroid nodules, obtaining filtered images at each scale. Then, each scale's filtered image is analyzed separately. Based on the degree of change in high-frequency information content in different preset directions within each region of the filtered image, the target direction for each region is determined. The target direction is the direction with the largest change in high-frequency information content. Combining the proportion of high-frequency information content in the target direction across all regions, the high-frequency information change direction of the filtered image at each scale is determined, and the consistency between the high-frequency information change direction of a single region and the overall change direction is evaluated. Next, by integrating the consistency factor of direction change, the target direction, and the differences in high-frequency information content of the same region in filtered images of adjacent scales, segmentation parameters for each region in the ultrasound image are obtained. The grayscale values ​​of pixels in the original ultrasound image are then corrected to obtain the target image. The target image shows a greater difference between the thyroid nodule region and the normal region, solving the problem of ultrasound images containing a large amount of stray noise in existing technologies. This improves the accuracy of neck ultrasound image segmentation results for patients with thyroid nodules, allowing doctors to focus more on the target region during subsequent examinations. Attached Figure Description

[0039] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of an adaptive segmentation method for ultrasound images provided in an embodiment of the present invention. Detailed Implementation

[0041] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description of an adaptive segmentation method for ultrasound images based on the present invention is provided in conjunction with the accompanying drawings and preferred embodiments.

[0042] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0043] The following description, in conjunction with the accompanying drawings, details the specific scheme of the adaptive segmentation method for ultrasound images provided by this invention.

[0044] An embodiment of an adaptive segmentation method for ultrasound images:

[0045] The specific scenario addressed in this embodiment is as follows: During the capture of neck ultrasound images of patients with thyroid nodules, multiplicative speckle patterns are generated due to the superposition of diffusion function and medium scattering. The presence of these speckles reduces the contrast and clarity of the image, blurring the boundaries of the thyroid nodule region and interfering with the segmentation and identification results. When obtaining the target thyroid nodule tissue in the ultrasound image through threshold segmentation, it is usually difficult to directly segment the target region in the neck ultrasound image of the patient with thyroid nodules due to the interference of noise regions, thus affecting the doctor's judgment of the lesion tissue. This embodiment analyzes the feature differences between the target thyroid region and the multiplicative speckle interference, and adaptively adjusts the grayscale of pixels in different regions to improve the accuracy of the target region extraction results in the ultrasound image.

[0046] This embodiment proposes an adaptive segmentation method for ultrasound images, such as... Figure 1 As shown, an adaptive segmentation method for ultrasound images in this embodiment includes the following steps:

[0047] Step S1: Obtain a neck ultrasound image of a patient with thyroid nodules, and process the ultrasound image using multiple high-pass filters of different scales to obtain filtered images of each scale.

[0048] First, the neck region of patients with thyroid nodules is scanned using a B-mode ultrasound diagnostic instrument to obtain neck ultrasound images. Then, the acquired ultrasound images are converted to grayscale. High-pass filtering is a filtering method that allows high-frequency signals to pass through normally while blocking or weakening low-frequency signals. However, the degree of blocking or weakening varies depending on the frequency and the filtering program. Since the grayscale changes of pixels within the speckle area are relatively complex, and the pixels of thyroid nodules exhibit a certain regularity along the nodule's growth direction, a series of high-pass filters at different scales can be used to process the patient's thyroid ultrasound images. Based on the analysis of boundary information changes during multi-scale filtering, the boundary information of the thyroid region is enhanced, enabling accurate segmentation of thyroid nodules. Therefore, multiple high-pass filters of different scales are then used to filter the grayscale ultrasound images, obtaining filtered images at various scales. The scale and number of high-pass filters are set by the implementer according to specific circumstances. Image grayscale processing and high-pass filtering are existing technologies and will not be elaborated further here.

[0049] Thus, this embodiment has obtained filtered images at multiple scales.

[0050] Step S2: For each scale of the filtered image, the target direction of each region is selected based on the degree of change of high-frequency information content in different preset directions in each region of the filtered image; the high-frequency information change direction of the filtered image is determined by combining the proportion of high-frequency information content in the target direction of all regions; and the direction change consistency factor of each region is obtained based on the difference between the target direction and the high-frequency information change direction of a single region.

[0051] Because the grayscale changes of pixels at the edge of thyroid nodules are within a certain range and the degree of change is relatively regular, while multiplicative speckle interference is caused by interference from ultrasound imaging attributes, resulting in disordered and random grayscale values ​​of pixels in non-target tissue areas of the image. Therefore, in the image processed by a series of multi-scale high-pass filters, the changes in high-frequency information within multiplicative speckle interference are significant and continuous, and the changes are similar, while the changes in high-frequency information of thyroid nodules decrease sharply with scale changes. Therefore, this step divides the multi-scale filtered image into blocks and analyzes several block regions by calculating the consistency of pixel change patterns in different blocks to obtain the probability that different block regions belong to multiplicative speckle interference, thereby achieving adaptive segmentation of the thyroid nodule region.

[0052] Thyroid nodules are usually a relatively homogeneous connective tissue structure. In ultrasound images, they show uniform information changes in different growth directions. However, speckle interference, due to the complex and varied internal information changes, shows chaotic information changes in different growth directions. Therefore, by analyzing the degree of change of high-frequency information in different directions within a rectangular area, the direction with the greatest change in high-frequency information content is selected as the main direction for the current region analysis.

[0053] First, the ultrasound image is divided into several regions of equal size. In this embodiment, the number of regions in the ultrasound image is 16. In specific applications, the implementer can set this number according to the specific situation. Since the filtered image is obtained by filtering the ultrasound image, there is a one-to-one correspondence between the filtered image and the pixels in the ultrasound image. That is, it is equivalent to dividing each filtered image into multiple regions, and the regions in the filtered image and the ultrasound image have a one-to-one correspondence.

[0054] Next, this embodiment will use a filtered image as an example for illustration. Other filtered images can be processed using the method provided in this embodiment.

[0055] Specifically, the filtered image at any scale is denoted as the image to be analyzed.

[0056] Multiple analysis directions are set for the center pixel of each region of the image to be analyzed. Specifically, a rectangular region is defined with the horizontal direction as the starting direction and an angle of 10 degrees as the interval. There are multiple analysis directions, each of which is a preset direction. In other words, multiple preset directions are obtained. Filters (such as Gabor filters) are used to extract the high-frequency information content of pixels in different analysis directions in each region.

[0057] Any region in the image to be analyzed is designated as a candidate region. First, the standard deviation and average value of the high-frequency information content in each preset direction within the candidate region are calculated. Then, based on the standard deviation and average value of the high-frequency information content in each preset direction within the candidate region, the degree of change of the high-frequency information content in each preset direction of the candidate region is obtained. Both the standard deviation and the average value are positively correlated with the degree of change.

[0058] Among them, a positive correlation means that the dependent variable increases as the independent variable increases, and the dependent variable decreases as the independent variable decreases. It can be an additive relationship, a multiplicative relationship, etc., which is determined by the actual application.

[0059] As a specific example, the product of the standard deviation and the average high-frequency information content in each preset direction within the candidate region of the image to be analyzed is calculated. The normalized result of this product is taken as the degree of change of high-frequency information content in each preset direction within the candidate region of the image to be analyzed. It should be noted that there is a degree of change of high-frequency information content in each preset direction within the candidate region. In this embodiment, the maximum-minimum normalization method is used to normalize the product of the standard deviation and the average value. The maximum-minimum normalization method is existing technology and will not be elaborated on here. As other implementation methods, other existing data normalization methods can also be used for processing, which will not be elaborated on here either.

[0060] Furthermore, the direction corresponding to the maximum value of the change in high-frequency information content in all preset directions of the candidate region is taken as the target direction of the candidate region.

[0061] Using the above method, each region in the image to be analyzed is processed separately, and the target direction of each region in the image to be analyzed can be obtained. The target direction is the preset direction in which the high-frequency information content changes the most in the corresponding region.

[0062] For ultrasound images, the same set of instrument probes and imaging principles are used during the scanning process. Therefore, the directions of the maximum changes in high-frequency information in multiple multiplicative speckle interference regions are relatively consistent. However, the nodule region accounts for a relatively small proportion and shows a large directional difference in different regions. Therefore, this step obtains the consistency of image directional changes by weighted averaging, and then calculates the degree of difference between the directional changes of different sub-regions and the overall directional changes, as a screening factor for the consistency of directional changes between speckle interference and nodule regions.

[0063] Specifically, by using the proportion of high-frequency information content in the target direction of each region in the image to be analyzed, the target direction is weighted and averaged to obtain the direction of high-frequency information change in the image to be analyzed; wherein, the proportion of high-frequency information content in the target direction of each region is the ratio between the high-frequency information content in the target direction of each region and the total content of high-frequency information in the image to be analyzed.

[0064] In this embodiment, a specific method for calculating the direction of high-frequency information change in the image to be analyzed is given. The direction of high-frequency information change in the image to be analyzed can be expressed as:

[0065]

[0066] in, This indicates the direction of high-frequency information changes in the image to be analyzed. This indicates the number of regions within the image to be analyzed. This represents the proportion of high-frequency information content in the target direction of the i-th region in the image to be analyzed. This indicates the target direction of the i-th region in the image to be analyzed. The symbol indicates rounding up.

[0067] It should be noted that in this embodiment, the direction in the calculation formula for the direction of high-frequency information change is represented by the included angle. That is, for any direction, the included angle between the direction and the preset direction is substituted into the above calculation formula for the direction of high-frequency information change. In this embodiment, the preset direction is the horizontal direction to the right.

[0068] After obtaining the high-frequency information change direction of the image to be analyzed, the candidate region will still be used as an example for explanation. Other regions in the image to be analyzed can be processed using the method provided in this embodiment. The consistency factor between the direction change of speckle interference and the thyroid nodule region is determined based on the consistency between the target direction of the candidate region and the high-frequency information change direction of the image to be analyzed. Specifically, the difference between the target direction of the candidate region and the high-frequency information change direction of the image to be analyzed is calculated and recorded as the first difference. The smaller the first difference, the higher the consistency between the target direction of the candidate region and the high-frequency information change direction of the image to be analyzed. Then, the consistency factor of the direction change of the candidate region is obtained based on the first difference. The first difference and the consistency factor of the direction change are negatively correlated.

[0069] Among them, a negative correlation means that the dependent variable decreases as the independent variable increases, and the dependent variable increases as the independent variable decreases. It can be a subtraction relationship, a division relationship, etc., which is determined by the actual application.

[0070] In this embodiment, a specific formula for calculating the direction change consistency factor is given. The direction change consistency factor of the i-th region in the image to be analyzed can be expressed as:

[0071]

[0072] in, This represents the direction change consistency factor of the i-th region in the image to be analyzed. This indicates the target direction of the i-th region in the image to be analyzed. This indicates the direction of high-frequency information changes in the image to be analyzed. This indicates the preset first adjustment parameter. Indicates the absolute value sign.

[0073] In this embodiment, a preset first adjustment parameter is introduced into the calculation formula of the direction change consistency factor to prevent the denominator from being 0. In this embodiment, the preset first adjustment parameter is 0.01. In specific applications, the implementer can set it according to the specific situation. This represents the first difference corresponding to the i-th region. The larger the value, the lower the consistency between the two directions. The more obvious the difference between the information change direction in the i-th region and the overall change direction of the ultrasound image, the more independent the high-frequency information change direction in the i-th region is. In this case, the i-th region is more likely to belong to a thyroid nodule, and the smaller the directional change consistency factor of the i-th region is.

[0074] Using the above method, it is possible to obtain the direction change consistency factor of each region in the filtered image at each scale.

[0075] Step S3: Combine the differences in the direction change consistency factor, target direction, and high-frequency information content of the same region in the filtered images of adjacent scales to obtain the segmentation parameters of each region.

[0076] As the patient's neck ultrasound images underwent a series of high-pass filtering at different scales, the content of high-frequency information in the images gradually changed. At this point, due to the relatively homogeneous internal structure of the thyroid nodule, the information at a certain filter boundary position appeared abruptly changed during multi-scale processing; while the multiplicative speckle interference region, due to its complex texture information, showed relatively uniform information changes under multiple scales of filtering. Therefore, further analysis was conducted by considering the fluctuations in the directional change factor within the same region during a series of high-pass filtering processes.

[0077] For any region in an ultrasound image:

[0078] The product of the direction change consistency factor of the region and the target direction of the region in the filtered image at each scale is recorded as the first feature value of the region in the filtered image at each scale. It should be noted that the target direction involved in the calculation of the first feature value is also the angle between the target direction and the preset direction. There is a corresponding first feature value for the region in the filtered image at each scale.

[0079] The filtered images at all scales are sorted in ascending order of filtering scale to obtain a filtered image sequence. It should be noted that the filtered images at adjacent scales mentioned later refer to two adjacent filtered images in the filtered image sequence. The filtered image at the previous scale in the adjacent scale sequence is designated as the first image, and the filtered image at the next scale in the adjacent scale sequence is designated as the second image. The difference between the first feature value of a region in the first image and the first feature value of the region in the second image is calculated and designated as the second difference. The ratio between the second difference and the first feature value of the region in the first image is designated as the first ratio. The negative correlation mapping result of the first ratio is used as the change index of the region in the first image.

[0080] In this embodiment, a specific formula for calculating the change index is given. The change index of the j-th region in the filtered image at the x-th scale can be expressed as:

[0081]

[0082] in, This represents the change index of the j-th region in the filtered image at the x-th scale. Let represent the orientation change consistency factor of the j-th region in the filtered image at scale x+1. This represents the target orientation of the j-th region in the filtered image at scale x+1. Let $\frac{ ... This represents the target orientation of the j-th region in the filtered image at the x-th scale. Indicates the absolute value sign. This represents an exponential function with the natural constant as its base.

[0083] This represents the first feature value of the j-th region in the filtered image at the (x+1)-th scale; This represents the first feature value of the j-th region in the filtered image at the x-th scale; Indicates the second difference; The first ratio represents the degree of change in the direction of the greatest change in high-frequency information content at adjacent scales. The smaller the degree of change, the smaller the degree of change in the region after filtering at different scales, and the less likely the region is to contain thyroid nodule information. It should be noted that if the denominator is 0, 0.01 is added to the denominator to prevent the formula from being meaningless.

[0084] Using the above method, the change index of the region in the filtered image at each scale can be obtained. Since the high-frequency information content of the thyroid nodule region decreases as it changes along the direction of maximum high-frequency information change, the region segmentation parameters are constructed by combining the fluctuation pattern of high-frequency information content changes in scale filtering.

[0085] Specifically, the difference in high-frequency information content within the region in the filtered images of adjacent scales, summed with a preset adjustment parameter, is denoted as the second feature value of the region in the filtered images of adjacent scales; where the preset adjustment parameter is greater than 0. The normalized result of the sum of the ratios between the change index of the region and the second feature value of the region in the filtered images of adjacent scales is determined as the segmentation parameter of the region.

[0086] In this embodiment, a specific formula for calculating the segmentation parameters is given. The segmentation parameters of the j-th region in the ultrasound image can be expressed as:

[0087]

[0088] in, This represents the segmentation parameters of the j-th region in the ultrasound image. Indicates the number of regions in an ultrasound image. This represents the high-frequency information content of the j-th region in the filtered image at the x-th scale. This represents the high-frequency information content of the j-th region in the filtered image at the (x+1)-th scale. This represents the change index of the j-th region in the filtered image at the x-th scale.

[0089] The preset adjustment parameter is introduced into the calculation formula of the segmentation parameter to prevent the denominator from being 0. In this embodiment, the preset adjustment parameter is 1. In specific applications, the implementer can set it according to the specific situation. This represents the second feature value of the j-th region in the filtered images at scales x and (x+1)-th. A normalization function is used to... The calculation results are mapped to [-1, 1], that is, the range of the segmentation parameter is [-1, 1].

[0090] In the same region, the larger the value of the change index obtained after high-pass filtering at multiple scales and the smaller the change in high-frequency information content in adjacent directions, the greater the possibility that the region contains thyroid nodule information. Therefore, when performing adaptive segmentation on the pixels in the region, it is more necessary to mark it as a thyroid nodule region.

[0091] Using the above method, segmentation parameters for each region in a neck ultrasound image can be obtained.

[0092] Step S4: Adjust the grayscale values ​​of pixels in the ultrasound image using segmentation parameters to obtain the target image, and perform adaptive segmentation on the target image.

[0093] The segmentation parameters of each region in the ultrasound image were obtained using the above method. The larger the segmentation parameter, the greater the probability that the pixel in the region belongs to thyroid nodule tissue; the smaller the segmentation parameter, the greater the probability that the pixel in the region belongs to multiplicative speckle interference. Therefore, in the process of segmenting the ultrasound image, the grayscale information of the pixels in the ultrasound image is weighted by combining the segmentation parameters to obtain complete and clear thyroid nodule tissue.

[0094] Specifically, for any pixel in any region of an ultrasound image: calculate the product of the pixel's grayscale value in the ultrasound image and the segmentation parameters of the region where the pixel is located, and denote this product as the first product; use the sum of the reference grayscale value corresponding to the pixel and the first product as the grayscale feature value of the pixel; wherein, the reference grayscale value corresponding to the pixel is the minimum grayscale value of all pixels in the region where the pixel is located in the ultrasound image. Using this method, the grayscale feature value of each pixel in the ultrasound image can be obtained.

[0095] Furthermore, the feature gray values ​​of all pixels in the ultrasound image are mapped to the interval [0, 255] to obtain new gray values, which are then used as the gray values ​​of the corresponding pixels in the target image. It should be noted that the method of mapping a set of numbers to a set interval is existing technology and will not be elaborated upon further in this embodiment.

[0096] After obtaining the target image, the Otsu thresholding method can be used to obtain a complete and clear thyroid nodule tissue, or a semantic segmentation network can be used to extract the complete and clear thyroid nodule tissue. These methods are existing methods and will not be described in detail in this embodiment. Subsequently, doctors can use the extracted thyroid nodule tissue region to assess the patient's condition.

[0097] Thus, the adaptive segmentation of ultrasound images of a patient's thyroid nodules was completed using the method provided in this embodiment.

[0098] This embodiment first uses multiple high-pass filters of different scales to filter the neck ultrasound images of patients with thyroid nodules, obtaining filtered images at each scale. Then, each scale of filtered image is analyzed separately. Based on the degree of change in high-frequency information content in different preset directions within each region of the filtered image, the target direction of each region is determined. The target direction is the direction with the largest change in high-frequency information content. Combining the proportion of high-frequency information content in the target direction of all regions, the high-frequency information change direction of the filtered image at each scale is determined, and the consistency between the high-frequency information change direction of a single region and the overall change direction is evaluated. Then, by combining the consistency factor of direction change, target direction, and high-frequency information content of the same region in filtered images of adjacent scales, the segmentation parameters of each region in the ultrasound image are obtained. Subsequently, the gray values ​​of pixels in the original ultrasound image are corrected to obtain the target image. The difference between the thyroid nodule region and the normal region in the target image is greater, which solves the problem of ultrasound images containing a large amount of stray noise information in the prior art, improves the accuracy of the segmentation results of neck ultrasound images of patients with thyroid nodules, and allows doctors to give higher attention to the target region during subsequent further examinations.

[0099] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive segmentation method for ultrasound images, characterized in that, The method includes the following steps: Neck ultrasound images of patients with thyroid nodules were acquired, and the ultrasound images were processed using multiple high-pass filters of different scales to obtain filtered images of each scale. For each scale of the filtered image, the target direction of each region is selected based on the degree of change of high-frequency information content in different preset directions within each region of the filtered image; the high-frequency information change direction of the filtered image is determined by combining the proportion of high-frequency information content in the target direction of all regions; and the direction change consistency factor of each region is obtained based on the difference between the target direction and the high-frequency information change direction of a single region. By combining the differences in the direction change consistency factor, target direction, and high-frequency information content of the same region in filtered images of adjacent scales, the segmentation parameters of each region are obtained; The target image is obtained by adjusting the gray values ​​of pixels in the ultrasound image using segmentation parameters, and then adaptive segmentation is performed on the target image. The acquisition of the degree of variation in high-frequency information content in different preset directions within each region includes: Based on the standard deviation and average value of the high-frequency information content in each preset direction within the candidate region of the image to be analyzed, the degree of change of the high-frequency information content in each preset direction of the candidate region is obtained, and both the standard deviation and the average value are positively correlated with the degree of change. The image to be analyzed is a filtered image of any scale, and the candidate region is any region in the image to be analyzed; Determine the direction of high-frequency information changes in the filtered image, including: By utilizing the proportion of high-frequency information content in the target direction of each region in the image to be analyzed, a weighted average is calculated on the target direction to obtain the direction of high-frequency information change in the image to be analyzed. The proportion of high-frequency information content in the target direction of each region is the ratio between the high-frequency information content in the target direction of each region and the total high-frequency information content in the image to be analyzed. The direction of high-frequency information change in the image to be analyzed is represented as follows: in, This indicates the direction of high-frequency information changes in the image to be analyzed. This indicates the number of regions within the image to be analyzed. This represents the proportion of high-frequency information content in the target direction of the i-th region in the image to be analyzed. This indicates the target direction of the i-th region in the image to be analyzed. Indicates the rounding up symbol; Obtain the consistency factor of directional change in each region, including: Calculate the first difference between the target orientation of the candidate region and the high-frequency information change direction of the image to be analyzed; Based on the first difference, a direction change consistency factor of the candidate region is obtained, and the first difference is negatively correlated with the direction change consistency factor. The process of obtaining segmentation parameters for each region includes: For any region in an ultrasound image: Based on the change consistency factor of the direction change and the change of the target direction in any region of the filtered images of adjacent scales, the change index of any region is obtained. Based on the change index of any region and the difference in high-frequency information content within any region in all adjacent scale filtered images, the segmentation parameters of any region are obtained. The step of obtaining the change index of any region based on the change consistency factor of the direction change and the change of the target direction in the filtered images of adjacent scales includes: The product of the direction change consistency factor of any region in the filtered image at each scale and the target direction is denoted as the first feature value of any region in the filtered image at each scale. The filtered image at the previous scale among the filtered images at adjacent scales is designated as the first image, and the filtered image at the next scale among the filtered images at adjacent scales is designated as the second image. Calculate the second difference between the first feature value of any region in the first image and any region in the second image; record the ratio between the second difference and the first feature value of any region in the first image as the first ratio. The negative correlation mapping result of the first ratio is used as a change index for any region in the first image.

2. The adaptive segmentation method for ultrasound images according to claim 1, characterized in that, Filter the target directions for each region, including: The direction corresponding to the maximum value of the change in high-frequency information content in all preset directions of each region is taken as the target direction of each region.

3. The adaptive segmentation method for ultrasound images according to claim 1, characterized in that, The step of obtaining segmentation parameters for any region based on the change index of any region and the difference in high-frequency information content within that region in all adjacent scale filtered images includes: The difference in high-frequency information content within any region of the filtered images at adjacent scales, summed with a preset adjustment parameter, is denoted as the second feature value of any region of the filtered images at adjacent scales; wherein the preset adjustment parameter is greater than 0. The normalized result of the sum of the ratios between the change index of any region and the second feature value of any region in the filtered image of adjacent scales is determined as the segmentation parameter of any region.

4. The adaptive segmentation method for ultrasound images according to claim 1, characterized in that, The method of adjusting the grayscale values ​​of pixels in an ultrasound image using segmentation parameters to obtain the target image includes: For any pixel in any region of an ultrasound image: calculate the first product of the gray value of the pixel in the ultrasound image and the segmentation parameters of the region where the pixel is located; and use the sum of the reference gray value corresponding to the pixel and the first product as the gray feature value of the pixel. The feature gray values ​​of all pixels in the ultrasound image are mapped to the interval [0, 255] to obtain the target image.

5. The adaptive segmentation method for ultrasound images according to claim 4, characterized in that, The reference gray value corresponding to any pixel is the minimum gray value of all pixels in the region where the pixel is located in the ultrasound image.

Images