A gynecological intrauterine fetus ultrasound image collaborative analysis method and system

Abstract

The application relates to the technical field of image enhancement, in particular to a gynecological and obstetrical intrauterine fetus ultrasonic image collaborative analysis method and system, which comprises the following steps: acquiring a fetus ultrasonic image and a reference ultrasonic image; determining the noise distribution range of the reference ultrasonic image according to the gray value of each pixel point in the reference ultrasonic image; determining the noise intensity of each pixel point in the fetus ultrasonic image according to the gradient value of each pixel point in the fetus ultrasonic image and the reference ultrasonic image, and combining the noise distribution range; determining the noise coupling index of each pixel point in the fetus ultrasonic image according to the gradient direction and the edges of each pixel point in the fetus ultrasonic image; and performing noise reduction processing on the fetus ultrasonic image according to the noise intensity and the noise coupling index, so as to obtain a fetus ultrasonic image after noise reduction, thereby assisting doctors in collaborative analysis. The application can ensure the noise reduction effect, avoid the loss of the cardiac texture information in the fetus ultrasonic image, and is beneficial to collaborative analysis.

Description

Technical Field

[0001] This invention relates to the field of image enhancement technology, specifically to a method and system for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology. Background Technology

[0002] Ultrasound imaging technology is a non-invasive examination method that uses high-frequency sound waves to generate real-time images of internal structures. It is widely used in obstetrics and gynecology, especially in fetal development monitoring, facilitating real-time observation of the development of various fetal organs. The four-chamber view is one of the most common views in fetal echocardiography, typically used to assess the left and right atria, ventricles, heart valves, and the symmetry of cardiac structures. Since congenital heart disease is one of the most common congenital diseases, early screening is crucial for fetal health. Through four-chamber view analysis, medical personnel can initially screen for heart malformations or other abnormalities, allowing for timely further diagnosis.

[0003] In the process of analyzing intrauterine fetal ultrasound images in obstetrics and gynecology, the fetal ultrasound images acquired by existing technologies are affected by noise. During the noise reduction process, the anatomical structure of the fetus is highly variable, such as the microstructure of the heart valves. Furthermore, there is a strong coupling between noise and the texture of the heart structure. Therefore, directly performing noise reduction processing on fetal ultrasound images will affect the accuracy of the heart texture information in the ultrasound images and the noise reduction effect, which may lead to misjudgment. Summary of the Invention

[0004] To address the technical problem that existing noise reduction methods cannot simultaneously ensure noise reduction effectiveness and avoid the loss of cardiac texture information in fetal ultrasound images, the present invention aims to provide a method and system for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology. The specific technical solution adopted is as follows: One embodiment of the present invention provides a method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology, the method comprising the following steps: Acquire a fetal ultrasound image and a reference ultrasound image, wherein the reference ultrasound image is the fetal ultrasound image in overgain mode; The noise distribution characteristics of the reference ultrasound image are analyzed based on the gray value of each pixel in the reference ultrasound image to determine the noise distribution breadth of the reference ultrasound image. The noise distribution difference between the fetal ultrasound image and the reference ultrasound image is analyzed based on the gradient value of each pixel in the fetal ultrasound image and the reference ultrasound image. Combined with the noise distribution breadth, the noise intensity of each pixel in the fetal ultrasound image is determined. Based on the gradient direction of each pixel in the fetal ultrasound image and each edge in the fetal ultrasound image, the noise coupling index of each pixel in the fetal ultrasound image is determined. Based on the noise intensity of each pixel in the fetal ultrasound image and the noise coupling index, the fetal ultrasound image is denoised to obtain a denoised fetal ultrasound image, which can assist doctors in collaborative analysis.

[0005] Further, determining the noise distribution breadth of the reference ultrasound image includes: Based on the gray value of each pixel in the reference ultrasound image, determine the highest gray value peak and the target gray value peak on the fitted curve of the gray value histogram; Based on the highest grayscale peak value and the target grayscale peak value, the grayscale peak right shift rate of the fitted curve is determined as the noise distribution breadth of the reference ultrasound image.

[0006] Further, determining the target grayscale peak value includes: Determine the maximum value point on the fitted curve of the grayscale histogram of the reference ultrasound image; The fitted curve is divided into two curve segments using the maximum value point, and the maximum value of the curve segment located to the right of the maximum value point is taken as the target grayscale peak value.

[0007] Further, determining the right shift rate of the grayscale peak of the fitted curve based on the highest grayscale peak and the target grayscale peak includes: The grayscale peak shift amount is determined based on the grayscale difference between the highest grayscale peak and the target grayscale peak. Determine the minimum point located between the highest grayscale peak and the target grayscale peak, use the minimum point to divide the fitted curve into two curve segments, calculate the definite integral for each of the two curve segments, and obtain the absolute difference between the two definite integrals; The grayscale peak shift and the absolute difference are fused and analyzed to obtain the grayscale peak right shift rate of the fitted curve; the grayscale peak right shift rate is positively correlated with the grayscale peak shift and negatively correlated with the absolute difference.

[0008] Further, determining the noise intensity of each pixel in the fetal ultrasound image includes: Based on the gradient value of each pixel in the fetal ultrasound image and the reference ultrasound image, determine the distance variance of each pixel in the fetal ultrasound image and the reference ultrasound image; Based on the difference between the distance variance corresponding to each pixel in the fetal ultrasound image and the reference ultrasound image, and in combination with the noise distribution breadth, the noise intensity of each pixel in the fetal ultrasound image is determined. The difference between the noise intensity and the distance variance is negatively correlated, while it is positively correlated with the noise distribution breadth.

[0009] Further, determining the distance variance corresponding to each pixel point in the fetal ultrasound image and the reference ultrasound image includes: The fetal ultrasound image and the reference ultrasound image are referred to as the ultrasound image to be analyzed, and any pixel in the ultrasound image to be analyzed is taken as the target pixel. Determine the gradient value of the target pixel in the ultrasound image to be analyzed. If the gradient value of the target pixel is the largest in its eight neighborhood, then the target pixel is taken as the maximum value point, and then all the maximum value points in the ultrasound image to be analyzed are obtained. Based on the distance between every two maximum points within the preset neighborhood of each pixel, the distance variance within the preset neighborhood of each pixel is calculated and used as the distance variance of the corresponding pixel in the ultrasound image to be analyzed. The preset neighborhood range is a circular region centered on the target pixel and with the absolute straight-line distance between the target pixel and its nearest k-th maximum point as the radius, where k is a preset constant.

[0010] Further, determining the noise coupling index for each pixel in the fetal ultrasound image includes: For each pixel in the fetal ultrasound image, the pixel with the largest gray level difference among the pixels within the eight neighborhoods of the pixel is taken as the directional pixel, and the unit vector from the pixel to the directional pixel is taken as the gradient direction. The fetal ultrasound image is divided into local regions based on the gradient direction of each pixel in the fetal ultrasound image. The noise coupling index of each pixel's local region is determined based on the gradient direction of each pixel in each local region and the number of pixels located on the edge.

[0011] Furthermore, the step of dividing the fetal ultrasound image into various local regions based on the gradient direction of each pixel in the fetal ultrasound image includes: Calculate the gradient direction variance in the eight-neighbor area of ​​each pixel; The gradient direction variance is used to perform superpixel segmentation on the fetal ultrasound image to obtain several pixel clusters; By clustering the pixels into local regions, we can obtain the various local regions corresponding to the fetal ultrasound image.

[0012] Further, determining the noise coupling index of the local region to which each pixel belongs based on the gradient direction of each pixel in each local region and the number of pixels located on the edge includes: Calculate the average value of the gradient direction variance of all pixels in each local region, and use it as the first noise coupling factor of the corresponding local region; The number of pixels located on the edge in each local region is used as the second noise coupling factor for the corresponding local region. The first noise coupling factor and the second noise coupling factor of each local region are fused and analyzed to determine the noise coupling index of the local region to which each pixel belongs. The noise coupling index is positively correlated with both the first noise coupling factor and the second noise coupling factor.

[0013] Another embodiment of the present invention provides a collaborative analysis system for intrauterine fetal ultrasound images in obstetrics and gynecology, including a processor and a memory. The processor is used to process instructions stored in the memory to implement a collaborative analysis method for intrauterine fetal ultrasound images in obstetrics and gynecology.

[0014] The present invention has the following beneficial effects: This invention provides a method and system for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology. It acquires fetal ultrasound images in over-gain and normal modes. By analyzing the image features of the fetal ultrasound images in over-gain mode, the noise distribution breadth is obtained. Then, by combining the noise distribution differences between the fetal ultrasound images and reference ultrasound images, the noise intensity of each pixel in the fetal ultrasound image is determined. By determining the noise intensity for each pixel, the influence of the variable anatomical structure of the fetus during the noise reduction process can be effectively overcome, improving the accuracy and robustness of fetal ultrasound image noise reduction processing. By analyzing the gradient direction of each pixel and the edges in the fetal ultrasound images, the cardiac texture in the fetal ultrasound images is analyzed, and the noise coupling index of each pixel is determined. By determining the noise coupling index for each pixel, the loss of cardiac texture information caused by strong coupling between noise and cardiac structure texture can be effectively avoided, which is beneficial to improving the reliability of fetal ultrasound image noise reduction processing. By fusing the noise intensity and noise coupling index of each pixel in the fetal ultrasound images for noise reduction processing, the noise suppression effect of the fetal ultrasound images can be improved while preserving cardiac texture details, thereby improving the accuracy of the collaborative analysis results obtained based on fetal ultrasound images. Attached Figure Description

[0015] 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.

[0016] Figure 1 A flowchart of a method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology, provided as an embodiment of the present invention; Figure 2 This is a flowchart illustrating the implementation of step S2 in an embodiment of the present invention; Figure 3 This is a flowchart illustrating the implementation of step S3 in an embodiment of the present invention; Figure 4 This is a flowchart illustrating the implementation of step S4 in an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the specific implementation methods, structures, features, and effects of the technical solution proposed according to the present invention are described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] 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.

[0019] It should be noted that, in the embodiments of the present invention, in order to facilitate subsequent data processing and avoid differences in units and numerical magnitudes between data, the data is standardized to eliminate the influence of dimensions in data calculations.

[0020] The application scenario targeted by this invention can be as follows: When performing collaborative analysis of fetal ultrasound images, in order to improve the image quality of fetal ultrasound images, it is necessary to perform noise reduction processing on fetal ultrasound images. However, due to the variable anatomical structure of the fetus and the strong coupling between noise and cardiac structure texture, direct noise reduction will affect the accurate acquisition of fetal cardiac texture information and the noise reduction effect, which may cause misjudgment when performing four-chamber cardiac section analysis.

[0021] One embodiment of the present invention provides a method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology, such as... Figure 1 As shown, it includes the following steps: S1, Acquire fetal ultrasound images and reference ultrasound images.

[0022] Here, the reference ultrasound image is a fetal ultrasound image in overgain mode.

[0023] For example, acquiring fetal ultrasound images during collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology includes: Pregnant women should adopt a suitable position, such as a lateral or semi-recumbent position, to reduce the obstruction of sound waves by the fat layer and improve penetration efficiency. The operator uses a transabdominal ultrasound probe suitable for fetal imaging at different gestational weeks and acquires fetal ultrasound video under optimized instrument parameters. Each frame in the fetal ultrasound video can be regarded as an independent image data source, and the denoising process is the same for each frame. To facilitate understanding of the scheme, this embodiment takes any frame as an example for analysis, as the fetal ultrasound image for denoising.

[0024] In the collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology, the planar localization of the four-chamber view of the heart is the core link of fetal heart screening. Therefore, in order to accurately extract fetal heart image features, fetal ultrasound images that have completed fetal heart identification can be used for subsequent image denoising analysis.

[0025] In this embodiment, the heart portion of the fetus in utero is the most dynamic part in the image, and the heartbeat is periodic. Based on this, the image features of the fetal heart in the fetal ultrasound image are analyzed.

[0026] For example, acquiring fetal ultrasound images during gynecological intrauterine fetal ultrasound image co-analysis includes: The first step is to acquire several consecutively distributed fetal ultrasound images from the fetal ultrasound video, such as acquiring 10 frames of fetal ultrasound images; corner point detection is performed on each frame of fetal ultrasound images, and the obtained corner points are used as location points. The second step is to use optical flow detection to obtain optical flow field data in continuous frame images and track the positional change of each point in the image, that is, the change in image distance at the same point per unit time, which can reflect the movement of the fetus in the uterus at different times in the image. The third step is to obtain the change fitting curve for each location point. The horizontal axis of the change fitting curve is the acquisition time of the ultrasound image at the location point, and the vertical axis is the change in position of the location point at each acquisition time. The fourth step is to use the autocorrelation function to obtain the autocorrelation function value of the curve fitted for the change at each location point, and record it as the periodicity of the movement of the location point. The larger the autocorrelation function value, the stronger the periodicity of the curve. The fifth step is to use the Otsu threshold to segment the periodicity of all location points and filter out the location points that are greater than or equal to the threshold as the target location points. The sixth step is to obtain the delay value in the autocorrelation function of the fitting curve of the change of each target location point, and use the delay value to perform cluster analysis on the target location points to obtain the clusters of each location point; Step 7: Calculate the intercept of the fitted curve of the change of each target location point in each location point cluster, and take the inverse of the variance of all intercepts in the same location point cluster to obtain the heart coefficient of each location point cluster. In this embodiment, discrete location change points are fitted to establish a continuous function of location change over time. Exemplarily, a first-order linear regression model is used for fitting, the model parameters are solved using the least squares method, and the resulting constant term is used as the intercept of the fitted curve. In other implementations, if the time series exhibits nonlinear drift, a second-order polynomial model can also be used for fitting, with the constant term as the intercept; however, it is necessary to ensure that the fitting order is much smaller than the number of sampling points to prevent overfitting.

[0027] If an extreme case exists where the intercept variance is 0, the heart coefficient of each location point cluster can be obtained by adding a preset non-zero anti-surge constant to the variance of all intercepts in the same location point cluster and then performing inverse processing. The non-zero anti-surge constant can be empirically set to 0.01.

[0028] Step 8: Take each target location point in the cluster corresponding to the maximum heart coefficient as the heart location point; then determine the fetal heart region in each frame of fetal ultrasound image based on each heart location point, and take any frame of fetal ultrasound image with a known fetal heart region as the fetal ultrasound image for noise analysis.

[0029] It is worth noting that the fetal ultrasound images used for image noise analysis in this embodiment can be the original fetal ultrasound images, or fetal ultrasound images with fetal heart identification completed. If only pathological analysis of the fetal heart is performed, the extracted fetal heart region images can also be used. Of course, since this embodiment combines the image characteristics of the heart for noise intensity analysis, it is also applicable to noise reduction analysis of cardiac-related images of other types of patients.

[0030] The process of acquiring fetal ultrasound images in over-gain mode is existing technology and will not be described in detail here.

[0031] It should be noted that after digital gain amplification of the acquired fetal ultrasound images, a reference ultrasound image in simulated overgain mode is obtained, i.e., the fetal ultrasound image in overgain mode. In the fetal ultrasound images acquired in normal mode, there is a mixture of snow noise and real texture, while the snow noise in the fetal ultrasound images in overgain mode is enhanced, which helps to highlight the noise characteristics.

[0032] It should be further explained that pairing and analyzing fetal ultrasound images under normal conditions and in gain mode helps to distinguish noise from the true structure of the heart.

[0033] Thus, the basic data for image noise analysis were obtained, namely the fetal ultrasound image and the reference ultrasound image.

[0034] S2, based on the grayscale value of each pixel in the reference ultrasound image, analyze the grayscale distribution characteristics of the noise pixels to determine the noise distribution breadth of the reference ultrasound image.

[0035] First, it should be noted that fetal ultrasound images in overgain mode are pure noise. By analyzing reference ultrasound images, the characteristics of the noise can be determined. Subsequently, the noise feature points can be compared and matched with the fetal ultrasound images to obtain the noise distribution in the fetal ultrasound images.

[0036] In this embodiment, under normal circumstances, the grayscale histogram of a fetal ultrasound image contains concentrated grayscale peaks. However, due to the over-gain mode, snowflake noise is enhanced, causing the concentrated grayscale peaks in the grayscale histogram to be widened and tail to the right. Therefore, by analyzing the rightward shift of the grayscale peaks in the grayscale histogram of the reference ultrasound image, the extent of noise distribution in the reference ultrasound image is determined.

[0037] As an exemplary implementation, step S2 described above can be achieved through... Figure 2 Steps S21 to S22 shown are implemented as follows: S21, Based on the gray value of each pixel in the reference ultrasound image, determine the highest gray value peak and the target gray value peak on the fitted curve of the gray value histogram.

[0038] Here, the target grayscale peak is the grayscale peak located to the right of the highest grayscale peak.

[0039] For example, a grayscale histogram is determined based on the grayscale value of each pixel in the reference ultrasound image. The horizontal axis of the grayscale histogram represents each grayscale level, and the vertical axis represents the frequency of occurrence of each grayscale level in the reference ultrasound image. A fitting curve of the grayscale histogram is obtained, and the maximum value point on the fitting curve is determined. The maximum value point is the highest grayscale peak of the fitting curve. The fitting curve is divided into two curve segments using the maximum value point. Since the maximum value in the fitting curve is the highest grayscale peak in the histogram, the maximum value in the curve segment located to the right of the maximum value point is taken as the right-shifted peak of the grayscale histogram, i.e., the target grayscale peak value.

[0040] S22, based on the highest gray value peak and the target gray value peak, determine the gray value peak right shift rate of the fitted curve, which serves as the reference noise distribution breadth of the ultrasound image.

[0041] Here, the noise distribution breadth reflects the overall distribution breadth of snowflake noise in the reference ultrasound image.

[0042] For example, determining the noise distribution breadth of a reference ultrasound image includes: The first step is to determine the grayscale peak shift based on the grayscale difference between the highest grayscale peak and the target grayscale peak.

[0043] Specifically, the difference between the target grayscale peak and the highest grayscale peak on the horizontal axis is obtained as the grayscale peak shift of the grayscale histogram of the reference ultrasound image.

[0044] The second step is to determine the minimum point between the highest grayscale peak and the target grayscale peak, and use the minimum point to divide the fitted curve into two curve segments. Calculate the definite integral for each of the two curve segments and obtain the absolute difference between the two definite integrals.

[0045] By analyzing the difference in peak volume between the highest grayscale peak and the target grayscale peak, the degree of grayscale peak shift can be determined.

[0046] Specifically, first determine the definite integrals of the two curve segments, then calculate the definite integrals separately, and obtain the absolute difference between the two definite integrals.

[0047] The third step is to perform a fusion analysis on the grayscale peak shift and absolute difference to obtain the grayscale peak right shift rate of the fitted curve.

[0048] Among them, the grayscale peak right shift rate is positively correlated with the grayscale peak shift amount, and negatively correlated with the absolute difference.

[0049] Specifically, the ratio of grayscale peak shift to absolute difference is calculated, and the ratio is normalized using a linear normalization function (such as percentile normalization). The normalized value is then used as the rightward shift rate of the grayscale peak in the fitted curve, which represents the noise distribution breadth of the ultrasound image. Generally, the absolute difference is not zero. If an extreme case exists, a non-zero constant, such as 0.001, is added to the denominator of the fraction.

[0050] It should be noted that the greater the shift of the gray-level peak in the gray-level histogram of the reference ultrasound image, and the smaller the difference in the size of the two peaks, the greater the possibility that the gray-level peak in the gray-level histogram is shifted to the right, and the more shifted the peak is, the wider the noise distribution of the reference ultrasound image.

[0051] Thus, one of the key indicators for determining the noise intensity of each pixel in a fetal ultrasound image was obtained: the noise distribution breadth of the reference ultrasound image.

[0052] S3. Analyze the noise distribution differences between the fetal ultrasound image and the reference ultrasound image based on the gradient values ​​of each pixel in the fetal ultrasound image and the reference ultrasound image, and determine the noise intensity of each pixel in the fetal ultrasound image based on the noise distribution breadth.

[0053] Here, noise intensity refers to the probability that a pixel is a noise point. Noise intensity can be used later to adjust the filter kernel size of the pixel to achieve noise reduction.

[0054] First, it should be noted that the rightward shift of the grayscale peak in normal fetal ultrasound images is not significant. Therefore, the noise severity of fetal ultrasound images is lower than that in over-gain mode, but they are similar in terms of noise characteristics. Thus, the similarity between the noise characteristics of different locations in normal mode and those in over-gain mode can indicate the noise situation in normal mode.

[0055] As an exemplary implementation, step S3 described above can be achieved through... Figure 3 Steps S31 to S32 shown are implemented as follows: S31, Based on the gradient value of each pixel in the fetal ultrasound image and the reference ultrasound image, determine the distance variance corresponding to each pixel in the fetal ultrasound image and the reference ultrasound image.

[0056] Here, distance variance is a quantitative indicator of the distribution of maxima within the region surrounding a pixel, determined by the gradient value.

[0057] For example, determining the distance variance of each pixel in the fetal ultrasound image and the reference ultrasound image includes: The first step is to refer to the fetal ultrasound image and the reference ultrasound image as the ultrasound image to be analyzed, and to take any pixel in the ultrasound image to be analyzed as the target pixel.

[0058] In this embodiment, the distance variance of pixels in the fetal ultrasound image and the reference ultrasound image is calculated in the same way. For ease of description, the ultrasound image to be analyzed and the target pixel are analyzed.

[0059] The second step is to determine the gradient value of the target pixel in the ultrasound image to be analyzed. If the gradient value of the target pixel is the largest in its eight neighborhood, then the target pixel is taken as the maximum value point, and then all the maximum value points in the ultrasound image to be analyzed are obtained.

[0060] In this embodiment, a maximum point refers to a pixel whose gradient value is still the largest compared to pixels within its eight-neighbor range, and which is located at the center of the eight-neighbor range. By obtaining the maximum point, the distribution of gradient values ​​in the ultrasound image to be analyzed can be made easier to determine the distance variance corresponding to each pixel.

[0061] The third step is to calculate the distance variance within the preset neighborhood of each pixel based on the distance between every two maximum points within the preset neighborhood of each pixel, and use this distance variance as the distance variance of the corresponding pixel in the ultrasound image to be analyzed.

[0062] In this embodiment, the preset neighborhood range is a circular region centered on the target pixel and with a radius equal to the absolute straight-line distance between the target pixel and its k-th nearest maximum point, where k is a preset constant. The preset constant k represents the minimum number of local maxima contained within the neighborhood of the target pixel, controlling the neighborhood scale and thus determining the balance between statistical smoothness and local sensitivity: a smaller k value results in a smaller neighborhood and greater sensitivity to local details and noise; a larger k value results in a larger neighborhood and smoother, more stable statistical results. In this embodiment, the value of k should be much smaller than the total number of maximum points in the entire image to ensure that the neighborhood has a clear local physical meaning. As an exemplary and non-limiting implementation, k can be any integer from 3 to 5, for example, k=3.

[0063] Additionally, if the number of maxima within the preset neighborhood is less than two, the distance variance of the corresponding pixel is set to a preset default constant. This default constant is typically 10 times the maximum normal distance variance. If there are no pixels in the entire image for which normal distance variance can be calculated (an extreme case), it is directly set to... .

[0064] S32, based on the difference between the distance variance of each pixel in the fetal ultrasound image and the reference ultrasound image, and combined with the noise distribution breadth, determine the noise intensity of each pixel in the fetal ultrasound image.

[0065] It should be noted that the noise distribution is the same under normal and overgain conditions, with the only difference being the noise intensity. Therefore, the noise intensity of each pixel in the fetal ultrasound image can be analyzed by the similarity of the distribution of the maximum points within the preset neighborhood of the pixel.

[0066] Among them, the difference between noise intensity and distance variance is negatively correlated, while it is positively correlated with the noise distribution breadth.

[0067] As an example, the formula for calculating the noise intensity of the u-th pixel in a fetal ultrasound image can be: In the formula, This represents the noise intensity of the u-th pixel in the fetal ultrasound image. This indicates the extent of noise distribution in the reference ultrasound image. This represents the absolute value of the difference between the distance variances of the u-th pixel in the fetal ultrasound image and the reference ultrasound image. This represents a non-zero constant, used to avoid the denominator of a fraction being zero. Its empirical value is consistent with the dimension of the denominator, such as 0.001.

[0068] In the formula for calculating noise intensity, The larger the value, the more pronounced the rightward shift of the grayscale peak in over-gain mode, resulting in a greater overall noise intensity in the fetal ultrasound image. Furthermore, with higher noise intensity... The smaller the value, the more similar the distribution of maxima around the pixel in the fetal ultrasound image is to the distribution of maxima in the overgain mode. Since there are noise points in the maxima, it can also be characterized as the similarity of the noise distribution around the pixel. The greater the similarity of the noise distribution, the greater the noise intensity of the u-th pixel in the fetal ultrasound image.

[0069] Referring to the above calculation process for the noise intensity of the u-th pixel in the fetal ultrasound image, the noise intensity of each pixel in the fetal ultrasound image can be obtained.

[0070] At this point, the noise intensity of each pixel in the fetal ultrasound image was obtained.

[0071] S4. Based on the gradient direction of each pixel in the fetal ultrasound image and each edge in the fetal ultrasound image, determine the noise coupling index of each pixel in the fetal ultrasound image.

[0072] Here, the noise coupling index refers to the degree of similarity between the image features of the image and the noise. In order to prevent the loss of the texture details of the heart itself in the image during the noise reduction process, the noise reduction intensity should be lower in the areas with strong noise coupling.

[0073] It should be noted that noise distribution in fetal ultrasound images requires denoising processing based on the noise distribution. However, due to the strong coupling between noise and the image, direct denoising may eliminate cardiac texture in the fetal ultrasound image, leading to inaccurate acquisition of cardiac texture information. To avoid this defect, a noise coupling index is determined for each pixel in the fetal ultrasound image, and this index is also used as a key indicator for image denoising.

[0074] It should be further explained that the gradient direction of the texture on the heart is relatively concentrated, while the gradient direction of noise is relatively dispersed. Therefore, the fetal ultrasound image can be segmented according to the gradient direction, and the noise coupling of the local area can be analyzed based on the segmentation results.

[0075] As an exemplary implementation, step S4 described above can be achieved through... Figure 4 Steps S41 to S43 shown are implemented as follows: S41, for each pixel in the fetal ultrasound image, the pixel with the largest gray level difference among the eight neighboring pixels is obtained as the direction pixel, and the unit vector from the pixel to the direction pixel is used as the gradient direction.

[0076] S42, the fetal ultrasound image is divided into various local regions based on the gradient direction of each pixel in the fetal ultrasound image.

[0077] For example, acquiring various local regions corresponding to fetal ultrasound images includes: The first step is to calculate the gradient direction variance in the eight neighborhood range of each pixel.

[0078] Here, the smaller the gradient direction variance, the more consistent the directions; the larger the gradient direction variance, the more dispersed the directions.

[0079] Specifically, for any pixel, the gradient direction of each pixel in the eight-neighbor range of the pixel is obtained; the average vector of the gradient directions of all pixels in the eight-neighbor range of the pixel is calculated, and then the length of the average vector is calculated, the value of which is between 0 and 1; the direction variance is calculated based on the length of the average vector, that is, the circular variance is calculated, which is used as the gradient direction variance.

[0080] The process of calculating the gradient direction variance is existing technology and will not be described in detail here.

[0081] The second step involves using gradient directional variance to perform superpixel segmentation on the fetal ultrasound image, resulting in several pixel clusters.

[0082] Superpixel segmentation can group pixels with consistent gradient direction distribution around a pixel into the same pixel cluster.

[0083] The implementation process of superpixel segmentation is existing technology and is not within the scope of protection of this invention, so it will not be described in detail here.

[0084] The third step is to cluster the pixels as local regions to obtain the local regions corresponding to the fetal ultrasound images.

[0085] S43, determine the noise coupling index of the local region to which each pixel belongs based on the gradient direction of each pixel in each local region and the number of pixels located on the edge.

[0086] It should be noted that the more discrete the gradient direction of a local region, the more noise interference there is in that local region. If there are still many edges in the local region, it means that there is a lot of texture information in the local region. The combination of the two will result in a large noise coupling in the corresponding local region.

[0087] For example, determining the noise coupling index of a local region includes: The first step is to calculate the average value of the gradient direction variance of all pixels in each local region, which is used as the first noise coupling factor for the corresponding local region.

[0088] The second step is to use the number of pixels located on the edge in each local region as the second noise coupling factor for that local region.

[0089] Specifically, in order to analyze the texture features of fetal ultrasound images, Canny edge detection is first used to obtain each edge in the fetal ultrasound image, and then the number of pixels located on the edges in each local region is counted and used as the second noise coupling factor of the local region.

[0090] The third step is to perform a fusion analysis on the first and second noise coupling factors of each local region to determine the noise coupling index of the local region to which each pixel belongs.

[0091] Among them, the noise coupling index is positively correlated with both the first noise coupling factor and the second noise coupling factor.

[0092] Specifically, the first noise coupling factor and the second noise coupling factor of the same local region are multiplied together, and the resulting product is used as the noise coupling index of the corresponding local region.

[0093] It is worth noting that the noise coupling index of each pixel located in the same local area is the same, so the noise coupling index of each pixel can be obtained.

[0094] Thus, this embodiment has determined the noise coupling index for each pixel in the fetal ultrasound image.

[0095] S5. Based on the noise intensity and noise coupling index of each pixel in the fetal ultrasound image, the fetal ultrasound image is denoised to obtain a denoised fetal ultrasound image to assist doctors in collaborative analysis.

[0096] It should be noted that in local areas with strong noise coupling, in order to prevent the loss of texture details of the heart itself in the fetal ultrasound image after noise reduction, the noise reduction intensity of pixels with strong noise coupling should be low.

[0097] As an exemplary implementation, acquiring a denoised fetal ultrasound image includes: The first step is to determine the noise reduction intensity of each pixel based on the noise intensity and noise coupling index of each pixel in the fetal ultrasound image.

[0098] The noise reduction intensity is positively correlated with the noise intensity and negatively correlated with the noise coupling index.

[0099] For example, the ratio of the noise intensity to the noise coupling index of the same pixel is used as the noise reduction intensity of the corresponding pixel, thereby obtaining the noise reduction intensity of each pixel in the fetal ultrasound image.

[0100] It should be noted that the noise coupling index is unlikely to be zero under normal circumstances. If there is an extreme case, a non-zero constant is added to the denominator of the fraction, such as 0.001.

[0101] The second step involves adjusting the smoothing coefficient of the corresponding pixel in the fetal ultrasound image during noise reduction using the noise reduction intensity of each pixel. Then, the improved nonlocal mean filter is used to denoise the fetal ultrasound image, resulting in a denoised fetal ultrasound image.

[0102] In this embodiment, the smoothing coefficient is a key parameter for controlling the noise reduction effect. A larger smoothing coefficient enhances noise reduction but easily blurs edges, such as the structure of a fetal heart. A smaller smoothing coefficient preserves details but weakens noise resistance. Therefore, the smoothing coefficient is adaptively determined by the noise reduction intensity of each pixel. Of course, another key parameter affecting the noise reduction effect can be the size of the search window.

[0103] An improved nonlocal mean filter suitable for fetal ultrasound images can be obtained by using an adaptive smoothing coefficient. Using the improved nonlocal mean filter to denoise fetal ultrasound images can meet the denoising requirements of different regions.

[0104] After obtaining the noise-reduced fetal ultrasound images, doctors combine standard sections to observe the integrity of the heart structure, measure key functional parameters (such as ejection fraction and ventricular wall thickness), and correlate them with the patient's medical history and other examination data (such as electrocardiogram and blood indicators) to comprehensively assess cardiac function and pathological risks.

[0105] Thus, more accurate results were obtained from the collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology.

[0106] Another embodiment of the present invention provides a collaborative analysis system for intrauterine fetal ultrasound images in obstetrics and gynecology, including a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement the above-described collaborative analysis method for intrauterine fetal ultrasound images in obstetrics and gynecology.

[0107] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology, characterized in that, Includes the following steps: Acquire a fetal ultrasound image and a reference ultrasound image, wherein the reference ultrasound image is the fetal ultrasound image in overgain mode; The noise distribution characteristics of the reference ultrasound image are analyzed based on the gray value of each pixel in the reference ultrasound image to determine the noise distribution breadth of the reference ultrasound image. The noise distribution difference between the fetal ultrasound image and the reference ultrasound image is analyzed based on the gradient value of each pixel in the fetal ultrasound image and the reference ultrasound image. Combined with the noise distribution breadth, the noise intensity of each pixel in the fetal ultrasound image is determined. Based on the gradient direction of each pixel in the fetal ultrasound image and each edge in the fetal ultrasound image, the noise coupling index of each pixel in the fetal ultrasound image is determined. Based on the noise intensity of each pixel in the fetal ultrasound image and the noise coupling index, the fetal ultrasound image is denoised to obtain a denoised fetal ultrasound image, which can assist doctors in collaborative analysis.

2. The method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology according to claim 1, characterized in that, Determining the noise distribution breadth of the reference ultrasound image includes: Based on the gray value of each pixel in the reference ultrasound image, determine the highest gray value peak and the target gray value peak on the fitted curve of the gray value histogram; Based on the highest grayscale peak value and the target grayscale peak value, the grayscale peak right shift rate of the fitted curve is determined as the noise distribution breadth of the reference ultrasound image.

3. The method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology according to claim 2, characterized in that, Determining the target grayscale peak includes: Determine the maximum value point on the fitted curve of the grayscale histogram of the reference ultrasound image; The fitted curve is divided into two curve segments using the maximum value point, and the maximum value of the curve segment located to the right of the maximum value point is taken as the target grayscale peak value.

4. The method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology according to claim 3, characterized in that, The step of determining the grayscale peak right shift rate of the fitted curve based on the highest grayscale peak value and the target grayscale peak value includes: The grayscale peak shift amount is determined based on the grayscale difference between the highest grayscale peak and the target grayscale peak. Determine the minimum point located between the highest grayscale peak and the target grayscale peak, use the minimum point to divide the fitted curve into two curve segments, calculate the definite integral for each of the two curve segments, and obtain the absolute difference between the two definite integrals; The grayscale peak shift and the absolute difference are fused and analyzed to obtain the grayscale peak right shift rate of the fitted curve; the grayscale peak right shift rate is positively correlated with the grayscale peak shift and negatively correlated with the absolute difference.

5. The method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology according to claim 1, characterized in that, Determining the noise intensity of each pixel in the fetal ultrasound image includes: Based on the gradient value of each pixel in the fetal ultrasound image and the reference ultrasound image, determine the distance variance of each pixel in the fetal ultrasound image and the reference ultrasound image; Based on the difference between the distance variance corresponding to each pixel in the fetal ultrasound image and the reference ultrasound image, and in combination with the noise distribution breadth, the noise intensity of each pixel in the fetal ultrasound image is determined. The difference between the noise intensity and the distance variance is negatively correlated, while it is positively correlated with the noise distribution breadth.

6. The method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology according to claim 5, characterized in that, Determining the distance variance of each pixel in the fetal ultrasound image and the reference ultrasound image includes: The fetal ultrasound image and the reference ultrasound image are referred to as the ultrasound image to be analyzed, and any pixel in the ultrasound image to be analyzed is taken as the target pixel. Determine the gradient value of the target pixel in the ultrasound image to be analyzed. If the gradient value of the target pixel is the largest in its eight neighborhood, then the target pixel is taken as the maximum value point, and then all the maximum value points in the ultrasound image to be analyzed are obtained. Based on the distance between every two maximum points within the preset neighborhood of each pixel, the distance variance within the preset neighborhood of each pixel is calculated and used as the distance variance of the corresponding pixel in the ultrasound image to be analyzed. The preset neighborhood range is a circular region centered on the target pixel and with the absolute straight-line distance between the target pixel and its nearest k-th maximum point as the radius, where k is a preset constant.

7. The method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology according to claim 1, characterized in that, Determining the noise coupling index for each pixel in the fetal ultrasound image includes: For each pixel in the fetal ultrasound image, the pixel with the largest gray level difference among the pixels within the eight neighborhoods of the pixel is taken as the directional pixel, and the unit vector from the pixel to the directional pixel is taken as the gradient direction. The fetal ultrasound image is divided into local regions based on the gradient direction of each pixel in the fetal ultrasound image. The noise coupling index of each pixel's local region is determined based on the gradient direction of each pixel in each local region and the number of pixels located on the edge.

8. The method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology according to claim 7, characterized in that, The process of dividing the fetal ultrasound image into local regions based on the gradient direction of each pixel in the fetal ultrasound image includes: Calculate the gradient direction variance in the eight-neighbor area of ​​each pixel; The gradient direction variance is used to perform superpixel segmentation on the fetal ultrasound image to obtain several pixel clusters; By clustering the pixels into local regions, we can obtain the various local regions corresponding to the fetal ultrasound image.

9. A method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology according to claim 8, characterized in that, The step of determining the noise coupling index of the local region to which each pixel belongs based on the gradient direction of each pixel in each local region and the number of pixels located on the edge includes: Calculate the average value of the gradient direction variance of all pixels in each local region, and use it as the first noise coupling factor of the corresponding local region; The number of pixels located on the edge in each local region is used as the second noise coupling factor for the corresponding local region. The first noise coupling factor and the second noise coupling factor of each local region are fused and analyzed to determine the noise coupling index of the local region to which each pixel belongs. The noise coupling index is positively correlated with both the first noise coupling factor and the second noise coupling factor.

10. A collaborative analysis system for intrauterine fetal ultrasound images in obstetrics and gynecology, characterized in that, It includes a processor and a memory, the processor being used to process instructions stored in the memory to implement a method for collaborative analysis of intrauterine fetal ultrasound images in obstetrics and gynecology as described in any one of claims 1-9.

Images