CN111192251B - Follicle ultrasonic processing method and system based on level set image segmentation - Google Patents

Abstract

The invention provides a follicle ultrasonic processing method and a follicle ultrasonic processing system based on level set image segmentation, which are characterized in that a follicle ultrasonic image is subjected to image enhancement through pretreatment to obtain a high-quality pretreatment image; dividing the preprocessed image into a plurality of regions according to the gray distribution characteristics of the ultrasonic image, respectively calculating the region threshold of each region, and performing gray self-adaptive segmentation on each region based on the region threshold; performing secondary segmentation on the pre-segmentation area by adopting level set image segmentation, and enhancing the difference between a follicle area and a non-follicle area in the pre-segmentation area to obtain a suspected follicle area; based on a decision tree and a bagging algorithm, partitioning the suspected follicle area, extracting an independent complete follicle area, and performing follicle quality grading evaluation. The problems of strong subjectivity, few quantitative indexes and the like in ultrasonic follicle monitoring are solved; the automatic segmentation of the follicle ultrasonic image is realized, the diagnosis process is optimized, the diagnosis quality is improved, and the labor intensity of medical workers is reduced.

Description

Follicle ultrasound processing method and system based on level set image segmentation

Technical Field

The present invention relates to the fields of image enhancement, image denoising and image segmentation, and in particular, to a method and system for ultrasonic processing of follicles based on level set image segmentation, and in particular to an image processing method for ultrasonic images of follicles based on image denoising and adaptive histogram based on bilateral filtering, pre-segmentation based on grayscale adaptive threshold and secondary segmentation based on level set method-CV model, and adopts least square method to perform digital quantitative evaluation on follicle quality.

Background Art

Medical image segmentation is a process of development from completely relying on manual segmentation to semi-manual and semi-automatic segmentation and then slowly moving towards fully automatic segmentation. Early medical image segmentation relied on manual segmentation to draw the boundaries of the desired area on the image. Semi-automatic segmentation combines computer technology and human knowledge to complete the image segmentation. Compared with simple manual segmentation, to a certain extent, it reduces the labor cost and the influence of human factors, but the segmentation of ultrasound images is still affected by the operator's technical ability and experience. In recent years, with the development of fuzzy technology and artificial intelligence technology, this type of hormone has been widely used in the field of image segmentation, which has also made the segmentation of some medical images completely free from human intervention and achieved fully automatic segmentation.

For ultrasound images, after searching: Ou Yue et al. realized the segmentation of liver ultrasound images in the article "Ultrasound Image Segmentation with Improved Geodesic Active Contour Model". Jin Lian et al. combined genetic algorithm and level set to effectively segment the tumor in liver ultrasound images in the article "Level Set Ultrasound Image Segmentation Based on Genetic Algorithm". Chen Guanrao et al. applied a neural network with biological simulation structure in the segmentation of cardiac ultrasound images in the article "Application of a Neural Network Based on Biological Simulation Structure in Ultrasound Medical Image Segmentation".

However, there is still a lack of research on follicle ultrasound image processing. Compared with other medical ultrasound images, follicle ultrasound images have smaller target areas, inconsistent numbers, and different shapes, making it difficult to directly apply other ultrasound image processing methods.

The prior art related to the present application is patent document 105496453A, which provides a yellow cattle follicle ultrasound monitoring system and a monitoring method thereof, including the following units: a reading and displaying ultrasound image unit, an image cropping unit, an image denoising unit, a follicle detection unit, a surface extraction unit, a three-dimensional reconstruction unit, a volume calculation unit, a volume detection unit, and a point set rendering unit. A platform for processing and calculating yellow cattle follicle ultrasound images is provided, and the program can perform denoising, follicle detection, follicle surface extraction, follicle three-dimensional reconstruction, follicle volume calculation and monitoring functions on ultrasound images. It can monitor the development of yellow cattle follicles under different development states, and can diagnose and track diseases for monitoring and treatment, with the advantages of being fast, efficient, highly accurate, and intuitive.

Summary of the invention

In view of the defects in the prior art, the object of the present invention is to provide a method and system for follicle ultrasound processing based on level set image segmentation.

According to the present invention, a method for ultrasonic processing of follicles based on level set image segmentation comprises:

Preprocessing step: identifying the follicle ultrasound image, and performing image enhancement on the follicle ultrasound image through preprocessing to obtain a high-quality preprocessed image;

Pre-segmentation step: according to the grayscale distribution characteristics of the ultrasound image, the pre-processed image is divided into multiple regions, the regional threshold of each region is calculated respectively, and the grayscale adaptive segmentation of each region is performed based on the regional threshold to obtain multiple pre-segmented regions;

Secondary segmentation step: level set image segmentation is used to perform secondary segmentation on the pre-segmented area, so as to enhance the distinction between the follicle area and the non-follicle area in the pre-segmented area and obtain the suspected follicle area;

Regional identification steps: extract image features of suspected follicle areas, partition suspected follicle areas based on decision trees and bagging algorithms, extract independent and complete follicle areas, and calculate characteristic indicators of follicle quality assessment in this area. Use the follicle quality assessment system model established by least squares fitting combined with expert quality assessment scoring to perform computer scoring assessment of follicle quality.

Preferably, the pretreatment step comprises:

Noise suppression step: after identifying the follicle ultrasound image, bilateral filtering is used to suppress the image noise while retaining the edge detail information of the follicle ultrasound image to obtain a denoised image;

Image enhancement step: Based on adaptive histogram equalization, different enhancement schemes are used for different parts of the denoised image to enhance the contrast while retaining the image details to obtain a preprocessed image.

Preferably, the pre-segmentation step comprises:

Determine the threshold step: take the area with small grayscale change in the preprocessed image as the central area, take the average grayscale value of the central area as the basic threshold, and reduce the threshold value of the peripheral area outside the central area;

Region division step: compare the basic threshold of the adjacent area of the pixel with the threshold value of the pixel, perform region division adaptively, and determine the pre-segmentation area.

Preferably, the secondary segmentation step comprises:

Area exclusion step: according to the size of mature follicles, a margin is set as the area threshold of the pre-segmented area to exclude areas that are too small or too large, and obtain a rough outline of the suspected follicle area;

Contour control step: Apply the CV model to segment and control the rough contour of the suspected follicle area to obtain the suspected follicle area.

Preferably, the region identification step comprises:

Partitioning step: extract the multi-level features of the suspected follicle area, apply principal component analysis to reduce the dimension of the multi-level features, apply decision tree and bagging algorithm to partition the suspected follicle area, and identify the non-independent complete follicle area in the partition;

Evaluation steps: extract characteristic indicators of independent complete follicle areas in the partition, evaluate the characteristic indicators in combination with the follicle evaluation system of expert quality evaluation, and fit the evaluation results using the least squares method to obtain a digital quantitative evaluation of follicle quality.

According to the present invention, a follicle ultrasound processing system based on level set image segmentation is provided, comprising:

Preprocessing module: identifies the follicle ultrasound image, and enhances the follicle ultrasound image through preprocessing to obtain a high-quality preprocessed image;

Pre-segmentation module: According to the grayscale distribution characteristics of the ultrasound image, the pre-processed image is divided into multiple regions, the regional threshold of each region is calculated respectively, and the grayscale adaptive segmentation of each region is performed based on the regional threshold to obtain multiple pre-segmented regions;

Secondary segmentation module: The level set image segmentation is used to perform secondary segmentation on the pre-segmented area, enhancing the distinction between the follicle area and the non-follicle area in the pre-segmented area, and obtaining the suspected follicle area;

Regional identification module: Extract image features of suspected follicle areas, partition suspected follicle areas based on decision trees and bagging algorithms, extract independent and complete follicle areas and conduct expert quality evaluation.

Preferably, the preprocessing module comprises:

Noise suppression module: After identifying the follicle ultrasound image, bilateral filtering is used to suppress the noise of the image while retaining the edge detail information of the follicle ultrasound image to obtain a denoised image;

Image enhancement module: Based on adaptive histogram equalization, different enhancement schemes are used for different parts of the denoised image to enhance the contrast while retaining the image details to obtain a preprocessed image.

Preferably, the pre-segmentation module comprises:

Determine the threshold module: take the area with small grayscale change in the preprocessed image as the central area, take the average grayscale value of the central area as the basic threshold, and reduce the threshold value of the peripheral area outside the central area;

Region division module: compares the basic threshold of the adjacent area of the pixel with the threshold value of the pixel, adaptively divides the area, and determines the pre-segmentation area.

Preferably, the secondary segmentation module comprises:

Area exclusion module: A margin is set according to the size of mature follicles as the area threshold of the pre-segmented area, which excludes areas that are too small or too large to obtain a rough outline of the suspected follicle area;

Contour control module: The CV model is used to segment and control the rough contour of the suspected follicle area to obtain the suspected follicle area.

Preferably, the region identification module comprises:

Partitioning module: extract the multi-level features of the suspected follicle area, apply principal component analysis to reduce the dimensionality of the multi-level features, apply decision tree and bagging algorithm to partition the suspected follicle area, and identify the non-independent complete follicle area in the partition;

Evaluation module: Calculate the characteristic indicators of follicle quality evaluation in independent and complete follicle areas, and perform computer scoring evaluation of follicle quality through the follicle quality evaluation system model established by least squares fitting combined with expert quality evaluation scoring.

Compared with the prior art, the present invention has the following beneficial effects:

The present invention is based on the level set-CV model method, through pre-processing including bilateral filtering and adaptive histogram equalization, pre-segmentation based on adaptive grayscale threshold, segmentation based on the CV model-level set method, and suspected follicle area identification based on decision tree and bagging algorithm, overcomes the influence of follicle speckle noise and complex texture, obtains a relatively clear and accurate follicle boundary contour, proposes a CV model-level set method segmentation speed control algorithm for the weak edge phenomenon of follicles appearing in follicle ultrasound images, and improves the segmentation effect. The least squares method is used in combination with expert quality assessment scoring for regression to achieve digital quantitative evaluation of follicle quality.

BRIEF DESCRIPTION OF THE DRAWINGS

Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

Fig. 1 is a flow chart of follicle ultrasound image processing;

Figure 2 shows the changes in the image processing process;

FIG3 is a diagram showing the effect of an ovarian follicle ultrasound image and pre-segmentation based on a grayscale adaptive threshold, a is an ovarian follicle ultrasound image, and b is a diagram showing the effect of a ovarian follicle ultrasound image pre-segmentation;

Figure 4 is a diagram showing the effect of secondary segmentation based on the level set-CV model;

Figure 5 is a diagram showing the follicle area division and follicle quality assessment results.

DETAILED DESCRIPTION

The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

The present invention adopts bilateral filtering and adaptive histogram equalization preprocessing; then, according to the intensity change characteristics of follicular ultrasound images, grayscale adaptive threshold segmentation is used for pre-segmentation; then, the CV model-level set method is used to perform secondary segmentation on the suspected follicle area to further improve the segmentation effect, and the segmentation speed of the algorithm is adjusted to address the problem that the algorithm is prone to leakage in the weak edge area of multiple follicles; finally, based on the morphology and texture characteristics of the follicles, a decision tree and bagging algorithm are used to identify the suspected follicle area, and finally, a follicle evaluation system is established based on indicators such as the roundness, echo uniformity, internal and external grayscale differences, fullness and size of the follicles, and the least squares method is used in combination with expert evaluation for regression, thereby realizing digital quantitative evaluation of follicles. A follicle ultrasound processing method based on level set image segmentation provided by the present invention includes:

Preprocessing step: identifying the follicle ultrasound image, and performing image enhancement on the follicle ultrasound image through preprocessing to obtain a high-quality preprocessed image;

Pre-segmentation step: according to the grayscale distribution characteristics of the ultrasound image, the pre-processed image is divided into multiple regions, the regional threshold of each region is calculated respectively, and the grayscale adaptive segmentation of each region is performed based on the regional threshold to obtain multiple pre-segmented regions;

Secondary segmentation step: level set image segmentation is used to perform secondary segmentation on the pre-segmented area, so as to enhance the distinction between the follicle area and the non-follicle area in the pre-segmented area and obtain the suspected follicle area;

Regional identification steps: extract image features of suspected follicle areas, partition suspected follicle areas based on decision trees and bagging algorithms, extract independent and complete follicle areas, and calculate characteristic indicators of follicle quality assessment in this area. Use the follicle quality assessment system model established by least squares fitting combined with expert quality assessment scoring to perform computer scoring assessment of follicle quality.

Specifically, the pretreatment step includes:

Noise suppression step: after identifying the follicle ultrasound image, bilateral filtering is used to suppress the image noise while retaining the edge detail information of the follicle ultrasound image to obtain a denoised image;

Image enhancement step: Based on adaptive histogram equalization, different enhancement schemes are used for different parts of the denoised image to enhance the contrast while retaining the image details to obtain a preprocessed image.

Specifically, the pre-segmentation step includes:

Determine the threshold step: take the area with small grayscale change in the preprocessed image as the central area, take the average grayscale value of the central area as the basic threshold, and reduce the threshold value of the peripheral area outside the central area;

Region division step: compare the basic threshold of the adjacent area of the pixel with the threshold value of the pixel, perform region division adaptively, and determine the pre-segmentation area.

Specifically, the secondary segmentation step includes:

Area exclusion step: according to the size of mature follicles, a margin is set as the area threshold of the pre-segmented area to exclude areas that are too small or too large, and obtain a rough outline of the suspected follicle area;

Contour control step: Apply the CV model to segment and control the rough contour of the suspected follicle area to obtain the suspected follicle area.

Specifically, the region identification step includes:

Partitioning step: extract the multi-level features of the suspected follicle area, apply principal component analysis to reduce the dimension of the multi-level features, apply decision tree and bagging algorithm to partition the suspected follicle area, and identify the non-independent complete follicle area in the partition;

Evaluation steps: Extract the characteristic indicators of follicle quality evaluation of independent complete follicle areas in the partition, and perform computer scoring evaluation of follicle quality through the follicle quality evaluation system model established by combining the least squares method fitting with expert quality evaluation.

According to the present invention, a follicle ultrasound processing system based on level set image segmentation is provided, comprising:

Preprocessing module: identifies the follicle ultrasound image, and enhances the follicle ultrasound image through preprocessing to obtain a high-quality preprocessed image;

Pre-segmentation module: According to the grayscale distribution characteristics of the ultrasound image, the pre-processed image is divided into multiple regions, the regional threshold of each region is calculated respectively, and the grayscale adaptive segmentation of each region is performed based on the regional threshold to obtain multiple pre-segmented regions;

Secondary segmentation module: The level set image segmentation is used to perform secondary segmentation on the pre-segmented area, enhancing the distinction between the follicle area and the non-follicle area in the pre-segmented area, and obtaining the suspected follicle area;

Regional identification module: Extract image features of suspected follicle areas, partition suspected follicle areas based on decision trees and bagging algorithms, extract independent and complete follicle areas and conduct expert quality evaluation.

Specifically, the preprocessing module includes:

Noise suppression module: After identifying the follicle ultrasound image, bilateral filtering is used to suppress the noise of the image while retaining the edge detail information of the follicle ultrasound image to obtain a denoised image;

Image enhancement module: Based on adaptive histogram equalization, different enhancement schemes are used for different parts of the denoised image to enhance the contrast while retaining the image details to obtain a preprocessed image.

Specifically, the pre-segmentation module includes:

Determine the threshold module: take the area with small grayscale change in the preprocessed image as the central area, take the average grayscale value of the central area as the basic threshold, and reduce the threshold value of the peripheral area outside the central area;

Region division module: compares the basic threshold of the adjacent area of the pixel with the threshold value of the pixel, adaptively divides the area, and determines the pre-segmentation area.

Specifically, the secondary segmentation module includes:

Area exclusion module: A margin is set according to the size of mature follicles as the area threshold of the pre-segmented area, which excludes areas that are too small or too large to obtain a rough outline of the suspected follicle area;

Contour control module: The CV model is used to segment and control the rough contour of the suspected follicle area to obtain the suspected follicle area.

Specifically, the region identification module includes:

Partitioning module: extract the multi-level features of the suspected follicle area, apply principal component analysis to reduce the dimensionality of the multi-level features, apply decision tree and bagging algorithm to partition the suspected follicle area, and identify the non-independent complete follicle area in the partition;

Evaluation module: Calculate the characteristic indicators of follicle quality evaluation in independent and complete follicle areas, and perform computer scoring evaluation of follicle quality through the follicle quality evaluation system model established by least squares fitting combined with expert quality evaluation scoring.

The follicle ultrasonic processing system based on level set image segmentation provided by the present invention can be implemented by the step flow of the follicle ultrasonic processing method based on level set image segmentation. Those skilled in the art can understand the follicle ultrasonic processing method based on level set image segmentation as a preferred example of the follicle ultrasonic processing system based on level set image segmentation.

In a specific implementation, as shown in FIG1 , the present invention processes the ultrasonic follicle image based on grayscale adaptive threshold segmentation and CV model-level set method, realizes the segmentation of the follicle area, and finally realizes the evaluation of the follicle quality by the least square method, including the following steps:

Step 100 image preprocessing, including step 110 and step 120, adopts bilateral filtering to suppress noise and adaptive histogram enhancement to enhance the image, improve image quality, and prepare for subsequent processing;

Step 110: noise suppression of the follicle ultrasound image, by using bilateral filtering to achieve image noise suppression on the basis of retaining edge detail information of the image.

Step 120: Image enhancement of the follicle ultrasound image, using different enhancement schemes for different parts of the image based on adaptive histogram equalization to enhance contrast while retaining image details.

Step 200 is based on the pre-segmentation of grayscale adaptive threshold, including steps 210, 220 and 230. The pre-segmentation of the follicle ultrasound image based on adaptive threshold segmentation calculates different thresholds for each region according to the grayscale distribution characteristics of the ultrasound image and processes the image by comprehensively considering the grayscale of the pixels surrounding the pixel point.

Step 210: Calculate a basic threshold for segmenting the follicle ultrasound image according to a central region of the follicle ultrasound image where the grayscale change is small.

Step 220: According to the characteristics that the grayscale of the central area of the follicle ultrasound image is higher and the grayscale of the peripheral area is lower, the threshold value of the peripheral area is reduced during segmentation.

Step 230: Irregular bright spots and weak edges of some follicles on the follicle ultrasound image are likely to affect the segmentation, especially making it easy for follicles that are close to each other to be connected during segmentation. Therefore, instead of taking the gray value of a pixel as the object, the average gray value of the adjacent area of a pixel is compared with the threshold as the basis for segmentation.

The calculation of different thresholds by region refers to selecting the threshold for follicle segmentation in the central region based on the average grayscale of the central region with smaller grayscale changes, and then adjusting the threshold of the peripheral region according to the characteristics of the larger grayscale in the middle region of the ultrasound image and the smaller grayscale in the peripheral region. In addition, in order to reduce the influence of irregular bright spots and weak edges of some follicles on the follicle ultrasound image, the average grayscale of the adjacent region of a pixel is compared with the threshold as the basis for segmentation. Compared with other segmentation methods, this method effectively avoids the problem of segmenting multiple follicles into a whole during segmentation due to the follicles being too close and the edges being weak, and effectively avoids the erroneous segmentation caused by the texture and bright spots of the follicle ultrasound image, and relatively speaking, the proportion of non-follicle areas mistakenly divided into follicle areas is relatively small. The disadvantage is that the segmented target area is generally smaller than the actual area of the follicle, and this problem can be solved in the subsequent secondary segmentation based on the level set.

Step 300 is a secondary segmentation based on the level set-CV model, including steps 310 and 320, using the secondary segmentation based on the level set method-CV model, adopting the level set method, and applying the CV model to perform secondary segmentation on the image based on the grayscale adaptive segmentation result.

Step 310: After pre-segmentation, the main areas obtained are: the area where the follicle is located, the wrong segmentation caused by background texture and bright spots, and part of the background area. For these three types of areas, the size of the segmented area and the distance from the edge are mainly used to determine whether to perform the next segmentation based on the area. The wrongly segmented background area is generally larger in area and closer to the ultrasound boundary, while the area obtained by the wrong segmentation of background texture and bright spots is smaller in area. Therefore, a certain margin is set according to the size of the mature follicle as the area size threshold to exclude areas that are too small and too large.

Step 320: After grayscale adaptive threshold segmentation, the non-follicle area is excluded, and the rough outline of the suspected follicle area is obtained. However, since the shape of each outline is different, it is difficult to express the initial outline. Therefore, the ellipse closest to the outline is selected as the initial motion curve. The center of gravity of the area is taken as the center of the ellipse, and the projection length of the area on the x-axis and y-axis is calculated as the major axis and minor axis of the ellipse.

Step 330: Apply the CV model-level set method to perform secondary segmentation on the follicle ultrasound image. During the segmentation process, the results obtained in the comprehensive pre-segmentation are used to adjust the CV model segmentation speed. The closest ellipse based on the grayscale adaptive threshold pre-segmentation result is used as the initial motion curve for the secondary segmentation, and in the secondary segmentation process using the level set method-CV model, the pre-segmentation result is compared with the secondary segmentation result. When the secondary segmentation result is larger than the pre-segmentation area, the segmentation speed is reduced. Suppose the segmentation speed of the CV model-level set algorithm at time t is v(t), and the area of the inner contour at time t is S(t). In the pre-segmentation, the area of the area obtained is S ₀ , and the actual area of the follicle is S _1. Generally, S ₀ <S ₁ , and S ₀ is the area of the area surrounded by the initial motion curve. During segmentation, the following adjustments are made:

Where λ is the speed control coefficient, λ≤1.

Step 400 is based on the identification of suspected follicle regions by decision tree and bagging algorithm, including step 410, step 420, step 430, extracting the features of the suspected follicle regions, applying PCA for dimensionality reduction, and then using decision tree for classification, and combining bagging algorithm to improve the performance of decision tree;

Step 410: Extract the grayscale, texture, shape contour and space of the suspected follicle area. Establish a follicle evaluation system based on the roundness, echo uniformity, grayscale difference between inside and outside, fullness and size of the follicle, and use the least squares method combined with expert evaluation for regression to achieve digital quantitative evaluation of the follicle.

Step 420: Apply the PCA (Principal Component Analysis) algorithm to process the above feature data to achieve dimensionality reduction.

Step 430: Apply decision tree and bagging algorithm to identify suspected follicle regions, and divide the suspected follicle regions into 4 categories: non-follicle region, independent and complete follicle region, multiple follicle region, and others, and mark the non-independent and complete follicle region on the image.

Step 500 of the follicle quality evaluation is to establish a follicle evaluation system based on indicators such as the roundness of the follicle, uniformity of the internal echo, difference in brightness inside and outside the follicle, fullness, size, etc., and to use the least squares method combined with expert evaluation for regression to achieve a digital quantitative evaluation of the follicle, which includes steps 510 and 520.

Step 510: Extract the circumference, area, inner and outer area brightness and other indicators of the independent and complete follicle area, and calculate the roundness, internal echo uniformity, follicle inner and outer brightness difference, fullness, size and other indicators based on the above indicators to establish a follicle evaluation system.

Circumference calculation formula:

L = ∫dl

Area calculation formula:

S = ∫∫ds

The calculation formula of regional average brightness is:

Among them, g(i,j) is the grayscale value of the pixel in the region.

The calculation formula of pixel coordinates of the center of gravity of the region is:

X＝∑x/n

Y＝∑y/n

The formula for calculating the maximum average diameter of follicles is:

Among them, d and _dτ are a pair of diameter line segments passing through the center of gravity of the follicle area and perpendicular to each other.

The roundness index is calculated using factors such as the area S and perimeter L of the follicle area. The calculation formula is:

Among them, S is the area of the follicle area, and L is the circumference of the follicle area.

等要素计算，计算公式：The echo uniformity index uses the pixel gray value g(i,j) in the follicle area and the internal gray value average

The calculation formula is:

为卵泡所在区域平均亮度。Among them, g ⁱⁿ (i, j) is the pixel gray value of the follicle area.

is the average brightness of the area where the follicle is located.

The brightness difference index of the inside and outside of the follicle is calculated by taking the center of gravity of the follicle area as the center and twice the average radius r of the follicle as the radius. The area is divided into two parts, the inside and outside of the follicle, according to the follicle segmentation results. The average brightness of each area is calculated, and the brightness difference index of the inside and outside of the follicle is calculated using the following formula:

The fullness index is calculated using the area S of the follicle area and the area _SC of the minimum circumscribed circle covering the follicle area. The calculation formula is:

θ＝S/S _C

The size index is calculated using the maximum average diameter of the follicles.

Step 520: Based on the expert's evaluation of the follicle quality, the least squares method is used to fit the indicators such as the roundness of the follicle, the uniformity of the internal echo, the difference in brightness inside and outside the follicle, the fullness, and the size, so as to obtain a digital quantitative evaluation of the follicle quality.

As shown in Figure 2, during the implementation of the present invention, the change state of the follicle ultrasound image is finally displayed as a follicle with a relatively clear and accurate boundary contour, and three clear follicles are obtained in the figure. As shown in Figure 5, the suspected follicle area identification effect diagram based on the decision tree and bagging algorithm and the follicle quality assessment result based on the least squares method, the figure contains a specific mark for the follicle score, which is convenient for evaluation and identification.

As shown in Figure 3, the effect of pre-segmentation based on follicular ultrasound image and grayscale adaptive threshold, where a is the follicular ultrasound image and b is the effect of pre-segmentation of follicular ultrasound image. As shown in Figure 4, the effect of secondary segmentation based on level set-CV model, the non-independent complete follicle area is marked in the figure.

The present invention solves the problems that the evaluation of ultrasonic follicle monitoring is mainly completed by physicians, which is highly subjective, has high requirements for physicians, and has few quantitative indicators. It realizes automatic segmentation of follicle ultrasound images, optimizes the diagnosis process, improves the diagnosis quality, and reduces the labor intensity of medical workers.

Those skilled in the art know that, in addition to implementing the system, device and its various modules provided by the present invention in a purely computer-readable program code, it is entirely possible to implement the same program in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers and embedded microcontrollers by logically programming the method steps. Therefore, the system, device and its various modules provided by the present invention can be considered as a hardware component, and the modules included therein for implementing various programs can also be considered as structures within the hardware component; the modules for implementing various functions can also be considered as both software programs for implementing the method and structures within the hardware component.

The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims (10)

1. A method for ultrasonic processing of follicles based on level set image segmentation, characterized by comprising: Preprocessing step: identifying the follicle ultrasound image, and performing image enhancement on the follicle ultrasound image through preprocessing to obtain a high-quality preprocessed image; Pre-segmentation step: according to the grayscale distribution characteristics of the ultrasound image, the pre-processed image is divided into multiple regions, the regional threshold of each region is calculated respectively, and the grayscale adaptive segmentation of each region is performed based on the regional threshold to obtain multiple pre-segmented regions; Secondary segmentation step: level set image segmentation is used to perform secondary segmentation on the pre-segmented area, so as to enhance the distinction between the follicle area and the non-follicle area in the pre-segmented area and obtain the suspected follicle area; Regional identification step: Extract the image features of the suspected follicle area, partition the suspected follicle area based on the decision tree and bagging algorithm, extract the independent and complete follicle area, and perform computer scoring and evaluation of the follicle quality. 2. The method for follicle ultrasonic processing based on level set image segmentation according to claim 1, characterized in that the preprocessing step comprises: Noise suppression step: after identifying the follicle ultrasound image, bilateral filtering is used to suppress the image noise while retaining the edge detail information of the follicle ultrasound image to obtain a denoised image; Image enhancement step: Based on adaptive histogram equalization, different enhancement schemes are used for different parts of the denoised image to enhance the contrast while retaining the image details to obtain a preprocessed image. 3. The method for follicle ultrasonic processing based on level set image segmentation according to claim 1, characterized in that the pre-segmentation step comprises: Determine the threshold step: take the area with small grayscale change in the preprocessed image as the central area, take the average grayscale value of the central area as the basic threshold, and reduce the threshold value of the peripheral area outside the central area; Region division step: compare the basic threshold of the adjacent area of the pixel with the threshold value of the pixel, perform region division adaptively, and determine the pre-segmentation area. 4. The method for follicle ultrasonic processing based on level set image segmentation according to claim 1, characterized in that the secondary segmentation step comprises: Area exclusion step: according to the size of mature follicles, a margin is set as the area threshold of the pre-segmented area to exclude areas that are too small or too large, and obtain a rough outline of the suspected follicle area; Contour control step: Apply the CV model to segment and control the rough contour of the suspected follicle area to obtain the suspected follicle area. 5. The method for follicle ultrasonic processing based on level set image segmentation according to claim 1, characterized in that the region identification step comprises: Partitioning step: extract the multi-level features of the suspected follicle area, apply principal component analysis to reduce the dimensionality of the multi-level features, apply decision tree and bagging algorithm to partition the suspected follicle area, and identify the non-independent complete follicle area in the partition; Evaluation steps: Calculate the characteristic indicators of follicle quality evaluation of independent complete follicle areas in the extracted partitions, and perform computer scoring evaluation of follicle quality through the follicle quality evaluation system model established by combining the least squares method fitting with expert quality evaluation. 6. A follicle ultrasound processing system based on level set image segmentation, characterized by comprising: Preprocessing module: identifies the follicle ultrasound image, and enhances the follicle ultrasound image through preprocessing to obtain a high-quality preprocessed image; Pre-segmentation module: According to the grayscale distribution characteristics of the ultrasound image, the pre-processed image is divided into multiple regions, the regional threshold of each region is calculated respectively, and the grayscale adaptive segmentation of each region is performed based on the regional threshold to obtain multiple pre-segmented regions; Secondary segmentation module: The level set image segmentation is used to perform secondary segmentation on the pre-segmented area, enhancing the distinction between the follicle area and the non-follicle area in the pre-segmented area, and obtaining the suspected follicle area; Regional identification module: Extract image features of suspected follicle areas, partition suspected follicle areas based on decision trees and bagging algorithms, extract independent and complete follicle areas, and perform computer scoring and evaluation of follicle quality. 7. The follicle ultrasound processing system based on level set image segmentation according to claim 6, characterized in that the preprocessing module comprises: Noise suppression module: After identifying the follicle ultrasound image, bilateral filtering is used to suppress the noise of the image while retaining the edge detail information of the follicle ultrasound image to obtain a denoised image; Image enhancement module: Based on adaptive histogram equalization, different enhancement schemes are used for different parts of the denoised image to enhance the contrast while retaining the image details to obtain a preprocessed image. 8. The follicle ultrasound processing system based on level set image segmentation according to claim 6, characterized in that the pre-segmentation module comprises: Determine the threshold module: take the area with small grayscale change in the preprocessed image as the central area, take the average grayscale value of the central area as the basic threshold, and reduce the threshold value of the peripheral area outside the central area; Region division module: compares the basic threshold of the adjacent area of the pixel with the threshold value of the pixel, adaptively divides the area, and determines the pre-segmentation area. 9. The follicle ultrasound processing system based on level set image segmentation according to claim 6, characterized in that the secondary segmentation module comprises: Area exclusion module: A margin is set according to the size of mature follicles as the area threshold of the pre-segmented area, which excludes areas that are too small or too large to obtain a rough outline of the suspected follicle area; Contour control module: The CV model is used to segment and control the rough contour of the suspected follicle area to obtain the suspected follicle area. 10. The follicle ultrasound processing system based on level set image segmentation according to claim 6, characterized in that the region identification module comprises: Partitioning module: extract the multi-level features of the suspected follicle area, apply principal component analysis to reduce the dimensionality of the multi-level features, apply decision tree and bagging algorithm to partition the suspected follicle area, and identify the non-independent complete follicle area in the partition; Evaluation module: Calculate and extract the characteristic indicators of follicle quality evaluation of independent complete follicle areas in the partition, and perform computer scoring evaluation of follicle quality through the follicle quality evaluation system model established by least squares fitting combined with expert quality evaluation scoring.

Images