Double-ventricle automatic outlining method and system based on echocardiography

Abstract

The invention relates to the technical field of image segmentation, in particular to a double-ventricle automatic outlining method and system based on echocardiography, and the method comprises the steps: obtaining a plurality of target regions according to the dynamic features of the same pixel point in continuous frames of echocardiography; screening out a ventricular region according to the characteristic change of the target region in the continuous frame echocardiogram; determining a transition path from each edge pixel point on the initial edge of the ventricular region to the center of the ventricular region; the reliability degree of each edge pixel point is obtained according to the number of the pixel points of the transition path and the gray value variation amplitude, the reliability degree of each edge pixel point is used as a weight, the edge of the ventricular region is fitted, the final edge of the ventricular region is obtained, the final edge is clearer and more reliable, smoothness and stability of the contour of the ventricular region are achieved, and the ventricular region is more accurate. Jittering and distortion are suppressed, the extraction precision of the double-ventricular region in the echocardiogram is improved, and then the outlining accuracy of the double-ventricular region in the echocardiogram is improved.

Description

Technical Field

[0001] This invention relates to the field of image segmentation technology, specifically to an automatic biventricular delineation method and system based on echocardiography. Background Technology

[0002] Echocardiography is an imaging technique that uses high-frequency sound waves to observe the structure and function of the heart. In analyzing echocardiograms, it is necessary to identify and delineate the biventricular region within the image.

[0003] Existing techniques typically employ edge detection, thresholding, or machine learning models based on single-frame echocardiography to directly identify ventricular boundaries. However, due to motion blur caused by cardiac pulsation, image noise, and the gray-scale similarity between the atria and ventricles, these methods relying on the static features of single-frame echocardiography face two fundamental challenges: first, it is difficult to reliably distinguish between true tissue boundaries and artifacts caused by blurring; second, it is prone to misidentifying atrial regions as ventricles, leading to segmentation errors. As a result, the ventricular contours extracted by existing methods often exhibit unreasonable jitter, breaks, or morphological distortions between different frames, resulting in insufficient accuracy. This affects the extraction accuracy of the biventricular region in echocardiography, and consequently, the accuracy of delineating the biventricular region in echocardiography. Summary of the Invention

[0004] To address the low accuracy of existing echocardiographic delineation of the biventricular region, this invention aims to provide an automatic biventricular delineation method and system based on echocardiography. The specific technical solution adopted is as follows: In a first aspect of the present invention, a method for automatically delineating two ventricles based on echocardiography is provided, comprising: Acquire consecutive frames of echocardiograms covering at least one cardiac cycle, acquired by an ultrasound probe; Based on the dynamic characteristics of the same pixel in consecutive frames of echocardiography, several target regions in the echocardiogram are obtained; wherein, the dynamic characteristics characterize the motion state or grayscale change pattern of the pixel in the time dimension. Based on the characteristic changes of each target region in consecutive frames of echocardiography, two ventricular regions are selected from each target region. For each of the two ventricular regions, the initial edge of the ventricular region in each frame of echocardiography is determined, and the transition path from each edge pixel on the initial edge to the center of the ventricular region is determined; wherein, the center of the ventricular region is the geometric centroid of the ventricular region; The reliability of each edge pixel is obtained from the number of pixels in the transition path and the magnitude of grayscale value change on the transition path. The reliability of each edge pixel is used as a weight to fit the edge of the ventricular region, thereby obtaining the final edge of the ventricular region for outlining the dual ventricular contour, which is then displayed on the ultrasound device's display interface.

[0005] In an exemplary embodiment, the process of acquiring the target region includes: Determine the degree of fluctuation in the area of ​​the initial connected region where the same pixel is located in each frame of echocardiography; the initial connected region is obtained by threshold segmentation of each frame of echocardiography. Determine the mean gray value and the range of gray values ​​of the same pixel in each frame of echocardiography; Based on the fluctuation level, the mean gray value, and the gray value range, a fixed evaluation index is obtained for each pixel; the fixed evaluation index is positively correlated with the fluctuation level and negatively correlated with the mean gray value and the gray value range. The target region is obtained based on the fixed evaluation index of each pixel.

[0006] In an exemplary embodiment, obtaining the target region based on a fixed evaluation index for each pixel includes: Adjacent target pixels constitute the target region; the target pixels are pixels with a fixed evaluation index greater than a preset evaluation index threshold.

[0007] In an exemplary embodiment, the step of selecting two ventricular regions from each target region based on the characteristic changes of each target region in consecutive frames of echocardiography includes: Determine the degree of deviation of the area change of the initial connected region corresponding to the target region in consecutive frames of echocardiography relative to the cardiac cycle; Based on the overall magnitude of the area change of the initial connected domain corresponding to the target region and the difference in area change in adjacent echocardiogram frames, combined with the degree of deviation, the probability that the target region belongs to the ventricular region is obtained. Based on the stated possibilities, two ventricular regions are selected from the target region.

[0008] In one exemplary embodiment, the process of obtaining the degree of deviation includes: Determine the area change curve of the initial connected component corresponding to the target region for consecutive frames of echocardiography; Obtain the time interval between the maximum and minimum values ​​in the area change curve; The degree of deviation is determined by the difference in duration between the time interval and half a cardiac cycle.

[0009] In one exemplary embodiment, the process of obtaining the possibility includes: Using the extreme points in the area change curve as segmentation points, the area change curve is divided into several curve segments; the extreme points include maximum points and minimum points; By fusing the differences in area change of the target region in adjacent echocardiogram frames, the overall difference in area change is obtained; The regularity of the target region is obtained based on the degree of deviation, the number of curve segments, and the overall difference in area change; the regularity is inversely correlated with the degree of deviation, the number of curve segments, and the overall difference in area change. Based on the degree of regularity and the overall magnitude of area change, the probability that the target region belongs to the ventricular region is obtained; the probability is positively correlated with both the degree of regularity and the overall magnitude of area change, and the overall magnitude of area change is the difference between the maximum and minimum values.

[0010] In an exemplary embodiment, the process of obtaining the transition path includes: Determine the line connecting the edge pixel to the center of the ventricular region; Starting from the edge pixel, traverse each pixel on the connecting line until the gray value is lower than the preset chamber threshold for the first time to obtain the transition path.

[0011] In an exemplary embodiment, the grayscale value change range is obtained by the difference in grayscale values ​​between the first pixel and the last pixel on the transition path; The process of obtaining the reliability level includes: The reliability of edge pixels is obtained based on the number of pixels in the transition path and the magnitude of grayscale value change; the reliability is inversely correlated with the number of pixels in the transition path and positively correlated with the magnitude of grayscale value change.

[0012] In an exemplary embodiment, fitting the edge of the ventricular region using the reliability of each edge pixel as a weight to obtain the final edge of the ventricular region includes: Using the reliability of each edge pixel as a weight, a weighted fitting method is used to fit the edge of the ventricular region to obtain the final edge of the ventricular region.

[0013] In a second aspect of the present invention, an automatic biventricular delineation system based on echocardiography is provided, comprising: a memory and a processor; the memory is connected to the processor; the memory is used to store program instructions; the processor is used to implement the above-described automatic biventricular delineation method based on echocardiography when the program instructions are executed.

[0014] The present invention has the following beneficial effects: First, the present invention analyzes the dynamic features of the same pixel in consecutive frames of echocardiography to obtain several target regions in the echocardiogram. This overcomes the dependence on static features of a single frame and utilizes dynamic information. Moreover, by analyzing the feature changes of each target region, regions that conform to the kinematic characteristics of the ventricle can be accurately screened out, thereby reliably distinguishing the ventricle from the atrium and other tissues at the region level, solving the misjudgment caused by gray-scale similarity in traditional methods. The transition path reflects the spatial transition structure from the edge to the center of the ventricular region. Analyzing the transition path determines the reliability of each edge pixel. The reliability can directly quantify the reliability of the edge pixel. Using this as a weight, the edge of the ventricular region is fitted to obtain the final edge of the ventricular region. The final edge is clearer and more reliable, achieving smoothness and stability of the ventricular region contour, suppressing jitter and distortion, improving the extraction accuracy of the biventricular region in echocardiography, and thus improving the delineation accuracy of the biventricular region in echocardiography. The technical solution provided by this invention, through a progressive process of coarse localization of dynamic features → fine screening of temporal changes → reliability assessment of transition paths → weighted fitting optimization, can fundamentally reduce the risk of misidentifying the atrium as the ventricle, effectively suppress interference caused by motion blur and noise, and obtain a more accurate and reliable ventricular contour. Attached Figure Description

[0015] Figure 1 This is a flowchart of an automatic biventricular delineation method based on echocardiography provided in one embodiment of the present invention; Figure 2 This is a flowchart of the target area acquisition process provided in one embodiment of the present invention; Figure 3 This is a flowchart illustrating the implementation of step S2 provided in one embodiment of the present invention; Figure 4 This is a flowchart illustrating the possibilities provided by one embodiment of the present invention. Detailed Implementation

[0016] 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 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.

[0017] 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. All data and information collected in this application have been obtained with full consent.

[0018] This embodiment provides an automatic biventricular delineation method based on echocardiography, which is used to identify the biventricular region (including the left and right ventricular regions) of the subject's echocardiogram and realize the automatic delineation of the biventricular region.

[0019] First, acquire consecutive frames of echocardiograms covering at least one cardiac cycle, obtained by an ultrasound probe.

[0020] This embodiment provides an automatic biventricular delineation method based on echocardiography, which processes the echocardiogram of the subject. As a specific application scenario, the echocardiogram of the subject is first acquired. The subject is in the left lateral decubitus position for ultrasound examination. A phased array probe is used to scan the heart region of the subject through the intercostal space. A professional physician places the probe at various standard acoustic window positions to obtain a series of two-dimensional sections. This invention mainly performs automatic biventricular delineation, so it analyzes based on the series of sections at the apex acoustic window, so that multiple consecutive frames of echocardiograms are obtained at each acoustic window position.

[0021] It should be understood that the duration of the continuous frame echocardiograms obtained in this embodiment is one complete cardiac cycle (i.e., one complete heartbeat cycle). That is, the continuous frame echocardiograms are acquired within one complete cardiac cycle, facilitating subsequent data processing. In an exemplary embodiment, continuous frame echocardiograms with a duration longer than one complete cardiac cycle can be acquired, and electrocardiogram (ECG) data can be acquired simultaneously. The times of the continuous frame echocardiograms and ECG data are matched, and then a time interval of one complete cardiac cycle is determined based on the ECG data. Continuous frame echocardiograms within this time interval are then extracted as a complete cardiac cycle continuous frame echocardiogram.

[0022] The time interval between two adjacent echocardiogram frames is set according to actual needs. Since the typical frame rate range in routine transthoracic echocardiography is 50-100 frames / second, the typical range of the time interval between two adjacent frames is 20 milliseconds-10 milliseconds, with the specific value set according to actual needs. In this embodiment, the number of echocardiogram acquisitions is set to T, where T is greater than or equal to 2. The specific value is calculated from the length of a complete cardiac cycle and the time interval between two adjacent echocardiogram frames. It should be understood that each echocardiogram frame has the same size. Pixel matching is performed on each echocardiogram frame to ensure that the same pixel point has a corresponding relationship in each frame.

[0023] This embodiment can also preprocess the obtained echocardiogram frames, such as performing Gaussian filtering and contrast enhancement, to make the boundaries of the heart structures clearer. It should be understood that this embodiment constructs a two-dimensional coordinate system using the length and width of the echocardiogram as the horizontal and vertical axes, respectively, and maps each echocardiogram into this system. Since the echocardiogram is a grayscale image, to facilitate subsequent data processing, the grayscale value of each pixel in the echocardiogram is divided by 255, making the grayscale value of each pixel range from 0 to 1, and eliminating the dimension. Therefore, the grayscale value of each pixel is dimensionless data, and all grayscale values ​​mentioned below are dimensionless data.

[0024] like Figure 1 As shown in the figure, this embodiment provides an automatic biventricular delineation method based on echocardiography, which includes the following steps: Step S1: Based on the dynamic characteristics of the same pixel in consecutive frames of echocardiography, obtain several target regions in the echocardiogram; Step S2: Based on the characteristic changes of each target region in consecutive frames of echocardiography, select two ventricular regions from each target region; Step S3: For any ventricular region in the two ventricular regions, determine the initial edge of the ventricular region in each frame of echocardiography, and determine the transition path from each edge pixel on the initial edge to the center of the ventricular region. Step S4: The reliability of each edge pixel is obtained from the number of pixels in the transition path and the gray value change range on the transition path; Step S5: Using the reliability of each edge pixel as a weight, fit the edge of the ventricular region to obtain the final edge of the ventricular region, so as to outline the contour of the two ventricles and display it on the display interface of the ultrasound device.

[0025] The following detailed explanation of each step, in conjunction with the accompanying drawings, is provided.

[0026] Before step S1 begins, this embodiment first performs image registration preprocessing on the acquired consecutive frame echocardiograms. Since cardiac pulsation and respiratory movements cause displacement and deformation of myocardial tissue in the image coordinate system, directly performing cross-frame analysis based on fixed pixel coordinates lacks physical meaning. Therefore, the first or intermediate frame in the consecutive frames is used as the reference frame, and optical flow or deformable image registration algorithms are employed to calculate the deformation field of each subsequent frame relative to the reference frame. Thus, the "same pixel" or "features of a pixel in each frame of echocardiogram" mentioned in subsequent steps refers to the corresponding pixel in other frames corresponding to a pixel in the reference frame, within the registered coordinate system. This registration step ensures that the object of temporal analysis is the same myocardial tissue point, providing a foundation for the accuracy of subsequent calculations.

[0027] Step S1: Based on the dynamic characteristics of the same pixel in consecutive frames of echocardiography, obtain several target regions in the echocardiogram.

[0028] Because the continuous beating of the heart causes motion blurring of the boundaries of structures such as the ventricles and atria in echocardiography, current techniques may result in errors of over- or under-delineation when automatically outlining these blurred areas. During a heartbeat, there are periodic contractions and relaxations; the myocardium contracts and thins, or relaxes and thickens. This is reflected in consecutive echocardiogram frames by a periodic increase and decrease in the proportion of myocardial structures, corresponding to the compression and expansion of the atrium / ventricle volume. While the atrium / ventricle volume decreases during this process, there are also regions that remain constant (i.e., the size of the region when it shrinks to its minimum volume). Therefore, it is necessary to first obtain identifiable chamber regions (here, chambers refer to the ventricles and atria) that consistently appear in consecutive echocardiogram frames, which could be the left and right ventricles or the left and right atria.

[0029] In echocardiography, the pixel values ​​of the chamber regions corresponding to the atria and ventricles are low, and the area of ​​the connected regions they occupy changes significantly in consecutive echocardiogram frames. Therefore, based on the changes in gray values ​​and connected region features in consecutive echocardiogram frames, target regions in the echocardiogram are obtained, and these target regions serve as the basis for subsequent judgment of ventricular regions. This embodiment uses any one pixel as the analysis object, and based on the dynamic characteristics of that pixel in consecutive echocardiogram frames, that is, based on the dynamic characteristics of the same pixel in consecutive echocardiogram frames, several target regions in the echocardiogram are obtained. In an exemplary embodiment, such as... Figure 2 As shown, the following is a specific process for obtaining the target area: Step S11: Determine the degree of fluctuation in the area of ​​the initial connected region where the same pixel is located in each frame of echocardiography.

[0030] Because the atria and ventricles have lower grayscale values ​​in echocardiograms, while other regions have higher grayscale values, an adaptive thresholding segmentation algorithm is used to segment any given echocardiogram frame, resulting in several initial connected components. Since adaptive thresholding is a standard technique for image segmentation, it will not be elaborated further. It should be understood that at least four initial connected components should be obtainable in an echocardiogram.

[0031] To facilitate subsequent data processing, the area of ​​each initial connected region in each frame of echocardiography (area refers to the number of pixels contained in the initial connected region) is normalized. In an exemplary embodiment, the ratio of the area of ​​the initial connected region to the area of ​​the echocardiography is calculated, and the result is used as the normalized area of ​​the initial connected region, thus eliminating the influence of dimensions. The areas of the initial connected regions mentioned below are all normalized results.

[0032] For any given pixel, the key is its position in the echocardiogram (i.e., its two-dimensional coordinates in a two-dimensional coordinate system). The initial connected region of the pixel in each echocardiogram frame is determined, thus obtaining the area of ​​the initial connected region in each frame. Finally, the degree of fluctuation of the area of ​​the initial connected region in each echocardiogram frame is obtained. In this embodiment, the standard deviation of the area of ​​the initial connected region of the pixel in each echocardiogram frame is calculated, and this standard deviation is used as the degree of fluctuation. Since regions such as the ventricle and atrium exhibit continuous changes with heartbeats, a larger standard deviation of the area of ​​the initial connected region of the pixel in each echocardiogram frame indicates a higher degree of fluctuation in the area of ​​the initial connected region. This means the pixel is more likely to belong to a region such as the ventricle or atrium, and the pixel's fixed evaluation index is higher. The fixed evaluation index is positively correlated with the degree of fluctuation.

[0033] Step S12: Determine the mean gray value and the range of gray values ​​of the same pixel in each frame of echocardiography.

[0034] The grayscale values ​​of the pixel are obtained in each frame of echocardiography, and the average grayscale value of the pixel in each frame is calculated as the mean grayscale value of the pixel in each frame of echocardiography. Simultaneously, the maximum and minimum grayscale values ​​of the pixel in each frame of echocardiography are obtained, and the difference between the maximum and minimum grayscale values ​​is calculated as the grayscale range of the pixel in each frame of echocardiography. The smaller the mean grayscale value, the more likely the pixel is located in the ventricle, atrium, or other regions of the heart, and the higher the fixed evaluation index of the pixel. The fixed evaluation index is inversely correlated with the mean grayscale value. A smaller grayscale range indicates that the grayscale value of a pixel is more stable across echocardiogram frames. This means the pixel is more likely to belong to the region where the atrium / ventricle contracts to its minimum volume. Because if it belongs to this region, even with continuous changes due to heartbeats, the grayscale value of pixels within this region will not change significantly. Therefore, the fixed evaluation index of this pixel is higher, and the fixed evaluation index is inversely correlated with the grayscale range. Thus, a small mean grayscale value and a small grayscale range indicate that the pixel's grayscale value is small and stable across echocardiogram frames.

[0035] Step S13: Based on the degree of fluctuation, the mean gray value, and the range of gray values, obtain the fixed evaluation index for each pixel.

[0036] Based on the standard deviation of the area of ​​the initial connected region where the pixel is located in each frame of echocardiography, the mean gray value of the pixel in each frame of echocardiography, and the range of gray values ​​of the pixel in each frame of echocardiography, a fixed evaluation index for the pixel is obtained. Based on the above logical analysis, the following is a specific calculation method for the fixed evaluation index of the pixel: ; in, This represents a fixed evaluation metric for the i-th pixel. Let represent the area of ​​the initial connected region where the i-th pixel is located in the echocardiogram of frame t. This represents the standard deviation of the area of ​​the initial connected region where the i-th pixel is located in each frame of echocardiography. This represents the average grayscale value of the i-th pixel across all frames of echocardiography. This represents the grayscale range of the i-th pixel in each frame of echocardiography.

[0037] Step S14: Obtain the target region based on the fixed evaluation index of each pixel.

[0038] Step S13 obtains a fixed evaluation index for each pixel. This fixed evaluation index characterizes the probability that a pixel is a fixed region within one of the four chambers of the heart. A larger fixed evaluation index indicates that the pixel is more likely to be located in the fixed region corresponding to a ventricle or atrium. This embodiment presets an evaluation index threshold, the specific value of which is set according to actual judgment needs; for example, it is set to 0.6. The fixed evaluation index of each pixel is compared with the preset evaluation index threshold. Pixels with a fixed evaluation index greater than the preset threshold are identified as target pixels. Target regions are generated based on these target pixels, i.e., regions formed by consecutive target pixels are used as target regions (i.e., connectivity analysis is performed on consecutive target pixels to obtain target regions), thus obtaining several target regions. Target regions represent chamber regions that remain unchanged in each frame of echocardiography, i.e., local regions that remain essentially unchanged during automatic delineation of the two ventricles.

[0039] The preset evaluation index threshold is used for initial screening of target pixels. Its specific value can be determined based on a test dataset containing multiple samples by statistically analyzing the distribution of fixed evaluation indices for each sample. For example, the Otsu method can be used to adaptively find the optimal segmentation threshold, or it can be set to a fixed value such as 0.6 based on experience.

[0040] It should be understood that the target area has a certain size. That is, in the application scenario provided in this embodiment, the area of ​​the target area cannot be too small. Therefore, this embodiment can preset a lower limit threshold for the area. If the area of ​​the target area is less than the lower limit threshold, it means that the target area is a noise area and it is discarded.

[0041] Step S2: Based on the characteristic changes of each target region in consecutive frame echocardiograms, select two ventricular regions from each target region.

[0042] The four chambers of the heart move differently during a heartbeat. The left and right ventricles pump blood into the systemic and pulmonary circulations, respectively, while the atria primarily store venous return blood. The ventricles need to pump blood at relatively high pressure to participate in the systemic or pulmonary circulation; therefore, compared to the atria, the overall range of motion of the ventricles is greater, which is reflected in echocardiography as a relatively larger change in the area of ​​the connected regions.

[0043] Existing technologies often provide a hard prediction result when performing automatic ventricular segmentation on echocardiograms, ignoring the image blurring caused by the dynamic feature changes in consecutive frames of echocardiograms. This may result in oversegmentation of other regions as ventricles in the obtained automatic biventricular delineation results, or undersegmentation of ventricular parts as other structures.

[0044] Because the ventricles and atria exhibit continuous periodic expansion and contraction in echocardiography, the area of ​​their corresponding connected regions also periodically increases and decreases. However, the changes in the connected regions undergoing micro-motion do not necessarily follow this pattern. Therefore, invalid fixed regions can be excluded based on the regularity of the connected region area. Thus, this step filters out the ventricular region from each target region based on the characteristic changes of each target region in consecutive frames of echocardiography. In an exemplary embodiment, such as... Figure 3 As shown, the following is a specific implementation process for step S2: Step S21: Determine the degree of deviation of the area change of the initial connected region corresponding to the target region in consecutive frame echocardiograms relative to the cardiac cycle.

[0045] The target region represents a definite local area within the chambers to be divided, that is, the area corresponding to when the chamber contracts to its minimum level. Because of the heart's motion, the areas of motion in echocardiography are not limited to the ventricles; other connected regions may also exhibit micro-motions in the acquired echocardiogram. In other words, the target region may also include areas of the atria or other regions that remain unchanged during cardiac motion.

[0046] Taking any target region as an example, for any frame of echocardiography, the initial connected component in which the target region is located in that frame of echocardiography is obtained as the initial connected component corresponding to the target region in that frame of echocardiography, thereby obtaining the initial connected component corresponding to the target region in each frame of echocardiography.

[0047] The area of ​​the initial connected region corresponding to the target region in each frame of echocardiography is obtained and arranged chronologically. Curve fitting (e.g., using least squares polynomial fitting) is then performed to obtain the area change curve of the initial connected region corresponding to the target region across consecutive echocardiogram frames. The area change curve is plotted with time on the horizontal axis (i.e., the frame number of the echocardiogram) and the area of ​​the connected region on the vertical axis. The area change curve represents the change pattern of the connected region by the number of pixels in the initial connected region corresponding to the target region in each frame of echocardiography.

[0048] The maximum and minimum values ​​in the area change curve are obtained. These extreme values ​​(i.e., the maximum and minimum values) serve as the turning points of the initial connected domain corresponding to the target region in its expansion and contraction states. Specifically, they represent the states where the chamber contracts to its minimum area and expands to its maximum area, thus representing the possible periodic characteristics of the target region. It should be understood that since the area of ​​the target region is less than or equal to the area of ​​its corresponding initial connected domain, the minimum value in the area change curve is greater than or equal to the area of ​​the corresponding target region.

[0049] The time interval between the maximum and minimum values ​​in the area change curve is obtained. Ideally, the time interval between the maximum and minimum values ​​is equal to half the length of a cardiac cycle. The greater the difference between the time interval between the maximum and minimum values ​​and the length of half a cardiac cycle, the higher the deviation of the area change of the initial connected region corresponding to the target region in consecutive frame echocardiograms relative to the cardiac cycle. The deviation degree is obtained by calculating the time difference between the time interval between the maximum and minimum values ​​in the area change curve and the length of half a cardiac cycle (the time difference is the absolute value of the difference in duration). The deviation degree is positively correlated with the time difference. In an exemplary embodiment, the deviation degree is obtained by the following calculation method: ; in, This represents the degree of deviation of the area change of the initial connected region corresponding to the k-th target region in consecutive echocardiogram frames relative to the cardiac cycle. The time interval between the maximum and minimum values ​​of the initial connected component corresponding to the k-th target region in the area change curve of consecutive echocardiogram frames is represented by T, which represents the number of frames of the echocardiogram, i.e., the length of a complete cardiac cycle.

[0050] Step S22: Based on the overall magnitude of the area change of the target region and the difference in area change in adjacent echocardiogram frames, combined with the degree of deviation, the probability that the target region belongs to the ventricular region is obtained.

[0051] The area of ​​the initial connected region corresponding to the target region in consecutive echocardiogram frames will vary to some extent. By determining the overall magnitude of the area change and the difference in area change in adjacent echocardiogram frames, and combining this with the obtained deviation, the probability that the target region belongs to the ventricular region can be determined. In an exemplary embodiment, such as... Figure 4 As shown, the following is a specific process for obtaining the possibilities: Step S221: Using the extreme points in the area change curve as segmentation points, divide the area change curve into several curve segments.

[0052] Identify the extreme points in the area change curve, including maximum and minimum points. Use these extreme points as segmentation points to divide the area change curve into several curve segments.

[0053] Since the changes in the heart chambers are typically a continuous cyclical process of expansion-contraction-expansion..., the area change curve corresponds to a complete cardiac cycle. Ideally, the number of curve segments obtained by dividing the area change curve should be as small as possible, and ideally, the maximum point should be the maximum point, and the minimum point should be the maximum point. Therefore, the more curve segments there are, the lower the regularity of the target region, the less the area change of the initial connected region corresponding to the target region conforms to the characteristics of the heart chambers, and the less the initial connected region corresponding to the target region belongs to the ventricular region. The regularity of the target region is inversely correlated with the number of curve segments.

[0054] In addition, this embodiment can determine the minimum number of ideal curve segments for segmenting the area change curve under ideal conditions. Specifically, if the first frame of echocardiography happens to be the moment of expansion or contraction, that is, the first frame of echocardiography corresponds to the maximum or minimum value position point in the area change curve, then the minimum number of ideal curve segments for segmenting the area change curve is 2; otherwise, the minimum number of ideal curve segments for segmenting the area change curve is 3.

[0055] Step S222: Fuse the differences in area change of the target region in adjacent echocardiogram frames to obtain the overall difference in area change.

[0056] For the area of ​​the initial connected region corresponding to the target region in consecutive echocardiogram frames, obtain the area change difference of the initial connected region corresponding to the target region in two adjacent echocardiogram frames. Specifically, the area change difference is the area of ​​the initial connected region corresponding to the target region in the later echocardiogram frame minus the area of ​​the initial connected region corresponding to the previous echocardiogram frame.

[0057] By iterating through all adjacent echocardiogram frames in a continuous sequence, the average difference in area change between adjacent frames of the initial connected region corresponding to the target region is calculated. The absolute value of this average is then taken as the overall difference in area change corresponding to the target region. For a complete cardiac cycle, the expansion and contraction of a cardiac ventricle is a continuous process, and the expansion and contraction processes are corresponding processes. That is, the amount of expansion and contraction of the cardiac ventricle remains essentially consistent in each cycle. Therefore, the closer the average difference in area change between adjacent frames of the initial connected region corresponding to the target region is to 0, the higher the regularity of the target region, the more the area change of the initial connected region corresponding to the target region conforms to the pulsating characteristics of the cardiac ventricle, and the more closely the initial connected region corresponding to the target region belongs to the ventricular region. Therefore, the regularity of the target region is inversely correlated with the overall difference in area change corresponding to the target region.

[0058] Step S223: Based on the degree of deviation, the number of curve segments, and the overall difference in area change, the regularity of the target area is obtained.

[0059] The greater the deviation of the area change of the initial connected region corresponding to the target region in consecutive echocardiogram frames from the cardiac cycle, the lower the regularity of the target region. Furthermore, the less the area change of the initial connected region corresponding to the target region conforms to the pulsating characteristics of the heart chambers, and the less likely the initial connected region corresponding to the target region belongs to the ventricular region. Therefore, the regularity of the target region is inversely correlated with the degree of deviation.

[0060] The regularity of the target region is determined by the deviation of its area change relative to the cardiac cycle from the area change curve of the initial connected region in consecutive echocardiogram frames, the number of curve segments into which the initial connected region corresponds to the target region in the consecutive echocardiogram frames, and the overall difference in the area change of the target region. Based on the above logical analysis, the following is a method for calculating the regularity of the target region: ; in, This indicates the degree of regularity in the k-th target region. This represents the number of curve segments into which the area change curve corresponding to the k-th target region is divided. This represents the minimum number of ideal curve segments required to divide the area change curve. Let represent the area of ​​the initial connected region corresponding to the k-th target region in the t-th frame of echocardiography. This represents the area of ​​the initial connected region corresponding to the k-th target region in the (t-1)-th frame of echocardiography.

[0061] This represents the difference between the number of curve segments in the area change curve corresponding to the k-th target region and the minimum ideal number of curve segments. Ideally, this represents the difference between the number of curve segments in the curve and the minimum ideal number of curve segments. equal , The larger the difference, the lower the regularity of the k-th target region. The denominator is set to T, which is used to... Normalization.

[0062] This represents the difference in area change of the initial connected component corresponding to the k-th target region in the echocardiograms of frame t-1 and frame t. This represents the overall difference in area change corresponding to the k-th target region.

[0063] Similarly, the regularity of each target region is obtained. The regularity reflects the degree of conformity between the area change of the initial connected domain corresponding to the target region and the characteristics of ventricular motion. The greater the regularity, the more likely the target region is to be a heart ventricular chamber.

[0064] Step S224: Based on the regularity and the overall magnitude of the area change, determine the probability that the target area belongs to the ventricular region.

[0065] Because the ventricles need to pump blood at relatively high pressure to participate in systemic or pulmonary circulation, their overall motion amplitude is greater than that of the atria. This is reflected in frame echocardiography as a larger overall amplitude of the area change of the initial connected domain corresponding to the target region. Therefore, the difference between the maximum and minimum values ​​in the area change curve corresponding to the target region is calculated as the overall amplitude of the area change. The larger the overall amplitude of the area change, the greater the probability that the initial connected domain corresponding to the target region belongs to the ventricle; that is, the greater the probability that the target region belongs to the ventricle. Thus, the probability that the target region belongs to the ventricle is positively correlated with the overall amplitude of the area change. Furthermore, the higher the regularity of the target region, the greater the probability that the initial connected domain corresponding to the target region belongs to the ventricle; that is, the greater the probability that the target region belongs to the ventricle. Therefore, the probability that the target region belongs to the ventricle is positively correlated with the regularity of the target region. Based on the above logic, the following is a specific method for calculating the probability that the target region belongs to the ventricle: ; in, This indicates the probability that the k-th target region belongs to the ventricular region. This represents the maximum value in the area change curve corresponding to the k-th target region. This represents the minimum value in the area change curve corresponding to the k-th target region.

[0066] Step S23: Based on probability, select two ventricular regions from the target region.

[0067] According to step S22, the probability that each target region belongs to the ventricular region is obtained. The higher the probability, the more likely the region corresponding to the target region in the initial connected domain is to belong to the ventricular region. Therefore, the two target regions with the highest probability are selected, and their corresponding initial connected domains are taken as the ventricular regions. The two target regions with the highest probability are taken as the fixed regions of the ventricular region.

[0068] To improve the accuracy of the screening, especially in scenarios with atypical dynamic characteristics under pathological conditions, this embodiment further introduces anatomical location prior constraints for verification after selecting the two most likely target regions. Specifically, in a standard apical four-chamber view, the two ventricles are usually located in the lower half of the image (closer to the ultrasound probe), while the two atria are located in the upper half. Therefore, it is determined whether the centroids of the two selected regions are both located in the preset lower half of the image. If they are, they are confirmed as ventricular regions; if not, the next candidate target region is selected sequentially from high to low probability scores and combined with the regions that have met the conditions for location constraint judgment, until two regions that both meet the location constraints are found as the final ventricular regions.

[0069] Step S3: For any ventricular region in the two ventricular regions, determine the initial edge of the ventricular region in each frame of echocardiography, and determine the transition path from each edge pixel on the initial edge to the center of the ventricular region.

[0070] During heartbeat, the ventricular region in an echocardiogram may appear blurred. The longer and smoother the grayscale transition band, the stronger the blurring.

[0071] For any ventricular region, an edge detection algorithm is used to obtain the edges of the ventricular region in each frame of echocardiography, which are defined as initial edges. It should be understood that the ventricular region is an independent region, and its initial edges form a closed contour.

[0072] As a specific implementation method, the edge detection algorithm can employ image edge detection methods known in the art. For example, a wavelet transform-based edge detection method can be used, which extracts the ventricular edge -1 while suppressing speckle noise in the ultrasound image through wavelet multi-scale analysis and fuzzy set processing. Alternatively, a method combining Markov random fields and level sets can be used to process ambiguous edges in the image using local statistics, thereby obtaining a continuous and smooth ventricular boundary.

[0073] It should be noted that the edge detection algorithms described above are all conventional techniques used in ultrasound medical image processing in this field. Those skilled in the art can choose appropriate algorithms to extract the initial edges of the ventricular region according to actual needs, as long as the edges of the ventricular region in each frame of echocardiography can be identified. The specific implementation methods will not be elaborated here.

[0074] For the initial edge of the ventricular region in any frame of echocardiography, the transition path from each edge pixel on the initial edge to the center of the ventricular region is obtained. The center of the ventricular region is obtained by: obtaining the mean of the abscissas of each pixel in the ventricular region as the abscissa of the center of the ventricular region, and obtaining the mean of the ordinates of each pixel in the ventricular region as the ordinate of the center of the ventricular region, thus obtaining the two-dimensional coordinates of the center of the ventricular region.

[0075] For any edge pixel on the initial edge, the line connecting that edge pixel to the center of the ventricular region is obtained, thus acquiring each pixel on the line in the echocardiogram. It should be understood that the number of pixels on the line cannot be zero, and the grayscale values ​​of the pixels on the line generally show a decreasing trend. Therefore, in order to determine the transition path of the edge pixel from the line, in this embodiment, the edge pixel is used as the starting position, and each pixel on the line is compared with a preset chamber threshold along the direction from the starting position to the center of the ventricular region. The preset chamber threshold represents the grayscale value within the ventricular chamber, indicating that the corresponding pixel belongs to the ventricular chamber and is no longer a pixel in the edge transition stage. Since the grayscale value within the ventricular chamber is usually small, its value can be set based on the average grayscale value inside the ventricular chambers statistically from a large number of samples. For example, within a normalized grayscale range of 0-1, it can be set to 0.1. Iterate through all pixels along the connecting line until the grayscale value first falls below a preset chamber threshold. Identify the pixel whose grayscale value first falls below the preset chamber threshold, and use the portion between the preceding pixel and the edge pixel on the connecting line as the transition path from that edge pixel to the center of the ventricular region. This yields the transition paths from each edge pixel on the initial edge to the center of the ventricular region.

[0076] Step S4: The reliability of each edge pixel is obtained from the number of pixels in the transition path and the gray value change range on the transition path.

[0077] For any edge pixel on the initial edge, obtain the number of pixels in the transition path of that edge pixel. The number of pixels represents the length of the transition path. The more pixels there are, the longer the transition path is, and the higher the blurriness of the edge where the edge pixel is located. In turn, the lower the reliability of the edge pixel is. The two are inversely correlated.

[0078] The grayscale value change amplitude along the transition path of the edge pixel is obtained. The grayscale value change amplitude characterizes the overall grayscale value change amplitude along the transition path. In an exemplary embodiment, the grayscale value of the first pixel along the transition path is subtracted from the grayscale value of the last pixel, and the resulting grayscale value difference is normalized using the sigmoid function. The normalized result is the grayscale value change amplitude along the transition path of the edge pixel. The grayscale value change amplitude characterizes the smoothness of the transition along the transition path. The smaller the grayscale value change amplitude, the smoother the transition, and the lower the reliability of the edge pixel; the two are positively correlated.

[0079] The reliability of an edge pixel is determined by the number of pixels in its transition path and the magnitude of grayscale value change along that path. Based on the above logical analysis, a specific method for calculating reliability is given below: ; in, This indicates the reliability of the j-th edge pixel. This represents the total number of pixels along the line connecting the j-th edge pixel to the center of the ventricular region. This represents the number of pixels in the transition path of the j-th edge pixel. This represents the magnitude of grayscale value change along the transition path of the j-th edge pixel. This represents the weighting coefficient. Negative correlation normalization of the number of pixels in the transition path of the j-th edge pixel. This is a preset value, which can be adjusted based on prior knowledge or by optimizing model performance on the validation set to focus on metrics that have a greater impact on the final result.

[0080] It should be understood that the number of pixels in the transition path is at least 1. If the number of pixels in the transition path is 1, it means that the transition path only contains the edge pixel itself. In this case, the blurriness of the edge pixel is very small, and the reliability of the edge pixel is directly set to 1.

[0081] The reliability of each edge pixel on the initial edge of the ventricular region in each frame of echocardiography is obtained using the above method. High reliability indicates that the position of the corresponding edge pixel is relatively accurate, while low reliability indicates that the position of the corresponding edge pixel may be affected by noise and is not very accurate.

[0082] Step S5: Using the reliability of each edge pixel as a weight, fit the edge of the ventricular region to obtain the final edge of the ventricular region, so as to outline the contour of the two ventricles and display it on the display interface of the ultrasound device.

[0083] For the initial connected component of any ventricular region in any frame of echocardiography, the reliability of each edge pixel is used as a weight, and a weighted fitting method is employed to fit the edge of the ventricular region in that frame of echocardiography, resulting in the final edge of the ventricular region in that frame of echocardiography. The final edge is smoother, thus achieving automatic delineation of the dual ventricular contours, which is then displayed on the ultrasound device's display interface.

[0084] Based on the weights of each edge pixel, a weighted fitting method is used to fit the edges of the image region. This is a conventional technique, and a brief explanation follows. It should be understood that for the initial connected component of any ventricular region in any frame of echocardiography, this embodiment can normalize the reliability of each edge pixel by weighting it. Specifically, the sum of the reliability of each edge pixel is calculated, and then the ratio of the reliability of each edge pixel to the sum of its reliability is calculated. The result is used as the weight of each edge pixel.

[0085] Record the coordinates of each edge pixel on the initial edge of the initial connected component. Each edge pixel is associated with a corresponding weight, that is, the coordinates of the j-th edge pixel are... The weight is There are J edge pixels on the initial edge.

[0086] This embodiment uses a weighted parametric B-spline curve to fit the edge. Specifically, it uses each edge pixel on the initial edge. As a control point, and its corresponding reliability level As weight By solving the standard weighted B-spline curve fitting problem, a smooth B-spline curve is obtained that best approximates the high-weight (high-reliability) control points while ignoring the influence of low-weight (low-reliability) control points. This curve represents the final edge of the ventricular region, thus achieving accurate and robust automatic delineation of the ventricular contour.

[0087] This embodiment also provides an automatic biventricular delineation system based on echocardiography, including: a memory and a processor; the memory is connected to the processor, and the memory is used to store program instructions; the processor is used to implement the steps in the above-described automatic biventricular delineation method embodiment based on echocardiography when the program instructions are executed.

[0088] In one exemplary embodiment, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the embodiments of the automatic biventricular delineation method based on echocardiography.

[0089] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0090] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. An automatic biventricular delineation method based on echocardiography, characterized in that, include: Acquire consecutive frames of echocardiograms covering at least one cardiac cycle, acquired by an ultrasound probe; Based on the dynamic characteristics of the same pixel in consecutive frames of echocardiography, several target regions in the echocardiogram are obtained; wherein, the dynamic characteristics characterize the motion state or grayscale change pattern of the pixel in the time dimension. Based on the characteristic changes of each target region in consecutive frames of echocardiography, two ventricular regions are selected from each target region. For each of the two ventricular regions, the initial edge of the ventricular region in each frame of echocardiography is determined, and the transition path from each edge pixel on the initial edge to the center of the ventricular region is determined; wherein, the center of the ventricular region is the geometric centroid of the ventricular region; The reliability of each edge pixel is obtained from the number of pixels in the transition path and the magnitude of grayscale value change on the transition path. The reliability of each edge pixel is used as a weight to fit the edge of the ventricular region, thereby obtaining the final edge of the ventricular region for outlining the dual ventricular contour, which is then displayed on the ultrasound device's display interface.

2. The method for automatic biventricular delineation based on echocardiography as described in claim 1, characterized in that, The process of acquiring the target region includes: Determine the degree of fluctuation in the area of ​​the initial connected region where the same pixel is located in each frame of echocardiography; the initial connected region is obtained by threshold segmentation of each frame of echocardiography. Determine the mean gray value and the range of gray values ​​of the same pixel in each frame of echocardiography; Based on the fluctuation level, the mean gray value, and the gray value range, a fixed evaluation index is obtained for each pixel; the fixed evaluation index is positively correlated with the fluctuation level and negatively correlated with the mean gray value and the gray value range. The target region is obtained based on the fixed evaluation index of each pixel.

3. The method for automatic biventricular delineation based on echocardiography as described in claim 2, characterized in that, The target region is obtained based on the fixed evaluation index of each pixel, including: Adjacent target pixels constitute the target region; the target pixels are pixels with a fixed evaluation index greater than a preset evaluation index threshold.

4. The method for automatic biventricular delineation based on echocardiography as described in claim 2, characterized in that, The step of selecting two ventricular regions from each target region based on the characteristic changes of each target region in consecutive frames of echocardiography includes: Determine the degree of deviation of the area change of the initial connected region corresponding to the target region in consecutive frames of echocardiography relative to the cardiac cycle; Based on the overall magnitude of the area change of the initial connected domain corresponding to the target region and the difference in area change in adjacent echocardiogram frames, combined with the degree of deviation, the probability that the target region belongs to the ventricular region is obtained. Based on the stated possibilities, two ventricular regions are selected from the target region.

5. The method for automatic biventricular delineation based on echocardiography as described in claim 4, characterized in that, The process of obtaining the degree of deviation includes: Determine the area change curve of the initial connected component corresponding to the target region for consecutive frames of echocardiography; Obtain the time interval between the maximum and minimum values ​​in the area change curve; The degree of deviation is determined by the difference in duration between the time interval and half a cardiac cycle.

6. The method for automatic biventricular delineation based on echocardiography as described in claim 5, characterized in that, The process of obtaining the possibility includes: Using the extreme points in the area change curve as segmentation points, the area change curve is divided into several curve segments; the extreme points include maximum points and minimum points; By fusing the differences in area change of the target region in adjacent echocardiogram frames, the overall difference in area change is obtained; The regularity of the target region is obtained based on the degree of deviation, the number of curve segments, and the overall difference in area change; the regularity is inversely correlated with the degree of deviation, the number of curve segments, and the overall difference in area change. Based on the degree of regularity and the overall magnitude of area change, the probability that the target region belongs to the ventricular region is obtained; the probability is positively correlated with both the degree of regularity and the overall magnitude of area change, and the overall magnitude of area change is the difference between the maximum and minimum values.

7. The method for automatic biventricular delineation based on echocardiography as described in claim 1, characterized in that, The process of obtaining the transition path includes: Determine the line connecting the edge pixel to the center of the ventricular region; Starting from the edge pixel, traverse each pixel on the connecting line until the gray value is lower than the preset chamber threshold for the first time to obtain the transition path.

8. The method for automatic biventricular delineation based on echocardiography as described in claim 1, characterized in that, The grayscale value change range is obtained by the difference in grayscale value between the first pixel and the last pixel on the transition path; The process of obtaining the reliability level includes: The reliability of edge pixels is obtained based on the number of pixels in the transition path and the magnitude of grayscale value change; the reliability is inversely correlated with the number of pixels in the transition path and positively correlated with the magnitude of grayscale value change.

9. The method for automatic biventricular delineation based on echocardiography as described in claim 1, characterized in that, The step of fitting the edge of the ventricular region with the reliability of each edge pixel as a weight to obtain the final edge of the ventricular region includes: Using the reliability of each edge pixel as a weight, a weighted fitting method is used to fit the edge of the ventricular region to obtain the final edge of the ventricular region.

10. An automated biventricular delineation system based on echocardiography, characterized in that it comprises: Memory and processor; The memory is connected to the processor; The memory is used to store program instructions; The processor is configured to implement, when program instructions are executed, the automatic biventricular delineation method based on echocardiography as described in any one of claims 1-9.

Images