Ultrasonic image analysis method and device based on image processing technology, and electronic equipment

Abstract

The invention provides an ultrasonic image analysis method and device based on an image processing technology and electronic equipment. The method comprises the following steps: denoising an ultrasonic image by using an image denoising technology; positioning a mitral valve and a regurgitation area in the ultrasonic image; key structure parameters in the ultrasonic image are quantized; and an automatic analysis system is established so as to realize rapid processing of ultrasonic images and extraction of parameter information related to functional mitral valve regurgitation. By implementing the method, manual measurement operation of doctors can be reduced, diagnosis of the doctors is assisted, and contributions are made to further research on pathological mechanisms of the doctors.

Description

Technical Field

[0001] The present invention relates to the field of ultrasonic image processing, and in particular to an ultrasonic image analysis, device and electronic equipment based on image processing technology. Background Art

[0002] Functional mitral regurgitation is a common pathological phenomenon of heart disease, usually caused by high pressure during ventricular contraction or mitral regurgitation. Accurate diagnosis of functional mitral regurgitation is of great significance for the treatment and prognosis of patients. Traditional diagnostic methods mainly rely on the doctor's experience and visual inspection, which are easily affected by subjective factors and have certain limitations.

[0003] With the continuous development of ultrasound technology, ultrasound images have become a common means of diagnosing functional mitral regurgitation. Ultrasound images provide intuitive observation of cardiac structure and function in a non-invasive and real-time manner. However, due to the characteristics of ultrasound images, such as noise, artifacts, and resolution limitations, it is still a challenging task to directly obtain parameter information related to functional mitral regurgitation from ultrasound images and perform accurate analysis.

[0004] In response to the above problems, some related studies have been conducted on ultrasound image processing methods and systems to improve the diagnostic accuracy and efficiency of functional mitral regurgitation. Among them, some methods use image segmentation algorithms based on machine learning, such as the Unet network segmentation method, which can automatically segment key structures such as the mitral valve from ultrasound images. In addition, optical flow calculation positioning technology is used to analyze the motion changes of the image centerline, which can help determine the location of functional mitral regurgitation.

[0005] However, there are still some problems with the existing ultrasound image processing methods. First, in order to address the noise and artifact problems of ultrasound images, the effect of noise reduction processing needs to be further improved to provide clearer image quality. Secondly, accurate segmentation and positioning of ultrasound images are essential for subsequent data extraction and analysis. In addition, the development of automated analysis systems can provide fast and accurate processing results and reduce the manual measurement work of doctors, which has important clinical significance. Summary of the invention

[0006] The present invention provides an ultrasonic image processing method and system for assisting physicians in diagnosing functional mitral regurgitation. The method is based on ultrasonic image processing technology, and extracts parameter information related to functional mitral regurgitation by performing a series of processing and analysis on ultrasonic images, thereby providing doctors with reliable diagnostic assistance and supporting the study of pathological mechanisms.

[0007] The solution of the embodiment of the present invention is implemented as follows:

[0008] In a first aspect, an embodiment of the present invention provides an ultrasonic image analysis method based on image processing technology, the method comprising:

[0009] Use image denoising technology to denoise the ultrasound image;

[0010] Locate the mitral valve and regurgitant areas on ultrasound images;

[0011] Quantification of key structural parameters in ultrasound images;

[0012] An automated analysis system was established to rapidly process ultrasound images and extract parameter information related to functional mitral regurgitation.

[0013] In a second aspect, an embodiment of the present invention provides an ultrasonic image analysis device based on image processing technology, the device comprising:

[0014] An image acquisition module, used for acquiring ultrasound images;

[0015] An image processing module, used for denoising, segmenting and motion calculation of ultrasound images;

[0016] An analysis module for extracting parameter information related to functional mitral regurgitation;

[0017] Display module, used to display analysis results and generate reports.

[0018] In a third aspect, an embodiment of the present invention provides a computer program product, the product comprising:

[0019] receiving and processing ultrasound image data;

[0020] Apply denoising AutoEncoder image denoising technology to denoise image data;

[0021] Run the Unet network structure segmentation technology to accurately locate the mitral valve and regurgitation area in the ultrasound image;

[0022] Use optical flow calculation technology to quantify key structural parameters; generate analysis result reports or provide diagnostic suggestions.

[0023] The embodiments of the present invention have the following beneficial effects:

[0024] This paper provides a method and system based on ultrasound image processing, which uses advanced image processing technology and automated analysis systems to quickly and accurately extract parameter information related to functional mitral regurgitation, providing important auxiliary support for doctors' diagnostic decisions. At the same time, this method also helps to deeply understand the pathological mechanism of functional mitral regurgitation, thereby promoting the research and treatment of related diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a summary diagram of Embodiment 1 of the present invention.

[0026] Figure 2 This is a network structure diagram of the AutoEncoder provided in Example 1 of the present invention.

[0027] Figure 3 A schematic diagram of the UNet network structure provided in Example 1 of the present invention.

[0028] Figure 4 This is a schematic diagram of the optical flow algorithm structure provided in Example 1 of the present invention.

[0029] Figure 5 This is a structural diagram of the automatic quantitative analysis system of mitral valve transthoracic ultrasound provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0030] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the embodiments provided herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the embodiments provided below are partial embodiments for implementing the present invention, rather than providing all embodiments for implementing the present invention. In the absence of conflict, the technical solutions recorded in the embodiments of the present invention can be implemented in any combination.

[0031] Before further describing the present invention in detail, Figure 2 The structural diagram of the mitral valve transthoracic ultrasound automatic quantitative analysis system is shown to illustrate the nouns and terms designed in the embodiments of the present invention. The nouns and terms involved in the embodiments of the present invention are subject to the following explanations.

[0032] Mitral regurgitation: It is a common heart valve disease in the world. The incidence rate in the general population in Western countries is 1.7%. The incidence rate increases with age, accounting for about 10% of people over 75 years old. Functional mitral regurgitation (FMR) accounts for about 70% of MR and about 2.5% of the total population. In people over 75 years old, its incidence rate can be as high as 13%. It is an independent driving factor for poor prognosis in patients with heart failure (HF). The mortality rate of untreated symptomatic severe MR patients is 30% to 50% within 5 years, and the readmission rate due to decompensated HF is as high as 90%.

[0033] FMR: It is caused by global or segmental left ventricular (LV) dysfunction or left atrial (LA) dilatation, which leads to mitral ring dilatation, papillary muscle displacement, mitral valve leaflet tethering and leaflet remodeling. FMR does not have organic lesions of the leaflets, chordae tendineae and papillary muscles, and usually includes atrial FMR (AFMR) and ventricular FMR (VFMR).

[0034] AFMR: related to LA dilation, common in patients with long-term atrial fibrillation (AF) and heart failure with preserved ejection fraction. LA dilation pulls the annulus and leaflet commissure upward, flattens the anterior leaflet movement and restricts leaflet apposition.

[0035] VFMR occurs due to global or segmental LV deformation and dysfunction, which causes apical and posterolateral displacement of the papillary muscles, leading to tethering of the valve leaflets and resulting in malapposition.

[0036] In order to solve the problem of fuzzy parameter information related to functional mitral regurgitation obtained directly from ultrasound images, the inventors found that the following method can be used.

[0037] 1) Combining multimodal information: In addition to ultrasound images, other imaging techniques (such as MRI, CT, etc.) can also be used to obtain more comprehensive cardiac structural information. By registering and fusing image data from different modalities, the accuracy and reliability of parameters related to functional mitral regurgitation can be improved.

[0038] 2) Introducing machine learning methods: Using machine learning algorithms, models can be trained using a large amount of labeled image data to analyze and extract relevant features of functional mitral regurgitation. For example, the convolutional neural network (CNN) architecture in deep learning can be used to learn and automatically extract features from images, thereby reducing the interference of human subjective factors on parameter extraction.

[0039] 3) Establish database and reference standards: Collect and establish a large-scale ultrasound image database, and establish reference standards in conjunction with the professional knowledge of clinicians. By analyzing and statistically analyzing the data of known cases, a correlation model between functional mitral regurgitation parameters and the severity of the disease can be established, thereby achieving prediction and evaluation of unknown cases.

[0040] 4) Further optimize ultrasound image acquisition and processing technology: improve the performance of ultrasound equipment, optimize image acquisition parameters and imaging modes, and reduce the impact of noise and artifacts. At the same time, continuously improve ultrasound image processing algorithms to improve the accuracy of positioning and segmentation of functional mitral regurgitation, as well as the accuracy of parameter extraction. In view of the above problems, an embodiment of the present invention provides an ultrasonic image analysis method and device based on image processing technology.

[0041] Embodiment 1

[0042] The present embodiment provides an ultrasound image analysis method based on image processing technology, the method comprising: using image denoising technology to denoise the ultrasound image; locating the mitral valve and regurgitation area in the ultrasound image; quantifying key structural parameters in the ultrasound image; and establishing an automated analysis system to quickly process ultrasound images and extract parameter information related to functional mitral regurgitation.

[0043] The method of using image denoising technology to denoise ultrasound images includes but is not limited to the following implementation methods:

[0044] 1) Statistical filter: Use statistical methods, such as mean filter, median filter, etc., to smooth the pixels of ultrasound images and reduce the impact of noise in the image. These methods perform noise estimation and filtering operations based on the local statistical characteristics of pixel values ​​to achieve ultrasound image denoising.

[0045] 2) Adaptive filter: Use adaptive filters, such as adaptive median filter, adaptive Gaussian filter, etc., to filter according to the local features of the image. These filters can automatically select appropriate filter parameters according to the changes in pixel values, and retain more detail information while smoothing the image, thereby achieving the purpose of ultrasound image denoising.

[0046] 3) Wavelet denoising: The ultrasound image is converted to the wavelet domain, and the sparse nature of the wavelet coefficients is used for denoising. The wavelet coefficients are thresholded to filter out the noise component, and then the denoised wavelet coefficients are inversely transformed back to the image domain to obtain a clear ultrasound image.

[0047] 4) Total Variation-based method: Total variation regularization is a commonly used image denoising method. By minimizing the total variation of the image, the noise in the ultrasound image can be removed while maintaining the edge information of the image. This method can effectively balance noise suppression and detail preservation during the denoising process.

[0048] 5) Deep learning-based methods: Using deep learning models such as deep convolutional neural networks (CNNs), the characteristics and noise distribution of ultrasound images can be learned through large-scale training data sets, thereby achieving more accurate denoising effects. An end-to-end ultrasound image denoising model can be designed, or a pre-trained model can be used for transfer learning.

[0049] 6) Non-local Means Method: The non-local means algorithm is based on the principle of similarity and performs noise estimation and filtering by comparing the similarity of a pixel with its surrounding pixels. This method uses the structural information of the image itself and can reduce the noise level while maintaining the image details.

[0050] In one possible implementation, the image enhancement of the ultrasound image includes: denoising AutoEncoder image denoising technology adopts an autoencoder network structure to remove noise from the ultrasound image and improve image quality and clarity by learning the mapping relationship between input and output.

[0051] Specifically, Figure 2 The AutoEncoder network structure diagram is shown as an example.

[0052] based on Figure 2 The denoising function structure of the ultrasonic image processing part in the first embodiment is described. However, the denoising function structure of the ultrasonic image processing part may not be the same as Figure 3 Same functional structure.

[0053] Ultrasound images are affected by strong speckle noise, which makes the imaging effect poor. AutoEncoder is a powerful method of unsupervised training mapping function, which can extract the most significant features of the data through the hidden layer of lower latitude while minimizing the reconstruction error between the encoding layer and the data layer, thereby achieving the purpose of denoising.

[0054] It is proposed to use denoising AutoEncoder technology as image preprocessing technology to achieve denoising of transthoracic ultrasound images.

[0055] AutoEncoder can be seen as three parts: Encoder, Latent, and Decoder.

[0056] The encoder learns a spatial representation of the implicit features of the input data; the decoder reconstructs the original input data from the learned low-dimensional features.

[0057] The encoder h encodes the original input X to obtain the feature X', X' = h(X); the decoder f decodes the feature X' to generate Y, that is, Y = f(X) = f(h(X)). The closer X and Y are, the better the extracted X' feature represents the row.

[0058] The method for locating the mitral valve and the regurgitation area in the ultrasound image includes but is not limited to the following implementations:

[0059] 1) Morphological-based methods: Use morphological operations in image processing, such as corrosion, dilation, opening and closing operations, and edge detection algorithms, such as Canny edge detection, to extract edges and specific shapes in ultrasound images. Positioning can be achieved by analyzing the morphological features of the mitral valve and regurgitation area, such as size, shape, and position.

[0060] 2) Threshold segmentation method: The ultrasound image is segmented by setting an appropriate brightness threshold to separate the mitral valve and regurgitation area from other tissues or structures. Simple threshold segmentation methods such as global threshold method or adaptive threshold method can be used, or more complex segmentation algorithms such as clustering or watershed algorithms can be used.

[0061] 3) Feature extraction-based methods: Identify and locate the mitral valve and regurgitation area by extracting features from ultrasound images, such as texture, shape, color, etc. Machine learning algorithms such as support vector machines (SVM), random forests, or deep learning models (such as convolutional neural networks) can be used for feature extraction and classification tasks to achieve positioning.

[0062] 4) Template matching-based method: Use the predefined mitral valve and regurgitation area shapes as templates and match them with the ultrasound image to achieve positioning. Template matching algorithms, such as normalized cross correlation or phase correlation, can be used to find the best matching position.

[0063] 5) Region growing-based method: select one or more seed points, perform region growing in the ultrasound image, expand the region by judging the similarity and connectivity between pixels, and finally obtain the mitral valve and regurgitation regions. Positioning and segmentation can be performed based on the specific implementation of the region growing algorithm, such as threshold-based, gradient-based, etc.

[0064] In one possible implementation, the Unet network structure segmentation technology adopts an encoder-decoder structure, and the use of the Unet network structure to perform binary segmentation on the ultrasound image includes: using the Unet network to learn the semantic information of the image, and accurately segmenting the mitral valve and regurgitation areas in the ultrasound image based on the semantic information.

[0065] Specifically, Figure 3 The schematic diagram of the UNet network structure is shown as an example.

[0066] based on Figure 3 The ultrasonic image processing part segmentation function structure in the first embodiment is described. However, the ultrasonic image processing part segmentation function structure may not be the same as Figure 4 Same functional structure.

[0067] After obtaining the transthoracic ultrasound images that have undergone preprocessing such as data desensitization and denoising, it is planned to use UNet, an image segmentation algorithm that has achieved wide success in medical image processing, to perform pixel-level recognition of the upper and lower leaflets of the mitral valve, thereby realizing subsequent intelligent quantitative analysis of the mitral valve.

[0068] UNet is a symmetrical network structure that looks like the English letter U, so it is called UNet. Overall, UNet is an Encoder-Decoder structure (the same as FCN). The first half is feature extraction, and the second half is upsampling. The network structure is as follows: Figure 4 shown.

[0069] Encoder: The left half consists of two 3x3 convolutional layers (ReLU) + a 2x2 maxpooling layer to form a downsampling module;

[0070] Decoder: The right half consists of an upsampling convolution layer + feature concat + two 3x3 convolution layers (ReLU) to form an upsampling module.

[0071] The encoder consists of convolution operations and downsampling operations. The convolution structure used is unified as a 3x3 convolution kernel, padding = 0, striding = 1. There is no padding, so the H and W of the feature map become smaller after each convolution. When skipping, pay attention to the dimension of the feature map.

[0072] The decoder is used to restore the original resolution of the feature map. In addition to convolution, the key steps of this process are upsampling and skip-layer connection. Upsampling is usually implemented by transposed convolution and interpolation. Among the interpolation implementation methods, bilinear interpolation has better overall performance and is more common.

[0073] In order to achieve good network performance, skip-layer connections are essential. The skip-layer connections in UNet fuse the underlying position information with the deep semantic information through splicing. It should be noted that in FCN, feature maps are fused in an additive manner.

[0074] U-Net segmentation was originally used for the segmentation of biomedical images. Until now, a large part of the segmentation networks for medical images use U-Net as the backbone of the network. Therefore, we use this segmentation method here to identify the upper and lower valves of the mitral valve.

[0075] Methods for quantifying key structural parameters in ultrasound images include but are not limited to the following implementations:

[0076] 1) Shape analysis-based methods: Quantitatively describe key structural parameters by analyzing the shape characteristics of the structure, such as area, perimeter, diameter, aspect ratio, etc. Shape descriptors, such as Hu moments, Zernike moments, or Fourier descriptors, can be used to represent the shape of the structure and calculate the corresponding eigenvalues ​​or indices.

[0077] 2) Texture analysis-based methods: Quantitatively describe key structural parameters by analyzing the texture features of the structure, such as gray-level co-occurrence matrix (GLCM), gray-level difference matrix (GLDM) or local binary pattern (LBP). These texture features can provide information about the internal organization and texture of the structure, such as granularity, roughness or uniformity.

[0078] 3) Intensity analysis-based methods: quantify key structural parameters by analyzing the image intensity information of the structure, such as average gray value, gray histogram, gradient, etc. Statistical features such as mean, variance, kurtosis and skewness can be used to describe the intensity distribution characteristics of the structure.

[0079] 4) Edge detection-based methods: quantify key structural parameters by detecting the edges or contours of the structure. Edge detection algorithms such as Canny edge detection, Sobel operator or Laplace operator can be used to extract the edge information of the structure and calculate related parameters such as edge length, edge density, etc.

[0080] 5) Machine learning-based methods: By using machine learning algorithms such as support vector machines (SVM), random forests, or deep learning models (such as convolutional neural networks), features related to key structural parameters can be learned and predicted. Image feature extraction algorithms such as local binary patterns, HOG (histogram of oriented gradients), etc. can be used to extract image features and use them as input to machine learning models for training and prediction.

[0081] In one possible implementation mode, the use of optical flow calculation technology to analyze structural parameters in ultrasound images includes: calculating motion information of key structural parameters in ultrasound images based on the correspondence between adjacent frames in the image sequence.

[0082] Figure 4 The schematic diagram of the optical flow algorithm structure is shown as an example.

[0083] based on Figure 4 The ultrasonic image processing part positioning function structure in the first embodiment is described. However, the ultrasonic image processing part positioning function structure may not be the same as Figure 5 Same functional structure.

[0084] Another key to achieving automatic quantitative analysis of the mitral valve lies in the measurement of the mitral valve movement velocity and the accurate positioning of key structures such as the upper and lower valve leaflets and the valve ring.

[0085] It is planned to use the optical flow algorithm combined with the image segmentation results to calculate the average swing speed of the upper and lower leaflets of the mitral valve. At the same time, based on the characteristics that the mitral valve tip moves faster while the position of the valve ring is relatively fixed, these key positions are located.

[0086] The basic principle of the optical flow method for detecting moving objects is to assign a velocity vector to each pixel in the image, which forms an image motion field. At a specific moment of motion, the points on the image correspond one-to-one to the points on the three-dimensional object. This correspondence can be obtained from the projection relationship. According to the velocity vector characteristics of each pixel, the image can be dynamically analyzed.

[0087] If there is no moving object in the image, the optical flow vector changes continuously in the entire image area. When there is a moving object in the image, there is relative motion between the target and the image background, and the velocity vector formed by the moving object must be different from the velocity vector of the neighborhood background, so the moving object and its position can be detected.

[0088] The main problem with using optical flow methods for moving object detection is that most optical flow methods are time-consuming to calculate, and their real-time and practicality are poor. However, the advantage of optical flow methods is that they not only carry the motion information of moving objects, but also carry rich information about the three-dimensional structure of the scene. It can detect moving objects without knowing any information about the scene.

[0089] The movement speed of the leaflet structure is calculated by the optical flow algorithm, and the fastest moving cusp and the slowest moving annulus structure are located based on this. On this basis, data such as the leaflet diameter, annular diameter, and upper and lower annular sections can be calculated.

[0090] Directly using optical flow technology will focus more on the anterior leaflet and chordae tendineae with a larger range of motion, while ignoring the posterior leaflet, thus affecting the calibration and measurement of parameters. By combining optical flow technology with image segmentation, the range of motion of the anterior and posterior leaflets and the annulus can be more accurately identified, so this requires the comprehensive processing of optical flow technology and image segmentation technology.

[0091] Methods for establishing an automated analysis system include but are not limited to the following implementation methods:

[0092] Single-layer architecture: This is the simplest form of architecture, where all functional modules are on one level. Input data is processed by the preprocessing module, then passed to the analysis module, and finally outputs the results through the output module. This architecture is suitable for simple analysis tasks and systems with relatively simple logic.

[0093] Layered architecture: The system is divided into multiple layers according to its functions, and each layer is responsible for different functions. Common layers include data access layer, preprocessing layer, analysis layer and output layer. Data enters the system from the access layer, undergoes multiple processing and analysis, and finally outputs the results. The layered architecture makes the system more flexible and scalable, and modules at different levels can be developed and updated independently.

[0094] Client-server architecture: The system is divided into two parts: client and server. The client is responsible for the user interface and interaction, sending requests to the server and receiving results. The server is responsible for data processing and analysis, and returns the results to the client. This architecture allows the system to support multiple users and remote access, and the server can perform distributed computing and resource management.

[0095] In one possible implementation, the automated analysis system includes: performing image enhancement, denoising and standardization on ultrasound images; performing binary segmentation on ultrasound images using a Unet network structure to extract the mitral valve and regurgitation areas; using optical flow calculation technology to analyze structural parameters in ultrasound images, such as flow velocity and flow; and generating a report or providing evaluation suggestions based on the analysis results.

[0096] Specifically, Figure 5 The structure diagram of the automatic quantitative analysis system of mitral valve transthoracic ultrasound is shown as an example.

[0097] based on Figure 5 The structure of the automatic quantitative analysis system of mitral valve transthoracic ultrasound in the first embodiment is described. It is understandable that the structure of the automatic quantitative analysis system of mitral valve transthoracic ultrasound may not be limited to the Figure 2 Same structure.

[0098] The structure of the mitral valve transthoracic ultrasound automatic quantitative analysis system includes the entire system operation logic structure.

[0099] The system includes data cleaning and calibration (preliminary work of system input), ultrasound image processing (internal operating function of the system), mitral valve quantitative analysis (internal operating function of the system), and system development (latter work of system output).

[0100] Data cleaning and calibration were performed by collecting transthoracic ultrasound image data of mitral valves from 50 patients, manually selecting video frames that can clearly see quantitative indicators such as valve leaflets, and calibrating them using relevant calibration software, such as Labelme calibration software.

[0101] Ultrasonic image processing is performed through image denoising, image segmentation, and optical flow calculation and positioning, but not limited to such technologies, such as denoising AutoEncoder image denoising and other technologies.

[0102] Mitral valve quantitative analysis calculates quantitative indicators through relevant medical mathematical formulas, such as the calculation formula for the change rate of the valve ring diameter.

[0103] System development first designs a visualization method for diagnostic indicators, then software development, and finally software testing to form a visual software interface for human-computer interaction, such as UI system interface and other programs.

[0104] Embodiment 2

[0105] An ultrasonic image analysis device based on image processing technology, the device comprising: an image acquisition module for acquiring ultrasonic images; an image processing module for performing denoising, segmentation and motion calculation on ultrasonic images; an analysis module for extracting parameter information related to functional mitral regurgitation; and a display module for displaying analysis results and generating reports.

Claims

1. An ultrasonic image analysis method based on image processing technology, characterized in that: The method includes: using image denoising technology to denoise ultrasound images; locating the mitral valve and regurgitation area in the ultrasound image; quantifying key structural parameters in the ultrasound image; and establishing an automated analysis system to quickly process ultrasound images and extract parameter information related to functional mitral regurgitation.

2. The method according to claim 1, characterized in that: The automated analysis system includes: performing image enhancement, denoising and standardization on ultrasound images; performing binary segmentation on ultrasound images using a Unet network structure to extract mitral valve and regurgitation areas; using optical flow calculation technology to analyze structural parameters in ultrasound images, such as flow velocity and flow; and generating a report or providing evaluation suggestions based on the analysis results.

3. The method according to claim 2, characterized in that The image enhancement of the ultrasonic image includes: denoisingAutoEncoder image denoising technology adopts an autoencoder network structure to remove noise in the ultrasonic image and improve image quality and clarity by learning the mapping relationship between input and output.

4. The method according to claim 2, characterized in that: The Unet network structure segmentation technology adopts an encoder-decoder structure, and the use of the Unet network structure to perform binary segmentation on the ultrasound image includes: using the Unet network to learn the semantic information of the image, and accurately segmenting the mitral valve and regurgitation areas in the ultrasound image according to the semantic information.

5. The method according to claim 2, characterized in that: The use of optical flow calculation technology to analyze the structural parameters in the ultrasound image includes: calculating the motion information of the key structural parameters in the ultrasound image according to the corresponding relationship between adjacent frames in the image sequence.

6. An ultrasonic image analysis device based on image processing technology, characterized in that: The device comprises: an image acquisition module for acquiring ultrasound images; an image processing module for performing denoising, segmentation and motion calculation on ultrasound images; an analysis module for extracting parameter information related to functional mitral regurgitation; and a display module for displaying analysis results and generating reports.

7. A computer program product for executing an ultrasonic image analysis method based on image processing technology, characterized in that: The product includes: receiving and processing ultrasound image data; applying denoisingAutoEncoder image denoising technology to denoise image data; running Unet network structure segmentation technology to accurately locate the mitral valve and regurgitation area in the ultrasound image; using optical flow calculation technology to quantify key structural parameters; generating analysis result reports or providing diagnostic suggestions.

8. An electronic device, characterized in that: The method comprises a processor and a memory, wherein a computer program is stored in the memory, and the processor executes the computer program to implement the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by at least one processor, the method according to any one of claims 1 to 7 is implemented.

Images