Blood fat monitoring system based on blood vessel ultrasonic image

Abstract

The invention discloses a blood fat monitoring system based on a vascular ultrasound image, and belongs to the technical field of medical image recognition. The problems that an existing system cannot automatically recognize the blood vessel structure and evaluate the blood fat level and cannot ensure the stability and real-time performance of data transmission are solved, the intima-media membrane and lumen contours in the blood vessel ultrasonic image are automatically recognized in a short time through the convolutional neural network model, multi-dimensional features are extracted from the intima-media membrane and lumen contours, and the accuracy of the blood vessel ultrasonic image is improved. The identification precision and the diagnosis accuracy are improved; a multi-node gateway communication channel is arranged, and an optimal gateway is selected for data feedback in combination with a network distance threshold value, so that a blood fat prediction result is fed back to a medical interaction terminal in real time, it is ensured that early warning is triggered in time when the blood fat level exceeds a normal range, and the situation that a certain gateway breaks down or the communication distance exceeds the threshold value is prevented. And the system can be automatically switched to other available gateways, so that the stability and the reliability of the data in the transmission process are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image recognition, and in particular to a blood lipid monitoring system based on vascular ultrasound images. Background Art

[0002] In medical diagnosis, blood lipid monitoring is a common method used to assess cardiovascular health. Traditional blood lipid testing methods rely primarily on biochemical analysis of blood samples. However, due to the time delay and inconvenience of blood collection, real-time monitoring of patients' blood lipid levels is impossible.

[0003] In recent years, vascular ultrasound imaging has been widely used in the diagnosis of vascular diseases. Ultrasound images can be used to visualize the structure of blood vessel walls and blood flow. While existing technologies can provide two-dimensional or three-dimensional images of blood vessels, they lack effective intelligent analysis systems, making it impossible to automatically identify vascular structure and assess blood lipid levels. Furthermore, the relatively closed environment of hospital departments often leads to data transmission interruptions due to network communication failures. This makes the stability and real-time nature of data transmission a critical issue that needs to be addressed in practical applications.

[0004] Therefore, the existing needs are not met, so we propose a blood lipid monitoring system based on vascular ultrasound images. Summary of the Invention

[0005] The purpose of the present invention is to provide a blood lipid monitoring system based on vascular ultrasound images, which automatically identifies the vascular intima and lumen contours in vascular ultrasound images in a short time through a convolutional neural network model, and extracts multi-dimensional features from them, thereby improving the accuracy of recognition and diagnosis; by setting up a multi-node gateway communication channel, the optimal gateway is selected for data feedback in combination with the network distance threshold, so as to facilitate the real-time feedback of blood lipid prediction results to the medical interaction terminal, ensuring that an early warning is triggered in time when the blood lipid level exceeds the normal range, so as to prevent the system from automatically switching to other available gateways when a gateway fails or the communication distance exceeds the threshold, thereby ensuring the stability and reliability of data during transmission, and solving the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions: A blood lipid monitoring system based on vascular ultrasound images, the system comprising: an image acquisition unit, an image processing unit and a data transmission unit, the image processing unit comprising an image recognition module and a blood lipid assessment module; The image acquisition unit is configured to acquire a patient's vascular ultrasound image based on ultrasound imaging technology, and perform denoising and enhancement processing on the vascular ultrasound image; The image recognition module is configured to train a convolutional neural network model based on historical vascular ultrasound images and clinical blood lipid test results, use the model to identify the vascular intima and media, and lumen contours in the vascular ultrasound images, and segment the recognition results to extract multi-dimensional features; the multi-dimensional features include: vascular wall thickness, lumen diameter, grayscale distribution, texture characteristics, and hemodynamic simulation parameters; The blood lipid assessment module is configured to input the multi-dimensional features into the trained convolutional neural network model and output the blood lipid prediction value; The data transmission unit is configured to set up a multi-node gateway communication channel for the blood lipid monitoring system based on a wireless communication protocol.

[0007] Furthermore, the image recognition module includes: A model building module, configured to build a convolutional neural network model, including: an input layer, a hidden layer, a fully connected layer, and an output layer; an image annotation module configured to annotate the vascular intima-media and lumen contours in historical vascular ultrasound images, and train a convolutional neural network model using the annotated historical vascular ultrasound images; The image segmentation module is configured to segment the annotated historical vascular ultrasound image through a convolutional neural network model and output the segmentation result.

[0008] Furthermore, the image recognition module further includes a contrast adjustment module for performing contrast adjustment on the historical vascular ultrasound images before marking the vascular intima-media and lumen contours in the historical vascular ultrasound images. The specific execution steps include: extracting a central grayscale value for each of the historical vascular ultrasound images as a first central grayscale value; Segmenting each vascular ultrasound image to obtain a plurality of image blocks; wherein the size of each image block includes but is not limited to 3×3, 5×5, and 9×9; extracting a central grayscale value of each image block contained in the vascular ultrasound image as a second central grayscale value; comparing the second center grayscale value with the first center grayscale value; According to the comparison result of the second center grayscale value and the first center grayscale value, an image block whose second center grayscale value is smaller than the first center grayscale value is selected as a target image block; Obtaining a center grayscale standard deviation corresponding to the target image block according to the center grayscale value of the target image block as a first grayscale standard deviation; Obtaining a grayscale standard deviation of the vascular ultrasound image according to the grayscale value of each pixel point in the vascular ultrasound image as a second grayscale standard deviation; The contrast of the historical vascular ultrasound image is adjusted using the first grayscale standard deviation and the second grayscale standard deviation.

[0009] Furthermore, contrast adjustment is performed on the historical vascular ultrasound image using the first grayscale standard deviation and the second grayscale standard deviation, including: Performing ratio processing on the first grayscale standard deviation and the second grayscale standard deviation to obtain a first standard deviation ratio; Extract the grayscale value of the pixels contained in each target image block; Obtaining a grayscale standard deviation corresponding to each target image block using the grayscale values of the pixels contained in each target image block as a third grayscale standard deviation; Performing ratio processing on the third grayscale standard deviation and the second grayscale standard deviation to obtain a second standard deviation ratio; Comparing the second standard deviation ratio with the first standard deviation ratio to determine whether there is a target image block whose second standard deviation ratio is not lower than the first standard deviation ratio; When there is no target image block whose second standard deviation ratio is not lower than the first standard deviation ratio, the contrast of the historical vascular ultrasound image is adjusted using the first grayscale standard deviation and the second grayscale standard deviation; According to the comparison result of the second standard deviation ratio and the first standard deviation ratio, a target image block having a second standard deviation ratio not less than the first standard deviation ratio is selected from the target image block as an observation image block; The contrast of the historical vascular ultrasound image is adjusted using the third grayscale standard deviation corresponding to the observed image block.

[0010] Furthermore, the image recognition module further includes: The feature extraction module is configured to identify the thickness of the blood vessel wall by measuring the distance between the blood vessel intima-media and the lumen boundary based on the segmented blood vessel intima-media and lumen contours; identify the maximum diameter of the lumen by measuring the shortest distance of the lumen contour; calculate the grayscale value distribution of the blood vessel wall and lumen area and identify the texture features of the blood vessel wall using a gray-level co-occurrence matrix texture analysis method; and calculate the hemodynamic parameters using computational fluid dynamics simulation based on the segmented blood vessel geometric parameters; The feature fusion module is configured to integrate vascular wall thickness, lumen diameter, grayscale distribution, texture features, and hemodynamic simulation parameters into multi-dimensional features, thereby obtaining a model for automatically identifying multi-dimensional features in vascular ultrasound images.

[0011] Furthermore, the data transmission unit includes: a data aggregation module configured to receive data from the image acquisition unit and the image processing unit and convert the data into a unified format; A gateway setting module is configured to configure multiple communication gateways between the image acquisition unit, the image processing unit, and the medical interaction terminal based on a wireless communication protocol, and to set a network distance threshold based on a standard gateway communication distance parameter; The gateway selection module is configured to determine whether the actual communication distance parameter of each gateway is within the threshold range based on the network distance threshold; if the communication distance parameter of the current gateway does not exceed the network distance threshold, data feedback is performed through the current gateway; if the communication distance parameter of the current gateway exceeds the network distance threshold, it is determined whether the communication distance parameter of the adjacent gateway is within the threshold range, and a neighboring gateway within the network distance threshold range is selected for data feedback; The gateway management module is configured to adjust the network management and configuration of multiple nodes according to clinical needs, including: adding, deleting and parameter setting of nodes.

[0012] Furthermore, the blood lipid assessment module includes: The model training module is configured to divide the multi-dimensional features and annotated historical vascular ultrasound images and clinical blood lipid test results into training and test sets, import the training set into the input layer, further process the multi-dimensional features through the hidden layer and the fully connected layer to obtain the blood lipid analysis results; and output the blood lipid prediction value based on the output layer. The prediction value includes: blood lipid level value and blood lipid level classification result; The model validation module is configured to compare the training results with the clinical blood lipid test results to verify whether the model's prediction results are accurate; if not, the model parameters are adjusted and retrained; if accurate, the model is tested on the test set to optimize the model's generalization ability.

[0013] Furthermore, the image processing unit further includes a blood lipid early warning module; The blood lipid early warning module is configured to set up an early warning mechanism for the blood lipid monitoring system according to clinical guidelines; and when the early warning is triggered, timely feedback is provided to the medical interactive terminal.

[0014] Furthermore, the blood lipid early warning module includes: a threshold setting module configured to set warning thresholds for different blood lipid indicators according to clinical guidelines; The early warning trigger module is configured to compare the blood lipid prediction value with the early warning threshold. If the prediction value exceeds the threshold range, the early warning mechanism is triggered; and the early warning information is fed back to the medical interaction terminal based on the data transmission unit. The early warning information includes: the blood lipid prediction value and the recommended intervention measures.

[0015] Furthermore, the image acquisition unit includes: An image acquisition module is configured to acquire two-dimensional ultrasound images of blood vessels using high-frequency ultrasound imaging equipment and generate Doppler vascular images based on clinical needs; and collect historical ultrasound vascular images and clinical blood lipid test results of clinical patients as data samples for building and training the prediction model; a denoising processing module configured to perform multi-scale wavelet transform on the ultrasound image based on the wavelet transform method, perform soft threshold processing on the high-frequency coefficients, enhance the low-frequency coefficients in the fuzzy domain, and obtain a denoised ultrasound image through wavelet reconstruction; an enhancement processing module configured to adjust the contrast of the denoised ultrasound image using gamma correction; The data storage module is configured to store the pre-processed vascular ultrasound image in the system and display it through the medical interactive terminal.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. In the present invention, a convolutional neural network model is used to automatically identify the intima-media and lumen contours in vascular ultrasound images in a short period of time, thereby reducing human error. Multi-dimensional features are extracted from them, providing more comprehensive information for blood lipid assessment, improving the accuracy and reliability of recognition and diagnosis, so as to adapt to different monitoring scenarios and clinical needs.

[0017] 2. In the present invention, a multi-node gateway communication channel is set up for the system through a wireless communication protocol, the communication distance parameters of the gateway are monitored in real time, and the optimal gateway is selected for data feedback in combination with the network distance threshold, so that the blood lipid prediction results can be fed back to the medical interaction terminal in real time, ensuring that an early warning is triggered in time when the blood lipid level exceeds the normal range, reminding doctors or patients to take intervention measures, providing important support for clinical decision-making; to prevent a gateway from failing or the communication distance exceeding the threshold, the system can automatically switch to other available gateways, thereby ensuring the stability and reliability of data during transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a diagram showing the module composition of the blood lipid monitoring system based on vascular ultrasound images of the present invention. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0020] To solve the existing problems, although vascular ultrasound imaging technology can provide two-dimensional or three-dimensional images of blood vessels, it lacks an effective intelligent analysis system and cannot automatically identify vascular structure and assess blood lipid levels. In addition, the environment of each department in the hospital is relatively closed, and network communication failures often lead to data transmission interruptions. Therefore, in its actual application, the stability and real-time performance of data transmission have become technical issues that need to be solved urgently. Figure 1 , this embodiment provides the following technical solutions: The blood lipid monitoring system based on vascular ultrasonic images includes: an image acquisition unit, an image processing unit and a data transmission unit.

[0021] An image acquisition unit is configured to acquire a patient's vascular ultrasound image based on ultrasound imaging technology and perform denoising and enhancement processing on the vascular ultrasound image; the image acquisition unit includes: The image acquisition module is configured to acquire two-dimensional ultrasound images of blood vessels based on high-frequency ultrasound imaging equipment. High-frequency ultrasound imaging can provide high-resolution images of vascular structure, so as to observe subtle changes in the blood vessel wall and help to detect vascular lesions at an early stage. Combined with the generation of Doppler vascular images based on clinical needs, it can provide information on blood flow velocity and direction, which helps to evaluate the hemodynamic state in the blood vessels and provide more comprehensive data support for blood lipid monitoring. It also collects historical ultrasound vascular images and clinical blood lipid test results of clinical patients as data samples for constructing and training the prediction model, thereby improving the accuracy and generalization ability of the model.

[0022] The denoising processing module is configured to perform multi-scale wavelet transform on the ultrasound image based on the wavelet transform method, thereby decomposing the image into sub-bands of different frequencies; first, soft threshold processing is performed on the high-frequency coefficients to remove noise components less than the threshold, effectively suppressing high-frequency noise while retaining the main features of the signal; then, the low-frequency coefficients are enhanced in the fuzzy domain to highlight important structural information in the image; finally, wavelet reconstruction is performed to significantly improve the image quality, thereby obtaining a denoised ultrasound image.

[0023] The enhancement processing module is configured to use gamma correction to adjust the contrast of the denoised ultrasound image, which can enhance the details of the image and make the vascular structure clearer, facilitating subsequent image analysis; and by adjusting the contrast and brightness of the image, the boundaries of the vascular wall and lumen are highlighted, thereby improving the readability and diagnostic value of the image.

[0024] The data storage module is configured to store the pre-processed vascular ultrasound images in the system and display them through the medical interactive terminal to facilitate further analysis by doctors or researchers.

[0025] The image processing unit includes: an image recognition module, a blood lipid assessment module and a blood lipid early warning module.

[0026] The image recognition module is configured to train a convolutional neural network model based on historical vascular ultrasound images and clinical blood lipid test results, use the convolutional neural network model to identify the vascular intima-media and lumen contours in the vascular ultrasound images, and segment the recognition results to extract multi-dimensional features. The multi-dimensional features include: vascular wall thickness, lumen diameter, grayscale distribution, texture characteristics, and hemodynamic simulation parameters. The image recognition module includes: The model building module is configured to use the Python framework to build a convolutional neural network model, including: input layer, hidden layer, fully connected layer and output layer; specifically, the input layer is used to receive preprocessed vascular ultrasound image data; the convolution layer is used to extract various types of features, such as edges, textures, etc., through multiple convolution kernels; the pooling layer is used to reduce the spatial size of the feature map, reduce the amount of calculation and prevent overfitting; the fully connected layer is used to integrate the features extracted by the convolution layer and the pooling layer to output the final feature vector; the output layer is used to output results according to task requirements, such as multi-dimensional features such as vascular wall thickness and lumen diameter.

[0027] The image annotation module is configured to annotate the vascular intima-media and lumen contours in historical vascular ultrasound images, use the annotated historical vascular ultrasound images and their corresponding clinical blood lipid test results as data samples for training the convolutional neural network model, and train the convolutional neural network model.

[0028] The image segmentation module is configured to segment the annotated historical vascular ultrasound image through a convolutional neural network model and output the segmentation results of the vascular intima-media and lumen contours.

[0029] The feature extraction module is configured to identify the thickness of the vascular wall based on the segmented endo-media and lumen contours by measuring the distance between the endo-media and lumen boundaries, so as to better identify vascular lesions, such as atherosclerosis; identify the maximum diameter of the lumen by measuring the shortest distance of the lumen contour, thereby reflecting the degree of vascular stenosis; calculate the grayscale value distribution of the vascular wall and lumen area, and use the grayscale co-occurrence matrix texture analysis method to identify the texture features of the vascular wall, such as contrast, correlation, energy, etc.; combine the segmented vascular geometric parameters, such as lumen diameter and vascular wall thickness, and use computational fluid dynamics simulation to calculate hemodynamic parameters, such as wall shear stress, to analyze the causes of vascular lesions; and oscillatory shear index, to evaluate blood flow disturbances in the blood vessels.

[0030] The feature fusion module is configured to integrate the vascular wall thickness, lumen diameter, grayscale distribution, texture features and hemodynamic simulation parameters into multi-dimensional features, and obtain a model for automatically identifying multi-dimensional features in vascular ultrasound images; specifically, the vascular wall thickness, lumen diameter, grayscale distribution, texture features and hemodynamic parameters are first normalized to ensure that the features are fused in the same dimension; the normalized feature vectors are spliced to form a multi-dimensional feature vector; for example: if the vascular wall thickness is a feature value, the lumen diameter is a feature value, the grayscale distribution is a feature vector, the texture feature is a feature vector, and the hemodynamic parameter is a feature vector, the above feature vectors are spliced into a long vector to obtain multi-dimensional features, which provide comprehensive feature input for subsequent blood lipid assessment.

[0031] The blood lipid assessment module is configured to input multi-dimensional features into a trained convolutional neural network model and output blood lipid prediction values; the blood lipid assessment module includes: The model training module is configured to divide the multi-dimensional features and annotated historical vascular ultrasound images and clinical blood lipid test results into 80% training sets and 20% test sets to ensure that the data distribution of the training sets and test sets is representative and avoid the impact of data bias on model performance; the training set is imported into the input layer, local features are extracted through the convolution layer in the hidden layer, and the pooling layer reduces the spatial size of the feature map to reduce the amount of calculation and prevent overfitting; the feature vector processed by the hidden layer is input into the fully connected layer to further integrate the feature information to obtain the blood lipid analysis results; and the blood lipid prediction value is output based on the output layer. The predicted value includes: blood lipid level values, such as: specific values of total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), etc.; and blood lipid level classification results, such as: normal, mildly abnormal, and severely abnormal.

[0032] The model validation module is configured to compare the training results with the clinical blood lipid test results to verify whether the model's prediction results are accurate; if not, the model parameters, such as learning rate, hidden layer structure, regularization parameters, etc., are adjusted and retrained until the model performance on the validation set reaches a satisfactory level; if accurate, the model is tested on the test set to optimize the model's generalization ability.

[0033] The beneficial effects achieved by the above content are: the convolutional neural network model can automatically identify the vascular intima-media and lumen contours in vascular ultrasound images in a short time, thereby reducing human errors; and extract multi-dimensional features from them, thereby providing more comprehensive information for blood lipid assessment, improving the accuracy and reliability of recognition and diagnosis, so as to adapt to different monitoring scenarios and clinical needs.

[0034] At the same time, the image recognition module also includes a contrast adjustment module for performing contrast adjustment on the historical vascular ultrasound images before marking the vascular intima-media and lumen contours in the historical vascular ultrasound images. The specific execution steps include: extracting a central grayscale value for each of the historical vascular ultrasound images as a first central grayscale value; Segmenting each vascular ultrasound image to obtain a plurality of image blocks; wherein the size of each image block includes but is not limited to 3×3, 5×5, and 9×9; extracting a central grayscale value of each image block contained in the vascular ultrasound image as a second central grayscale value; comparing the second center grayscale value with the first center grayscale value; According to the comparison result of the second center grayscale value and the first center grayscale value, an image block whose second center grayscale value is smaller than the first center grayscale value is selected as a target image block; Obtaining a center grayscale standard deviation corresponding to the target image block according to the center grayscale value of the target image block as a first grayscale standard deviation; Obtaining a grayscale standard deviation of the vascular ultrasound image according to the grayscale value of each pixel point in the vascular ultrasound image as a second grayscale standard deviation; The contrast of the historical vascular ultrasound image is adjusted using the first grayscale standard deviation and the second grayscale standard deviation.

[0035] The technical effect of the above technical solution is as follows: the first center grayscale value represents the average grayscale level of the entire vascular ultrasound image, reflecting the overall brightness of the image. The second center grayscale value is the center grayscale value of the image block, reflecting the brightness characteristics of the local area of the image. By comparing the two, image blocks with a second center grayscale value smaller than the first center grayscale value are selected as target image blocks. These target image blocks are often relatively dark local areas that may contain some details but are difficult to distinguish due to low brightness. Selecting them allows for targeted processing of these areas. The first grayscale standard deviation (the center grayscale standard deviation corresponding to the target image block) reflects the dispersion of grayscale values within the target image block, that is, the grayscale variation within the local area. The second grayscale standard deviation (the grayscale standard deviation of the vascular ultrasound image) reflects the dispersion of grayscale values across the entire image. Using these two standard deviations to adjust contrast is based on the relationship between standard deviations and image contrast: larger standard deviations indicate greater grayscale value variation within the image or image block, and higher contrast. Specifically, the grayscale of the target image block and surrounding areas in the image may be stretched or compressed based on the proportional relationship or difference between the two. By extracting and comparing the central grayscale values, the target image block is selected. The contrast is then adjusted based on the central grayscale standard deviation of the target image block (the first grayscale standard deviation) and the grayscale standard deviation of the entire image (the second grayscale standard deviation). This method can highlight areas of the image where grayscale differences are not obvious, making details such as the vascular endothelium and lumen contours clearer, facilitating subsequent annotation and improving the accuracy and reliability of image recognition. This method considers both local (image block) and global grayscale characteristics of the image, effectively balancing the contrast between different image regions, avoiding local overbrightness or overdarkness, and making the overall image visually more uniform and consistent, improving image quality and providing a higher-quality data foundation for subsequent image analysis. The method can process image blocks of different sizes and flexibly adapt to the characteristics and resolution of various vascular ultrasound images. It can effectively adjust the contrast of historical vascular ultrasound images from different sources and imaging conditions, enhancing the versatility and adaptability of the technical solution.

[0036] Specifically, using the first grayscale standard deviation and the second grayscale standard deviation to adjust the contrast of the historical vascular ultrasound image includes: Performing ratio processing on the first grayscale standard deviation and the second grayscale standard deviation to obtain a first standard deviation ratio; Extract the grayscale value of the pixels contained in each target image block; Obtaining a grayscale standard deviation corresponding to each target image block using the grayscale values of the pixels contained in each target image block as a third grayscale standard deviation; Performing ratio processing on the third grayscale standard deviation and the second grayscale standard deviation to obtain a second standard deviation ratio; Comparing the second standard deviation ratio with the first standard deviation ratio to determine whether there is a target image block whose second standard deviation ratio is not lower than the first standard deviation ratio; When there is no target image block whose second standard deviation ratio is not lower than the first standard deviation ratio, the contrast of the historical vascular ultrasound image is adjusted using the first grayscale standard deviation and the second grayscale standard deviation; The adjusted contrast is obtained by the following formula: Among them, E t represents the contrast after adjustment; E0 represents the contrast before adjustment; σ 01 and σ 02 Represent the first grayscale standard deviation and the second grayscale standard deviation respectively; specifically, The difference between the two is divided by the larger value, mapping the grayscale dispersion difference to the range [-1, 1]. This difference can be expressed as a relative ratio, facilitating integrated calculations with other components and accurately reflecting the impact of local and global grayscale variations on contrast adjustment. Adding 1 to the normalized difference ratio ensures that the calculated result is within the range [0, 2]. This result serves as an adjustment factor. A value of 1 indicates no contrast adjustment; a value greater than 1 increases contrast; and a value less than 1 decreases contrast. This formula determines the direction and magnitude of contrast adjustment based on the grayscale dispersion difference between the target image block and the entire image. This formula calculates the adjustment factor based on the grayscale dispersion difference between the target image block and the entire image, accurately adjusting contrast based on local and global grayscale variations. For images with large grayscale dispersion differences, it effectively increases contrast to highlight details. For images with small differences, it avoids over-adjustment, ensuring that contrast adjustment better meets actual image requirements and improves image clarity and readability. Through normalization and calculation of the adjustment factor, the formula adapts to the grayscale characteristics of diverse vascular ultrasound images. Regardless of the original grayscale discreteness of the image, it can be 01 and σ 02 Proper contrast adjustment enhances the method's adaptability to diverse images, effectively optimizing image quality under various imaging conditions. The formula's computational structure ensures that the adjustment factor varies within a reasonable range, avoiding situations where contrast adjustment can become uncontrolled due to excessively large or small grayscale dispersion differences. This ensures a stable and reliable contrast adjustment process, preventing extreme contrast enhancement or reduction, and maintaining the stability of the image's visual effects.

[0037] According to the comparison result of the second standard deviation ratio and the first standard deviation ratio, a target image block having a second standard deviation ratio not less than the first standard deviation ratio is selected from the target image block as an observation image block; The contrast of the historical vascular ultrasound image is adjusted using the third grayscale standard deviation corresponding to the observed image block; The adjusted contrast is obtained by the following formula: Among them, E t represents the contrast after adjustment; E0 represents the contrast before adjustment; σ 01 and σ 02 Represent the first grayscale standard deviation and the second grayscale standard deviation respectively; n represents the number of observed image blocks; σ 03i represents the third grayscale standard deviation of the i-th observed image block. Specifically, The grayscale difference adjustment items for n observation image blocks are summed to comprehensively consider the impact of all observation image blocks. The average is then divided by n to normalize the combined impact of multiple observation image blocks, resulting in a single value that represents the overall grayscale difference impact of all observation image blocks, reflecting the combined effect of the grayscale changes of the observation image blocks relative to the target image block and the entire image. Add 1 to the average impact value above, ensuring the result is greater than 0, to form an adjustment factor. When this factor is greater than 1, the contrast is increased; when it is less than 1, the contrast is decreased. The direction and magnitude of the contrast adjustment are determined based on the comprehensive grayscale differences of the observed image blocks.

[0038] The technical effect of the above-mentioned technical solution is as follows: A first standard deviation ratio (the ratio of the first grayscale standard deviation to the second grayscale standard deviation) is calculated to measure the relative relationship between the grayscale dispersion of a target image block (a relatively dark area) and the grayscale dispersion of the entire image. The first grayscale standard deviation reflects the grayscale variation within the target image block, while the second grayscale standard deviation reflects the grayscale variation of the entire image. This ratio indicates the position of the target image block within the grayscale variation system of the entire image. The third grayscale standard deviation (the grayscale standard deviation calculated from the grayscale values of each pixel in the target image block) and the second standard deviation ratio (the ratio of the third grayscale standard deviation to the second grayscale standard deviation) are calculated to further refine the comparison of the grayscale dispersion within the target image block and the grayscale dispersion of the entire image, analyzing the grayscale characteristics of each target image block from a more microscopic perspective. By comparing the second standard deviation ratio with the first standard deviation ratio, target image blocks with a second standard deviation ratio no less than the first standard deviation ratio are selected as observation image blocks. These observed image blocks have a relatively unique relationship with the overall image in terms of grayscale dispersion, resulting in more significant grayscale variations within them. These blocks may contain information critical to image analysis and will be subsequently targeted for processing. If no target image blocks exist whose second standard deviation ratio is at least as high as the first standard deviation ratio, this indicates that all target image blocks share relatively consistent grayscale dispersion characteristics. The first and second grayscale standard deviations are then used to perform contrast adjustment on the entire historical vascular ultrasound image. Based on their relationship, the image grayscale range is stretched or compressed to enhance overall contrast. If an observed image block exists, the third grayscale standard deviation corresponding to the observed image block is used to perform contrast adjustment on the historical vascular ultrasound image. Because the grayscale variations within the observed image blocks are unique, using the third grayscale standard deviation as a basis allows for more precise grayscale adjustment of these key image regions and their surrounding areas, highlighting their details while maintaining the visual balance of the overall image. By screening observed image blocks and performing targeted adjustments, areas containing important information can be effectively highlighted, making key features such as the vascular intima-media and lumen contours more distinct. This improves the effective information extraction rate and image recognition accuracy for medical diagnosis and other analysis. By comprehensively considering the grayscale dispersion of both the local (target image block) and overall image, and adjusting contrast both overall and in key local areas, this approach can achieve a more reasonable grayscale distribution, avoid inappropriate local or overall contrast, and improve the image's visual quality and readability. This method analyzes image grayscale characteristics from multiple perspectives and performs differentiated processing based on the grayscale dispersion of different image blocks. This method accurately adapts to the characteristics of different historical vascular ultrasound images, accurately identifying and appropriately adjusting areas requiring contrast enhancement for images under various imaging conditions, thus enhancing the accuracy and adaptability of the technical solution.

[0039] The blood lipid early warning module is configured to set an early warning mechanism for the blood lipid monitoring system according to clinical guidelines and provide timely feedback to the medical interactive terminal when the early warning is triggered. The blood lipid early warning module includes: The threshold setting module is configured to set warning thresholds for different blood lipid indicators according to clinical guidelines; for example: the normal range of total cholesterol (TC) is 3.9-5.2 mmol / L, the normal range of triglycerides (TG) is 0.56-1.7 mmol / L, the normal range of low-density lipoprotein cholesterol (LDL-C) is 2.07-3.1 mmol / L, and the normal range of high-density lipoprotein cholesterol (HDL-C) is 1.03-2.07 mmol / L. If the above indicators exceed the corresponding range, an early warning will be triggered.

[0040] The early warning trigger module is configured to compare the blood lipid prediction value with the early warning threshold. If the prediction value exceeds the threshold range, the early warning mechanism is triggered; and based on the data transmission unit, the early warning information is fed back to the medical interaction terminal, which can be a doctor's computer, a display screen at a nurse station, or a patient's mobile device, to ensure that the early warning information can be conveyed to relevant personnel in a timely manner; the early warning information includes: blood lipid prediction value, that is, specific numerical values, such as the predicted values of total cholesterol, triglycerides, etc.; and according to the specific circumstances of dyslipidemia, corresponding intervention suggestions are provided, such as: dietary adjustment, exercise suggestions or drug treatment, etc., to remind doctors or patients to take corresponding intervention measures, thereby improving the early detection and intervention capabilities of dyslipidemia.

[0041] In one embodiment, for example, if the predicted total cholesterol value is 6.0 mmol / L, it is outside the normal range, and the warning information will prompt the patient to adjust his diet and increase exercise.

[0042] The data transmission unit is configured to be based on a wireless communication protocol, such as LoRa, to set up a multi-node gateway communication channel for the blood lipid monitoring system, enabling data exchange between the image acquisition unit, the image processing unit, and the data transmission unit. Specifically, the vascular ultrasound image is fed back to the image processing unit for image recognition and blood lipid assessment, and the blood lipid prediction value is fed back to the medical interactive terminal in real time for display. The data transmission unit includes: The data aggregation module is configured to receive data from the image acquisition unit and the image processing unit, including: vascular ultrasound images, multi-dimensional features such as: vascular wall thickness, lumen diameter, grayscale distribution, texture features, hemodynamic parameters, etc.; and convert the data into a unified format, such as: standardized JSON format, to facilitate subsequent processing and analysis.

[0043] The gateway setting module is configured based on a wireless communication protocol, such as the LoRa gateway, which configures multiple communication gateways between the image acquisition unit, the image processing unit, and the medical interaction terminal, and sets the network distance threshold based on the standard gateway communication distance parameters to ensure that the gateway is within the effective communication range.

[0044] The gateway selection module is configured to monitor the actual communication distance parameters of each gateway in real time based on the network distance threshold, and determine whether the actual communication distance parameters of each gateway are within the threshold range; if the communication distance parameters of the current gateway do not exceed the network distance threshold, data feedback is performed through the current gateway; if the communication distance parameters of the current gateway exceed the network distance threshold, it is determined whether the communication distance parameters of the adjacent gateways are within the threshold range, and a neighboring gateway within the network distance threshold range is selected for data feedback; during the communication process, if the communication quality of the current gateway deteriorates or exceeds the network distance threshold, the system automatically switches to the backup gateway to ensure the continuity of data transmission.

[0045] The gateway management module is configured to adjust the network management and configuration of multiple nodes according to clinical needs, including: adding, deleting and setting parameters of nodes; by supporting dynamic management and configuration of multi-node networks, the system can flexibly adjust the network structure according to clinical needs and adapt to different application scenarios.

[0046] In one embodiment, for example: a LoRa gateway is configured to support long-distance, low-power wireless communication; based on the communication distance parameters of the LoRa gateway, a network distance threshold is set to 2 kilometers; and the actual communication distance of the LoRa gateway is monitored in real time. If the communication distance of the current LoRa gateway exceeds the 2-kilometer threshold, it is automatically switched to an adjacent LoRa gateway.

[0047] The beneficial effects achieved by the above content are: setting up multi-node gateway communication channels for the system through wireless communication protocols, monitoring the communication distance parameters of the gateway in real time, and selecting the optimal gateway for data feedback based on the network distance threshold, so as to facilitate the real-time feedback of blood lipid prediction results to the medical interactive terminal, ensuring that when the blood lipid level exceeds the normal range, an early warning is triggered in time, reminding doctors or patients to take intervention measures, and providing important support for clinical decision-making; to prevent when a gateway fails or the communication distance exceeds the threshold, the system can automatically switch to other available gateways, thereby ensuring the stability and reliability of data during transmission.

[0048] Working principle: The convolutional neural network model is used to automatically identify the vascular intima-media and lumen contours in vascular ultrasound images, and multi-dimensional features are extracted from them, providing more comprehensive information for blood lipid assessment; according to clinical guidelines, warning thresholds are set for different blood lipid indicators; if the predicted value exceeds the threshold range, the warning mechanism is triggered, and the warning information is fed back to the medical interaction terminal to remind doctors or patients to take corresponding intervention measures; a multi-node gateway communication channel is constructed based on the wireless communication protocol, and the optimal gateway is selected in combination with the network distance threshold to feed back the blood lipid prediction results to the medical interaction terminal to prevent the system from automatically switching to other available gateways when a gateway fails or the communication distance exceeds the threshold, thereby ensuring the stability and reliability of data during transmission.

[0049] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "including," "having," or any other variations thereof are intended to cover non-exclusive possessors, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or includes elements that are inherent to such process, method, article, or apparatus.

[0050] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions, and alterations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A blood lipid monitoring system based on vascular ultrasound images, characterized in that: The system comprises: an image acquisition unit, an image processing unit and a data transmission unit, wherein the image processing unit comprises an image recognition module and a blood lipid assessment module; The image acquisition unit is configured to acquire a patient's vascular ultrasound image based on ultrasound imaging technology, and perform denoising and enhancement processing on the vascular ultrasound image; The image recognition module is configured to train a convolutional neural network model based on historical vascular ultrasound images and clinical blood lipid test results, use the model to identify the vascular intima and media, and lumen contours in the vascular ultrasound images, and segment the recognition results to extract multi-dimensional features; the multi-dimensional features include: vascular wall thickness, lumen diameter, grayscale distribution, texture characteristics, and hemodynamic simulation parameters; The blood lipid assessment module is configured to input the multi-dimensional features into the trained convolutional neural network model and output the blood lipid prediction value; The data transmission unit is configured to set up a multi-node gateway communication channel for the blood lipid monitoring system based on a wireless communication protocol.

2. The blood lipid monitoring system based on vascular ultrasound images according to claim 1, characterized in that: The image recognition module includes: A model building module, configured to build a convolutional neural network model, including: an input layer, a hidden layer, a fully connected layer, and an output layer; an image annotation module configured to annotate the vascular intima-media and lumen contours in historical vascular ultrasound images, and train a convolutional neural network model using the annotated historical vascular ultrasound images; The image segmentation module is configured to segment the annotated historical vascular ultrasound image through a convolutional neural network model and output the segmentation result.

3. The blood lipid monitoring system based on vascular ultrasound images according to claim 1 or 2, characterized in that: The image recognition module further includes a contrast adjustment module for performing contrast adjustment on the historical vascular ultrasound images before marking the vascular intima-media and lumen contours in the historical vascular ultrasound images. The specific steps of performing the adjustment include: extracting a central grayscale value for each of the historical vascular ultrasound images as a first central grayscale value; Perform image segmentation on each vascular ultrasound image to obtain multiple image blocks; extracting a central grayscale value of each image block contained in the vascular ultrasound image as a second central grayscale value; comparing the second center grayscale value with the first center grayscale value; According to the comparison result of the second center grayscale value and the first center grayscale value, an image block whose second center grayscale value is smaller than the first center grayscale value is selected as a target image block; Obtaining a center grayscale standard deviation corresponding to the target image block according to the center grayscale value of the target image block as a first grayscale standard deviation; Obtaining a grayscale standard deviation of the vascular ultrasound image according to the grayscale value of each pixel point in the vascular ultrasound image as a second grayscale standard deviation; The contrast of the historical vascular ultrasound image is adjusted using the first grayscale standard deviation and the second grayscale standard deviation.

4. The blood lipid monitoring system based on vascular ultrasound images according to claim 3, characterized in that: The contrast of the historical vascular ultrasound image is adjusted using the first grayscale standard deviation and the second grayscale standard deviation, including: Performing ratio processing on the first grayscale standard deviation and the second grayscale standard deviation to obtain a first standard deviation ratio; Extract the grayscale value of the pixels contained in each target image block; Obtaining a grayscale standard deviation corresponding to each target image block using the grayscale values of the pixels contained in each target image block as a third grayscale standard deviation; Performing ratio processing on the third grayscale standard deviation and the second grayscale standard deviation to obtain a second standard deviation ratio; Comparing the second standard deviation ratio with the first standard deviation ratio to determine whether there is a target image block whose second standard deviation ratio is not lower than the first standard deviation ratio; When there is no target image block whose second standard deviation ratio is not lower than the first standard deviation ratio, the contrast of the historical vascular ultrasound image is adjusted using the first grayscale standard deviation and the second grayscale standard deviation; According to the comparison result of the second standard deviation ratio and the first standard deviation ratio, a target image block whose second standard deviation ratio is not lower than the first standard deviation ratio is screened out from the target image blocks as an observation image block.

5. The blood lipid monitoring system based on vascular ultrasound images according to claim 2, characterized in that: The image recognition module further includes: The feature extraction module is configured to identify the thickness of the blood vessel wall by measuring the distance between the blood vessel intima-media and the lumen boundary based on the segmented blood vessel intima-media and lumen contours; identify the maximum diameter of the lumen by measuring the shortest distance of the lumen contour; calculate the grayscale value distribution of the blood vessel wall and lumen area and identify the texture features of the blood vessel wall using a gray-level co-occurrence matrix texture analysis method; and calculate the hemodynamic parameters using computational fluid dynamics simulation based on the segmented blood vessel geometric parameters; The feature fusion module is configured to integrate vascular wall thickness, lumen diameter, grayscale distribution, texture features, and hemodynamic simulation parameters into multi-dimensional features, thereby obtaining a model for automatically identifying multi-dimensional features in vascular ultrasound images.

6. The blood lipid monitoring system based on vascular ultrasound images according to claim 1, characterized in that: The data transmission unit includes: a data aggregation module configured to receive data from the image acquisition unit and the image processing unit and convert the data into a unified format; A gateway setting module is configured to configure multiple communication gateways between the image acquisition unit, the image processing unit, and the medical interaction terminal based on a wireless communication protocol, and to set a network distance threshold based on a standard gateway communication distance parameter; The gateway selection module is configured to determine whether the actual communication distance parameter of each gateway is within the threshold range based on the network distance threshold; if the communication distance parameter of the current gateway does not exceed the network distance threshold, data feedback is performed through the current gateway; if the communication distance parameter of the current gateway exceeds the network distance threshold, it is determined whether the communication distance parameter of the adjacent gateway is within the threshold range, and a neighboring gateway within the network distance threshold range is selected for data feedback; The gateway management module is configured to adjust the network management and configuration of multiple nodes according to clinical needs, including: adding, deleting and parameter setting of nodes.

7. The blood lipid monitoring system based on vascular ultrasound images according to claim 2, characterized in that: The blood lipid assessment module includes: The model training module is configured to divide the multi-dimensional features and annotated historical vascular ultrasound images and clinical blood lipid test results into training and test sets, import the training set into the input layer, further process the multi-dimensional features through the hidden layer and the fully connected layer to obtain the blood lipid analysis results; and output the blood lipid prediction value based on the output layer. The prediction value includes: blood lipid level value and blood lipid level classification result; The model validation module is configured to compare the training results with the clinical blood lipid test results to verify whether the model's prediction results are accurate; if not, the model parameters are adjusted and retrained; if accurate, the model is tested on the test set to optimize the model's generalization ability.

8. The blood lipid monitoring system based on vascular ultrasound images according to claim 1, characterized in that: The image processing unit further includes a blood lipid early warning module; The blood lipid early warning module is configured to set up an early warning mechanism for the blood lipid monitoring system according to clinical guidelines; and when the early warning is triggered, timely feedback is provided to the medical interactive terminal.

9. The blood lipid monitoring system based on vascular ultrasound images according to claim 8, characterized in that: The blood lipid early warning module includes: a threshold setting module configured to set warning thresholds for different blood lipid indicators according to clinical guidelines; The early warning trigger module is configured to compare the blood lipid prediction value with the early warning threshold. If the prediction value exceeds the threshold range, the early warning mechanism is triggered; and the early warning information is fed back to the medical interaction terminal based on the data transmission unit. The early warning information includes: the blood lipid prediction value and the recommended intervention measures.

10. The blood lipid monitoring system based on vascular ultrasound images according to claim 1, characterized in that: The image acquisition unit includes: An image acquisition module is configured to acquire two-dimensional ultrasound images of blood vessels using high-frequency ultrasound imaging equipment and generate Doppler vascular images based on clinical needs; and collect historical ultrasound vascular images and clinical blood lipid test results of clinical patients as data samples for building and training the prediction model; a denoising processing module configured to perform multi-scale wavelet transform on the ultrasound image based on the wavelet transform method, perform soft threshold processing on the high-frequency coefficients, enhance the low-frequency coefficients in the fuzzy domain, and obtain a denoised ultrasound image through wavelet reconstruction; an enhancement processing module configured to adjust the contrast of the denoised ultrasound image using gamma correction; The data storage module is configured to store the pre-processed vascular ultrasound image in the system and display it through the medical interactive terminal.

Images