Ultrasonic image acquisition and analysis system combining edge calculation and attention analysis

Abstract

The invention relates to the technical field of image acquisition, in particular to an ultrasonic image acquisition and analysis system combining edge calculation and attention analysis, which comprises the following steps of: receiving original image data acquired by ultrasonic equipment in real time through an edge calculation node, and performing multi-scale spatial filtering processing on the original image data to generate a preprocessed image; inputting the preprocessed image into attention analysis, and generating an attention thermodynamic diagram through an anatomical positioning module and a dynamic tracking module which are cascaded; and dynamically adjusting scanning parameters of the ultrasonic equipment and generating an operation guide instruction based on brightness distribution characteristic analysis of the attention thermodynamic diagram. According to the method, the basic correction and the compensation coefficient are adaptively calculated based on historical experience or a gradient optimization model, pressure-depth cooperative compensation is achieved, the space offset is further calculated through estimation, the three-dimensional direction guide vector and the tactile vibration feedback instruction are generated, and the pressure-depth cooperative compensation is achieved. Therefore, an operator can complete the probe pose optimization operation without frequently watching a screen.

Description

Technical Field

[0001] The present invention relates to the field of image acquisition technology, and in particular to an ultrasound image acquisition and analysis system combining edge computing and attention analysis. Background Art

[0002] Ultrasound imaging is widely used in clinical examinations of cardiovascular diseases, abdominal organs, and obstetrics and gynecology due to its real-time performance, lack of radiation, and portability. However, ultrasound images inherently suffer from low signal-to-noise ratios, blurred structural boundaries, and susceptibility to interference from operator manipulation. In particular, in areas with small differences in acoustic impedance between tissues (such as blood vessels and surrounding soft tissue), the accuracy of structural recognition relies heavily on operator experience. To improve image quality and operational efficiency, some systems have introduced technologies such as filtering enhancement, attention guidance, and motion compensation in recent years. However, the following issues still exist:

[0003] Filtering processing lacks structural dependence and parallel optimization mechanism: Traditional image preprocessing methods mostly use fixed-scale spatial filtering operations, fail to adaptively adjust parameters according to image content, and the parameter coupling between processing modules is weak, resulting in over-smoothing of key anatomical boundaries and serious redundancy in filtering links, making it difficult to meet the real-time requirements under edge computing conditions.

[0004] Attention modeling is limited to static image frames and lacks a temporal consistency enhancement mechanism: Existing ultrasound intelligent analysis methods are mostly based on extracting spatial attention or channel weights from single-frame images, ignoring the dynamic consistency of structures between consecutive frames. They are easily affected by interference such as probe jitter and tissue deformation, causing structural positioning drift or instability.

[0005] Operators lack a refined feedback mechanism, making it difficult to achieve active guided scanning of anatomical structures: Existing systems mainly provide two-dimensional image feedback for ultrasound probe operations, lacking focus adjustment, frame rate control, and motion guidance mechanisms based on structural distribution. Operators need to frequently check the screen and manually adjust parameters while moving the probe, resulting in low efficiency and high errors, especially in high-speed scanning or deep imaging scenarios. Summary of the Invention

[0006] The present invention provides an ultrasound image acquisition and analysis system that combines edge computing and attention analysis. It integrates edge computing capabilities, a structure-aware filtering chain, attention heat map dynamic modeling, and feedback closed-loop control to improve the real-time clarity, structural stability, and operator guidance efficiency of ultrasound images.

[0007] An ultrasound image acquisition and analysis system combining edge computing and attention parsing, including:

[0008] The edge computing node receives the raw image data collected by the ultrasound device in real time, performs multi-scale spatial filtering on the raw image data, and generates a pre-processed image;

[0009] Inputting the preprocessed image into an attention parsing unit, generating an attention heat map through a cascaded anatomical positioning module and a dynamic tracking module, wherein the high-brightness area in the attention heat map represents the real-time spatial distribution of key anatomical structures;

[0010] Based on the brightness distribution feature analysis of the attention heat map, the scanning parameters of the ultrasound device are dynamically adjusted and operation guidance instructions are generated. The operation guidance instructions include probe movement direction suggestions and focus depth correction amounts.

[0011] Optionally, the generation of the preprocessed image specifically includes:

[0012] Construct a three-stage filtering pipeline consisting of an adaptive Gaussian kernel, anisotropic diffusion kernel, and directional gradient kernel;

[0013] The three-level filtering process is performed synchronously through the parallel computing unit of the edge computing node, in which the technical features of the output of the previous level of filtering are used as the input parameters of the next level of filtering to form a cascade optimization link.

[0014] Optionally, in the three-stage filtering pipeline:

[0015] The adaptive Gaussian kernel dynamically adjusts the filter radius based on the signal-to-noise ratio of the original image data and outputs a primary denoised image;

[0016] The anisotropic diffusion kernel suppresses the diffusion intensity of non-vascular tissue according to the local gradient direction of the primary denoised image and outputs a contrast-enhanced image;

[0017] The directional gradient kernel extracts edge features matching preset anatomical directions in the contrast-enhanced image, and generates a preprocessed image including organ contour markings.

[0018] Optionally, the anatomical localization module in S2 receives the preprocessed image, performs feature parsing through the parallel spatial attention submodule and channel attention submodule, generates an initial heat map and outputs a feature weight matrix;

[0019] The dynamic tracking module receives the initial heat map and feature weight matrix of N consecutive frames to generate a dynamic enhanced heat map;

[0020] The initial heat map and the dynamically enhanced heat map are weightedly fused through the heat map fusion unit, and the final output is an attention heat map that synchronizes the high-brightness area with the spatial distribution of key anatomical structures in real time.

[0021] Optionally, the spatial attention submodule uses a dilated convolutional layer to extract multi-receptive field features and generate an initial heat map, in which pixel values ​​represent the probability of the existence of anatomical structures;

[0022] The channel attention submodule analyzes the response strength of different tissue channels through a deformable convolution kernel and outputs a feature weight matrix of blood vessels and organ boundaries.

[0023] Optionally, the dynamic tracking module performs the following:

[0024] Calculate the displacement vector of feature points between adjacent frames based on the optical flow field prediction network and generate motion compensation coefficients;

[0025] The motion compensation coefficient is multiplied by the current frame feature weight matrix to generate a dynamic enhancement heat map.

[0026] Optionally, the weight value during the weighted fusion is dynamically adjusted according to the moving speed of the ultrasound probe.

[0027] Optionally, the analysis of the brightness distribution characteristics of the attention heat map includes segmenting the attention heat map using a region growing algorithm, extracting continuous regions with brightness values ​​exceeding a preset brightness threshold as key anatomical regions; calculating the brightness peak coordinates and coverage area of ​​each key anatomical region, and generating a structured parameter set including spatial position characteristics as an input basis for the scanning control strategy.

[0028] Optionally, the dynamic adjustment of the scanning parameters of the ultrasound device specifically includes:

[0029] adjusting the focal depth of the ultrasound transmission beam in real time according to the depth direction component of the brightness peak coordinate of the structured parameter set so that the focal depth matches the actual depth of the target anatomical structure;

[0030] Based on the ratio of the coverage area of ​​the key anatomical region in the structured parameter set to the preset standard area, the imaging frame rate is adjusted: when the ratio is less than 80%, the frame rate is increased to 1.5 times the original value to enhance the ability to capture details.

[0031] Optionally, the generation of the operation guidance instruction includes:

[0032] Calculate the offset between the current position of the ultrasound probe and the optimal imaging focus position, and generate a movement suggestion including a three-dimensional direction arrow;

[0033] Combining the focus depth correction amount and the current ultrasound probe pressure sensor data, a tactile feedback instruction is generated, which prompts the depth adjustment direction through the vibration pattern of the probe handle.

[0034] Beneficial effects of the present invention:

[0035] This invention constructs a multi-stage image filtering pipeline consisting of an adaptive Gaussian kernel, an anisotropic diffusion kernel, and a directional gradient kernel, forming a structured, step-by-step feature extraction chain. Dynamic parameter transfer is used as a coupling mechanism between the filtering modules. Signal-to-noise ratio estimates drive the diffusion intensity threshold setting, and the vascular orientation distribution map guides edge feature extraction. This avoids the over-smoothing of structural boundaries caused by parameter independence in traditional serial filtering. Deploying this chain in parallel on edge computing nodes reduces overall processing latency by approximately 48%, while improving the spatial clarity and anatomical legibility of low-contrast structures such as blood vessels and cardiac cavities.

[0036] The present invention proposes a cascaded anatomical positioning module and a dynamic tracking module, extracts multi-receptive field heat maps through the spatial attention submodule, differentially weights the responses of different tissues through the channel attention submodule, and introduces an optical flow field motion compensation mechanism between consecutive frames to generate a structural consistency enhanced heat map. It also introduces a speed-sensitive fusion strategy to automatically reduce the weight of low-confidence image segments when the probe moves too fast, thereby improving the robustness and time synchronization of structure tracking in high-dynamic scenes.

[0037] The present invention extracts the spatial distribution characteristics of the structural area based on the brightness heat map, sets the real-time focal depth based on its depth coordinates, and dynamically adjusts the frame rate in combination with the regional area to ensure that the imaging strategy matches the target anatomical features. It introduces focal depth correction. When the probe pressure exceeds the set threshold, the basic correction amount and compensation coefficient are adaptively calculated based on historical experience or a gradient optimization model to achieve pressure-depth collaborative compensation. The spatial offset is further estimated and calculated, and a three-dimensional direction guidance vector and tactile vibration feedback instruction are generated, so that the operator can complete the probe posture optimization operation without frequently looking at the screen. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 A schematic diagram of system logic of an embodiment of the present invention;

[0040] Figure 2 Schematic diagram of an attention parsing unit according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0042] It should be noted that references in the specification to "one embodiment," "an embodiment," "an exemplary embodiment," "some embodiments," etc. indicate that the described embodiments may include specific features, structures, or characteristics, but not every embodiment necessarily includes such specific features, structures, or characteristics. In addition, when specific features, structures, or characteristics are described in conjunction with an embodiment, it is within the knowledge of persons skilled in the relevant art to implement such features, structures, or characteristics in conjunction with other embodiments (whether or not explicitly described).

[0043] In general, terms can be understood, at least in part, from their use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending at least in part on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0044] like Figure 1-Figure 2 As shown, an ultrasound image acquisition and analysis system combining edge computing and attention parsing includes:

[0045] The original image data collected by the ultrasound equipment is received in real time through the edge computing node, and multi-scale spatial filtering is performed on the original image data to generate a pre-processed image;

[0046] The pre-processed image is input into the attention parsing, and the attention heat map is generated through the cascaded anatomical positioning module and dynamic tracking module. The high-brightness area in the attention heat map represents the real-time spatial distribution of key anatomical structures;

[0047] Based on the analysis of the brightness distribution characteristics of the attention heat map, the scanning parameters of the ultrasound equipment are dynamically adjusted and operation guidance instructions are generated. The operation guidance instructions include probe movement direction suggestions and focus depth correction amounts.

[0048] Pre-image processing specifically includes:

[0049] a. Based on the local signal-to-noise ratio of the original image, a Gaussian kernel filter with spatial adaptive capabilities is constructed. This filter automatically adjusts its spatial smoothing radius according to the noise level in different areas of the image to perform preliminary denoising on the image, preserving structural information while suppressing background interference. Specifically, an adaptive Gaussian kernel based on the local signal-to-noise ratio of the image is constructed to perform primary denoising on the original image I0(x,y):

[0050] I1(x,y)=G σ(x,y) *I0(x,y), where G σ(x,y) represents a Gaussian kernel centered at (x, y), the standard deviation σ varies with the local signal-to-noise ratio of the image, * represents a convolution operation, I1(x, y) is the primary denoised image, and σ(x, y) represents the local signal-to-noise ratio SNR. local Adaptive calculation, set to: Among them, α is the empirical weight factor (0.5 to 1.0).

[0051] b. Using the primary denoised image as input, perform anisotropic diffusion. This process non-uniformly regulates the diffusion intensity based on the local gradient direction in the image, strengthening the boundaries of anatomical structures, especially enhancing the recognizability of slender structures such as blood vessels, while suppressing the blurring of adjacent areas to background tissue. The anisotropic diffusion kernel process is specifically as follows: Construct an anisotropic diffusion process based on I1(x,y):

[0052] Where I2 is the contrast-enhanced image, is the divergence operator, is the image gradient, c(·) is the diffusion coefficient function, which controls the diffusion degree at different positions and is defined as: κ is the threshold value obtained by inverting the local signal-to-noise ratio output by the Gaussian kernel, which controls the diffusion suppression strength, t is the iterative diffusion time step, and I2(x,y) is the final output contrast-enhanced image. Represents the image gradient magnitude, which is used to represent edge strength.

[0053] c. Based on the contrast-enhanced image, a direction-selective edge response kernel is constructed to extract edge features in the image that are consistent with the set anatomical structure direction. This process uses a preset set of structural directions to filter the contour response in the image that best matches the target direction and outputs a final preprocessed image including organ boundary information for subsequent attention area determination. Specifically, a directional gradient kernel is constructed based on I2(x,y) to extract edge features consistent with the preset anatomical direction: in, Represents the gradient (directional derivative) along the direction θ, Θ is a set of directions related to the anatomical structure (such as the long axis direction of the heart), I3(x,y) is the output of the preprocessed image containing the edge contours of key organs, Indicates the result of selecting the direction with the strongest response among all preset directions.

[0054] d. The entire image preprocessing process is executed in parallel within the edge computing node using a three-stage filtering pipeline. The output of each filtering stage not only serves as the input image for the next stage, but also the structural parameters it generates are used to adaptively adjust the parameters of subsequent modules. For example, the local signal-to-noise ratio extracted in the denoising stage is used to set the diffusion intensity threshold, while the directional distribution of vascular structures defines the set of convolution directions in the edge extraction stage. This establishes a cascaded dependency relationship between technical features, improving the targetedness and stability of structural enhancement.

[0055] The generation of attention heatmap includes:

[0056] a. The anatomical localization module receives the image I3(x,y) output by the image preprocessing module and extracts key area features through the parallel structure of the spatial attention submodule and the channel attention submodule:

[0057] Spatial attention submodule: H init (x,y)=Conv dilated (I3), where I3(x,y) represents the pixel value at the spatial position (x,y) of the preprocessed image output by the image preprocessing module, and Conv dilated Represents the dilated convolution operation, which is used to extract multi-receptive field spatial features. init (x, y) represents the initial heat map, and each pixel value represents the probability of the current position being a key anatomical structure.

[0058] Channel attention submodule: W c =DeformableConv(I3), where DeformableConv is a deformable convolution operation, W c is a feature weight matrix that reflects the response intensity of each tissue channel (blood vessels, cardiac chambers, muscle layers, etc.) to the target task.

[0059] b. Dynamic tracking module: This module receives N consecutive frames of input. (initial heat map and channel weight matrix of frame t), performing temporal consistency enhancement:

[0060] Optical flow estimation and motion compensation coefficient calculation:

[0061] Dynamic enhancement of feature weights:

[0062] Among them, V t→t+1 Represents the feature point displacement vector field from the tth frame to the t+1th frame, represents the dynamic enhanced heat map, reflecting the spatiotemporal coupling result of spatial position and tissue structure response, FlowNet represents the optical flow prediction network, outputs the feature point displacement field between consecutive frames, and W c,t+1 (x,y) represents the optical flow displacement vector at position (x,y) from the tth frame to the t+1th frame, which is used to estimate the motion of the structure point.

[0063] c. Heatmap fusion unit: fuses the initial heatmap output by spatial attention with the dynamic enhanced heatmap to output the final attention heatmap: H final (x,y)=λ·H init (x,y)+(1-λ)·H dyn (x, y), where λ = f(v p ) is the fusion weight, which depends on the probe movement speed v corresponding to the current frame p , when v p >v th (5cm / s), λ is automatically reduced to λ∈[0,1] to suppress motion blur interference, H final is the final output attention heat map, which is used to guide the operation of the ultrasound probe. th is the probe movement speed threshold.

[0064] The generation of operation guidance instructions is as follows:

[0065] Based on the attention heat map H final (x,y), through brightness area analysis, scanning parameter adaptive adjustment and multi-modal operation guidance generation, closed-loop linkage control of image enhancement and operation feedback is achieved:

[0066] a.Brightness distribution analysis:

[0067] Region growing segmentation: set brightness threshold T h ,Perform the region growing algorithm on the heat map and extract all continuous regions that meet the following conditions: in, is the pixel set of the i-th key anatomical region, T h is the brightness distribution threshold (set to 0.6 of the maximum value of the heat map);

[0068] Spatial structure parameter extraction, for each continuous area Extract its brightness peak coordinates and area and construct a structured parameter set: in, Indicates area Internal brightness peak coordinate, A i is the area of ​​the region, expressed as is the i-th item in the structured parameter set, A i It will serve as the input variable for subsequent focus control and frame rate control.

[0069] b. Dynamic adjustment of scanning parameters:

[0070] Focus depth adjustment: Get the component of the brightness peak coordinate on the image depth axis (y or z axis) from the structured parameters and set it as the target depth d corresponding to the area i , which is used as the ultrasound focus depth f of the current frame d :f d =d i , d i is the focal depth position corresponding to the brightness peak (focus reference), which is then used in the dynamic correction of the depth compensation mechanism;

[0071] Frame rate adjustment rule: According to the actual structural area A i , let A s A is the preset standard area of ​​the target structure. i With A s Perform ratio judgment. If the area is insufficient, it means that the target structure details in the image are not sufficiently presented. It is necessary to increase the imaging resolution by increasing the frame rate. The ratio judgment is:

[0072] Among them, FrameRate base The basic frame rate of the device ranges from 20 to 30 frames per second (fps). The specific value depends on the device model. Common medical ultrasound devices default to 25fps. FrameRate is the dynamically adjusted frame rate.

[0073] c. Operation guide instruction generation:

[0074] Direction guidance generation (based on probe offset calculation): Use the current frame probe position r current and the optimal imaging focus position r target Calculate the 3D spatial direction offset: Δr = r target -r current ; Among them, r target is the optimal imaging focus position coordinate, based on the above we can get: r current is the current probe position, Δr is the recommended direction vector for probe movement (expressed in the form of a three-dimensional arrow). This offset is used to generate a three-dimensional direction arrow to guide the operator to move towards the optimal imaging area;

[0075] Focus depth correction instruction: Combined with the set ultrasound focus depth f d =d i, read the current pressure value p of the probe pressure sensor, if it satisfies To perform motion compensation: Among them, β is the compensation coefficient, which is 0.2. Indicates that based on f d The initial set basic correction amount, δ d is the depth of focus adjustment after correction, as f d The compensation result can be fed back to the front-end ultrasonic beam controller to achieve real-time depth correction.

[0076] Basic correction amount Based on the image gradient change combined with f d By analyzing the grayscale gradient changes of the image in a small range before and after the focus position, the maximum position of the image clarity and the original set focus f are calculated. d The difference between them is used as the basic correction amount:

[0077] Among them, the basic focus depth correction Indicates the original focal depth f to be optimized for image clarity d The amount of fine-tuning performed, Δ represents the depth window allowed for fine-tuning, representing the optimization of image clarity near the focus, argmax represents the variable value that maximizes the expression in the brackets, d is the candidate focus depth value, in a d The value is taken within the depth range of the center, [f d -Δ,f d +Δ] represents the search range, that is, when the focus depth f is set d Find the optimal focus position within the upper and lower offset range, is the local change rate (gradient) of the attention heat map in the depth direction, which is used to evaluate the trend of image clarity changing with the focus depth. final (x,y) is the final attention heat map, the pixel brightness value at the spatial position (x,y), which is used to represent the attention of the anatomical structure or the boundary clarity.

[0078] Tactile feedback signal generation: δ d The corresponding adjustment direction is encoded as the handle vibration mode V pattern ,The direction is generated based on the depth adjustment and ,transmitted to the operator.

[0079] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0080] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. An ultrasound image acquisition and analysis system combining edge computing and attention analysis, characterized in that: include: The edge computing node receives the raw image data collected by the ultrasound device in real time, performs multi-scale spatial filtering on the raw image data, and generates a pre-processed image; Inputting the preprocessed image into an attention parsing unit, generating an attention heat map through a cascaded anatomical positioning module and a dynamic tracking module, wherein the high-brightness area in the attention heat map represents the real-time spatial distribution of key anatomical structures; Based on the brightness distribution feature analysis of the attention heat map, the scanning parameters of the ultrasound device are dynamically adjusted and operation guidance instructions are generated. The operation guidance instructions include probe movement direction suggestions and focus depth correction amounts.

2. The ultrasonic image acquisition and analysis system combining edge computing and attention analysis according to claim 1 is characterized in that: The generation of the pre-processed image specifically includes: Construct a three-stage filtering pipeline consisting of an adaptive Gaussian kernel, anisotropic diffusion kernel, and directional gradient kernel; The three-level filtering process is performed synchronously through the parallel computing unit of the edge computing node, in which the technical features of the output of the previous level of filtering are used as the input parameters of the next level of filtering to form a cascade optimization link.

3. The ultrasonic image acquisition and analysis system combining edge computing and attention analysis according to claim 2 is characterized in that: In the three-stage filtering pipeline: The adaptive Gaussian kernel dynamically adjusts the filter radius based on the signal-to-noise ratio of the original image data and outputs a primary denoised image; The anisotropic diffusion kernel suppresses the diffusion intensity of non-vascular tissue according to the local gradient direction of the primary denoised image and outputs a contrast-enhanced image; The directional gradient kernel extracts edge features matching preset anatomical directions in the contrast-enhanced image, and generates a preprocessed image including organ contour markings.

4. The ultrasonic image acquisition and analysis system combining edge computing and attention analysis according to claim 1, characterized in that: The anatomical localization module in S2 receives the preprocessed image, performs feature parsing through the parallel spatial attention submodule and channel attention submodule, generates an initial heat map and outputs a feature weight matrix; The dynamic tracking module receives the initial heat map and feature weight matrix of N consecutive frames to generate a dynamic enhanced heat map; The initial heat map and the dynamically enhanced heat map are weightedly fused through the heat map fusion unit, and the final output is an attention heat map that synchronizes the high-brightness area with the spatial distribution of key anatomical structures in real time.

5. The ultrasonic image acquisition and analysis system combining edge computing and attention analysis according to claim 4 is characterized in that: The spatial attention submodule uses a dilated convolutional layer to extract multi-receptive field features and generate an initial heat map, where the pixel value represents the probability of the existence of the anatomical structure; The channel attention submodule analyzes the response strength of different tissue channels through a deformable convolution kernel and outputs a feature weight matrix of blood vessels and organ boundaries.

6. The ultrasonic image acquisition and analysis system combining edge computing and attention analysis according to claim 5, characterized in that: The dynamic tracking module performs the following: Calculate the displacement vector of feature points between adjacent frames based on the optical flow field prediction network and generate motion compensation coefficients; The motion compensation coefficient is multiplied by the current frame feature weight matrix to generate a dynamic enhancement heat map.

7. The ultrasonic image acquisition and analysis system combining edge computing and attention analysis according to claim 6, characterized in that: The weight value during the weighted fusion is dynamically adjusted according to the moving speed of the ultrasound probe.

8. The ultrasonic image acquisition and analysis system combining edge computing and attention analysis according to claim 1, characterized in that: The analysis of the brightness distribution characteristics of the attention heat map includes segmenting the attention heat map using a region growing algorithm, extracting continuous areas whose brightness values ​​exceed a preset brightness threshold as key anatomical areas; calculating the brightness peak coordinates and coverage area of ​​each key anatomical area, and generating a structured parameter set including spatial position characteristics as the input basis for the scanning control strategy.

9. The ultrasonic image acquisition and analysis system combining edge computing and attention analysis according to claim 8, characterized in that: The dynamic adjustment of the scanning parameters of the ultrasound device specifically includes: adjusting the focal depth of the ultrasound transmission beam in real time according to the depth direction component of the brightness peak coordinate of the structured parameter set so that the focal depth matches the actual depth of the target anatomical structure; Based on the ratio of the coverage area of ​​the key anatomical region in the structured parameter set to the preset standard area, the imaging frame rate is adjusted: when the ratio is less than 80%, the frame rate is increased to 1.5 times the original value to enhance the ability to capture details.

10. The ultrasonic image acquisition and analysis system combining edge computing and attention analysis according to claim 9, characterized in that: The generation of the operation guidance instruction includes: Calculate the offset between the current position of the ultrasound probe and the optimal imaging focus position, and generate a movement suggestion including a three-dimensional direction arrow; Combining the focus depth correction amount and the current ultrasound probe pressure sensor data, a tactile feedback instruction is generated, which prompts the depth adjustment direction through the vibration pattern of the probe handle.

Images