Wearable cardiac ultrasound equipment and self-adaptive control method

Abstract

The invention belongs to the technical field of wearable ultrasonic medical equipment. According to the wearable cardiac ultrasound equipment and the self-adaptive control method, a phased array probe is fixed to a flexible attaching structure, and the phased array probe is fixed to the side, used for being attached to the skin, of the flexible attaching structure; the phased array probe comprises a plurality of independent ultrasonic transmitting units, each ultrasonic transmitting unit has an independent excitation signal, and ultrasonic beams at different angles are dynamically synthesized by controlling each ultrasonic transmitting unit so as to focus and deflect ultrasonic waves, so that an imaging area is adjusted; the phased array probe is in communication connection with a control terminal, and the control terminal is configured to execute switching of ultrasonic modes according to a manual instruction or automatically, and control the phased array probe to adjust an imaging mode according to the switched ultrasonic modes. Through switching of the ultrasonic modes, the heart ultrasonic detection precision of the wearable equipment is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wearable ultrasonic medical equipment, and in particular to a wearable cardiac ultrasonic equipment and an adaptive control method. Background Art

[0002] The statements in this section merely provide background art related to the present invention and do not necessarily constitute prior art.

[0003] Normal cardiac function is crucial for maintaining systemic perfusion throughout the body. Because cardiac function is constantly changing, long-term continuous monitoring is crucial for providing reliable and accurate diagnoses. Existing clinical imaging methods are limited to short-term use due to cumbersome equipment setup and high costs, making it impossible to continuously monitor the heart over extended periods. To address these challenges with existing ultrasound examinations, researchers have designed a wearable cardiac ultrasound device.

[0004] However, existing wearable cardiac ultrasound devices still have the following problems: The standard clinical cardiac ultrasound examination process includes not only B-ultrasound, but also color ultrasound (color Doppler ultrasound) and optional M-ultrasound. Current wearable cardiac ultrasound devices cannot automatically switch between B-ultrasound, color ultrasound, and M-ultrasound modes, and cannot adaptively switch ultrasound modes quickly for the wearer. In addition, existing wearable devices also have some other problems. For example, the collected cardiac ultrasound images cannot be fully processed in real time to determine whether the ultrasound images are qualified. They still need to be transmitted to a remote terminal for medical personnel to make judgments. For example, current cardiac imaging recommendations mainly rely on expert evaluation, lacking automated analysis of cardiac ultrasound images and automatic optimization of imaging recommendations, which may lead to the omission of important medical information. Summary of the Invention

[0005] In order to solve the problem of poor adaptive ultrasound mode switching capability of existing wearable cardiac ultrasound devices, the present invention provides a wearable cardiac ultrasound device and an adaptive control method. While ensuring the ease of use of the wearable cardiac ultrasound device, the present invention realizes ultrasound mode switching, improves the accuracy of cardiac ultrasound detection, and can perform continuous and controllable adaptive ultrasound examinations on the wearer's heart, greatly improving the wearer's user experience.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a wearable cardiac ultrasound device.

[0007] A wearable cardiac ultrasound device includes a flexible attachment structure for attachment to a part of a human body to be detected, wherein a phased array probe is fixed to the flexible attachment structure, and the phased array probe is fixed to a side of the flexible attachment structure for attachment to the skin; The phased array probe consists of multiple independent ultrasonic transmitting units, each of which has an independent excitation signal. By controlling each ultrasonic transmitting unit, ultrasonic beams of different angles are dynamically synthesized to focus and deflect the ultrasound, thereby adjusting the imaging area. The phased array probe is communicatively connected to a control terminal, and the control terminal is configured to switch the ultrasound mode according to a manual instruction or automatically, and control the phased array probe according to the switched ultrasound mode to adjust the imaging mode.

[0008] In an implementation of the first aspect of the present invention, the control terminal automatically switches the ultrasound mode, including: Enter B-ultrasound mode, and after obtaining the standard B-ultrasound section, detect the region of interest. After the region of interest is detected continuously within the set time and the difference between the regions of interest detected multiple times is less than the set threshold, switch to color ultrasound mode, and perform disease prediction based on the color ultrasound image in color ultrasound mode. After the prediction is completed, switch back to B-ultrasound mode; Whether to start M-mode ultrasound is determined based on the current standard section indicators or disease prediction results. If M-mode ultrasound is required, the system enters M-mode ultrasound mode and makes a final diagnosis of the disease based on the M-mode ultrasound image in M-mode ultrasound mode. If M-mode ultrasound is not required, the system continues to use the current B-mode ultrasound image to guide the phased array probe to find the next standard section until the last standard section is fully detected.

[0009] In an implementation of the first aspect of the present invention, the control terminal is further configured to: enhance the detected ultrasonic echo signal to obtain an ultrasonic image, perform image segmentation on the ultrasonic image, and perform ultrasonic quality evaluation on the segmented feature image; if the evaluation is unqualified, control the phased array probe to re-detect for re-evaluation until a qualified ultrasonic image is obtained after the evaluation is qualified.

[0010] As a further limitation, performing enhancement processing on the detected ultrasonic echo signal to obtain an ultrasonic image, and when the ultrasonic signal is ultrasonic echo data of a single angle, includes the following process: Perform IQ demodulation on the single-angle ultrasonic echo data to obtain complex IQ data containing amplitude and phase information; Performing time delay compensation on the complex IQ data to obtain time delayed IQ data; Input the time-delayed IQ data into the encoder to obtain a low-dimensional feature representation; The low-dimensional feature representation is input into a time-frequency feature fusion module to obtain a reconstructed and enhanced ultrasound image.

[0011] As a further limitation, in the control terminal, performing image segmentation on the ultrasound image includes: For the ultrasound image to be segmented, the complexity of the image to be segmented is measured through edge detection and texture analysis. After allocating the denoising step size of the diffusion model according to the complexity, local features of different scales are extracted through the diffusion model. For the image to be segmented, global features of different scales are obtained through the deblurring mask autoencoder model; The local features and global features of the same scale are paired, the similarity between the two paired features is calculated, and the attention weight is generated. After the paired features are fused based on the attention weight, the segmentation representation is obtained through the activation function. The image to be segmented is segmented based on the segmentation representation to obtain the segmented feature image.

[0012] As a further limitation, in the control terminal, performing ultrasound quality assessment on the segmented feature image includes: Extracting multi-scale spatial features of the segmented feature image, and fusing the multi-scale spatial features to obtain multi-scale spatial fusion features; Perform spatial attention enhancement and channel attention enhancement on the multi-scale spatial fusion features to obtain multi-scale spatial fusion features with enhanced spatial attention and channel attention; The heart contour clarity and echo uniformity are calculated based on the multi-scale spatial fusion features enhanced by spatial attention, and the heart valve clarity is calculated based on the multi-scale spatial fusion features enhanced by channel attention. Extracting features of the feature image at different time points, and calculating the consistency of left ventricular wall motion based on the features of the feature image at different time points; Extract the signal-to-noise ratio of the feature image; Ultrasound quality assessment was performed based on cardiac outline clarity, echo homogeneity, cardiac valve clarity, left ventricular wall motion consistency, and signal-to-noise ratio.

[0013] In a second aspect, the present invention provides an adaptive control method for a wearable cardiac ultrasound device.

[0014] An adaptive control method for a wearable cardiac ultrasound device includes the following steps: Acquire an ultrasound image, and automatically switch the ultrasound mode according to the acquired ultrasound image; Performing enhancement processing on the ultrasonic images detected under different ultrasonic modes; Perform image segmentation on the enhanced ultrasound image; The ultrasound quality evaluation is performed on the segmented feature image. If the evaluation fails, the ultrasound image data is re-collected for re-evaluation until a qualified ultrasound image is obtained after the evaluation passes.

[0015] In a third aspect, the present invention provides an adaptive control system for a wearable cardiac ultrasound device.

[0016] An adaptive control system for a wearable cardiac ultrasound device, comprising: A mode switching unit is configured to: acquire an ultrasound image and automatically switch the ultrasound mode according to the acquired ultrasound image; An image enhancement unit is configured to: perform enhancement processing on the ultrasonic images detected in different ultrasonic modes; An image segmentation unit is configured to: perform image segmentation on the enhanced ultrasound image; The ultrasound evaluation unit is configured to: perform ultrasound quality evaluation on the segmented feature image; if the evaluation fails, re-collect ultrasound image data for re-evaluation until a qualified ultrasound image is obtained after the evaluation passes.

[0017] In a fourth aspect, the present invention provides a wearable cardiac ultrasound device, comprising: a processor and a computer-readable storage medium; a processor adapted to execute a computer program; A computer-readable storage medium having a computer program stored therein, wherein when the computer program is executed by the processor, the computer program implements the adaptive control method for a wearable cardiac ultrasound device according to the second aspect of the present invention.

[0018] In a fifth aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program is suitable for being loaded by a processor and executing the adaptive control method for a wearable cardiac ultrasound device as described in the second aspect of the present invention.

[0019] In a sixth aspect, the present invention provides a computer program product, comprising a computer program. When the computer program is executed by a processor, the computer program implements the adaptive control method for a wearable cardiac ultrasound device as described in the second aspect of the present invention.

[0020] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention innovatively develops a wearable cardiac ultrasound device, which controls the detection of a phased array probe according to a set ultrasound mode, enhances the detected ultrasound image, performs image segmentation on the enhanced ultrasound image, and performs ultrasound quality assessment on the segmented feature image. If the assessment is unqualified, the phased array probe is controlled to re-detect and re-evaluate until a qualified ultrasound image is obtained after the assessment is qualified. The present invention realizes adaptive control of the phased array probe according to the evaluation results through an integrated design of real-time acquisition, enhancement, segmentation and evaluation of ultrasound images. While ensuring the ease of use of the wearable cardiac ultrasound device, the accuracy of cardiac ultrasound detection is improved, the accuracy of the collected ultrasound image is guaranteed, and continuous, controllable and adaptive ultrasound examination of the wearer's heart can be performed, greatly improving the wearer's user experience.

[0021] 2. The present invention innovatively proposes a cardiac ultrasound mode switching strategy and a prompt-guided full-process automated cardiac ultrasound examination strategy. Through artificial intelligence (including but not limited to deep learning, deep reinforcement learning, etc.), it guides the automatic movement and adjustment of the phased array probe and the automatic switching between various modes (B-ultrasound, color ultrasound and M-ultrasound), thus realizing the full process and automation of cardiac ultrasound examination.

[0022] 3. This invention innovatively proposes an enhancement strategy for cardiac ultrasound images. Through a deep learning-based image reconstruction model, the imaging quality of wearable ultrasound devices is significantly improved. Under the limitations of low power consumption and single-angle plane wave emission, the generated images are close to the effects of composite multi-angle imaging (CPWC), while retaining the high frame rate characteristics of single-angle plane wave imaging. The time-frequency feature fusion module (TFF) is introduced, which effectively extracts and utilizes time domain and frequency domain information through multi-branch design and dynamic weight fusion mechanism. Compared with traditional models, the resolution and contrast of the image are further improved. The generated high-quality images are suitable for dynamic monitoring and clinical diagnosis, providing reliable support for the practical application of wearable ultrasound devices.

[0023] 4. The present invention innovatively proposes a segmentation strategy for color ultrasound images, which uses a dual-branch architecture for image segmentation. One branch uses a diffusion model to process images, and the other branch uses a deblurred mask autoencoder model to process images. Through the cross-modal attention mechanism, the features extracted by the two models are aligned, and the semantic representation is enhanced. There are significant improvements in segmentation accuracy, detail extraction of lesion boundaries, and the ability to capture multi-scale lesions, adapting to the challenges of different scales and complex backgrounds in medical images; through the cross-modal attention mechanism, the features of the diffusion model and the deblurred MIM model are aligned, and local and global features are fused at each layer, which is more coordinated in capturing details and overall structures. This mechanism enables the image segmentation model to pay attention to subtle boundary features and overall semantic consistency at the pixel level, thereby improving the segmentation effect.

[0024] 5. The present invention innovatively proposes a color ultrasound image evaluation strategy, which extracts multiple key cardiac features according to the set image quality evaluation criteria, evaluates the quality of cardiac ultrasound images based on the extracted multiple key cardiac features, and generates shooting suggestions based on the quality evaluation results. It can automatically evaluate the quality of cardiac ultrasound images and provide shooting optimization suggestions in real time, thereby achieving more efficient and accurate cardiac ultrasound image processing.

[0025] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0027] Figure 1 A schematic diagram of a wearable cardiac ultrasound device according to an exemplary embodiment of the present invention is provided; Figure 2 A cross-sectional view of a wearable cardiac ultrasound device according to an exemplary embodiment of the present invention; Figure 3 A schematic diagram of an adaptive control method for a wearable cardiac ultrasound device provided by an exemplary embodiment of the present invention; Figure 4 A schematic diagram of a cardiac ultrasound mode switching method provided by an exemplary embodiment of the present invention; Figure 5 A schematic diagram of an initial B-ultrasound image provided by an exemplary embodiment of the present invention; Figure 6 A schematic diagram of a standard B-ultrasound section provided by an exemplary embodiment of the present invention; Figure 7A schematic diagram of enabling color Doppler ultrasound mode for a region of interest provided by an exemplary embodiment of the present invention; Figure 8 A schematic diagram of an M super mode provided for an exemplary embodiment of the present invention; Figure 9 A schematic diagram of the default position of a sampling line provided by an exemplary embodiment of the present invention; Figure 10 A schematic diagram of a cardiac ultrasound image enhancement method provided by an exemplary embodiment of the present invention; Figure 11 A schematic diagram of a cardiac ultrasound image segmentation method provided by an exemplary embodiment of the present invention; Figure 12 A schematic diagram of a cardiac ultrasound image evaluation method provided by an exemplary embodiment of the present invention; Figure 13 A schematic diagram of an adaptive control system of a wearable cardiac ultrasound device provided by an exemplary embodiment of the present invention; Figure 14 A schematic diagram of a computer device provided for an exemplary embodiment of the present invention; Among them, 1. Power switch; 2. Imaging mode switching knob; 3. Status indicator light; 4. Flexible attachment structure; 5. Phased array probe. DETAILED DESCRIPTION

[0028] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0029] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0030] Wearable cardiac ultrasound technology is a major breakthrough in the medical and health field in recent years. Through miniaturization, flexibility and intelligent design, it has achieved real-time monitoring of cardiac function. This technology not only overcomes the limitations of traditional ultrasound equipment, which is large in size and relies on professional operators, but also provides a new solution for clinical diagnosis, intensive care and family health management. However, existing wearable cardiac ultrasound equipment has great limitations in monitoring accuracy and applicability. In view of this, the present invention proposes a wearable cardiac ultrasound device, such as Figure 1 and Figure 2 As shown, the wearable cardiac ultrasound device includes: a control terminal, a flexible attachment structure 4 (for example, a flexible patch), and a phased array probe 5. The phased array probe 5 is fixed to the side of the flexible attachment structure 4 for contact with the skin. The phased array probe 5 is in communication with the control terminal. The entire wearable cardiac ultrasound device is fixed to the user's measured part by a strap. The control terminal is configured to: switch the ultrasound mode according to manual instructions or automatic instructions, and control the phased array probe to adjust the imaging mode according to the switched ultrasound mode.

[0031] In the present invention, the phased array probe 5 transmits ultrasonic waves to the heart. The ultrasonic waves propagate and reflect back in the heart tissue and blood, and are received by the probe and converted into image information. This allows for clear observation of the structure and function of the heart, such as whether the size of each chamber of the heart is normal, whether the wall thickness is uniform, and whether there is myocardial hypertrophy or thinning. For heart valves, the morphology and opening and closing of the mitral valve, tricuspid valve, aortic valve, and pulmonary valve can be determined, as well as whether there is stenosis or regurgitation. The large blood vessels around the heart, such as the aorta and pulmonary artery, can also be detected, including their diameter, wall condition, and blood flow direction and velocity.

[0032] Phased array probes 5 have a variety of different array configurations. Based on the arrangement of the array elements, they can be categorized as one-dimensional linear arrays, two-dimensional matrix arrays, annular arrays, sector arrays, concave arrays, convex arrays, and dual-linear arrays. Different array configurations produce varying acoustic field characteristics, enabling phased arrays to be used for different inspection conditions. For example, annular arrays can achieve two-dimensional focusing of the beam profile, achieving greater energy concentration without requiring a large array number, and can maintain high penetration even at high frequencies. However, annular arrays cannot control beam deflection, and therefore are primarily used in medical imaging.

[0033] In the present invention, preferably, the control terminal is connected to an imaging mode switching knob 2 and a power switch 1, and the imaging mode switching knob 2 is used to switch the ultrasound mode; the control terminal is connected to a status indicator light 3, and the status indicator light 3 is used for the working status of the ultrasound equipment, and the working status includes power on, working and fault; the ultrasound mode includes: B-ultrasound mode, M-ultrasound mode and color ultrasound mode; the control terminal includes a display module; or the control terminal is connected to an external display module.

[0034] The core principle of the phased array in this invention is to change the beam direction by electronically controlling the timing of ultrasonic emission. This technology relies on multiple independent ultrasonic transmitting units (array elements), each of which can independently control the phase and amplitude of its excitation signal. By precisely adjusting these parameters, ultrasonic beams of different angles can be dynamically synthesized to achieve ultrasonic focusing and deflection, thereby adjusting the imaging area. The specific working method includes: Assume that the signal of the ultrasonic array element for: (1); in: For array element The signal amplitude, is the angular frequency of the ultrasonic wave, For array element phase delay.

[0035] If the target imaging direction is , the required phase delay It can be calculated by the following formula: (2); in: For array element The distance to the center of the phased array, is the wavelength of ultrasound.

[0036] During the adjustment process, the beam angle can be optimized using the following adaptive update strategy: (3); in: is the updated beam angle, is the current beam angle, Score the clarity of the goal, Rate the current clarity. is the adjustment coefficient (which can be determined by experiment, usually within the range of 5 degrees to 15 degrees).

[0037] The specific operating method of the wearable cardiac ultrasound device includes: (1) Prepare equipment.

[0038] Check the battery level: Make sure the device is charged or connected to a power source; Turn on the power: Press the power switch and confirm that the status indicator ring lights up, indicating that the device is turned on.

[0039] (2) Set the imaging mode.

[0040] Select mode: According to the body part and purpose to be examined, rotate the imaging mode switch knob to select the appropriate ultrasound imaging mode. The mode can also be selected adaptively.

[0041] (3) Prepare the skin and coupling agent.

[0042] Cleanse the skin: Cleanse the skin on the body area to be examined to remove oil and dirt; Apply coupling agent: Apply an appropriate amount of coupling agent evenly on the phased array probe and skin to ensure good sound wave transmission.

[0043] (4) Fixed equipment.

[0044] Device placement: Place the device body on the body part to be examined, ensuring that the phased array probe is in close contact with the skin; Attach the patch: Securely attach the device to the wearer's body via the flexible attachment structure 4, and gently press the flexible attachment structure 4 to allow it to adhere tightly to the skin.

[0045] (5) Start inspection.

[0046] Adjust position: Fine-tune the position and angle of the device as needed to obtain the best ultrasound image; Observe the image: Observe the ultrasound image through the connected external display or the display module of the control terminal, and perform necessary recording and analysis.

[0047] (6) Complete the inspection.

[0048] Turn off the power: After completing the inspection, turn off the power switch; Remove the device: Remove the flexible attachment structure 4, avoiding pulling on the skin; Clean the probe and skin: Use a clean cloth or tissue to wipe off the coupling gel on the phased array probe and skin.

[0049] In the present invention, preferably, the control terminal can automatically switch the ultrasound mode, including: Enter B-ultrasound mode, and after obtaining the standard B-ultrasound section, detect the region of interest. After the region of interest is detected continuously within the set time and the difference between the regions of interest detected multiple times is less than the set threshold, switch to color ultrasound mode, and perform disease prediction based on the color ultrasound image in color ultrasound mode. After the prediction is completed, switch back to B-ultrasound mode; Whether to start M-mode ultrasound is determined based on the current standard section indicators or disease prediction results. If M-mode ultrasound is required, the system enters M-mode ultrasound mode and makes a final diagnosis of the disease based on the M-mode ultrasound image in M-mode ultrasound mode. If M-mode ultrasound is not required, the system continues to use the current B-mode ultrasound image to guide the phased array probe to find the next standard section until the last standard section is fully detected.

[0050] In the present invention, preferably, the control terminal is also configured to: enhance the detected ultrasonic image, segment the enhanced ultrasonic image, and evaluate the ultrasonic quality of the segmented feature image. If the evaluation is unqualified, control the phased array probe 5 to re-detect and re-evaluate until a qualified ultrasonic image is obtained after the evaluation is qualified.

[0051] The ultrasonic data obtained by detection is enhanced, taking single-angle ultrasonic echo data as an example, including: IQ demodulation is performed on single-angle ultrasonic echo data to obtain complex IQ data containing amplitude and phase information; time delay compensation is performed on the complex IQ data to obtain time-delayed IQ data; the time-delayed IQ data is input into an encoder to obtain a low-dimensional feature representation; the low-dimensional feature representation is input into a time-frequency feature fusion module to obtain a reconstructed and enhanced ultrasonic image.

[0052] It is understandable that conventional image enhancement methods may be used to enhance the ultrasound image generated by multi-angle ultrasound echo data, such as a histogram equalization algorithm, a Gaussian filter algorithm, etc., which will not be described in detail here.

[0053] In the present invention, preferably, in the control terminal, image segmentation is performed on the enhanced ultrasound image, including: for the ultrasound image to be segmented, measuring the complexity of the image to be segmented through edge detection and texture analysis, allocating the denoising step size of the diffusion model according to the complexity, and then extracting local features of different scales through the diffusion model; for the image to be segmented, obtaining global features of different scales through a deblurring mask autoencoder model; pairing local features and global features of the same scale, calculating the similarity between the two paired features, generating attention weights, fusing the paired features based on the attention weights, obtaining segmentation representations through activation functions, segmenting the image to be segmented based on the segmentation representations, and obtaining feature images after segmentation.

[0054] In the present invention, preferably, in the control terminal, performing ultrasound quality assessment on the segmented feature image includes: Extract multi-scale spatial features from the segmented feature image, and fuse the multi-scale spatial features to obtain multi-scale spatial fusion features; perform spatial attention enhancement and channel attention enhancement on the multi-scale spatial fusion features to obtain multi-scale spatial fusion features with spatial attention enhancement and channel attention enhancement; calculate the heart contour clarity and echo uniformity based on the multi-scale spatial fusion features enhanced with spatial attention, and calculate the heart valve clarity based on the multi-scale spatial fusion features enhanced with channel attention; extract features of the feature image at different time points, and calculate the left ventricular wall motion consistency based on the features of the feature image at different time points; extract the signal-to-noise ratio of the feature image; and perform ultrasound quality assessment based on heart contour clarity, echo uniformity, heart valve clarity, left ventricular wall motion consistency, and signal-to-noise ratio.

[0055] The wearable cardiac ultrasound device proposed in the present invention realizes the full-process processing of real-time acquisition, enhancement, segmentation and evaluation of ultrasound images, and realizes adaptive control of the phased array probe based on the evaluation results. While ensuring the convenience of use of the wearable cardiac ultrasound device, the accuracy of cardiac ultrasound detection is improved.

[0056] Given that current wearable cardiac ultrasound devices cannot achieve the integration of real-time acquisition, enhancement, segmentation and evaluation of ultrasound images, Figure 3 An adaptive control method for a wearable cardiac ultrasound device is shown, comprising the following processes: S1: Acquire an ultrasound image and automatically switch the ultrasound mode according to the acquired ultrasound image; S2: performing enhancement processing on the ultrasonic images detected under different ultrasonic modes; S3: performing image segmentation on the enhanced ultrasound image; S4: Performing ultrasound quality assessment on the segmented feature image. If the assessment fails, re-collecting ultrasound image data for re-assessment until a qualified ultrasound image is obtained after the assessment passes.

[0057] In S1 of the present invention, an ultrasonic image obtained by a wearable cardiac ultrasound device is obtained, and the ultrasonic mode is automatically switched according to the obtained ultrasonic image. Specifically, Figure 4 Shown, including: S1.1: The B-ultrasound image acquired in real time by the phased array probe is used as the initial input, such as Figure 5 As shown (the image quality is poor and cannot be used for effective measurement and diagnosis of the heart).

[0058] S1.2: Input the initial B-ultrasound image into the B-ultrasound prompt guidance module, and control the emission angle of the phased array probe 5 for the initial B-ultrasound image so that the phased array probe moves to the first standard section of cardiac ultrasound examination, as shown in the attached figure. Figure 6 As shown, other index measurements and auxiliary diagnostic tasks are then performed based on the first standard section, such as whether there are defects in the atrial septum and ventricular septum, and whether the valve is adhered or thickened.

[0059] S1.3: After obtaining the first standard ultrasound section, the region of interest is detected. For example, to diagnose mitral regurgitation, the position of the mitral valve is detected. Here, the target detection method (such as the Yolo model) can be used to detect the position of the mitral valve using a rectangular frame. Since the mitral valve in the video is constantly moving, the largest rectangular frame is selected as the mitral valve position.

[0060] S1.4: After the region of interest, such as the mitral valve, is detected continuously and stably (for example, the region of interest remains basically unchanged for 2 consecutive seconds, that is, the range of change of the identification frame of the region of interest is less than the set threshold), switch to the color Doppler ultrasound mode, which simulates the doctor's operation of pressing a button, as shown in the attached figure. Figure 7 shown.

[0061] S1.5: Use deep learning methods to analyze color Doppler ultrasound images, detect color Doppler ultrasound indicators, and assist in the diagnosis of heart disease. For example, for mitral regurgitation, the color Doppler ultrasound image frame corresponding to the mitral valve location frame is used as the neural network input, with labels indicating the presence of regurgitation and the degree of regurgitation (mild, moderate, or severe). If the color Doppler ultrasound image at the mitral valve location is primarily red, but there is a significant blue area (not pseudo-color) at the mitral valve junction, this indicates moderate or even severe mitral regurgitation. The neural network is then used to predict the regurgitation results in the color Doppler ultrasound image.

[0062] S1.6: After the color Doppler ultrasound measurement and analysis is completed, the color Doppler ultrasound is turned off and the B-mode ultrasound is switched to. The M-mode ultrasound is then turned on based on the current standard section or the severity of the disease. For example, if the B-mode ultrasound of the ventricular septum shows abnormalities such as thickening or irregular movement in the parasternal left ventricular long axis section, the M-mode ultrasound is then turned on for further examination. Figure 8 To more accurately measure cardiac volume and ventricular wall thickness, see the figure below. If M-mode ultrasound is required, proceed to S1.7. If not, skip to S1.8.

[0063] S1.7: Based on the location of the region of interest, the position of the M-supersampling line is prompted and guided. If there is no doctor intervention, the line connecting the center point P2 and the vertex P1 of the region of interest is selected as the position of the M-supersampling line by default, as shown in the attached figure. Figure 9 As shown, M-ultrasound is turned on to obtain a continuous waveform of M-ultrasound, corresponding indicators are measured according to the waveform, and the disease is diagnosed according to the indicators.

[0064] S1.8: If this is the last standard section of the cardiac examination, then end; otherwise, jump to S1.2.

[0065] In S2 of the present invention, the ultrasonic images detected under different ultrasonic modes are enhanced, specifically, as follows: Figure 10 As shown, including: for the ultrasonic echo data RF, select any single angle , which is , that is, the echo data of the plane wave at a single angle is used as input, and the echo data of the plane wave at all angles is used as input. The CPWC method is used for imaging, and the high-quality image is taken as the target, which is recorded as HQI. An example of a dataset composed of HQI, which mainly includes two processes: echo data re-representation and image reconstruction model.

[0066] In the present invention, the echo data re-representation stage specifically includes: S2.1: Echo data for a single angle Perform IQ demodulation to obtain complex IQ data containing amplitude and phase information, which is recorded as: (4); in, Indicates the emission angle is , by The echo signal received by each array element is is the carrier frequency.

[0067] S2.2: Calculate the time of flight (TOF) based on the coordinates of the imaging point and the position of the transmitting array element. Then apply time delay compensation to the IQ data to obtain the time delayed IQ data (TDIQ): (5) ； (6); (7); (8); in, and represents the coordinates of the imaging point, and Indicates the The position coordinates of the probes, The default value is 0. represents the speed of light, The emission angle is IQ data at that time.

[0068] S2.3: Input to the autoencoder structure for data compression and dimensionality reduction. Specifically, The data first passes through the encoder for feature extraction and compression. The encoder is designed to significantly reduce the amount of data and redundant information while preserving the key features of the imaging point to the greatest extent possible. The output of the encoder is a compact low-dimensional feature representation, which will serve as the input for subsequent image reconstruction. At the same time, to ensure the effectiveness of the compressed features, the autoencoder structure also includes a decoder module (Decoder), which is used to restore the low-dimensional feature representation generated by the encoder to the original The purpose of this process is to evaluate the degree of information retention during compression and restoration, and to ensure that the low-dimensional features extracted by the encoder can cover the key information of the original data.

[0069] In the present invention, the image reconstruction model specifically includes: In order to fully extract the The present invention proposes a time-frequency feature fusion module (TFF) as a basic module in the network structure, which is applicable to network structures including but not limited to Unet and ResNet. The module extracts and fuses the time domain and frequency domain features of the input features through a multi-branch design to generate high-quality feature representations, providing key support for image reconstruction. The image reconstruction model of the present invention is based on the time-frequency feature fusion module. The time-frequency feature fusion module includes two main parts: a feature extraction module (TFE) and a feature fusion module (FFM). The feature extraction module consists of a time domain branch and a frequency domain branch, which is used to extract comprehensive time domain and frequency domain information from the input features, while the feature fusion module performs weighted fusion on the features output by multiple branches as the output of the time-frequency feature fusion module.

[0070] The feature extraction module first performs input features Enter the time domain branch and frequency domain branch for processing respectively. The time domain branch obtains the time domain feature representation by extracting the time domain feature Specifically, the input features first pass through a normalization layer to adjust the numerical range of the features and reduce the deviation of the data distribution. The normalized features pass through a convolution layer to capture the change pattern in the time dimension through the local receptive field. After the convolution operation, the features are nonlinearly transformed through the activation function to further enhance the expressive power of the features. Finally, the time domain branch outputs the time domain feature representation .

[0071] The frequency domain branch is further divided into a low-frequency branch and a high-frequency branch, which process the low-frequency and high-frequency components of the signal respectively. The low-frequency branch first converts the input features from the time domain to the frequency domain through a two-dimensional Fourier transform (2DFFT). The result of the Fourier transform contains the amplitude and phase information of the signal. Then, it passes through a low-pass filter with learnable parameters to extract the low-frequency components of the signal. Subsequently, the low-frequency components pass through the low-frequency feature extraction module to obtain the low-frequency feature representation. Specifically, the low-frequency components enter the complex domain normalization layer to adjust the amplitude distribution of the frequency domain data to make it more suitable for subsequent convolution operations. The normalized features pass through a complex convolution layer to further capture the local feature information in the low-frequency components. Subsequently, the convolution features are nonlinearly transformed through a complex activation function to enhance the expressive power of the features. Finally, the low-frequency features are restored to the spatiotemporal domain through a two-dimensional inverse Fourier transform (2DIFFT) to generate a low-frequency feature representation. .

[0072] The processing of the high-frequency branch is similar to that of the low-frequency branch. First, the input features are converted to the frequency domain through a two-dimensional Fourier transform; then, a high-pass filter with learnable parameters is used to extract the high-frequency components of the signal. Subsequently, the high-frequency components are passed through a high-frequency feature extraction module to obtain a high-frequency feature representation. Specifically, the high-frequency components pass through the complex domain normalization layer to adjust the distribution of features; the normalized features are passed through the complex convolution layer to capture the detailed characteristics of the high-frequency information. Subsequently, the convolution features pass through the complex activation function to further enhance the expressive power of the high-frequency features. Finally, the high-frequency features are restored to the spatiotemporal domain through the two-dimensional inverse Fourier transform to generate the final high-frequency feature representation. .

[0073] The feature fusion module receives the feature representations output by the time domain branch, low-frequency branch, and high-frequency branch, and performs weighted fusion on them. First, the three features are integrated through an adaptive weight learning module to generate three feature weights. These weights are normalized by a normalization function to ensure that the sum of the weights is 1. The normalized weights are multiplied by the time domain features, low-frequency features, and high-frequency features respectively to perform weighted processing on the features. Finally, the weighted features are linearly superimposed, then converted to the frequency domain through a two-dimensional Fourier transform, filtered through learnable parameters, and then transformed back to the time-space domain to obtain the output of the time-frequency feature fusion module .

[0074] It should be noted that the feature extraction structure of the time domain branch, the feature extraction structure of the frequency domain branch after filtering, and the adaptive weight learning module are not limited to the combination of the normalization layer, convolution layer, and nonlinear layer mentioned above.

[0075] Through the above structural design, the image reconstruction model combines and builds TFF as the basic module, and finally reconstructs a high-quality image, ensuring that the model output has good clarity and contrast.

[0076] Current cardiac ultrasound images often have complex background interference, subtle lesion boundaries, and multi-scale structural features. Traditional cardiac ultrasound image segmentation usually relies on feature extraction and segmentation prediction of a single model. A single model is difficult to simultaneously take into account semantic information of different scales and levels, resulting in limited segmentation results. In addition, noise and blur in medical images further increase the difficulty of accurate segmentation. In view of this, the present invention proposes a method for image segmentation of enhanced ultrasound images, specifically, as follows: Figure 11 As shown, the following process is included: S3.1: Measure the complexity of the image to be segmented through edge detection and texture analysis. After allocating the denoising step size of the diffusion model based on the current complexity, extract semantic features of different scales through the diffusion model.

[0077] The diffusion model works by gradually adding noise to a clean image, converting it into a noisy image. The original image is then restored through a reverse denoising process. The diffusion model generates image features over multiple time steps T and uses a noise predictor (i.e., a U-Net network model) to generate activation states at different stages. The U-Net model is the core module of the diffusion model, primarily performing iterative denoising on the Gaussian noise matrix in a diffusion loop. Using the diffusion model's reverse diffusion process, intermediate activation states of the labeled image are used to extract semantic features. During this reverse diffusion process, semantic features at different levels can be extracted at different levels of the U-Net architecture for subsequent analysis and processing.

[0078] In particular, this embodiment proposes an adaptive back-diffusion mechanism, which introduces dynamic step size adjustment based on image complexity during the back-diffusion process. The specific implementation method is to calculate the complexity of the current state of the image before each back-diffusion step and allocate a more detailed step size based on the complexity.

[0079] S3.1.1: Initial setup for back diffusion: First, a diffusion model is pre-trained based on a large-scale image dataset (e.g., ImageNet). Then, the pre-trained model is fine-tuned using a domain-specific ultrasound dataset to adapt to the characteristic distribution of medical images. After fine-tuning, the diffusion model begins the back diffusion process: the noisy image is gradually denoised using a noise predictor based on the U-Net architecture. This process gradually removes noise and restores the image from a random noise state to a clear image.

[0080] Before entering each back diffusion step, the diffusion model will dynamically calculate the complexity of the current image to determine the step size of the current denoising.

[0081] S3.1.2: Calculate the complexity of the current image: Before each back-diffusion step, extract the current image state and measure the image complexity through edge detection and texture analysis.

[0082] Based on the complexity judgment of edge detection, the edge can represent the outline of the lesion area. The edge density and edge strength of the image can be used to judge the image complexity. The Sobel operator or Canny edge detection operator is used to obtain the edge gradient of the image.

[0083] Specifically, for a given image 𝑥, the Sobel operators in the horizontal and vertical directions are used to obtain the gradients respectively: (9); (10); in, and are the horizontal and vertical Sobel kernels, represents the first pixels.

[0084] Edge Strength It can be expressed as: (11).

[0085] Average edge strength complexity for: (12); N represents the total number of edge pixels involved in the calculation, which can be expressed as: (13); Where W and H are the width and height of the image respectively; It is an indicator function that takes the value 1 when the condition is true and 0 otherwise; E(x,y)>ϵ means that the current pixel is considered to be an edge.

[0086] Based on the complexity judgment of texture features, on the basis of the Sobel operator, the texture complexity is expressed by calculating the average value of the local gradient intensity: (14).

[0087] The complexity of the two can be combined into a comprehensive complexity by linear combination : (15); Among them, α and β are weight coefficients.

[0088] S3.1.3: Adjust the denoising step size according to the overall complexity.

[0089] During back-diffusion, the diffusion model assigns a dynamic step size based on the current complexity. If the overall complexity is high (e.g., complex edges and textures, rich detail information), the step size is reduced to achieve more detailed denoising and preserve detailed features. If the overall complexity is low, the step size is increased, reducing the level of denoising to improve efficiency. Adaptive denoising with a dynamic step size enhances the preservation of local image details, resulting in richer and more accurate extraction of multi-scale semantic features.

[0090] The traditional diffusion model's reverse denoising process gradually restores the image using a fixed noise step size. This strategy can be inadequate for complex medical images or those with subtle features. By introducing an adaptive reverse diffusion mechanism, the diffusion model can dynamically adjust the denoising step size and intensity based on the image feature distribution and the current denoising state to more precisely preserve and extract semantic features. For example, if the diffusion model detects that the current image feature is very important, the step size can be reduced for more detailed denoising. Conversely, if the noise has little impact on the image semantics, the step size can be increased, reducing the number of denoising steps to improve efficiency.

[0091] S3.1.4: Perform dynamic step-size backdiffusion denoising: Perform denoising in each backdiffusion step using an adjusted step size. A small step size ensures that details in complex areas are gradually restored, avoiding information loss; a large step size allows for faster denoising in simpler areas.

[0092] In each denoising step, different layers of U-Net generate multi-scale activation states, through which semantic features at different levels are extracted.

[0093] S3.1.5: Multi-scale feature fusion and semantic feature extraction.

[0094] In the middle layer of each back-diffusion step, semantic features are extracted from the activation states of different scales of U-Net, and a multi-scale feature pyramid structure is used to ensure the comprehensiveness of feature information.

[0095] In order to extract higher-quality semantic features with multi-scale fine-grainedness, this embodiment introduces a multi-scale dynamic weight feature pyramid mechanism to improve the expressiveness of features at different scales in medical images. It adopts a multi-scale U-Net architecture to capture rich information in the image from global structure to local details by simultaneously performing feature extraction and noise prediction at different scales in the diffusion process.

[0096] Specifically, at the U-Net encoder stage, a multi-scale feature pyramid was constructed. Each scale feature was dynamically weighted by the scale assessment module, and the weight ratio was adjusted through soft normalization (Softmax) for further weighted fusion. To ensure that each scale feature can effectively represent information of different granularities, the feature pyramid adopts a convolution kernel design of different sizes during the generation process, thereby promoting the diffusion model to take into account both macroscopic structures and fine-grained features of subtle lesions when restoring features. In addition, to further improve feature selectivity, a hybrid weighted attention mechanism was introduced at the U-Net encoder stage, namely a dual-branch structure of channel attention and spatial attention. This can dynamically focus on key areas and channels with more diagnostic significance during the multi-scale feature fusion process, thereby improving the model's responsiveness to key information.

[0097] S3.2: For the image to be segmented, the MIM model and multi-scale features are fused to obtain a fused feature map.

[0098] The MIM model simulates missing image blocks by randomly occluding and blurring the image. The encoder learns to reconstruct these missing blocks, thereby extracting detailed representations of the image. This process uses convolution operations to extract local features and the Vision Transformer to capture global information. The Transformer is a deep learning model architecture that incorporates a self-attention mechanism. The Vision Transformer applies the Transformer to vision tasks, demonstrating its powerful ability to model global information.

[0099] The MIM model combines multi-scale convolution with the Vision Transformer structure. Local convolution operations help capture details and local information, while global Transformer operations effectively obtain long-distance dependencies and global information. This approach ensures that the model can process details while taking into account the coherence of global semantics.

[0100] S3.2.1: Random occlusion and blurring.

[0101] The input image is preprocessed by randomly blocking several image blocks to make the information in these areas missing. At the same time, the blocked areas are blurred to simulate the situation where image details are blurred or lost. This allows the MIM model to learn how to complete and restore the lost areas.

[0102] S3.2.2: MIM encoder structure.

[0103] Multi-scale convolution layer: Multi-scale convolution kernels are first applied in the encoder to obtain details of the image at different scales. These convolution operations capture local features at different scales, allowing the model to have a deeper understanding of the details of the occluded area and its neighboring parts.

[0104] Use convolution kernels of different sizes (such as 3×3, 5×5, etc.) to extract multi-scale features to ensure that the detailed changes in the image are captured.

[0105] Transformer layer: After the multi-scale convolutional layer, the Transformer module is introduced to extract global semantic information. The Transformer structure uses the self-attention mechanism to model the long-distance dependencies between various regions in the image, thereby obtaining global information.

[0106] In the attention mechanism, the features of multi-scale convolution output are considered so that the information of the occluded area can be better integrated with the surrounding areas.

[0107] Through the self-attention mechanism, the MIM model can complete and restore the semantic content of the occluded area at the global level.

[0108] S3.2.3: Multi-scale feature fusion module: First, the multi-scale feature fusion module receives the local features extracted by the multi-scale convolution layer (such as 3×3, 5×5, and 7×7 convolution outputs) and the global features of the Transformer layer. The dimensions of each feature may be different, so feature alignment and dimensionality reduction processing are required.

[0109] Among them, feature alignment and dimensionality reduction processing include: (1) Channel dimension reduction: Since the features from convolutions and Transformers of different scales may have different channel numbers, the channel numbers of these features are first aligned through 1×1 convolution to ensure that all features have consistent channel dimensions.

[0110] (2) Spatial alignment: If the spatial resolutions of multi-scale feature maps are different, bilinear interpolation or transposed convolution can be used to upsample or downsample the feature maps to align all feature maps to the same spatial size for subsequent fusion.

[0111] Next, multi-scale feature weight generation is performed, including: (1) Feature importance weight: In order to dynamically adjust the importance of features at different scales, a weight generator is introduced. Usually, a lightweight fully connected layer or attention mechanism is used to assign weights to features at each scale. By learning these weights, the MIM model can adaptively adjust the contribution of features at each scale according to the content of the input image.

[0112] (2) Calculation method: The features of each scale are input into the weight generator to generate the corresponding weight value, and then normalized by softmax (normalized exponential function) so that the sum of the weights of all scale features is 1.

[0113] Next, feature weighted fusion, including: (1) Scale-by-scale weighting: Multiply the features of different scales by their corresponding weights to ensure that the contribution ratio of features at each scale conforms to the weight distribution during fusion.

[0114] (2) Feature summation: The weighted features of all scales are summed element by element to obtain a fused feature map. At this time, the detail information and global semantic information are integrated to form a unified multi-scale feature representation.

[0115] S3.2.4: Output layer (feature enhancement after fusion): The fused feature map is further processed by a 3×3 convolution kernel to enhance local feature details and reduce information loss during the fusion process.

[0116] S3.2.5: Perform nonlinear transformations using an activation function (e.g., ReLU or Leaky ReLU) to further enhance the representational capabilities of the feature map and improve the segmentation accuracy of the MIM model. ReLU stands for Rectified Linear Unit, a commonly used activation function in neural networks. It generally refers to the ramp function in mathematics.

[0117] S3.3: Pair the semantic features with the fused feature map, calculate the similarity between the two paired features, generate attention weights, fuse the paired features based on the attention weights, obtain the segmentation representation through the activation function, and segment the image to be segmented based on the segmentation representation.

[0118] A cross-attention module is proposed to align features extracted from the diffusion model and the MIM model. Features extracted from different diffusion decoder layers are aligned with corresponding MIM features. The diffusion model and the MIM model each process different aspects of the image. The diffusion model focuses on capturing semantic features through progressive denoising, while the MIM model learns global and local information of the image by reconstructing occluded areas. The cross-attention module aligns the features extracted from different diffusion decoder layers with the corresponding MIM model features.

[0119] The local features extracted by each diffusion decoder layer are aligned with the corresponding hierarchical features of the MIM model to ensure that the fusion of local and global information can be fully utilized. This alignment operation not only enhances the mutual understanding of features from different models, but also improves segmentation accuracy.

[0120] To address the problem of insufficient context caused by attention calculation limited to the current level, the cross-modal attention module introduces a cross-level attention transfer mechanism. Through this mechanism, low-level detail features can effectively flow to high-level feature layers, helping to generate a more complete semantic representation.

[0121] The cross-level attention transfer mechanism specifically includes: Information transfer between layers: Setting up cross-layer connections to combine low-level features with high-level features so that low-level details can provide support for high-level features; By introducing hierarchical information into cross-modal data through layer-by-layer fusion, the alignment accuracy is improved.

[0122] In this embodiment, the specific processing steps of the cross-modal attention module are as follows: S3.3.1: Obtain the features of the diffusion model decoder layer and the MIM model encoder layer and pair them at the same level to form feature pairs.

[0123] S3.3.2: Cross-model attention calculation: For each pair of features, first generate attention weights by calculating the similarity between the features. These weights are used to measure the similarity between the current diffusion feature (semantic feature) and the MIM feature (fusion feature). The specific implementation method is: Calculate the feature similarity matrix through dot product: (16); in: and Represents semantic features and fusion features Middle Hedi The feature vector of the position; represents the dot product, represents the norm of the eigenvector; Indicates location and location The similarity between them is in the range of [−1,1].

[0124] Based on the similarity matrix S, the attention weight is calculated by the softmax function: (17); in: yes and The similarity of Represents the feature index; measure right The influence of is normalized by softmax to ensure that the sum of the weights of each position is 1.

[0125] (18); measure right The influence of is normalized by softmax to ensure that the sum of the weights of each position is 1.

[0126] S3.3.3: Weighted fusion feature output.

[0127] Each pair of features is weighted and summed according to the attention weights to obtain the final fused features. The fused features are processed through an activation function to form the final segmentation representation, ensuring that global and local information are integrated and optimized.

[0128] The two features are weighted fused according to the attention weight to obtain the fused feature representation The bidirectional attention mechanism ensures that the fusion process considers the interaction of the two features and avoids bias towards a single source.

[0129] (19); Among them, α and β are hyperparameters used to adjust the weights of diffusion features and MIM features.

[0130] S3.3.4: Output results.

[0131] The fused features output by the cross-modal attention module enter the downstream pixel classifier to further improve the segmentation accuracy.

[0132] In this embodiment, the diffusion model, the mask autoencoder model, the cross-modal attention module and the pixel classifier constitute the image segmentation model.

[0133] In this embodiment, two loss functions are used to train the image segmentation model: Pixel-level loss: The Dice coefficient is used to measure the quality of segmentation results and optimize the features extracted by the diffusion model and the MIM model. The Dice coefficient is a set similarity metric, typically used to calculate the similarity between two samples. It is a common evaluation metric for semantic segmentation and has a value range of [0, 1]. Cross-modal alignment loss: used to ensure that the features extracted by the diffusion model and the MIM model are well aligned and fused in the cross-modal attention mechanism.

[0134] Specifically, the feature fusion between the diffusion model and the MIM model is optimized through the cross-modal alignment loss. This loss function measures the similarity between the features of the two modalities, ensuring that they are fully aligned in the high-dimensional space. The cross-modal alignment loss can be divided into two main components: feature alignment loss and hierarchical reconstruction loss, each of which contributes to effective alignment between cross-modal features.

[0135] Feature alignment loss: measures the similarity between features from different modalities: (20); Where L is the total number of layers, The diffusion model and MIM model are l Layer characteristics, is a hyperparameter, and Cos represents cosine similarity.

[0136] Hierarchical reconstruction loss can be used to ensure that the cross-modal attention module maintains the integrity of information when fusing features. The mean squared error (MSE) can be used to measure the difference between the reconstructed features and the true labels: (twenty one); in, It is l The output of layer reconstruction, is the true label, is a hyperparameter used to control the importance of reconstruction loss at different levels, and N is the number of samples.

[0137] Image segmentation model training is performed through a joint optimization approach, gradually adjusting the feature extraction processes of the diffusion model and the MIM model, and ensuring feature alignment between the two through a cross-modal attention mechanism. The entire training process includes forward propagation, loss calculation, backpropagation, and gradient updates, and segmentation performance is gradually optimized through multiple rounds of iteration.

[0138] The image segmentation method based on a dual-branch network architecture provided in this embodiment uses a dual-branch architecture for image segmentation. One branch uses a diffusion model to process the image, while the other branch uses a pre-trained MIM network to capture details in the medical image. The cross-modal attention mechanism aligns the two features, enhancing the semantic representation, thereby enabling prediction through a pixel classifier. This has the following technical effects: Rich multi-level semantic feature extraction: By gradually restoring the image using diffusion model branches, the diffusion model can generate multi-level semantic features that include both detailed features of the lesion area and the boundary features between the background and the lesion, providing strong support for fine segmentation. Enhanced global structure representation: Through the MIM network branch, the MIM model can learn large-scale structural information in the image from a global perspective, and has stronger semantic understanding ability in distinguishing lesion areas from background areas, which helps improve the robustness of the image segmentation model in complex backgrounds; Cross-modal alignment enhances feature fusion: The cross-modal attention mechanism aligns the features of the diffusion model and the MIM model, fusing local and global features at each layer. This allows for a more coordinated capture of details and overall structure. This mechanism enables the image segmentation model to simultaneously focus on subtle boundary features and overall semantic consistency at the pixel level, thereby improving segmentation performance. Accurate segmentation prediction: After combining the features of the two branches, a pixel classifier is used to perform pixel-level classification on the enhanced semantic features. Through this fusion strategy, the image segmentation model significantly improves segmentation accuracy, extracts details of lesion boundaries, and captures multi-scale lesions, adapting to the challenges of different scales and complex backgrounds in medical images.

[0139] In step S4 of the present invention, the ultrasound quality evaluation is performed on the segmented feature image. If the evaluation fails, the ultrasound image data is re-collected for re-evaluation until a qualified ultrasound image is obtained after the evaluation passes. Figure 12 Shown, including: S4.1: Obtaining the segmented feature image , for the acquired segmented feature image performing preprocessing to obtain preprocessed cardiac ultrasound image data; S4.2: Extract multi-scale spatial features of cardiac ultrasound images and fuse the multi-scale spatial features to obtain multi-scale spatial fusion features; In this embodiment, a multi-scale convolutional neural network is used to extract the spatial features of cardiac ultrasound images. By using a variety of convolution kernels of different sizes, feature maps extracted by different convolution kernels are obtained, and then the feature maps of each scale are fused to obtain a fused feature map. , expressed as: (twenty two); in, For the The feature map extracted by the convolution kernel, is the number of convolution kernels, is the fusion weight, which is used to adjust the importance of each feature map.

[0140] S4.3: Perform spatial attention enhancement and channel attention enhancement on the multi-scale spatial fusion features to obtain multi-scale spatial fusion features with enhanced spatial attention and channel attention, specifically including the following steps: S4.3.1: Perform spatial attention enhancement on the multi-scale spatial fusion features to obtain spatial attention enhanced multi-scale spatial fusion features; In this embodiment, the importance of different positions in the image is calculated to focus on the edge area of ​​the heart. In particular, when evaluating the clarity of the cardiac cavity contour, spatial attention can help the model accurately focus on the edge part. The specific formula is: (twenty three); in, Indicates the importance of the pixel at each position, Indicates the weight distribution of pixels at different positions when calculating importance, Represents the convolution operation, through With the weight matrix Convolution captures the correlation information between different locations in the image. is the bias term, which is used to adjust the offset of the calculation result. Is the activation function, which maps the result of convolution and bias calculation to a set interval so that the output Able to indicate the importance of each position.

[0141] The final multi-scale spatial fusion feature map with spatial attention enhancement is obtained according to the importance of each position.

[0142] S4.3.2: Perform channel attention enhancement on the multi-scale spatial fusion features to obtain the channel attention enhanced multi-scale spatial fusion features; Channel weighting is used to enhance features related to cardiac structure and suppress noise and irrelevant parts.

[0143] The calculation formula for the importance score of each channel corresponding to the pixel is: (twenty four); in, Indicates the importance of each channel corresponding to the pixel, is the weight vector of channel attention, which weights the data of different channels. Represents the dot product operation, through With the weight vector The dot product can adjust the data according to the importance of different channels. is the bias term, which is used to adjust the offset of the calculation result. It is an activation function that maps the calculation results to the set interval.

[0144] The final channel attention-enhanced multi-scale spatial fusion feature map is obtained according to the importance of each channel; S4.4: Calculate heart contour clarity and echo uniformity based on multi-scale spatial fusion features enhanced by spatial attention, and calculate heart valve clarity based on multi-scale spatial fusion features enhanced by channel attention, specifically including the following steps: S4.4.1: Computing cardiac contour clarity using multi-scale spatial fusion features enhanced by spatial attention and echo uniformity ; Among them, the Canny edge detection method is used to detect the multi-scale spatial fusion features enhanced by spatial attention to identify the cardiac cavity boundary and obtain the cardiac contour clarity. ; When assessing the clarity of cardiac cavity contours, the model needs to accurately identify the edges of the cardiac cavity. The spatial attention mechanism uses this calculation to highlight the importance of edge regions in the image.

[0145] For example, pixels at the edge of a cardiac chamber may have higher values, while pixels further away from the edge may have lower values. This allows the model to focus more on features in the edge region, thereby more accurately quantifying the clarity of the cardiac chamber contour.

[0146] Among them, echo uniformity The calculation process includes: The multi-scale spatial fusion features enhanced by spatial attention are divided into multiple regions, and the standard deviation and average value of the pixel intensity in each region are calculated respectively. The echo uniformity is obtained based on the standard deviation and average value of the pixel intensity in each region. ; (25); in, M Indicates the number of regions, and Respectively represent The standard deviation and mean of the pixel intensities within a local area.

[0147] S4.4.2: Computing Heart Valve Clarity Based on Multi-Scale Spatial Fusion Features Enhanced by Channel Attention and Image Texture Features , specifically including the following steps: S4.4.2.1: Extract global and local features from an image using multi-scale spatial fusion feature extraction based on channel attention enhancement. In this embodiment, a CNN is used to extract features from the valve region to capture global and local features in the image. S4.4.2.1.2: Extract texture features from cardiac ultrasound images; In this embodiment, first, the cardiac ultrasound image is preprocessed, including denoising and enhancement, to ensure that the image quality is suitable for feature extraction. Then, the gray-level co-occurrence matrix (GLCM) method is used to extract texture features, such as contrast, correlation, and uniformity. In addition, the local binary pattern (LBP) is used to analyze the texture features of local areas of the image. By selecting these texture features, the most representative features are selected for subsequent analysis.

[0148] S4.4.2.3: Fusion of texture features with global and local features in the image to evaluate the clarity of the valve region.

[0149] When assessing heart valve clarity, data from different channels may contain different information. Some channels may be more relevant to the characteristics of the heart valve structure, while others may contain more noise or irrelevant information. The channel attention mechanism, through the above calculations, can increase the weight of channels related to the heart valve structure, allowing the model to pay more attention to the data of these important channels when extracting features, thereby more accurately quantifying heart valve clarity.

[0150] S4.5: Extract features of cardiac ultrasound images at different time points, and calculate left ventricular wall motion consistency based on the features of cardiac ultrasound images at different time points; The specific steps include: S4.5.1: Extract features of cardiac ultrasound images at different time points; In this embodiment, the TCN (Temporal Convolutional Network) model is used for analysis. TCN performs a convolution operation on the data at each time step, which is expressed as: (26); in, is the temporal convolution kernel weight, is the length of the convolution kernel, For the input data sequence At time step The value at the moment represents the current time step The historical data that the upconvolution operation depends on.

[0151] Through this convolution operation, TCN can capture the dependencies between data at different time points. For example, during a cardiac cycle, the states of the ventricular wall at different times are correlated, with the state at earlier times affecting the state at later times. By applying convolution to these time series data, TCN can tap into this inherent temporal dependency.

[0152] S4.5.2: Calculate left ventricular wall motion consistency based on echocardiographic features at different time points. In this embodiment, the TCN (Temporal Convolutional Network) model is used to analyze the consistency features of cardiac wall motion during the cardiac cycle. TCN is used to analyze the temporal characteristics of the image, especially the dynamic changes of the left ventricular wall. The motion consistency of the left ventricular wall is evaluated by extracting temporal dependent features. Left ventricular wall motion consistency characteristics The calculation formula is: (27); in, Indicates the motion consistency of the left ventricular wall, and Indicates a time point and The ventricular wall eigenvector at time .

[0153] S4.6: Extract the signal-to-noise ratio of cardiac ultrasound images.

[0154] The signal-to-noise ratio quantifies the ratio of signal to noise in an image, reflecting the clarity of effective information.

[0155] In this embodiment, the signal-to-noise ratio of the cardiac ultrasound image is calculated by calculating the signal and background noise of the selected area. , the calculation formula is: (28); in, is the average value of the signal, is the standard deviation of the noise.

[0156] S4.7: Maximize the similarity of each feature of similar samples through comparative learning to obtain optimized heart contour clarity, echo uniformity, heart valve clarity, left ventricular wall motion consistency, and signal-to-noise ratio; The specific steps include: S4.7.1: Define similar samples; For a pair of samples, if they are considered similar under some predefined similarity criteria, then ( and For samples and samples The similarity calculation result of the two samples can be based on certain attributes of the data itself or determined by some algorithm. For example, in cardiac ultrasound image data, the similarity between two samples can be judged based on certain basic features of the image (such as imaging angle, approximate similarity of cardiac structure, etc.); S4.7.2: Construct a loss function that maximizes the feature similarity of each feature of similar samples: (29); in, and Indicates different sample indices, Represents a sample and samples The inner product of the eigenvectors and , the value of the inner product reflects the similarity of the two eigenvectors in direction. and They are the modulus (length) of the eigenvectors and , and the cosine similarity of the two eigenvectors is obtained by calculation. Its value is between [-1,1]. The closer Indicates that the two feature vectors are more similar.

[0157] when When , we hope that the feature vectors of two similar samples are as similar as possible, that is, the cosine similarity is as close to 1 as possible, so the loss function uses To measure the degree of difference in features between similar samples, the loss function of contrastive learning is obtained by summing all satisfied sample pairs .

[0158] S4.7.3: By continuously adjusting model parameters (such as convolution kernel weights) to minimize the loss function, the model gradually learns feature representations that better distinguish similar from dissimilar samples. This allows the model to learn effective feature representations for cardiac ultrasound imaging data under unsupervised conditions, providing a more accurate feature foundation for subsequent image quality assessment and generation of shooting recommendations.

[0159] S4.8: Perform image quality assessment based on optimized cardiac outline clarity, echo uniformity, cardiac valve clarity, left ventricular wall motion consistency, and signal-to-noise ratio, and generate imaging recommendations based on the assessment results. The specific steps include: S4.8.1: Perform image quality assessment based on optimized cardiac contour clarity, echo uniformity, cardiac valve clarity, and left ventricular wall motion consistency, and obtain assessment results. In this embodiment, each image quality indicator is scored according to the set quantitative scoring standard, and the scoring result is 0-5 points; S4.8.2: Generate shooting suggestions based on the obtained results; In this embodiment, specific suggestions include: For heart definition: (1) If the score of cardiac cavity outline clarity is less than 2 points, it is recommended to adjust the probe angle by at least 15 degrees to improve the visualization of the cardiac cavity edge; (2) If the score of cardiac cavity outline clarity is between 2 and 3 points, it is recommended to fine-tune the probe angle by 5 to 10 degrees or increase the image contrast to improve the edge sharpness; (3) If the score of cardiac cavity outline clarity is greater than 3 points, the current setting is sufficient and no adjustment is required.

[0160] For left ventricular wall motion coherence: (1) If the left ventricular wall motion consistency score is lower than 2 points, it is recommended to adjust the probe position to ensure full capture of the cardiac cycle, or change the imaging mode; (2) If the left ventricular wall motion consistency score is between 2 and 3 points, it is recommended to fine-tune the probe position or slightly adjust the timing of cardiac cycle capture; (3) If the left ventricular wall motion consistency score is greater than 3 points, the current settings are sufficient and no adjustment is required.

[0161] For echo uniformity; (1) If the echo uniformity score is less than 2 points, it is recommended to adjust the probe focus setting or change the scanning depth to a more uniform area; (2) If the echo uniformity score is between 2 and 3 points, it is recommended to fine-tune the transmission power to optimize tissue reflection at different depths; (3) If the echo uniformity score is greater than 3 points, the current settings are sufficient and no adjustment is required.

[0162] For heart valve clarity: (1) If the echo uniformity score is less than 2 points, it is recommended to adjust the probe angle by at least 20 degrees to better align the valve area, or change the imaging frequency; (2) If the echo uniformity score is between 2 and 3 points, it is recommended to fine-tune the probe angle by about 10 degrees, or slightly adjust the imaging frequency; (3) If the echo uniformity score is greater than 3 points, the current settings are sufficient and no adjustment is required.

[0163] The scores of these indicators will be monitored in real time, and specific adjustment suggestions will be provided. According to the system prompts, the device settings can be gradually adjusted until the ideal image quality is achieved.

[0164] The above describes in detail the method of the embodiment of the present invention. To facilitate better implementation of the above method of the embodiment of the present application, a system of the embodiment of the present application is provided below.

[0165] Figure 13 An adaptive control system of a wearable cardiac ultrasound device provided by an exemplary embodiment of the present invention is shown, comprising: A mode switching unit is configured to: acquire an ultrasound image and automatically switch the ultrasound mode according to the acquired ultrasound image; An image enhancement unit is configured to: perform enhancement processing on the ultrasonic images detected in different ultrasonic modes; An image segmentation unit is configured to: perform image segmentation on the enhanced ultrasound image; The ultrasound evaluation unit is configured to: perform ultrasound quality evaluation on the segmented feature image; if the evaluation fails, re-collect ultrasound image data for re-evaluation until a qualified ultrasound image is obtained after the evaluation passes.

[0166] It can be understood that the above-mentioned units can be separately or completely combined into one or several other units to constitute, or one (or some) of the units can be further divided into multiple functionally smaller units to constitute, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application.

[0167] Figure 14Another exemplary embodiment of the present invention provides a wearable ultrasound device, which includes a processor, a communication interface, and a computer-readable storage medium. The processor, the communication interface, and the computer-readable storage medium may be connected via a bus or other means.

[0168] Among them, the communication interface is used to receive and send data, the computer-readable storage medium can be stored in the memory of the electronic device, the computer-readable storage medium is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer-readable storage medium.

[0169] A processor (or CPU (Central Processing Unit)) is the computing and control core of an electronic device. It is suitable for implementing one or more instructions, specifically for loading and executing one or more instructions to implement corresponding method processes or corresponding functions.

[0170] The processor is configured to perform the following process: Acquire ultrasound image data, and automatically switch ultrasound modes according to the acquired ultrasound image data; Performing enhancement processing on the ultrasonic images detected under different ultrasonic modes; Perform image segmentation on the enhanced ultrasound image; The ultrasound quality evaluation is performed on the segmented feature image. If the evaluation fails, the ultrasound image data is re-collected for re-evaluation until a qualified ultrasound image is obtained after the evaluation passes.

[0171] The present invention also provides a computer-readable storage medium (Memory). This computer-readable storage medium is a memory device in an electronic device for storing programs and data. It should be understood that the computer-readable storage medium herein may include both built-in storage media in the electronic device and, of course, extended storage media supported by the electronic device. The computer-readable storage medium provides storage space that stores the processing system of the electronic device.

[0172] Furthermore, the storage space also stores one or more instructions suitable for being loaded and executed by the processor. These instructions may be one or more computer programs (including program code). It should be noted that the computer-readable storage medium herein may be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device; alternatively, it may be at least one computer-readable storage medium located remotely from the processor.

[0173] In one embodiment, the computer-readable storage medium stores one or more instructions; the processor loads and executes the one or more instructions stored in the computer-readable storage medium to implement the following process: Acquire an ultrasound image, and automatically switch the ultrasound mode according to the acquired ultrasound image; Performing enhancement processing on the ultrasonic images detected under different ultrasonic modes; Perform image segmentation on the enhanced ultrasound image; The ultrasound quality evaluation is performed on the segmented feature image. If the evaluation fails, the ultrasound image data is re-collected for re-evaluation until a qualified ultrasound image is obtained after the evaluation passes.

[0174] The present invention also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the following process: Acquire an ultrasound image, and automatically switch the ultrasound mode according to the acquired ultrasound image; Performing enhancement processing on the ultrasonic images detected under different ultrasonic modes; Perform image segmentation on the enhanced ultrasound image; The ultrasound quality evaluation is performed on the segmented feature image. If the evaluation fails, the ultrasound image data is re-collected for re-evaluation until a qualified ultrasound image is obtained after the evaluation passes.

[0175] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A wearable cardiac ultrasound device, comprising a flexible attachment structure for attaching to a part of a human body to be detected, characterized in that: A phased array probe is fixed on the flexible attachment structure, and the phased array probe is fixed on a side of the flexible attachment structure for being close to the skin; The phased array probe consists of multiple independent ultrasonic transmitting units, each of which has an independent excitation signal. By controlling each ultrasonic transmitting unit, ultrasonic beams of different angles are dynamically synthesized to focus and deflect the ultrasound, thereby adjusting the imaging area. The phased array probe is communicatively connected to a control terminal, and the control terminal is configured to switch the ultrasound mode according to a manual instruction or automatically, and control the phased array probe according to the switched ultrasound mode to adjust the imaging mode.

2. The wearable cardiac ultrasound device according to claim 1, wherein The control terminal is connected to an imaging mode switching knob and a power switch, and the imaging mode switching knob is used to manually switch the ultrasound mode; The control terminal is connected to a status indicator light, which is used to indicate the working status of the ultrasound device, including power on, working and fault.

3. The wearable cardiac ultrasound device according to claim 1, wherein The ultrasound modes include: B-ultrasound mode, M-ultrasound mode and color ultrasound mode.

4. The wearable cardiac ultrasound device according to claim 1, wherein The control terminal includes a display module; or the control terminal is connected to an external display module.

5. The wearable cardiac ultrasound device according to any one of claims 1 to 4, wherein: The control terminal automatically switches the ultrasound mode, including: Enter B-ultrasound mode, and after obtaining the standard B-ultrasound section, detect the region of interest. After the region of interest is detected continuously within the set time and the difference between the regions of interest detected multiple times is less than the set threshold, switch to color ultrasound mode, and perform disease prediction based on the color ultrasound image in color ultrasound mode. After the prediction is completed, switch back to B-ultrasound mode; Whether to start M-mode ultrasound is determined based on the current standard section indicators or disease prediction results. If M-mode ultrasound is required, the system enters M-mode ultrasound mode and makes a final diagnosis of the disease based on the M-mode ultrasound image in M-mode ultrasound mode. If M-mode ultrasound is not required, the system continues to use the current B-mode ultrasound image to guide the phased array probe to find the next standard section until the last standard section is fully detected.

6. The wearable cardiac ultrasound device according to any one of claims 1 to 4, wherein: The control terminal is further configured to: enhance the detected ultrasonic echo signal to obtain an ultrasonic image, perform image segmentation on the ultrasonic image, and perform ultrasonic quality evaluation on the segmented feature image; if the evaluation is unqualified, control the phased array probe to re-detect and re-evaluate until a qualified ultrasonic image is obtained after the evaluation is qualified.

7. The wearable cardiac ultrasound device according to claim 6, wherein: The detected ultrasonic echo signal is enhanced to obtain an ultrasonic image. When the ultrasonic signal is ultrasonic echo data of a single angle, the following process is included: Perform IQ demodulation on the single-angle ultrasonic echo data to obtain complex IQ data containing amplitude and phase information; Performing time delay compensation on the complex IQ data to obtain time delayed IQ data; Input the time-delayed IQ data into the encoder to obtain a low-dimensional feature representation; The low-dimensional feature representation is input into a time-frequency feature fusion module to obtain a reconstructed and enhanced ultrasound image.

8. The wearable cardiac ultrasound device according to claim 6, wherein: In the control terminal, performing image segmentation on the ultrasound image includes: For the ultrasound image to be segmented, the complexity of the image to be segmented is measured through edge detection and texture analysis. After allocating the denoising step size of the diffusion model according to the complexity, local features of different scales are extracted through the diffusion model. For the image to be segmented, global features of different scales are obtained through the deblurring mask autoencoder model; The local features and global features of the same scale are paired, the similarity between the two paired features is calculated, and the attention weight is generated. After the paired features are fused based on the attention weight, the segmentation representation is obtained through the activation function. The image to be segmented is segmented based on the segmentation representation to obtain the segmented feature image.

9. The wearable cardiac ultrasound device according to claim 6, wherein: In the control terminal, performing ultrasound quality assessment on the segmented feature image includes: Extracting multi-scale spatial features of the segmented feature image, and fusing the multi-scale spatial features to obtain multi-scale spatial fusion features; Perform spatial attention enhancement and channel attention enhancement on the multi-scale spatial fusion features to obtain multi-scale spatial fusion features with enhanced spatial attention and channel attention; The heart contour clarity and echo uniformity are calculated based on the multi-scale spatial fusion features enhanced by spatial attention, and the heart valve clarity is calculated based on the multi-scale spatial fusion features enhanced by channel attention. Extracting features of the feature image at different time points, and calculating the consistency of left ventricular wall motion based on the features of the feature image at different time points; Extract the signal-to-noise ratio of the feature image; Ultrasound quality assessment was performed based on cardiac outline clarity, echo homogeneity, cardiac valve clarity, left ventricular wall motion consistency, and signal-to-noise ratio.

10. An adaptive control method for a wearable cardiac ultrasound device, characterized in that: The following processes are included: Acquire an ultrasound image, and automatically switch the ultrasound mode according to the acquired ultrasound image; Performing enhancement processing on the ultrasonic images detected under different ultrasonic modes; Perform image segmentation on the enhanced ultrasound image; The ultrasound quality evaluation is performed on the segmented feature image. If the evaluation fails, the ultrasound image data is re-collected for re-evaluation until a qualified ultrasound image is obtained after the evaluation passes.

11. The adaptive control method for a wearable cardiac ultrasound device according to claim 10, wherein: Automatically switch ultrasound modes based on acquired ultrasound image data, including: Enter B-ultrasound mode, and after obtaining the standard B-ultrasound section, detect the region of interest. After the region of interest is detected continuously within the set time and the difference between the regions of interest detected multiple times is less than the set threshold, switch to color ultrasound mode, and perform disease prediction based on the color ultrasound image in color ultrasound mode. After the prediction is completed, switch back to B-ultrasound mode; Whether to start M-mode ultrasound is determined based on the current standard section indicators or disease prediction results. If M-mode ultrasound is required, the system enters M-mode ultrasound mode and makes a final diagnosis of the disease based on the M-mode ultrasound image in M-mode ultrasound mode. If M-mode ultrasound is not required, the system continues to use the current B-mode ultrasound image to guide the phased array probe to find the next standard section until the last standard section is fully detected.

12. The adaptive control method for a wearable cardiac ultrasound device according to claim 11, wherein: According to the current B-ultrasound image, by controlling each ultrasonic transmitting unit of the phased array probe, ultrasonic beams with different detection angles are dynamically synthesized to find the next standard section.

13. The adaptive control method for a wearable cardiac ultrasound device according to claim 11, wherein: Disease prediction based on color ultrasound images in color ultrasound mode, including: A deep learning model is used to analyze the color ultrasound image to generate prediction results on whether a disease exists and the severity of the disease. When the disease is mitral valve regurgitation, the mitral valve position is framed according to the color ultrasound image frame, and the color ultrasound image frame corresponding to the mitral valve position frame is used as the input of the deep learning model. The deep learning model outputs whether there is regurgitation and the severity of regurgitation, and the regurgitation severity includes mild, moderate and severe.

14. The adaptive control method for a wearable cardiac ultrasound device according to claim 11, wherein: Determine whether to initiate M-mode ultrasound based on current standard section indicators or disease prediction results, including: When the disease severity in the disease prediction result is moderate or above, M-ultrasound is directly started; wherein the disease severity includes mild, moderate and severe; When one or more of the standard section indicators are abnormal, M-ultrasound is directly started; wherein, the abnormalities include the ventricular septum thickness being greater than the set threshold and the movement pattern of the ventricular septum being inconsistent with the standard movement pattern.

15. The adaptive control method for a wearable cardiac ultrasound device according to claim 11, wherein: The final diagnosis of the disease is made based on the M-mode ultrasound images, including: The line connecting the center point P2 and the vertex P1 of the region of interest is selected as the position of the M-ultrasound sampling line to obtain a continuous waveform of the M-ultrasound. The M-ultrasound index is measured based on the continuous waveform of the M-ultrasound, and the final diagnosis of the disease is made based on the M-ultrasound index.

16. The adaptive control method for a wearable cardiac ultrasound device according to claim 10, wherein: The ultrasonic image obtained in the M-mode is enhanced using single-angle ultrasonic echo data. The following processes are included: Perform IQ demodulation on the single-angle ultrasonic echo data to obtain complex IQ data containing amplitude and phase information; Performing time delay compensation on the complex IQ data to obtain time delayed IQ data; Input the time-delayed IQ data into the encoder to obtain a low-dimensional feature representation; The low-dimensional feature representation is input into a time-frequency feature fusion module to obtain a reconstructed and enhanced ultrasound image.

17. The adaptive control method for a wearable cardiac ultrasound device according to claim 16, wherein: Performing time delay compensation on the complex IQ data to obtain time-delayed IQ data includes: ,in, is the IQ data after time delay, ,in, , , and represents the coordinates of the imaging point, and Indicates the The position coordinates of the probes, represents the launch angle, Represents wave speed.

18. The adaptive control method for a wearable cardiac ultrasound device according to claim 16, wherein: The time-frequency feature fusion module includes: a time domain branch and a frequency domain branch, wherein the time domain branch obtains a time domain feature representation by extracting time domain features; The frequency domain branch includes a parallel low-frequency branch and a high-frequency branch, wherein the low-frequency branch is extracted by a low-frequency feature extraction module to obtain a low-frequency feature representation, and the high-frequency branch is extracted by a high-frequency feature extraction module to obtain a high-frequency feature representation; The feature fusion module performs weighted fusion on the time domain feature representation, the low-frequency feature representation and the high-frequency feature representation, and outputs a reconstructed and enhanced ultrasound image.

19. The adaptive control method for a wearable cardiac ultrasound device according to claim 18, wherein: The low-frequency branch is extracted by a low-frequency feature extraction module to obtain a low-frequency feature representation, including: The low-frequency branch converts the input features from the time domain to the frequency domain through a two-dimensional Fourier transform, and extracts the low-frequency components of the signal through a low-pass filter with learnable parameters. The low-frequency components are passed through a low-frequency feature extraction module and a two-dimensional inverse Fourier transform to obtain a low-frequency feature representation; The high-frequency branch is extracted by a low-frequency feature extraction module to obtain a high-frequency feature representation, including: The high-frequency branch converts the input features from the time domain to the frequency domain through a two-dimensional Fourier transform, and extracts the high-frequency components of the signal through a high-pass filter with learnable parameters. The high-frequency components are passed through a high-frequency feature extraction module and a two-dimensional inverse Fourier transform to obtain a high-frequency feature representation.

20. The adaptive control method for a wearable cardiac ultrasound device according to claim 18, wherein: The feature fusion module performs weighted fusion on the time domain feature representation, the low-frequency feature representation, and the high-frequency feature representation, including: Performing feature integration on the time domain feature representation, the low-frequency feature representation, and the high-frequency feature representation through an adaptive weight learning module to generate three feature weights, and normalizing the three feature weights through a normalization function to ensure that the sum of the weights is 1; The normalized weights are multiplied with the time domain features, low-frequency features, and high-frequency features respectively to perform weighted processing on the features. Finally, the weighted features are linearly superimposed, converted to the frequency domain through two-dimensional Fourier transform, filtered through learnable parameters, and then transformed back to the time and space domain to obtain the reconstructed and enhanced ultrasound image.

21. The adaptive control method for a wearable cardiac ultrasound device according to claim 10, wherein: Perform image segmentation on the enhanced ultrasound image, including: For the image to be segmented, the complexity of the image to be segmented is measured through edge detection and texture analysis. After allocating the denoising step size of the diffusion model according to the complexity, the semantic features are extracted through the diffusion model. After random occlusion and blurring of the segmented image, local and global features are extracted through the mask autoencoder model, and fused features are obtained through multi-scale feature fusion. The semantic features and fusion features are paired, the similarity between the two paired features is calculated, and attention weights are generated. After the paired features are fused based on the attention weights, a segmentation representation is obtained through an activation function, and the image to be segmented is segmented based on the segmentation representation.

22. The adaptive control method for a wearable cardiac ultrasound device according to claim 21, wherein: The complexity is: , , ,in, is the edge strength, and are the gradients in the horizontal and vertical directions respectively, N represents the total number of edge pixels involved in the calculation, α and β is the weight coefficient.

23. The adaptive control method for a wearable cardiac ultrasound device according to claim 10, wherein: Perform ultrasound quality assessment on the segmented feature image, including: Extracting multi-scale spatial features of the segmented feature image, and fusing the multi-scale spatial features to obtain multi-scale spatial fusion features; Perform spatial attention enhancement and channel attention enhancement on the multi-scale spatial fusion features to obtain multi-scale spatial fusion features with enhanced spatial attention and channel attention; The heart contour clarity and echo uniformity are calculated based on the multi-scale spatial fusion features enhanced by spatial attention, and the heart valve clarity is calculated based on the multi-scale spatial fusion features enhanced by channel attention. Extracting features of the feature image at different time points, and calculating the consistency of left ventricular wall motion based on the features of the feature image at different time points; Extract the signal-to-noise ratio of the feature image; Ultrasound quality assessment is performed based on cardiac contour clarity, echo uniformity, heart valve clarity, left ventricular wall motion consistency and signal-to-noise ratio. If the score is greater than the set assessment score, the assessment is qualified; otherwise, the assessment is unqualified.

24. The adaptive control method for a wearable cardiac ultrasound device according to claim 23, wherein: The calculation formula for spatial attention enhancement and channel attention enhancement of the multi-scale spatial fusion feature is: ； ； in, Indicates the importance of the pixel at each position, represents the multi-scale spatial fusion feature, Indicates the weight distribution of pixels at different positions when calculating importance, Represents the convolution operation, through With the weight matrix Convolution captures the correlation information between different locations in the image. and is the bias term, which is used to adjust the offset of the calculation result; Indicates the importance of each channel corresponding to the pixel, is the weight vector of channel attention, represents the dot product operation, is the activation function.

25. The adaptive control method for a wearable cardiac ultrasound device according to claim 23, wherein: The evaluation result score ranges from 0 to 5 points. Based on the final evaluation score, new ultrasound imaging recommendations are generated, including: If the cardiac cavity outline clarity score is less than 2 points, it is recommended to adjust the probe angle by at least 15 degrees to improve visualization of the cardiac cavity edge. If the cardiac cavity outline clarity score is between 2 and 3 points, it is recommended to fine-tune the probe angle by 5 to 10 degrees or increase the image contrast to improve edge sharpness. If the cardiac cavity outline clarity score is greater than 3 points, the current settings are sufficient and no adjustment is required. If the left ventricular wall motion consistency score is less than 2 points, it is recommended to adjust the probe position to ensure full capture of the cardiac cycle, or change the imaging mode; if the left ventricular wall motion consistency score is between 2 and 3 points, it is recommended to fine-tune the probe position or slightly adjust the timing of cardiac cycle capture; if the left ventricular wall motion consistency score is greater than 3 points, the current settings are sufficient and no adjustment is required; If the echo uniformity score is less than 2, it is recommended to adjust the probe focus setting or change the scanning depth to a more uniform area. If the echo uniformity score is between 2 and 3, it is recommended to fine-tune the transmit power to optimize tissue reflections at different depths. If the echo uniformity score is greater than 3, the current settings are sufficient and no adjustments are required. If the echo uniformity score is less than 2 points, it is recommended to adjust the probe angle by at least 20 degrees to better align the valve area, or change the imaging frequency. If the echo uniformity score is between 2 and 3 points, it is recommended to fine-tune the probe angle by about 10 degrees, or slightly adjust the imaging frequency. If the echo uniformity score is greater than 3 points, the current settings are sufficient and no adjustment is required.

26. An adaptive control system for a wearable cardiac ultrasound device, characterized in that: include: A mode switching unit is configured to: acquire an ultrasound image and automatically switch the ultrasound mode according to the acquired ultrasound image; An image enhancement unit is configured to: perform enhancement processing on the ultrasonic images detected in different ultrasonic modes; An image segmentation unit is configured to: perform image segmentation on the enhanced ultrasound image; The ultrasound evaluation unit is configured to: perform ultrasound quality evaluation on the segmented feature image; if the evaluation fails, re-collect ultrasound image data for re-evaluation until a qualified ultrasound image is obtained after the evaluation passes.

27. A wearable cardiac ultrasound device, characterized in that: include: a processor and a computer-readable storage medium; a processor adapted to execute a computer program; A computer-readable storage medium having a computer program stored therein, wherein when the computer program is executed by the processor, the adaptive control method for a wearable cardiac ultrasound device according to any one of claims 10 to 25 is implemented.

28. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor and executing the adaptive control method for a wearable cardiac ultrasound device according to any one of claims 11 to 25.

29. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the computer program implements the adaptive control method for a wearable cardiac ultrasound device according to any one of claims 10 to 25.

Images