Heart function automatic monitoring ultrasonic system and method with positioning and monitoring dual modes

Abstract

The application provides an automatic monitoring ultrasonic system and method with positioning and monitoring dual modes for heart function, which comprises a conformal ultrasonic probe, an electrocardiogram electrode patch, an ultrasonic front-end module, an electrocardiogram acquisition module, a master control and processing unit, a touch display and interaction module, and a local storage unit. The system of the application overcomes the defects of the prior art, such as being limited to occasional measurement mode, power consumption, transmission bandwidth, fixed frame rate acquisition, or lack of visualization basis.

Description

Technical Field

[0001] This invention relates to the field of ultrasound monitoring, specifically to an automatic cardiac function monitoring ultrasound system and method with dual modes of positioning and monitoring; more specifically, to a long-term automatic cardiac function monitoring ultrasound method and system with dual modes of positioning and monitoring. Background Technology

[0002] In clinical settings such as heart failure management and monitoring cardiotoxicity from chemotherapy for cancer, continuous dynamic assessment of left ventricular ejection fraction (LVEF) is crucial for assessing disease progression. However, existing LVEF monitoring methods have the following limitations: First, current echocardiographic assessments rely on manual operation and can only provide snapshots of individual measurements. Whether using large-scale departmental ultrasound or portable equipment, physicians must hold the probe, manually locate the slice, and measure frame by frame, obtaining only a single LVEF value. This cannot capture dynamic changes over several hours to days, making it difficult to identify early deterioration of cardiac function.

[0003] Second, indirect EF estimation methods based on non-imaging signals such as heart sounds and bioimpedance lack accuracy. These methods abandon direct visualization of cardiac ultrasound, inferring volume parameters only from one-dimensional signals. They rely on statistical models and population assumptions, have poor adaptability to individual anatomical variations, and their errors are insufficient to meet clinical needs such as precision medication decisions.

[0004] Third, the tightly coupled "acquisition-imaging-display" architecture of traditional ultrasound limits the temporal resolution of key phases. Synchronous pipelines require each frame to complete the entire process within a specified time. To obtain high frame rate imaging in short phases such as end-systole and end-diastole, either the overall processing power must be significantly increased, leading to increased cost and power consumption, or a uniform low frame rate must be maintained, causing the capacitive value frames to be easily missed and the accuracy of LVEF measurement to decrease.

[0005] Fourth, there is a lack of a long-term automatic cardiac function monitoring scheme that can accurately trigger and schedule acquisition and imaging resources on demand under the constraint of portable computing power. Existing technologies are either limited by manual intermittent measurement mode, or by the accuracy of indirect estimation, or by the limitation of the temporal resolution of key phases by tightly coupled architecture, and none of them can simultaneously meet the requirements of long-term automatic monitoring, direct ultrasound imaging and portable deployment.

[0006] In the field of echocardiographic cardiac function monitoring, existing technologies have mainly evolved along two routes: "portability" and "automation," but neither has been able to form a complete solution for the goal of long-term automatic monitoring.

[0007] Existing technology 1: Portable AI ultrasound diagnostic system Chinese patent application number CN202210017835.2 discloses "A method and apparatus for processing ejection fraction data based on cardiac ultrasound video". This solution combines image key point recognition and semantic segmentation models to automatically analyze cardiac ultrasound videos, filter end-systolic / end-diastolic images, and calculate LVEF. Applicant: Lepu Medical; Keywords: ultrasound, AI, ejection fraction, automatic calculation.

[0008] In addition, Chinese patent application number CN202210257197.1 discloses an "Artificial Intelligence Method and System Device for Echocardiographic Detection of Ejection Fraction," which uses a portable ultrasound device to acquire images, and an AI module automatically analyzes and evaluates the ejection fraction for home management of heart failure patients. Applicant: Suzhou Hermitz Health Technology Co., Ltd.

[0009] The main drawback of the above solutions is that, although they solve the problems of "portability" and "AI-automated calculation," their working mode is still essentially an occasional test—patients need to actively operate the device and align it with the probe to complete the test, making it impossible to achieve imperceptible long-term automatic monitoring or automatically generate time-series trend curves for LVEF. Portability and AI lower the barrier to entry for a single operation, but they do not change the assessment method of "single snapshot."

[0010] Existing technology 2: Wearable ultrasound patch technology The paper “A wearable cardiac ultrasound imager” (Hu et al., Nature, 2023, vol.613, pp.667–675) reported by Xu Sheng’s team at the University of California, San Diego, a wearable cardiac ultrasound imaging patch that integrates a flexible ultrasound transducer array. Its deep learning model can automatically extract the left ventricular volume and ejection fraction waveforms from continuous image recordings.

[0011] The main drawback of this approach is that the device relies on a cable connection to a computer for data download and real-time image display, making it unable to operate independently of an external host. Under this architecture, the device must operate at full power continuously or continuously transmit a large number of raw ultrasound images back to the terminal in real time. Its power consumption and transmission bandwidth exceed the limits for long-term wearability, thus limiting continuous wear time to less than 24 hours. Furthermore, the device lacks positioning imaging assistance and independent ECG synchronization design; probe placement relies on experience, making precise ECG gating trigger acquisition impossible.

[0012] Existing technology 3: ECG-gated triggered ultrasound acquisition technology European patent EP-2344042-A1 discloses "Methods for Acquisition and Display in Ultrasound Imaging." This patent provides a method for triggering ultrasound acquisition based on electrocardiogram (ECG) signals. Applicant: Fujifilm Visualsonics Inc.; Keywords: ECG, triggering, gating, ultrasound imaging.

[0013] Furthermore, US Patent Application No. US5099847 discloses a "High frame rate ultrasound system" that uses ECG R wave triggering to acquire images across multiple cardiac cycles and interweave them for high frame rate playback. A comparable company is Acuson (later acquired by Siemens).

[0014] The main drawbacks of the above scheme are: ① Gated triggering is only used to determine the start time of acquisition and is not combined with the dynamic adjustment mechanism of frame selection rate. The ultrasound system still works at a fixed frame rate, and the temporal resolution of key phases within a single cardiac cycle is not improved; ② The multi-cycle synthesis scheme is rhythm-dependent. When the patient has arrhythmia, the spatiotemporal misalignment of images from different heartbeats will occur; ③ It remains at the level of auxiliary functions of diagnostic ultrasound and is not combined with long-term automatic monitoring or LVEF automatic calculation system.

[0015] Existing Technology 4: Monitoring System for Indirect Estimation of EF by Multiple Sensors US Patent Application No. US11234601 discloses a multi-sensor cardiac function monitoring system that collects heart sound signals through multiple acoustic sensors on a wearable device and uses machine learning methods to calculate the ejection fraction. Keywords: heart sound, wearable, ejection fraction, machine learning.

[0016] The main drawback of this approach is that it abandons direct visualization of cardiac ultrasound, relying solely on one-dimensional heart sound signals to estimate ejection fraction (EF). Heart sound signals have a much lower signal-to-noise ratio and lower characteristic stability than ultrasound images, are highly sensitive to sensor location, patient posture, and environmental noise, have limited estimation accuracy, and lack direct visual evidence for result verification, posing a fundamental obstacle to clinical reliability. Summary of the Invention

[0017] To address the shortcomings of existing technologies, the purpose of this invention is to provide an automatic cardiac function monitoring ultrasound system and method with dual modes of positioning and monitoring.

[0018] An automatic cardiac function monitoring ultrasound system with dual modes of positioning and monitoring, provided by the present invention, is characterized in that it comprises: Fitted ultrasound probe: Connects to the ultrasound front-end module via a dedicated cable; ECG electrode patches: connected to the ECG acquisition module via dedicated lead wires; Ultrasonic front-end module: Receives mode switching and acquisition trigger commands from the main control and processing unit, outputs raw signal frame data streams to it; receives acquisition trigger timing signals generated by the hardware reference clock source of the main control and processing unit; ECG acquisition module: outputs ECG data stream and R-wave trigger signal to the main control and processing unit; receives sampling trigger timing signal generated by the hardware reference clock source of the main control and processing unit; Main control and processing unit: communicates bidirectionally with the ultrasound front-end module, ECG acquisition module, imaging processing unit, touch display and interaction module, and local storage unit through internal data channels; provides acquisition trigger timing signals directly generated from the same clock source to the ultrasound front-end module and ECG acquisition module through a hardware clock line; and issues working mode and imaging task scheduling instructions to the imaging processing unit. Imaging processing unit: receives raw signal frame data streams output by ultrasound front-end module; receives operating mode and imaging task scheduling instructions from main control and processing unit; outputs sequence of B-mode image frames that have been imaged to main control and processing unit. Touch display and interaction module: Receives display data from the main control and processing unit and sends touch operation events back to it; Local storage unit: connected to the main control and processing unit via an internal memory interface.

[0019] Preferably, in the fitted ultrasound probe: The lightweight probe, which integrates a small ultrasonic transducer array, can be moved by the operator in positioning mode to find the preset optimal acoustic window, and can be permanently fixed to the apex of the chest wall using coupling gel in monitoring mode. In the ECG electrode patch: The standard ECG electrode patch, independent of the ultrasound probe, is attached to the standard lead position on the patient's trunk to collect surface ECG signals; in monitoring mode, it can be attached for a preset time period to continuously provide ECG signals.

[0020] Preferably, in the ultrasonic front-end module: It completes ultrasonic beam transmission, reception, and beamforming, outputs raw signal frame data, and drives real-time imaging in positioning mode; in monitoring mode, it performs raw signal acquisition when the main control unit determines the key time phase based on the electrocardiogram signal; the acquisition trigger timing is directly driven and generated by the hardware reference clock source of the main control and processing unit. In monitoring mode, the system continuously monitors the electrocardiogram signal and predicts the arrival time of key phases in real time based on the RR interval. It then triggers the ultrasound front end to acquire raw signals within the preset key phase. In the ECG acquisition module: The system acquires surface electrocardiogram (ECG) signals through independent ECG electrode patches, and performs amplification, filtering, analog-to-digital conversion, and real-time R-wave detection. In positioning mode, it provides a reference waveform for the screen, and in monitoring mode, it continuously outputs ECG data streams and R-wave trigger signals to provide key phase prediction benchmarks for the main control unit. The sampling trigger timing is directly driven and generated by the hardware reference clock source of the main control and processing unit.

[0021] Preferably, in the main control and processing unit: It is responsible for the dual-mode switching logic, real-time imaging drive in positioning mode, key phase prediction and acquisition triggering based on ECG signals in monitoring mode, parallel scheduling of acquisition and imaging, lightweight AI endocardial segmentation and LVEF calculation, trend data management and retrospective response; it has a single hardware reference clock source, and the acquisition triggering timing of the ultrasound front-end module and the ECG acquisition module are directly driven and generated by the same clock source at the hardware level. The two share the same original oscillation signal, eliminating the relative drift between multiple clock sources at the circuit physical level. After completing the acquisition with a high selection rate in the preset key time phases, the system stops acquiring new data in the preset non-key time phases and concentrates all computing power on processing the previously backlogged raw data and completing the imaging processing in batches. In the imaging processing unit: The system receives raw signal frame data output from the ultrasound front end and performs imaging processing to generate B-mode image frames. In monitoring mode, the system continuously monitors ECG signals and predicts the arrival time of key phases in real time based on the RR interval. Within the preset key phases, the ultrasound front end is triggered to acquire raw signals. The imaging processing unit performs imaging processing in parallel, forming a parallel working architecture of acquisition production and imaging consumption. The imaging processing unit is configured according to the working mode: in positioning mode, it maintains real-time imaging capability; in monitoring mode, the preset key phases use a frame selection rate that is much higher than that in real-time display mode, so that the number of raw frames sent in exceeds its real-time processing throughput, and unprocessed frames are automatically backed up and temporarily stored; after the preset non-key phases stop acquisition, all computing power is concentrated to digest the backlog queue.

[0022] Preferably, in the touch display and interaction module: It provides a human-computer interaction interface with visual output and touch input. In positioning mode, it presents B-mode ultrasound images and ECG waveforms in real time to assist probe placement. In monitoring mode, it displays the LVEF time curve with a continuous trend graph and supports touch click to retrieve volumetric value frames, consecutive frame sequences, and AI segmentation contour overlay images. In the local storage unit: The database stores monitoring records, including calculated LVEF values, EDV / ESV plethysmogram frames and several consecutive frames before and after them, AI segmented contour overlay images, corresponding ECG bars and timestamps, and supports retrieval and review at any time.

[0023] According to the present invention, an automated cardiac function monitoring ultrasound method with dual modes of positioning and monitoring is provided, employing any one of the described automated cardiac function monitoring ultrasound systems with dual modes of positioning and monitoring, and performing the following: Step S1: Place the probe and confirm the ECG signal in positioning mode; Step S2: Initialize monitoring mode; Step S3: Collect ECG data with a high selection rate of preset key time phases; Step S4: Stop data acquisition during non-critical time phases and process the backlog of data; Step S5: Perform AI instant segmentation and LVEF calculation; Step S6: Update the trend and continue monitoring; Step S7: Perform interactive verification and backtracking.

[0024] Preferably, in step S1: Main control and processing unit: After the system is powered on, it enters positioning mode and works in real-time B-mode ultrasound, simultaneously displaying ultrasound sector images and electrocardiogram waveforms on the touch display interface. The ultrasound front-end module acquires raw signals and outputs raw frame data streams. The imaging processing unit operates in real-time imaging mode, performing imaging processing on the original frame with a conventional selection rate adapted to real-time display, generating B-mode image frames and outputting them to the display module in real time. The ECG acquisition module synchronously acquires ECG signals through independent ECG electrode patches, completes amplification, filtering, analog-to-digital conversion and R-wave detection, and outputs the ECG waveform to the display module for superposition and presentation. The operator moves the ultrasound probe in the apex of the patient's heart, finds the standard apical section based on the real-time image, and after confirming that the left ventricular endocardium is clearly displayed and the electrocardiogram waveform is stable, the probe is attached and fixed to the chest wall using coupling gel. Main control and processing unit: The operator clicks "Start Monitoring" on the touch interface, and the system completes the preparation for mode switching.

[0025] Preferably, in step S2: Main control and processing unit: Exit real-time B mode display, switch the imaging processing unit to non-real-time working mode, switch the touch display interface to the monitoring main interface, draw the LVEF time trend coordinate graph, and initialize the monitoring record database. The ultrasound front-end module enters standby mode, waiting for the ECG trigger acquisition command; The imaging processing unit switches to a non-real-time working mode, no longer requiring low-latency synchronous display, and prepares to receive background imaging tasks. The ECG acquisition module runs continuously, constantly outputting ECG data streams and R-wave trigger signals to the main control unit; In step S3: The main control and processing unit continuously analyzes the electrocardiogram (ECG) signal, and based on the ECG R wave and instantaneous heart rate (HR), predicts the start and end ranges of end-systolic and end-diastolic phases on the time axis; the prediction logic is as follows: It continuously receives digital ECG signals output from the ECG acquisition module and runs a real-time R-wave detection algorithm; whenever a new R-wave is detected... k k represents the k-th occurrence recorded, and the time of occurrence is T. k Then the interval between the current R wave and the previous R wave is calculated as RR. k The interval between the k-th R wave and the previous k-1 R waves is RR. k =T k -T k-1 Calculate the current instantaneous heart rate (HR). k =60 / RR k (beats / min), the instantaneous heart rate is the smoothed heart rate fluctuation, and the total smoothed heart rate HR is obtained by exponential weighted average. smooth For further prediction: HR smooth =α*HR k +(1-α)*HR smooth(last) Where α is the smoothing coefficient, ranging from 0.3 to 0.5, HR smooth(last) This is the smoothed heart rate value obtained from the previous calculation; Using the current R wave as the time origin and T=0, re-predict the end-systolic (ES) and end-diastolic (ED) windows within the current cardiac cycle, while terminating any unfinished prediction windows from the previous cycle; estimate the timing of the next R wave based on a smoothed heart rate (HR). smooth Estimate the length of the next R-wave interval (RR) in the current cycle. next RR next =60 / HR smooth The estimated value of the next R wave is T'. k+1 =T k +RR next ; Predicting the ES window, the center time of the end of contraction is located after the current R wave, and its relative offset is proportional to the RR interval; define the offset coefficient K. es The baseline value is 0.37, corresponding to a heart rate of 75 bpm; when the heart rate increases, K... es Decrease when heart rate decreases, increase when heart rate decreases, and keep the dynamic range between 0.30 and 0.45; then the ES center time: T ES_center =T k +K es *RR next ES window width W ESTaking 8% of the RR interval, and limiting it to between 40 ms and 120 ms, the start and end range of the ES window is [T]. ES_center -W ES / 2,T ES_center +W ES / 2]; The predicted end-diastolic (ED) window is defined as the time lead between the ED center and the expected next R wave; the time lead coefficient K between the ED center and the next R wave is defined. ed The baseline value is 0.15, and the dynamic range is 0.10 to 0.20; therefore, the center time of ED is: T ED_center =T' k+1 -K ed *RR next ES window width W ED The processing logic is the same as ES, so the start and end range of the ED window is [T]. ED_center -W ED / 2,T ED_center +W ED / 2]; When a critical phase is predicted to be approaching, an acquisition task is triggered: an acquisition command is sent to the ultrasound front-end module, and the frame selection rate of the raw signal during the critical phase is much higher than that of the real-time display mode. After receiving the acquisition command, the ultrasound front-end module immediately acquires the raw signal and outputs the raw frame data stream; the acquisition trigger timing is directly driven by the hardware reference clock source of the main control unit. The imaging processing unit operates in parallel to process the incoming raw frame data stream. Since the frame selection rate in monitoring mode is much higher than that in real-time display mode, the number of raw frames sent per unit time increases. The portion of raw frames that exceed the real-time processing throughput of the imaging processing unit is temporarily stored in an internal buffer, forming a backlog queue. The sampling trigger timing of the ECG acquisition module is driven by the same hardware reference clock source, maintaining the same timing source as the ultrasound acquisition at the circuit level; it continuously outputs ECG data stream and R-wave trigger signal, allowing the main control unit to accurately determine the start and end boundaries of key time phases.

[0026] Preferably, in step S4: Main control and processing unit: After the critical time phase window ends, a stop acquisition command is sent to the ultrasound front-end module; no new acquisitions are performed during the non-critical time phase period, and all computing power of the imaging processing unit is concentrated on processing the previously backlogged raw data; The ultrasound front-end module stops acquiring the original signal, and the high-voltage transmission circuit is partially shut down to reduce power consumption. The imaging processing unit continuously retrieves raw data frame by frame from the buffer backlog queue, completes imaging processing frame by frame, generates B-mode image frames and marks the original acquisition timestamp; the imaging data is generated and sent to the AI ​​model for analysis at the same time. In step S5: Main control and processing unit: After the imaging data is produced frame by frame, the pre-deployed lightweight deep learning segmentation network immediately performs automatic segmentation of the left ventricular endocardial boundary on the already imaged frames; the AI ​​continuously and dynamically searches for extreme frames of LVEDV and LVESV in the continuous frame sequence and uses the Simpson two-plane method to calculate the LVEF value. The LVEF calculation results, EDV / ESV capacity positive value frames and several preset consecutive frames before and after them, AI segmentation contour overlay map and corresponding ECG bar map, and acquisition timestamp are packaged and stored in the local storage unit. Preferably, in step S6: The main control and processing unit appends the current LVEF value to the trend data sequence and refreshes the LVEF time trend curve on the touch display interface; the current acquisition task is completed, and the ultrasound front-end module and imaging processing unit return to standby mode. The ECG acquisition module runs continuously to maintain continuous monitoring of ECG signals; The main control and processing unit continues to analyze the electrocardiogram signal and predicts the arrival time of the next preset key time phase based on the RR interval. When the preset key time phase is predicted to be about to enter again, a new round of acquisition task is triggered. In step S7: During the monitoring process, the touch display and interaction module allows users to touch and click any data point on the trend chart at any time, and a verification and retrospective window will pop up on the screen. The main control and processing unit retrieves the EDV / ESV tolerance value frame associated with the data point, as well as the consecutive frame sequences before and after it and the AI ​​segmentation contour overlay image from the local storage unit, and presents them on the touch screen for doctors to scroll through frame by frame. Main control and processing unit: By observing the continuous dynamic change of the internal cavity volume from large to small in key time phases, it independently judges whether the extreme value moment selected by the AI ​​is accurate by comparing the previous and next frames; this credibility verification mechanism ensures that each LVEF trend data is supported by original image evidence.

[0027] Compared with the prior art, the present invention has the following beneficial effects: This invention systematically overcomes the shortcomings of existing technologies, such as limitations imposed by occasional measurement modes, power consumption and transmission bandwidth, fixed frame rate acquisition, or lack of visualization data. Through a dual-mode architecture of positioning and monitoring, this invention transforms echocardiographic function assessment from manual occasional measurement to fully automated long-term monitoring based on ECG triggering. By decoupling acquisition and imaging in parallel, it removes the frame rate limitation imposed by tightly coupled pipelines. In monitoring mode, it predicts key time phases in real time based on ECG signals and triggers acquisition, using a frame selection rate far higher than in real-time display mode to densely acquire raw data. Acquisition of non-critical time phases is stopped, and computing power is concentrated to process backlogged data, achieving high frame density imaging of key time phases under the constraints of portable hardware computing power. Through AI-based real-time analysis and trend backtracking verification interaction modes, a complete result loop is formed. Accurate coverage of the ECG trigger acquisition window is ensured by a shared hardware clock. Attached Figure Description

[0028] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the system architecture; Figure 2 This is a schematic diagram of the overall process. Detailed Implementation

[0029] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0030] Example 1: 1. Extracting the invention point This invention relates to an automated ultrasound system and method for long-term cardiac function monitoring with dual modes of positioning and monitoring, such as... Figures 1-2 The invention shown is based on an integrated computing and interaction platform, a fitted ultrasound probe, and independent ECG electrode patches. Its core invention points include: First, the system employs a dual-mode architecture of real-time imaging positioning and fully automated monitoring. In positioning mode, the system outputs images in real-time using the standard B mode, assisting the operator in attaching and fixing the probe at the optimal acoustic window position. In monitoring mode, the system completely exits real-time imaging and enters a long-term working state of ECG-triggered acquisition and non-real-time imaging analysis.

[0031] Second, the monitoring operation mode of parallel decoupling between the underlying acquisition and imaging output. Traditional ultrasound imaging mainly consists of four serial processing stages: signal acquisition, beamforming, signal processing, and image processing. Signal acquisition is generally composed of a transducer and a front-end board, referred to as the ultrasound front-end acquisition module in this invention. Beamforming, signal processing, and image processing are generally implemented by high-performance computing hardware such as FPGAs or GPUs, collectively referred to as the imaging processing unit in this invention. In serial mode, the main performance bottleneck of ultrasound imaging lies in the beamforming stage. Using a producer-consumer analogy, the signal acquisition producer can generate a large number of raw data frames per unit time, but the number of beamforming frames that can be completed per unit time is far less than the number of raw data frames. Hence, there is the concept of a frame selection rate, which selects a portion of the raw frames to submit to subsequent units for beamforming synthesis. To achieve parallel decoupling between acquisition and imaging output, the device needs to be equipped with a certain amount of memory or high-speed storage as a cache to cache more raw frames beyond the real-time processing capacity of the imaging processing unit. Parallel mode allows the system's acquisition module and imaging processing unit to operate independently. The acquisition module receives a signal and switches to acquire data at intervals, storing it in a buffer. The imaging processing unit retrieves raw frames from the buffer and processes them one by one. While this compromises real-time performance, the buffer allows for more raw data to form the final image. In this invention, combined with the above approach, in monitoring mode, the system continuously monitors ECG signals, predicts the arrival time of critical phases in real time based on the RR interval, and triggers acquisition. The ultrasound front-end acquires raw signals during critical phases, acting as a high-speed data producer; the imaging processing workflow runs as an independent background task in parallel, processing data as it becomes available, without requiring real-time image output. This approach decouples the traditional tightly coupled synchronous pipeline of "acquisition-imaging-display" ultrasound into a parallel architecture of "acquisition production, imaging consumption," providing a foundation for allocating system computing power on demand to improve the temporal resolution of images during critical periods.

[0032] Third, a high frame selection rate acquisition method based on ECG-gated triggering. The system synchronizes ECG signals and triggers acquisition when the end-systolic and end-diastolic phases are predicted to begin. During critical time phases, a frame selection rate significantly higher than that of conventional real-time display mode is applied to the raw signals output from the ultrasound front end (in real-time display, the frequency of the raw signals acquired at the underlying layer is much higher than the display frame rate, so only a portion is selected for rendering; the selection probability is the frame selection rate, and a higher selection rate means more imaging frames are needed per unit time limit), significantly improving the temporal resolution of critical time phases. The high selection rate dramatically increases the amount of data sent to the imaging queue per unit time, and the portion exceeding the backend imaging processing capacity will result in raw data backlog. No new acquisitions are performed for non-critical time phases; only ECG monitoring is retained.

[0033] Fourth, the scheduling mechanism for high-selectivity acquisition during critical periods and subsequent centralized processing of backlogged data. After completing high-selectivity acquisition in critical time phases, the system stops new acquisition in non-critical time phases, concentrating all computing power on processing the previously backlogged raw data and completing imaging processing batch by batch. This "backlog-then-compensate" computing power scheduling enables the effective imaging frame density of critical time phases to be multiplied under the constraints of portable hardware computing power, ensuring the accuracy of LVEF measurements.

[0034] Fifth, AI-powered real-time analysis and an interactive mode combining trend charts and retrospective verification. In monitoring mode, after imaging data is generated, the system uses a lightweight deep learning model to perform automatic endocardial segmentation on key frame sequences in real time, locating the volumetric value in consecutive frames and calculating LVEF. The results are presented as a continuous trend curve. The system saves the volumetric value frame used for each calculation, along with several consecutive frames before and after it. Clicking on any data point on the trend chart allows the user to retrieve the corresponding volumetric value frame, the preceding and following consecutive frame sequences, and an overlay of the AI ​​segmentation contour. This allows the operator to retrospectively verify the accuracy of the AI's determination of extreme values, forming a closed loop of "trend overview—evidence retrospection—dynamic verification."

[0035] Sixth, a synchronous ultrasound-ECG acquisition mechanism driven by a single hardware clock. The system uses a single hardware reference clock source to directly drive the acquisition trigger timing of the ultrasound front end and the ECG acquisition module. The two share the same original oscillation signal, eliminating the relative drift between multiple clock sources at the circuit physical level, and ensuring that the ECG trigger acquisition window always accurately covers the target phase during long-term monitoring.

[0036] 2. Structural Relationships & Working Principle 2.1 Structural Composition and Connection Relationships The system of this invention consists of an adhesive ultrasound probe, ECG electrode patches, and an integrated computing and interactive platform. The platform internally includes the following functional modules: an ultrasound front-end module, an ECG acquisition module, a main control and processing unit, an imaging processing unit, a touch display and interaction module, and a local storage unit. The functions and connections of each module are as follows:

[0037] Key Explanation: In this system, during monitoring mode, the system continuously monitors ECG signals and predicts the arrival time of critical phases in real time based on the RR interval. Within the critical phase, the ultrasound front end is triggered to acquire raw signals. The imaging processing unit performs imaging processing in parallel, forming a parallel working architecture of "acquisition production and imaging consumption." The imaging processing unit can be configured according to the working mode: in positioning mode, it maintains real-time imaging capability; in monitoring mode, critical phases use a frame selection rate much higher than in real-time display mode, causing the number of raw frames sent to exceed its real-time processing throughput, and unprocessed frames are automatically backed up and temporarily stored; after non-critical phases stop acquisition, all computing power is concentrated to process the backlog queue.

[0038] In this system, the main control and processing unit has a single hardware reference clock source. The acquisition trigger timing of the ultrasound front-end module and the ECG acquisition module are directly driven and generated by the same clock source at the hardware level. The two share the same original oscillation signal and do not have their own independent clock domains. This eliminates the relative drift between multiple clock sources at the physical level, ensuring that the ECG trigger acquisition window always accurately covers the target time phase during long-term monitoring.

[0039] 2.2 Working Principle Upon power-up, the system defaults to positioning mode. The ultrasound front end acquires raw signals, while the imaging processing unit operates in real-time imaging mode. The touchscreen display simultaneously shows ultrasound sector images and ECG waveforms. The operator moves the probe across the patient's apex region, locating a standard apical section based on the real-time images. After confirming clear left ventricular structures and stable ECG signals, the operator secures the probe to the chest wall using coupling gel and then activates monitoring mode via the touchscreen interface.

[0040] After switching to monitoring mode, the display interface switches from real-time images to an LVEF-time trend graph. The imaging processing unit switches to non-real-time operation mode, and the system continuously monitors the ECG signal, predicting the arrival times of end-systole and end-diastole in real time based on the RR interval. When a critical phase is predicted to be approaching, an acquisition task is triggered: the ultrasound front end acquires raw signals during the critical phase, using a frame selection rate much higher than in real-time display mode, and the imaging processing unit processes them in parallel. Based on the same hardware clock, the ECG R wave and ultrasound frame timing are strictly consistent, ensuring accurate determination of the critical phase window. The high selection rate causes the number of raw frames sent to exceed the real-time processing throughput of the imaging processing unit; the excess is temporarily stored in a buffer to form a backlog queue. After the critical phase ends, the system stops acquiring data during non-critical phases, and the imaging processing unit concentrates all its computing power to process the backlog data, completing imaging in batches.

[0041] After the imaging data is generated, the lightweight AI model immediately performs automatic endocardial segmentation on the keyframe sequence, locates the extreme EDV / ESV frames in consecutive frames, and calculates LVEF. The calculation results, along with the positive volume frame and its preceding and following consecutive frames, the AI ​​contour overlay map, and the ECG bar graph, are stored locally. The LVEF value is appended to the trend database, and the curve is updated. The task ends, the system returns to standby, and continues to monitor ECG signals awaiting the next key phase trigger.

[0042] In routine monitoring, clinicians observe trend curves to understand the dynamic changes in LVEF. Clicking on any data point on the curve retrieves the calculated extreme value frame, the preceding and following frame sequences, and an AI-segmented contour overlay image, allowing operators to scroll through the dynamic changes in cardiac chamber volume. By comparing preceding and following frames, the accuracy of the AI's determination of extreme value moments can be independently verified.

[0043] 2.3 Workflow of software and hardware working together The following section uses a complete monitoring cycle as an example to detail the timing process of software and hardware collaborative execution. Step 1: Probe placement and ECG signal confirmation in positioning mode Software (Main Control and Processing Unit): After the system is powered on, it enters the positioning mode by default and works in the conventional real-time B-mode ultrasound. The ultrasound sector image and ECG waveform are displayed on the touch screen interface at the same time.

[0044] Hardware (ultrasound front-end module): Acquires raw signals and outputs raw frame data streams.

[0045] Hardware (Imaging Processing Unit): Operates in real-time imaging mode, performs imaging processing on the original frame at a standard selection rate adapted to real-time display, generates B-mode image frames, and outputs them to the display module in real time.

[0046] Hardware (ECG acquisition module): The ECG signal is acquired synchronously through independent ECG electrode patches, and amplification, filtering, analog-to-digital conversion and R-wave detection are completed. The ECG waveform is then output to the display module for superposition and presentation.

[0047] Operator: Hold the ultrasound probe and move it around the apex of the patient's heart. Find the standard apical section based on the real-time image. After confirming that the left ventricular endocardium is clearly displayed and the electrocardiogram waveform is stable, attach and fix the probe to the chest wall with coupling gel.

[0048] Software (main control and processing unit): The operator clicks "Start Monitoring" on the touch interface, and the system completes the mode switching preparation.

[0049] Step 2: Monitor mode initialization Software (Main Control and Processing Unit): Exit real-time B mode display, switch the imaging processing unit to non-real-time working mode, switch the touch display interface to the monitoring main interface, draw the LVEF-time trend coordinate graph, and initialize the monitoring record database.

[0050] Hardware (ultrasound front-end module): Enters standby mode, waiting for ECG trigger acquisition command.

[0051] Hardware (Imaging Processing Unit): Switches to non-real-time working mode, no longer requiring low-latency synchronous display, and prepares to receive background imaging tasks.

[0052] Hardware (ECG acquisition module): Continuously runs, continuously outputting ECG data stream and R-wave trigger signal to the main control unit.

[0053] Step 3: High selectivity acquisition of key phases triggered by ECG Software (main control and processing unit): Continuously analyzes ECG signals, predicting the start and end ranges of end-systolic and end-diastolic phases on the time axis based on the ECG R wave and instantaneous heart rate (HR). The prediction logic is as follows: It continuously receives digital ECG signals output from the ECG acquisition module and runs a real-time R-wave detection algorithm; whenever a new R-wave is detected... k k represents the k-th occurrence recorded, and the time of occurrence is T. k Then the interval between the current R wave and the previous R wave is calculated as RR. k The interval between the k-th R wave and the previous k-1 R waves is RR. k =T k -T k-1 Calculate the current instantaneous heart rate (HR). k =60 / RR k (beats / min), the instantaneous heart rate is the smoothed heart rate fluctuation, and the total smoothed heart rate HR is obtained by exponential weighted average. smooth For further prediction: HR smooth =α*HR k +(1-α)*HR smooth(last) Where α is the smoothing coefficient, ranging from 0.3 to 0.5, with a default value of 0.4, and HR smooth(last) This is the smoothed heart rate value obtained from the previous calculation; Using the current R wave as the time origin and T=0, re-predict the end-systolic (ES) and end-diastolic (ED) windows within the current cardiac cycle, while terminating any unfinished prediction windows from the previous cycle; estimate the timing of the next R wave based on a smoothed heart rate (HR). smooth Estimate the length of the next R-wave interval (RR) in the current cycle. next RR next =60 / HRsmooth The estimated value of the next R wave is T'. k+1 =T k +RR next ; Predicting the ES window, the center of the end of contraction is located after the current R wave, and its relative offset is proportional to the RR interval; define the offset coefficient K. es The baseline value is 0.37, corresponding to a heart rate of 75 bpm; when the heart rate increases, K... es Decrease when heart rate decreases, increase when heart rate decreases, and keep the dynamic range between 0.30 and 0.45; then the ES center time: T ES_center =T k +K es *RR next ES window width W ES Taking 8% of the RR interval, and limiting it to between 40 ms and 120 ms, the start and end range of the ES window is [T]. ES_center -W ES / 2,T ES_center +W ES / 2]; The predicted end-diastolic (ED) window is defined as the time lead between the ED center and the expected next R wave; the time lead coefficient K between the ED center and the next R wave is defined. ed The baseline value is 0.15, the dynamic range is 0.10–0.20, and the higher the heart rate, the closer the ED window is to the next R wave; therefore, the ED center time is: T ED_center =T' k+1 -K ed *RR next ES window width W ED The processing logic is the same as ES, so the start and end range of the ED window is [T]. ED_center -W ED / 2,T ED_center +W ED / 2]; When a critical time phase is predicted to be approaching, an acquisition task is triggered: an acquisition command is sent to the ultrasound front-end module, and the frame selection rate of the raw signal during that critical time phase is much higher than that of the real-time display mode.

[0054] Hardware (ultrasound front-end module): Upon receiving the acquisition command, it immediately acquires the raw signal and outputs the raw frame data stream. The acquisition trigger timing is directly driven by the hardware reference clock source of the main control unit.

[0055] Hardware (Imaging Processing Unit): Operates in parallel, processing the incoming raw frame data stream. Because the frame selection rate in monitoring mode is much higher than in real-time display mode, the number of raw frames sent per unit time increases significantly. Raw frames exceeding the real-time processing capacity of the imaging processing unit are temporarily stored in an internal buffer, forming a backlog queue.

[0056] Hardware (ECG acquisition module): The sampling trigger timing is driven by the same hardware reference clock source, maintaining the same timing source as ultrasound acquisition at the circuit level. It continuously outputs ECG data stream and R-wave trigger signal, allowing the main control unit to accurately determine the start and end boundaries of key time phases.

[0057] Step 4: Stop acquiring data during non-critical time phases and process backlogged data. Software (main control and processing unit): After the critical time phase window ends, a stop acquisition command is sent to the ultrasound front-end module. No new acquisitions are performed during non-critical time phases, and all computing power of the imaging processing unit is concentrated on processing the previously backlogged raw data.

[0058] Hardware (ultrasound front-end module): Stop the acquisition of raw signals and shut down the high-voltage transmission circuit to reduce power consumption.

[0059] Hardware (Imaging Processing Unit): Continuously retrieves raw data frame by frame from the buffer backlog queue, performs imaging processing frame by frame, generates B-mode image frames, and marks the original acquisition timestamp. Imaging data is simultaneously generated and fed into the AI ​​model for analysis.

[0060] Step 5: AI-based instantaneous segmentation and LVEF calculation Software (main control and processing unit): After the imaging data is generated frame by frame, the deployed lightweight deep learning segmentation network immediately performs automatic segmentation of the left ventricular endocardial boundary on the imaged frames. The AI ​​continuously and dynamically searches for extreme frames of LVEDV and LVESV in the continuous frame sequence, and then applies the Simpson two-plane method to calculate the LVEF value.

[0061] Software (main control and processing unit): Packages and stores the LVEF calculation results, EDV / ESV volumic value frames and several consecutive frames before and after them, AI segmentation contour overlay map and corresponding ECG bar map, and acquisition timestamp into the local storage unit.

[0062] Step 6: Trend Updates and Continued Monitoring Software (main control and processing unit): Appends the current LVEF value to the trend data sequence and refreshes the LVEF-time trend curve on the touch display interface. The acquisition task is now complete, and the ultrasound front-end module and imaging processing unit return to standby mode.

[0063] Hardware (ECG acquisition module): Continuously operates to maintain continuous monitoring of ECG signals.

[0064] Software (main control and processing unit): Continue to analyze the ECG signal, predict the arrival time of the next critical phase based on the RR interval, and when the critical phase is predicted to be about to enter again, return to step 3 to trigger a new round of acquisition tasks.

[0065] Step 7: Doctor Interaction Verification and Retrospection Software (touch display and interaction module): During the monitoring process, the doctor can touch and click any data point on the trend chart at any time, and a verification and retrospective window will pop up on the screen.

[0066] Software (main control and processing unit): Retrieves the EDV / ESV tolerance positive value frame associated with the data point, as well as the consecutive frame sequences before and after it and the AI ​​segmentation contour overlay image from the local storage unit, and presents them on the touch screen for doctors to scroll through frame by frame.

[0067] Software (main control and processing unit): By observing the continuous dynamic change in the intracardiac cavity volume from large to small during key time phases, doctors independently judge the accuracy of the extreme value moments selected by the AI ​​by comparing preceding and following frames. This credibility verification mechanism ensures that every LVEF trend data point is supported by original image evidence.

[0068] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0069] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. An automatic cardiac function monitoring ultrasound system with dual modes of positioning and monitoring, characterized in that, include: Fitted ultrasound probe: Connects to the ultrasound front-end module via a dedicated cable; ECG electrode patches: connected to the ECG acquisition module via dedicated lead wires; Ultrasonic front-end module: Receives mode switching and acquisition trigger commands from the main control and processing unit, outputs raw signal frame data streams to it; receives acquisition trigger timing signals generated by the hardware reference clock source of the main control and processing unit; ECG acquisition module: outputs ECG data stream and R-wave trigger signal to the main control and processing unit; receives sampling trigger timing signal generated by the hardware reference clock source of the main control and processing unit; Main control and processing unit: communicates bidirectionally with the ultrasound front-end module, ECG acquisition module, imaging processing unit, touch display and interaction module, and local storage unit through internal data channels; provides acquisition trigger timing signals directly generated from the same clock source to the ultrasound front-end module and ECG acquisition module through a hardware clock line; and issues working mode and imaging task scheduling instructions to the imaging processing unit. Imaging processing unit: receives raw signal frame data streams output by ultrasound front-end module; receives operating mode and imaging task scheduling instructions from main control and processing unit; outputs sequence of B-mode image frames that have been imaged to main control and processing unit. Touch display and interaction module: Receives display data from the main control and processing unit and sends touch operation events back to it; Local storage unit: connected to the main control and processing unit via an internal memory interface.

2. The automatic cardiac function monitoring ultrasound system with dual modes of positioning and monitoring as described in claim 1, characterized in that: In the aforementioned conformal ultrasound probe: The lightweight probe, which integrates a small ultrasonic transducer array, can be moved by the operator in positioning mode to find the preset optimal acoustic window, and can be permanently fixed to the apex of the chest wall using coupling gel in monitoring mode. In the ECG electrode patch: The standard ECG electrode patch, independent of the ultrasound probe, is attached to the standard lead position on the patient's trunk to collect surface ECG signals; in monitoring mode, it can be attached for a preset time period to continuously provide ECG signals.

3. The automatic cardiac function monitoring ultrasound system with dual modes of positioning and monitoring as described in claim 1, characterized in that: In the ultrasound front-end module: It completes ultrasonic beam transmission, reception, and beamforming, outputs raw signal frame data, and drives real-time imaging in positioning mode; in monitoring mode, it performs raw signal acquisition when the main control unit determines the key time phase based on the electrocardiogram signal; the acquisition trigger timing is directly driven and generated by the hardware reference clock source of the main control and processing unit. In monitoring mode, the system continuously monitors the electrocardiogram signal and predicts the arrival time of key phases in real time based on the RR interval. It then triggers the ultrasound front end to acquire raw signals within the preset key phase. In the ECG acquisition module: The system acquires surface electrocardiogram (ECG) signals through independent ECG electrode patches, and performs amplification, filtering, analog-to-digital conversion, and real-time R-wave detection. In positioning mode, it provides a reference waveform for the screen, and in monitoring mode, it continuously outputs ECG data streams and R-wave trigger signals to provide key phase prediction benchmarks for the main control unit. The sampling trigger timing is directly driven and generated by the hardware reference clock source of the main control and processing unit.

4. The automatic cardiac function monitoring ultrasound system with dual modes of positioning and monitoring as described in claim 1, characterized in that: In the main control and processing unit: It is responsible for the dual-mode switching logic, real-time imaging drive in positioning mode, key phase prediction and acquisition triggering based on ECG signals in monitoring mode, parallel scheduling of acquisition and imaging, lightweight AI endocardial segmentation and LVEF calculation, trend data management and retrospective response; it has a single hardware reference clock source, and the acquisition triggering timing of the ultrasound front-end module and the ECG acquisition module are directly driven and generated by the same clock source at the hardware level. The two share the same original oscillation signal, eliminating the relative drift between multiple clock sources at the circuit physical level. After completing the acquisition with a high selection rate in the preset key time phases, the system stops acquiring new data in the preset non-key time phases and concentrates all computing power on processing the previously backlogged raw data and completing the imaging processing in batches. In the imaging processing unit: The system receives raw signal frame data output from the ultrasound front end and performs imaging processing to generate B-mode image frames. In monitoring mode, the system continuously monitors ECG signals and predicts the arrival time of key phases in real time based on the RR interval. Within the preset key phases, the ultrasound front end is triggered to acquire raw signals. The imaging processing unit performs imaging processing in parallel, forming a parallel working architecture of acquisition production and imaging consumption. The imaging processing unit is configured according to the working mode: in positioning mode, it maintains real-time imaging capability; in monitoring mode, the preset key phases use a frame selection rate that is much higher than that in real-time display mode, so that the number of raw frames sent in exceeds its real-time processing throughput, and unprocessed frames are automatically backed up and temporarily stored; after the preset non-key phases stop acquisition, all computing power is concentrated to digest the backlog queue.

5. The automatic cardiac function monitoring ultrasound system with dual modes of positioning and monitoring as described in claim 1, characterized in that: In the touch display and interaction module: It provides a human-computer interaction interface with visual output and touch input. In positioning mode, it presents B-mode ultrasound images and ECG waveforms in real time to assist probe placement. In monitoring mode, it displays the LVEF time curve with a continuous trend graph and supports touch click to retrieve volumetric value frames, consecutive frame sequences, and AI segmentation contour overlay images. In the local storage unit: The database stores monitoring records, including calculated LVEF values, EDV / ESV plethysmogram frames and several consecutive frames before and after them, AI segmented contour overlay images, corresponding ECG bars and timestamps, and supports retrieval and review at any time.

6. An automated ultrasound method for monitoring cardiac function with dual modes of positioning and monitoring, characterized in that, Using the automatic cardiac function monitoring ultrasound system with dual modes of positioning and monitoring as described in any one of claims 1-5, the following functions are performed: Step S1: Place the probe and confirm the ECG signal in positioning mode; Step S2: Initialize monitoring mode; Step S3: Collect ECG data with a high selection rate of preset key time phases; Step S4: Stop data acquisition during non-critical time phases and process the backlog of data; Step S5: Perform AI instant segmentation and LVEF calculation; Step S6: Update the trend and continue monitoring; Step S7: Perform interactive verification and backtracking.

7. The automatic cardiac function monitoring ultrasound method with dual modes of positioning and monitoring according to claim 6, characterized in that, In step S1: Main control and processing unit: After the system is powered on, it enters positioning mode and works in real-time B-mode ultrasound, simultaneously displaying ultrasound sector images and electrocardiogram waveforms on the touch display interface. The ultrasound front-end module acquires raw signals and outputs raw frame data streams. The imaging processing unit operates in real-time imaging mode, performing imaging processing on the original frame with a conventional selection rate adapted to real-time display, generating B-mode image frames and outputting them to the display module in real time. The ECG acquisition module synchronously acquires ECG signals through independent ECG electrode patches, completes amplification, filtering, analog-to-digital conversion and R-wave detection, and outputs the ECG waveform to the display module for superposition and presentation. The operator moves the ultrasound probe in the apex of the patient's heart, finds the standard apical section based on the real-time image, and after confirming that the left ventricular endocardium is clearly displayed and the electrocardiogram waveform is stable, the probe is attached and fixed to the chest wall using coupling gel. Main control and processing unit: The operator clicks "Start Monitoring" on the touch interface, and the system completes the preparation for mode switching.

8. The automatic cardiac function monitoring ultrasound method with dual modes of positioning and monitoring according to claim 6, characterized in that: In step S2: Main control and processing unit: Exit real-time B mode display, switch the imaging processing unit to non-real-time working mode, switch the touch display interface to the monitoring main interface, draw the LVEF time trend coordinate graph, and initialize the monitoring record database. The ultrasound front-end module enters standby mode, waiting for the ECG trigger acquisition command; The imaging processing unit switches to a non-real-time working mode, no longer requiring low-latency synchronous display, and prepares to receive background imaging tasks. The ECG acquisition module runs continuously, constantly outputting ECG data streams and R-wave trigger signals to the main control unit; In step S3: The main control and processing unit continuously analyzes the electrocardiogram (ECG) signal, and based on the ECG R wave and instantaneous heart rate (HR), predicts the start and end ranges of end-systolic and end-diastolic phases on the time axis; the prediction logic is as follows: It continuously receives digital ECG signals output from the ECG acquisition module and runs a real-time R-wave detection algorithm; whenever a new R-wave is detected... k k represents the k-th occurrence recorded, and the time of occurrence is T. k Then the interval between the current R wave and the previous R wave is calculated as RR. k The interval between the k-th R wave and the previous k-1 R waves is RR. k =T k -T k-1 Calculate the current instantaneous heart rate (HR). k =60 / RR k (beats / min), the instantaneous heart rate is the smoothed heart rate fluctuation, and the total smoothed heart rate HR is obtained by exponential weighted average. smooth For further prediction: HR smooth =α*HR k +(1-α)*HR smooth(last) Where α is the smoothing coefficient, ranging from 0.3 to 0.5, HR smooth(last) This is the smoothed heart rate value obtained from the previous calculation; Using the current R wave as the time origin and T=0, re-predict the end-systolic (ES) and end-diastolic (ED) windows within the current cardiac cycle, while terminating any unfinished prediction windows from the previous cycle; estimate the timing of the next R wave based on a smoothed heart rate (HR). smooth Estimate the length of the next R-wave interval (RR) in the current cycle. next RR next =60 / HR smooth The estimated value of the next R wave is T'. k+1 =T k +RR next ; Predicting the ES window, the center time of the end of contraction is located after the current R wave, and its relative offset is proportional to the RR interval; define the offset coefficient K. es The baseline value is 0.37, corresponding to a heart rate of 75 bpm; when the heart rate increases, K... es Decrease when heart rate decreases, increase when heart rate decreases, and keep the dynamic range between 0.30 and 0.45; then the ES center time: T ES_center =T k +K es *RR next ES window width W ES Taking 8% of the RR interval, and limiting it to between 40 ms and 120 ms, the start and end range of the ES window is [T]. ES_center -W ES / 2,T ES_center +W ES / 2]; The predicted end-diastolic (ED) window is defined as the time lead between the ED center and the expected next R wave; the time lead coefficient K between the ED center and the next R wave is defined. ed The baseline value is 0.15, and the dynamic range is 0.10 to 0.20; therefore, the center time of ED is: T ED_center =T’ k+1 -K ed *RR next ES window width W ED The processing logic is the same as ES, so the start and end range of the ED window is [T]. ED_center -W ED / 2,T ED_center +W ED / 2]; When a critical phase is predicted to be approaching, an acquisition task is triggered: an acquisition command is sent to the ultrasound front-end module, and the frame selection rate of the raw signal during the critical phase is much higher than that of the real-time display mode. After receiving the acquisition command, the ultrasound front-end module immediately acquires the raw signal and outputs the raw frame data stream. The acquisition trigger timing is directly driven by the hardware reference clock source of the main control unit; The imaging processing unit operates in parallel to process the incoming raw frame data stream. Since the frame selection rate in monitoring mode is much higher than that in real-time display mode, the number of raw frames sent per unit time increases. The portion of raw frames that exceed the real-time processing throughput of the imaging processing unit is temporarily stored in an internal buffer, forming a backlog queue. The sampling trigger timing of the ECG acquisition module is driven by the same hardware reference clock source, maintaining the same timing source as the ultrasound acquisition at the circuit level; it continuously outputs ECG data stream and R-wave trigger signal, allowing the main control unit to accurately determine the start and end boundaries of key time phases.

9. The automatic cardiac function monitoring ultrasound method with dual modes of positioning and monitoring according to claim 6, characterized in that: In step S4: Main control and processing unit: After the critical time phase window ends, a stop acquisition command is sent to the ultrasound front-end module; no new acquisitions are performed during the non-critical time phase period, and all computing power of the imaging processing unit is concentrated on processing the previously backlogged raw data; The ultrasound front-end module stops acquiring the original signal, and the high-voltage transmission circuit is partially shut down to reduce power consumption. The imaging processing unit continuously retrieves raw data frame by frame from the buffer backlog queue, completes imaging processing frame by frame, generates B-mode image frames and marks the original acquisition timestamp; the imaging data is generated and sent to the AI ​​model for analysis at the same time. In step S5: Main control and processing unit: After the imaging data is produced frame by frame, the pre-deployed lightweight deep learning segmentation network immediately performs automatic segmentation of the left ventricular endocardial boundary on the already imaged frames; the AI ​​continuously and dynamically searches for extreme frames of LVEDV and LVESV in the continuous frame sequence and uses the Simpson two-plane method to calculate the LVEF value. The LVEF calculation results, EDV / ESV positive value frames and several preset consecutive frames before and after them, AI segmentation contour overlay map and corresponding ECG bar map, and acquisition timestamp are packaged and stored in the local storage unit.

10. The automatic cardiac function monitoring ultrasound method with dual modes of positioning and monitoring according to claim 6, characterized in that: In step S6: The main control and processing unit appends the current LVEF value to the trend data sequence and refreshes the LVEF time trend curve on the touch display interface; the current acquisition task is completed, and the ultrasound front-end module and imaging processing unit return to standby mode. The ECG acquisition module runs continuously to maintain continuous monitoring of ECG signals; The main control and processing unit continues to analyze the electrocardiogram signal and predicts the arrival time of the next preset key time phase based on the RR interval. When the preset key time phase is predicted to be about to enter again, a new round of acquisition task is triggered. In step S7: During the monitoring process, the touch display and interaction module allows users to touch and click any data point on the trend chart at any time, and a verification and backtracking window will pop up on the screen. The main control and processing unit retrieves the EDV / ESV tolerance value frame associated with the data point, as well as the consecutive frame sequences before and after it and the AI ​​segmentation contour overlay image from the local storage unit, and presents them on the touch screen for doctors to scroll through frame by frame. Main control and processing unit: By observing the continuous dynamic change of the internal cavity volume from large to small in key time phases, it independently judges whether the extreme value moment selected by the AI ​​is accurate by comparing the previous and next frames; this credibility verification mechanism ensures that each LVEF trend data is supported by original image evidence.

Images