Myocardial function evaluation method and device, storage medium and computer equipment

Abstract

The invention relates to the technical field of myocardial function evaluation, and discloses a myocardial function evaluation method and device, a storage medium and computer equipment, and the method comprises the steps: applying acupuncture stimulation to a target object, and collecting a multi-channel magnetocardiogram signal and various physiological data of the target object; fusing the various physiological data and the multichannel magnetocardiogram signal to generate a physiological signal set, so as to dynamically adjust the load parameter of acupuncture stimulation; performing time alignment and space fusion on the multi-channel magnetocardiogram signal by using a multi-channel deep fusion network to obtain a multi-dimensional magnetocardiogram signal, and extracting feature information by using a deep learning feature extraction network; after the feature information and the multiple physiological data are fused, a cardiac load evaluation model is used for evaluating the myocardial function state of the target object, and the cardiac load capacity and / or the cardiac blood pumping efficiency are / is obtained. According to the method, the acupuncture load and the magnetocardiogram technology are combined, the myocardial function is dynamically, continuously and accurately evaluated, and a reliable basis is provided for clinical decision making.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of myocardial function evaluation, and in particular to a myocardial function evaluation method and device, a storage medium and a computer device. BACKGROUND

[0002] In the field of myocardial function evaluation, the existing technology still has certain limitations. Specifically, clinicians mainly rely on conventional medical examination methods, such as electrocardiogram and echocardiogram, to evaluate the myocardial function of patients. Although these methods can reflect the basic state of the myocardium to some extent, the data obtained is mostly static information at a certain time point or specific state, making it difficult to achieve continuous and dynamic monitoring of myocardial function.

[0003] Meanwhile, in the actual diagnosis and treatment process, the myocardial function of patients is often in a state of continuous dynamic adjustment, especially when receiving intervention or load stimulation, the cardiac electrical activity and pumping capacity may change rapidly and subtly. However, the existing examination methods are often limited by sampling frequency, detection window and operation process, and cannot capture these dynamic changes in real time, making it difficult for doctors to fully and accurately grasp the real-time evolution trend of the myocardial function of patients, thereby affecting the accuracy of clinical evaluation and the timeliness of treatment decisions. SUMMARY

[0004] Therefore, the present application provides a myocardial function evaluation method and device, a storage medium and a computer device, which mainly aims to solve the technical problem that the existing myocardial function evaluation methods cannot achieve continuous and dynamic monitoring of myocardial function, affecting the accuracy of clinical evaluation and the timeliness of treatment decisions.

[0005] According to a first aspect of the present application, a myocardial function evaluation method is provided, comprising: In the process of applying acupuncture stimulation to the target object to form cardiac load, a plurality of magnetocardiography sensors are used to collect multi-channel magnetocardiography signals of the target object, and a physiological monitoring device is used to collect a plurality of physiological data of the target object; The plurality of physiological data are standardized and fused, and the fused physiological data and the multi-channel magnetocardiography signals are secondarily fused to generate a physiological signal set, and the load parameters of the acupuncture stimulation are dynamically adjusted based on the physiological signal set and the basic physiological data of the target object; The multi-channel magnetocardiography signals are time-aligned and spatially fused using a multi-channel deep fusion network to obtain multi-dimensional magnetocardiography signals, and the multi-dimensional magnetocardiography signals are denoised and enhanced, and then a deep learning feature extraction network is used to extract feature information related to myocardial function; The feature information and the plurality of physiological data are processed in real time to obtain fused features, and a preset cardiac load evaluation model is used to evaluate the myocardial function state of the target object based on the fused features to generate an evaluation result, wherein the evaluation result includes cardiac load capacity and / or cardiac pumping efficiency.

[0006] Optionally, the load parameter of the acupuncture stimulation is dynamically adjusted based on the physiological signal set and the basic physiological data of the target object, including: based on the basic physiological data, individual feature information and historical medical data of the target object, the load parameter of the acupuncture stimulation is calculated by using a preset optimization algorithm, wherein the load parameter includes the intensity, frequency and stimulation time of the acupuncture stimulation; in the process of applying acupuncture stimulation to the target object, the heart rate change value and the blood pressure change value are calculated in real time according to the physiological signal set; when the heart rate change value exceeds a preset heart rate change safety threshold, and / or when the blood pressure change value exceeds a preset blood pressure change safety threshold, a first difference value between the heart rate change value and the heart rate change safety threshold is calculated, and / or a second difference value between the blood pressure change value and the blood pressure change safety threshold is calculated; the acupuncture load adjustment amount is calculated according to the first difference value and / or the second difference value, and the load parameter is dynamically adjusted according to the acupuncture load adjustment amount; based on the load response data of the target object in the historical acupuncture stimulation process, the adjusted load parameter is optimized by using a preset machine learning algorithm.

[0007] Optionally, the multi-channel deep fusion network is used to time-align and spatially fuse the multi-channel magnetocardiogram signals to obtain multi-dimensional magnetocardiogram signals, and after denoising and enhancing the multi-dimensional magnetocardiogram signals, a deep learning feature extraction network is used to extract feature information related to myocardial function, including: the multi-channel magnetocardiogram signals are input into the multi-channel deep fusion network, the multi-channel magnetocardiogram signals are time-synchronized, and the time-synchronized multi-channel magnetocardiogram signals are weighted and fused based on a preset weight to output multi-dimensional magnetocardiogram signals with unified spatial dimensions; the multi-dimensional magnetocardiogram signals are input into a multi-level denoising neural network, the multi-dimensional magnetocardiogram signals are convolved and recursively time-series modeled to layer-by-layer filter out low-frequency environmental interference and high-frequency random noise, and the denoising and enhancing of the multi-dimensional magnetocardiogram signals are completed, wherein the multi-level denoising neural network includes a convolutional neural network and a recurrent neural network connected in series; the processed multi-dimensional magnetocardiogram signals are input into a deep learning feature extraction network to extract feature information related to myocardial function, wherein the deep learning feature extraction network combines a convolutional neural network and a long short-term memory network, the convolutional neural network is used to extract local frequency domain and morphological features of the processed multi-dimensional magnetocardiogram signals, and the long short-term memory network is used to capture long-range time-series dependence of the processed multi-dimensional magnetocardiogram signals.

[0008] Optionally, the real-time synchronization and fusion processing of the feature information and the multiple physiological data to obtain fusion features comprises: inputting the feature information and the multiple physiological data into a multi-modal fusion neural network, performing spatio-temporal alignment and feature-level fusion on the feature information and the multiple physiological data by using the multi-modal fusion neural network, and extracting and outputting fusion features, wherein the fusion features are used to reflect details of cardiac electrical activity and physiological state of the target object.

[0009] Optionally, the evaluation result is the cardiac load capacity; and the evaluation of the myocardial function state of the target object based on the fusion features and by using a preset cardiac load evaluation model to generate an evaluation result comprises: obtaining real-time heart rate values, real-time blood pressure values and real-time oxygen saturation values in the fusion features, and extracting a magnetocardiogram signal feature in the fusion features; calculating a product of the real-time heart rate values and a heart rate influence coefficient to obtain a first load contribution value; calculating a product of the real-time blood pressure values and a blood pressure influence coefficient to obtain a second load contribution value; calculating a product of the real-time oxygen saturation values and an oxygen saturation influence coefficient to obtain a third load contribution value; based on the magnetocardiogram signal feature, obtaining a QRS wave length value and an ST segment change value, and calculating a product of the sum of the QRS wave length value and the ST segment change value and an electrocardiogram feature influence coefficient to obtain a fourth load contribution value; and performing weighted summation on the first load contribution value, the second load contribution value, the third load contribution value and the fourth load contribution value based on a preset weight to obtain a cardiac load capacity index used to represent the cardiac load capacity.

[0010] Optionally, the evaluation result is the cardiac pumping efficiency; and the evaluation of the myocardial function state of the target object based on the fusion features and by using a preset cardiac load evaluation model to generate an evaluation result comprises: extracting a magnetocardiogram signal feature in the fusion features, and extracting an absolute value of a QRS wave, a time interval of the QRS wave and an electrocardium conduction velocity in the magnetocardiogram signal feature; calculating a sum of absolute values of all QRS waves to obtain a first sum value; calculating a sum of time intervals of all QRS waves to obtain a second sum value, and calculating a product of the second sum value and the electrocardium conduction velocity to obtain a product value; and calculating a ratio of the first sum value to the product value to obtain a cardiac pumping efficiency index used to represent the cardiac pumping efficiency.

[0011] Optionally, after the generating the evaluation result, the method further comprises: acquiring and integrating diversified individual characteristic information of the target object, and constructing a comprehensive decision feature vector based on the diversified individual characteristic information, wherein the diversified individual characteristic information comprises a static health record, genetic and family history, dynamic lifestyle data and the evaluation result; inputting the comprehensive decision feature vector into a preset integrated learning model, outputting a heart risk score, and comparing the heart risk score with a multi-level risk threshold, wherein the multi-level risk threshold comprises a high risk threshold and a medium risk threshold; when the heart risk score exceeds the high risk threshold, triggering an emergency warning and generating an adjustment scheme, wherein the adjustment scheme comprises a medical intervention and a drug adjustment suggestion; when the heart risk score is between the medium risk threshold and the high risk threshold, triggering a regular warning and generating a management scheme, wherein the management scheme comprises a lifestyle reinforcement intervention and a rehabilitation exercise suggestion; when the heart risk score is lower than the medium risk threshold, generating a preventive guidance scheme, wherein the preventive guidance scheme comprises a health maintenance and optimization suggestion.

[0012] According to a second aspect of the present application, a myocardial function evaluation device is provided, comprising: a multi-source acquisition module, configured to acquire multi-channel magnetocardiogram signals of a target object by using a plurality of magnetocardiogram sensors in a process of applying acupuncture stimulation to the target object to form a heart load, and simultaneously acquire a plurality of physiological data of the target object by using a physiological monitoring device; a load adjustment module, configured to perform standardization and fusion processing on the plurality of physiological data, and then perform secondary fusion on the fused physiological data and the multi-channel magnetocardiogram signals to generate a physiological signal set, and dynamically adjust a load parameter of the acupuncture stimulation based on the physiological signal set and basic physiological data of the target object; a signal processing module, configured to perform time alignment and spatial fusion on the multi-channel magnetocardiogram signals by using a multi-channel deep fusion network to obtain multi-dimensional magnetocardiogram signals, and then perform denoising and enhancement processing on the multi-dimensional magnetocardiogram signals, and extract feature information related to myocardial function by using a deep learning feature extraction network; a feature evaluation module, configured to perform real-time synchronization and fusion processing on the feature information and the plurality of physiological data to obtain fused features, and evaluate a myocardial function state of the target object based on the fused features and by using a preset heart load evaluation model to generate an evaluation result, wherein the evaluation result comprises a heart load capacity and / or a heart pumping efficiency.

[0013] According to a third aspect of the present application, a storage medium having a computer program stored thereon is provided, wherein the program is executed by a processor to implement the myocardial function evaluation method.

[0014] According to a fourth aspect of the present application, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned myocardial function evaluation method when executing the program.

[0015] The myocardial function evaluation method, device, storage medium and computer device provided by the present application apply acupuncture stimulation as a controllable and non-invasive load application method, continuously and synchronously collect cardiac magnetism signals and various physiological data in the whole process of load application, and compared with traditional electrocardiogram and cardiac ultrasound data acquisition methods, the complete dynamic response curve of the heart from rest to load and then to recovery can be continuously depicted through real-time data flow, so that the evolution trend of myocardial function can be grasped in real time, and the timeliness and continuity of evaluation are improved. The multi-channel cardiac magnetism sensing technology is introduced, the multi-channel deep fusion network and the adaptive denoising enhancement algorithm are combined, and the weak cardiac magnetism signals are super-synchronously collected and high-fidelity processed. The cardiac magnetism signals are sensitive to the changes of the electrical activity of the heart, can reflect the deep electrical physiological details that cannot be captured by traditional electrocardiogram, and then accurately identify and quantify the subtle and rapid changes of the electrical activity of the heart and the blood pumping function under acupuncture load. Moreover, the present application does not rely on a single signal, but through the dual fusion architecture of secondary fusion of physiological data and real-time synchronous fusion of feature information, the cardiac magnetism signals and macro physiological indexes are deeply integrated, the effective features are extracted by using the deep learning feature extraction network and the cardiac load evaluation model, and intelligent evaluation is performed, so that the myocardial function can be comprehensively evaluated from multiple aspects, and a comprehensive and accurate evaluation result is obtained, which provides a reliable basis for clinical decision-making. Finally, the present application constructs an intelligent closed-loop system, which can not only be passively monitored, but also dynamically adjust the acupuncture load parameters according to the real-time generated physiological signal set, so as to ensure that the load is always within the safe and effective range of the target object, and the evaluation result can feed back the signal processing and load adjustment strategy, so that the whole evaluation process is safe and reliable, and provides direct guidance for precise treatment. In summary, the above-mentioned method combines acupuncture load and cardiac magnetism technology, integrates signal processing, data fusion and artificial intelligence algorithm, and dynamically, continuously and accurately evaluates myocardial function, so as to provide a reliable basis for clinical decision-making.

[0016] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the specific embodiments of the present application can be implemented according to the content of the description, and in order to make the above and other purposes, characteristics and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0017] The drawings described herein are used to provide further understanding of the present application, and form a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation on the present application. In the drawings: Figure 1A flowchart illustrating a method for assessing myocardial function provided by an embodiment of the present invention is shown; Figure 2 This diagram illustrates the layout of a central magnetic sensor and physiological monitoring device for a myocardial function assessment method according to an embodiment of the present invention. Figure 3 A flowchart illustrating another method for assessing myocardial function provided by an embodiment of the present invention is shown; Figure 4 This invention illustrates a schematic diagram of the structure of a real-time closed-loop regulation system for physiological data in another method for assessing myocardial function provided by an embodiment of the present invention. Figure 5 This invention illustrates a flowchart of the dynamic adjustment of acupuncture stimulation load in another method for assessing myocardial function provided in an embodiment of the present invention. Figure 6 This diagram illustrates the organizational structure of the central magnetic signal processing and feature extraction process in another myocardial function assessment method provided by an embodiment of the present invention. Figure 7 A flowchart illustrating the calculation of the cardiac pumping efficiency index in another myocardial function assessment method provided by an embodiment of the present invention is shown. Figure 8 This illustration shows a flowchart of another myocardial function assessment method provided by an embodiment of the present invention, which integrates and analyzes diverse individual characteristic information and performs risk assessment. Figure 9 This invention illustrates a schematic diagram of the overall system architecture in another method for assessing myocardial function provided by an embodiment of the present invention. Figure 10 A schematic diagram of the structure of a myocardial function assessment device provided in an embodiment of the present invention is shown; Figure 11 A schematic diagram of another myocardial function assessment device provided in an embodiment of the present invention is shown; Figure 12 A schematic diagram of the device structure of a computer device provided in an embodiment of the present invention is shown. Detailed Implementation

[0018] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0019] This application provides a method for assessing myocardial function, such as... Figure 1 As shown, the method includes the following steps: 101. In the process of applying acupuncture stimulation to the target object to form cardiac load, a plurality of cardiac magnetic sensors are used to collect multi-channel cardiac magnetic signals of the target object, and a physiological monitoring device is used to collect a plurality of physiological data of the target object.

[0020] Wherein, the acupuncture stimulation refers to applying physical stimulation to specific acupoints through traditional Chinese medicine acupuncture means as a controllable and quantifiable external intervention to actively and non-invasively regulate the working load of the heart; the cardiac load refers to the total amount of work that the heart needs to bear in unit time to maintain blood circulation, in this application, it specifically refers to a controllable physiological stress state induced by acupuncture stimulation for testing the functional reserve and response capacity of the heart; the multi-channel cardiac magnetic signals refer to the time sequence signals reflecting the weak magnetic field changes generated by the electrical activity of the heart collected synchronously by a plurality of arrayed high-sensitivity cardiac magnetic sensors, each sensor constitutes an independent channel, and the multi-channel data can provide spatial distribution information of the electrical activity of the heart, and the signals have higher positioning accuracy and sensitivity to deep currents than traditional electrocardiogram (ECG); the physiological monitoring device is a medical instrument for real-time and continuous collection of basic vital signs of the human body, in this application, it specifically refers to a device for synchronous collection of one or more physiological parameters such as heart rate, blood pressure, blood oxygen saturation, respiratory rate, and surface electrocardiogram.

[0021] Specifically, at the beginning of the evaluation, first, the acupuncture device is used to apply pre-set or dynamically adjusted acupuncture stimulation to the specific acupoints of the target object, and the purpose of the stimulation is to serve as a standardized and controllable physiological load source to actively and safely stimulate the heart, simulate the state under different working intensities, and in the whole process of load application, the system synchronously starts two types of high-precision data collection, one of which uses a plurality of cardiac magnetic sensors worn on the chest of the target object to continuously collect multi-channel cardiac magnetic signals generated by the heart, and the other uses the connected physiological monitoring device to continuously record a plurality of basic physiological data of the target object. The collection of these two types of data is strictly synchronized in time and complementary in space, and together forms the basis of the original data for subsequent analysis.

[0022] Wherein, as shown in Figure 2 , a layout mode of cardiac magnetic sensors and physiological monitoring devices is provided, i.e. when the acupuncture stimulation is applied to the target object, the acupuncture needles are respectively placed at the Neiguan acupoint and the Xinshu acupoint, and at the same time, the respiratory signal is collected by a chest belt type respiratory sensor, the multi-channel cardiac magnetic signal collector is arranged on the chest corresponding to the heart region, and the electrocardiogram (ECG) electrode, the multi-parameter physiological monitor and the wrist type blood pressure / oxygen saturation meter synchronously collect physiological data.

[0023] In the embodiments of the present application, the acupuncture stimulation is defined as a load means of a heart function test, replacing the traditional risky exercise load or drug load, providing a safe, non-invasive and precisely adjustable evaluation approach for patients who cannot tolerate traditional load; by synchronously collecting multi-channel cardiac magnetic signals and various physiological data, a time-synchronized and spatially multi-dimensional data set is obtained at the beginning of the evaluation, the cardiac magnetic signals provide high-resolution electrical activity details, and the physiological data provide macroscopic circulatory system status, providing a data basis for subsequent comprehensive evaluation of myocardial function status; in addition, since the data collection is synchronized during the continuous process of load application, the obtained data is essentially a continuous time sequence rather than a static snapshot, enabling the technical solution of the present application to fundamentally support real-time and dynamic monitoring and analysis of heart function, thereby capturing the instantaneous changes and evolution trends of the heart state under load.

[0024] 102. Standardize and fuse the various physiological data, and then fuse the fused physiological data with the multi-channel cardiac magnetic signals to generate a physiological signal set, and dynamically adjust the load parameters of the acupuncture stimulation based on the physiological signal set and the basic physiological data of the target object.

[0025] Among them, the standardization processing refers to transforming the original physiological data collected from different physiological monitoring devices, having different dimensions and numerical ranges, into standard data with uniform scale through mathematical transformation, the purpose is to eliminate the dimensional difference, make different physiological parameters comparable, and facilitate subsequent fusion and calculation; data fusion refers to integrating, correlating and combining information from multiple independent data sources to generate a more complete and accurate consistent description than a single data source; the physiological signal set refers to the data set formed after standardization and fusion processing, having a uniform dimension, specifically integrating the original various physiological data information, and providing an input interface for further fusion with the cardiac magnetic signals; the basic physiological data refers to the stable physiological index measurement value of the target object in the resting state without load, such as resting heart rate, basic blood pressure, etc., these data serve as individualized reference benchmarks for evaluating the degree of change of physiological indicators under load; the load parameter refers to the physical variable controlling the acupuncture stimulation, mainly including stimulation intensity, stimulation frequency and stimulation time, and dynamically adjusting the load parameter is real-time control of acupuncture load.

[0026] Specifically, the system time-aligns multiple physiological data collected in real time, ensures that all signals are analyzed under the same time reference, then normalizes each physiological data to eliminate differences in dimensions and numerical ranges, and then uses a state estimation algorithm (such as a Kalman filter) to first fuse the normalized multiple physiological data, to obtain an intermediate feature that can comprehensively reflect the current overall physiological state, and then secondarily fuse the intermediate feature with the multi-channel cardiac magnetic signal to generate a unified and multi-dimensional physiological signal set; the system takes the physiological signal set as the core input, simultaneously retrieves the basic physiological data of the target object as an individualized benchmark, calculates the change amount of key physiological indicators such as heart rate and blood pressure relative to the basic value in real time through a preset intelligent algorithm, and compares it with the safety threshold; once the change amount exceeds the safety range, the system calculates the adjustment amount of the load parameter according to the algorithm, and immediately controls the acupuncture device to dynamically adjust the stimulation intensity or stimulation time, thereby forming a real-time closed loop from monitoring, analysis to adjustment.

[0027] In the embodiments of the present application, through the processing of standardization and twice fusion, the physiological data with different characteristics and the cardiac magnetic signal are unified into an organic whole, not only solving the problem of difficult direct comparative analysis of multi-source information, but also generating a signal set containing rich physiological state through fusion, providing a high-quality data basis for accurate evaluation; by comparing the real-time physiological signal set with the individual basic data, the system can perceive the response of each target object to the load, and the introduction of the dynamic adjustment mechanism makes the acupuncture load no longer fixed, but can be adaptively adjusted according to the real-time tolerance of the individual, ensuring the safety of the evaluation process and effectively avoiding the risk of excessive stress caused by improper load, especially suitable for patient groups with different physiological states; coupling data acquisition and processing with intervention means adjustment to form a closed-loop control system, not only can observe the cardiac response, but also can intervene in the load intensity, so that the entire evaluation process is carried out under controlled conditions, improving the safety of the evaluation, and making the final evaluation result better reflect the real functional potential of the heart under optimal load challenge.

[0028] 103. The multi-channel cardiac magnetic signal is time-aligned and spatially fused using a multi-channel deep fusion network to obtain a multi-dimensional cardiac magnetic signal, and after denoising and enhancing the multi-dimensional cardiac magnetic signal, a deep learning feature extraction network is used to extract feature information related to myocardial function.

[0029] The multi-channel deep fusion network is a deep learning neural network architecture, the input layer of which can receive signal data from multiple sensors in parallel, and through multi-layer nonlinear transformation, modeling and fusion of complex correlation between signals are realized in the deep layer of the network. In the present application, it is specifically used to solve the problems of multi-sensor data time alignment and spatial feature fusion. Time alignment refers to the process of eliminating the time misalignment between channels caused by differences in sensor hardware, transmission delay, etc., to ensure that all channel data points correspond strictly at the same physical time. Spatial fusion refers to further integrating information of signals collected at different spatial positions after time alignment to synthesize a multi-dimensional signal that can fully reflect the spatial distribution and global characteristics of the signal source. The multi-dimensional magnetocardiogram signal refers to the magnetocardiogram data obtained after time alignment and spatial fusion processing. It not only retains the original time sequence information, but also enhances the spatial dimension characteristics of the signal through fusion, forming a tensor containing more rich information. The deep learning feature extraction network is a model constructed based on a deep neural network, and its design goal is to automatically learn and extract effective and high-level abstract representations from input data for subsequent tasks.

[0030] Specifically, first, the original multi-channel magnetocardiogram signals collected from multiple magnetocardiogram sensors are input into a pre-trained multi-channel deep fusion network. Part of the function of the multi-channel deep fusion network is to achieve high-precision time alignment, automatically correct the subtle time delay between channels, and achieve sub-sampling point level time synchronization. The subsequent layers of the network perform spatial fusion, which generates a more consistent and rich multi-dimensional magnetocardiogram signal in time and space dimensions by weighting and combining the features of the aligned channel signals. Then, the multi-dimensional electromagnetic signal is sent to a multi-level denoising and enhancement processing module, which uses an architecture combining convolutional neural networks and recurrent neural networks to model the signal at multiple scales, adaptively filters out environmental electromagnetic interference, power frequency noise, motion artifacts, etc., and simultaneously performs adaptive gain adjustment to enhance the components of the effective signal. Finally, a high-quality, high signal-to-noise ratio signal is output. Finally, the processed multi-dimensional signal is fed into a deep learning feature extraction network, which automatically learns and extracts high-dimensional abstract feature information representing myocardial electrical activity patterns, conduction rules and abnormal states from complex signals, providing data representation for subsequent accurate evaluation.

[0031] In the embodiments of the present application, the quality and availability of the magnetocardiogram signal are significantly improved through intelligent processing. Specifically, a multi-level denoising enhancement technology is adopted, which can adaptively process the complex noise environment of the magnetocardiogram signal, effectively denoising while preserving or even enhancing weak pathological characteristics. A deep learning feature extraction network is used to automatically mine complex patterns and deep rules closely related to the myocardial function state, extract high-dimensional abstract features with high discriminability, and greatly improve the efficiency and objectivity of feature extraction, as well as the evaluation capability.

[0032] 104. Real-time synchronization and fusion processing of the feature information and the multiple physiological data are performed to obtain fusion features, and a myocardial function state of the target object is evaluated based on the fusion features and by using a preset cardiac load evaluation model to generate an evaluation result, wherein the evaluation result includes a cardiac load capacity and / or a cardiac pumping efficiency.

[0033] The fusion features refer to unified data representations generated by integrating and associating the feature information extracted from the magnetocardiogram signal and the multiple physiological data collected in real time, and simultaneously containing abstract patterns of cardiac electrical activity and quantitative indicators of macroscopic physiological state. The real-time synchronization processing refers to aligning or interpolating two types of data representing different physiological meanings and possibly having different collection periods to the same time reference point for processing to ensure strict correspondence in time. The cardiac load evaluation model is a mathematical model constructed based on a machine learning or deep learning algorithm, with the fusion features as input and the quantitative evaluation of the cardiac function state as output. The cardiac load capacity refers to the ability of the heart to maintain pumping function when responding to external load, usually represented by a quantitative index, which comprehensively reflects the comprehensive changes of heart rate, blood pressure, and cardiac electrical activity under load. The cardiac pumping efficiency refers to an index measuring the efficiency of pumping blood out of the heart with each contraction.

[0034] Specifically, first, the feature information is synchronously and fused with the continuously collected multiple physiological data in real time, which is specifically realized by a multi-modal fusion network, in which the two are aligned, associated and integrated at the feature level, and finally a unified fusion feature is output, which can comprehensively depict the electrical and physiological combined state of the heart; then, the fusion feature is input into a preset heart load evaluation model, which internally embeds the knowledge of the complex mapping relationship between myocardial function and various features, after receiving the fusion feature, the model performs two core computing tasks, one is to calculate a quantitative heart load capacity index based on the heart rate, blood pressure, blood oxygen and electrocardiogram waveform parameters contained in the fusion feature through weighted fusion or more complex algorithms, the other is to calculate the heart pumping efficiency based on the specific parameters derived from the magnetocardiogram signal in the fusion feature through a specific physical and mathematical model, and finally the system outputs at least one core index as the evaluation result.

[0035] In the embodiments of the present application, by synchronously and fusing the deep features reflecting the details of electrical activity with the physiological data reflecting the overall state in real time, the evaluation is no longer limited to a single signal level, but is based on the overall performance of the heart as a complete functional organ, the physiological significance of the evaluation conclusion is more complete, and the clinical reference value is higher; the standardized heart load evaluation model based on big data can automatically, quickly and quantitatively output the core function index from the complex fusion feature, eliminating the difference of human interpretation, so that the evaluation result has high objectivity, repeatability and comparability, providing a stable and reliable quantitative basis for clinical diagnosis; the finally generated evaluation result can include the heart load capacity and the heart pumping efficiency, which are two core clinical parameters in cardiovascular function evaluation, directly quantifying the reserve function and contraction efficiency of the heart, providing intuitive and accurate data-based decision support for doctors to judge the heart health status of patients, evaluate the treatment effect and develop rehabilitation programs, improving the practicality and clinical transformation ability of the present application, which can directly serve precision medicine.

[0036] The application provides a myocardial function evaluation method, which applies acupuncture stimulation as a controllable and non-invasive load application method, continuously and synchronously collects magnetocardiogram signals and various physiological data in the whole process of load application, and compared with traditional electrocardiogram and cardiac ultrasound data acquisition methods, can continuously depict a complete dynamic response curve of the heart from rest to load and then to recovery through real-time data flow, so that the evolution trend of myocardial function can be grasped in real time, and the timeliness and continuity of evaluation are improved; the multi-channel magnetocardiogram sensing technology is introduced, a multi-channel deep fusion network and an adaptive denoising enhancement algorithm are combined, and weak magnetocardiogram signals are super-synchronously collected and high-fidelity processed, wherein the magnetocardiogram signals are extremely sensitive to changes in cardiac electrical activity and can reflect deep electrophysiological details that cannot be captured by traditional electrocardiogram, and then the subtle and rapid changes of cardiac electrical activity and blood pumping function under acupuncture load can be accurately identified and quantified; and the application does not rely on a single signal, but through a double fusion architecture of secondary fusion of physiological data and real-time synchronous fusion of feature information, the magnetocardiogram signals and macro physiological indexes are deeply integrated, effective features are extracted by using a deep learning feature extraction network and a cardiac load evaluation model, and intelligent evaluation is performed, so that the myocardial function can be comprehensively evaluated from multiple aspects, and a comprehensive and accurate evaluation result is obtained, thereby providing a reliable basis for clinical decision-making; finally, the application constructs an intelligent closed-loop system, which can not only be passively monitored, but also dynamically adjust the acupuncture load parameters according to the real-time generated physiological signal set, so as to ensure that the load is always within the safe and effective range of the target object, and the evaluation result can feed back the signal processing and load adjustment strategy, so that the whole evaluation process is safe and reliable, and provides direct guidance for precise treatment. In summary, the above method combines acupuncture load and magnetocardiogram technology, integrates signal processing, data fusion and artificial intelligence algorithm, and dynamically, continuously and accurately evaluates myocardial function, thereby providing a reliable basis for clinical decision-making.

[0037] The application embodiment provides another myocardial function evaluation method, as shown in the following formula (I): Figure 3 The application embodiment provides another myocardial function evaluation method, as shown in the following formula (I): 201, collection and real-time monitoring of multi-channel magnetocardiogram signals and physiological data.

[0038] Specifically, the collected multi-channel magnetocardiogram signals are pre-processed to a certain extent, and in view of the deficiency that traditional methods based on wavelet or Kalman filtering are limited by signal frequency characteristics and noise types, the application proposes and implements a multi-dimensional magnetocardiogram signal collection and adaptive denoising strategy, which specifically combines a time-frequency adaptive filter (STAF) and an adaptive denoising waveform reconstruction algorithm, wherein the STAF algorithm introduces adaptive time-frequency analysis, decomposes the magnetocardiogram signals into a time-frequency domain through fusion of a short-time Fourier transform (STFT) and a wavelet transform, so as to adaptively select and apply an optimal filter for different frequency characteristic noise components (such as low-frequency interference and high-frequency noise) in the signals, and the transformation process can be represented as:

[0039] wherein, is a window function, is a time-frequency domain representation, is a frequency, is a time.

[0040] Subsequently, the adaptive noise reduction waveform reconstruction algorithm is refined by multi-scale wavelet packet analysis (WPD, Wavelet Packet Decomposition) to reconstruct the signal, the process can be represented as:

[0041] wherein, is a reconstruction coefficient, is a wavelet basis function, is a denoised signal.

[0042] Based on this method, the wavelet basis function can be dynamically adjusted to adapt to the characteristics of different patients' hearts, accurately stripping the frequency overlapping noise, thereby significantly improving the signal-to-noise ratio of the magnetocardiogram at the root; secondly, in view of the problem that the intensity of the magnetocardiogram fluctuates due to individual differences, such as different body types and heart positions, this step also integrates an adaptive signal amplification and gain adjustment method, discarding the traditional fixed gain amplifier, and adopting a gain adjustment algorithm based on real-time feedback, by continuously monitoring the amplitude of the input signal and the noise level, according to the formula:

[0043] wherein, is a real-time gain, is a basic gain, is a gain adjustment coefficient, is an input signal amplitude.

[0044] Based on this dynamic adjustment of real-time gain, it ensures that the signal is always processed within the optimal amplification range, effectively avoiding distortion caused by signal oversaturation or excessive attenuation, and implementing high-precision multi-sensor data synchronization and time alignment to ensure the consistency and comparability of multi-sensor data, specifically using a correction method based on the phase difference of the synchronization signal, by broadcasting a unified time synchronization marker signal (P-wave marker) in the system, each sensor can calculate the phase difference between the received signal and the local clock according to the built-in clock: wherein, is a time difference, and are signals from two sensors, and PhaseSync is a synchronization correction algorithm.

[0045] According to the real-time calibration data timestamp, strict space-time alignment of multi-channel data is realized at the sampling level, overcoming the asynchronous error caused by the traditional dependence on a single timestamp. In addition, to realize reliable data aggregation and flexible system deployment, the module adopts a low-delay and high-reliability wireless data transmission and remote monitoring design. Specifically, a distributed self-organizing network is constructed through an optimized Wi-Fi Mesh network protocol, and a 5G communication protocol is combined to ensure high-bandwidth and low-delay data transmission, especially in the multi-sensor concurrent scenario, effectively avoiding transmission bottlenecks. At the same time, low-power communication modules such as LoRa and NB-IoT are integrated to balance transmission rate and energy consumption. The power consumption model can be described as:

[0046] wherein, is the total power consumption, is the data transmission power consumption, is the idle power consumption, thereby supporting long-term, stable continuous work and remote monitoring of the system in a clinical environment.

[0047] In this embodiment, by combining adaptive time-frequency analysis with wavelet packet reconstruction algorithm, complex environmental noise can be filtered out and the true details of cardiac electrical activity can be preserved, thereby significantly improving the signal-to-noise ratio and quality of the original magnetocardiogram signal, providing a more pure and reliable data input basis for subsequent deep analysis and functional evaluation. Secondly, the dynamic gain adjustment mechanism realizes real-time intelligent control of signal intensity, ensuring that the collected signal is in the best processing interval regardless of individual differences, fundamentally avoiding processing distortion problems caused by improper signal amplitude. Moreover, the high-precision multi-channel synchronization technology realizes sub-sampling level time alignment accuracy through active phase difference calibration, ensuring that the data from sensors in different spatial positions are strictly consistent on the time axis, providing an indispensable accurate data basis for subsequent accurate spatial fusion and functional analysis. Finally, the optimized wireless transmission network and low-power design together realize the rapid and stable transmission and remote monitoring of magnetocardiogram signals and physiological data, meeting the real-time requirements and ensuring the long-term operation stability and deployment flexibility of the system in a clinical environment. In summary, the present application not only improves the accuracy, reliability and adaptability of the data acquisition link, but also provides data support for subsequent myocardial function intelligent evaluation and precise adjustment of acupuncture load, ensuring efficient, accurate and stable operation of the entire system.

[0048] 202, physiological data monitoring and load input.

[0049] In the embodiment, firstly, real-time physiological data monitoring is performed, and key physiological data of a target object is continuously collected by connecting a standard physiological monitoring device, the physiological data including but not limited to heart rate, electrocardiogram (ECG), blood pressure, respiratory rate and the like, and the above data constitutes an important index for evaluating the instant state and reaction of the heart; meanwhile, the system comprehensively utilizes high-precision magnetocardiogram signals and conventional ECG signals to jointly monitor and evaluate the electrical physiological activity of the heart and its dynamic change under load from different dimensions, so as to obtain specific reaction characteristics of the heart under different stimulation conditions such as acupuncture.

[0050] Secondly, the system performs standardization processing and data fusion on the collected raw data, in order to ensure that physiological signals from different devices, having different physical units and magnitudes, can be effectively analyzed and compared in a unified framework, the system strictly time-aligns all signals, and then respectively performs standardization processing on each kind of physiological data, the standardization process is according to the formula:

[0051] wherein, represents the collected raw data, is the mean value of the data set, is the standard deviation.

[0052] Finally, on the basis of completing the standardization preprocessing, a key multi-source data fusion operation is continued, aiming to organically integrate multi-channel, heterogeneous data respectively from the magnetocardiogram sensor and various physiological monitoring devices, and finally form a unified physiological signal set which is consistent in time and space and complementary in information, in order to realize high-quality fusion, an advanced state estimation algorithm is adopted, for example, a fusion model based on Kalman filter or Bayesian network, taking the Kalman filter as an example, the core recursive formula of the data fusion process is:

[0053] wherein, is the current estimated value, is the Kalman gain, is the observed measurement value, is the observation matrix, through such an algorithm, the system can effectively filter out noise, compensate for the shortcomings of a single sensor, and comprehensively utilize the uncertainty of multi-source information, so as to significantly improve the overall quality, consistency and reliability of the finally generated physiological signal set, thereby providing a solid data basis for subsequent load precise adjustment and myocardial function evaluation.

[0054] Based on this, the application discloses a closed-loop control process for real-time collection, standardization, fusion, analysis and feedback adjustment of physiological data, as shown in Figure 4As shown, it specifically includes the following levels: data acquisition layer, which time-aligns and standardizes the raw data from the multi-channel magnetocardiograph sensor and physiological monitoring device, ensuring that the data is analyzed under a unified time reference and dimension; fusion processing layer, which uses state estimation algorithms (such as Kalman filter) to first fuse the standardized physiological data, generating intermediate features, and then secondarily fuses the features with multi-channel magnetocardiograph signals to form a unified multi-dimensional physiological signal set; core analysis layer, which uses intelligent algorithms to calculate key physiological indicators (such as heart rate, blood pressure) in real time based on the multi-dimensional physiological signal set, and compares them with preset personalized safety thresholds; execution layer, which determines the current state as normal or abnormal based on the comparison results, and generates adjustment instructions to dynamically adjust the intensity or time of acupuncture stimulation if an abnormality is detected, thereby achieving real-time closed-loop regulation of cardiac load and ensuring the safety and effectiveness of the evaluation process.

[0055] 203. Generating a personalized load scheme.

[0056] Specifically, based on the target object's basic physiological data, individual characteristic information, and historical medical data, the preset optimization algorithm is used to calculate the load parameters of acupuncture stimulation, including the intensity, frequency, and stimulation time of acupuncture stimulation.

[0057] In this embodiment, a load generation algorithm is disclosed, which specifically uses the target object's basic physiological data (such as resting heart rate, basic blood pressure), detailed individual characteristic information (including age, gender, weight, specific cardiovascular disease history, etc.), and related historical medical data as key inputs. The algorithm uses a multi-objective optimization strategy for processing, and the core optimization objective function can be formally represented as:

[0058] wherein, P represents the personal characteristics of the target object, S represents the intensity and frequency of the load, T represents the needle application duration, and the goal is to maximize the load effect while maintaining safety. Performance(P,S,T) is a comprehensive evaluation function that quantifies the balance between the cardiac load effect and physiological safety achieved under given personal characteristics P using stimulation parameters (S, T).

[0059] The optimization goal of the algorithm is to find the optimal parameter combination that maximizes the stress effect on the heart (i.e., load effect) while ensuring that all load parameters are strictly within the preset safety boundaries, thereby achieving the best balance between safety and effectiveness.

[0060] 204. Real-time dynamic adjustment and intelligent optimization.

[0061] Specifically, in the process of applying acupuncture stimulation to the target object, the heart rate variation value and the blood pressure variation value are calculated in real time according to the physiological signal set; when the heart rate variation value exceeds the preset heart rate variation safety threshold, and / or when the blood pressure variation value exceeds the preset blood pressure variation safety threshold, a first difference value between the heart rate variation value and the heart rate variation safety threshold is calculated, and / or a second difference value between the blood pressure variation value and the blood pressure variation safety threshold is calculated; the acupuncture load adjustment amount is calculated according to the first difference value and / or the second difference value, and the load parameter is dynamically adjusted according to the acupuncture load adjustment amount; and the adjusted load parameter is optimized by using a preset machine learning algorithm based on the load reaction data of the target object in the historical acupuncture stimulation process.

[0062] In the embodiment, first, the key physiological indicators, in particular, the heart rate variation value (ΔHR) and the blood pressure variation value (ΔBP), are continuously extracted and calculated based on the real-time generated physiological signal set, so as to quantify the immediate reaction of the heart to the current load, and then the dynamic feedback and adjustment mechanism is started, that is, the real-time calculated heart rate variation value (ΔHR) and blood pressure variation value (ΔBP) are continuously compared with the preset personalized safety threshold, that is, the heart rate variation safety threshold (ΔHR threshold ) and the blood pressure variation safety threshold (ΔBP threshold ), and the specific calculation method is as follows:

[0063] In the formula, ΔHR represents the load adjustment amount, and ΔBP are control parameters, and ΔBP are the heart rate and blood pressure variation values, respectively, and ΔBP are the set heart rate and blood pressure safety thresholds.

[0064] When the heart rate variation value exceeds the preset heart rate variation safety threshold, and / or when the blood pressure variation value exceeds the preset blood pressure variation safety threshold, it indicates that the current load may exceed the instantaneous bearing range of the patient, and the load parameter is immediately dynamically adjusted, and a specific load adjustment strategy is executed, for example, when the heart load is too high, for example, the heart rate is too fast and the blood pressure is too high, the system will accordingly reduce the intensity of acupuncture stimulation or reduce the stimulation time; on the contrary, when the load reaction is insufficient, the intensity will be appropriately increased or the stimulation time will be prolonged.

[0065] In addition, the application also provides a fine adjustment based on a baseline deviation feedback formula:

[0066] In the formula, ΔHR and ΔBP are adjustment factors, and current heart rate and blood pressure, and base heart rate and blood pressure, thus ensuring that the load level can be dynamically adapted according to real-time physiological feedback, always maintaining within the individualized safe bearing range of the patient.

[0067] Further, based on the load response data of the target object in the historical acupuncture stimulation process, that is, the load parameters applied each time and the physiological change records generated, the adjusted load parameters are optimized using a preset machine learning algorithm, that is, the accumulated "stimulation-response" historical data is continuously analyzed, the key parameters in the feedback adjustment mechanism can be continuously optimized, for example, the safety threshold, control parameters or adjustment factors can be optimized, and even the initial load generation algorithm can be optimized. Through continuous machine learning and data analysis, the system can make the load adjustment strategy more and more accurate with the increase of the number of uses, thereby continuously improving the accuracy and efficacy of load application, and generating a highly personalized optimized load scheme that is more suitable for individual physiological state changes.

[0068] On the basis of the dynamic adjustment mechanism based on heart rate and blood pressure core feedback, more dimensional physiological parameters can also be integrated for auxiliary decision-making and optimization to achieve safer and more accurate personalized load adjustment. In addition to heart rate and blood pressure, the following parameters can be included in the personalized load scheme generation and dynamic feedback mechanism to provide a more comprehensive physiological state evaluation, specifically including: electrocardiogram signal, which provides direct information of cardiac electrical activity, can be used to analyze the morphology, time limit and rhythm of key waveforms such as QRS complex, P wave and T wave, and during load adjustment, the system can analyze the electrocardiogram waveform changes (such as ST segment depression or elevation, QT interval prolongation) in real time as sensitive indicators for judging myocardial ischemia, electrolyte imbalance or electrical conduction abnormalities. If abnormal waveform changes occur, early warning and active load reduction can be performed, even when heart rate and blood pressure have not exceeded the standard, intervention can be taken, thereby achieving advanced protection of electrical physiology safety; respiratory rate, which reflects the autonomic nervous system tension and ventilation-blood flow matching condition, abnormal increase of respiratory rate during load application may indicate that the body is in a state of excessive stress or there is potential cardiopulmonary functional compensation insufficiency. Respiratory rate is used as an auxiliary safety threshold, and when its change exceeds the reasonable range, the load adjustment amount based on heart rate is weighted and corrected to ensure that the load does not cause excessive respiratory work and systemic stress; blood oxygen saturation is a key indicator for evaluating cardiopulmonary function and tissue oxygenation efficiency. Monitoring blood oxygen saturation during the load process can timely detect hidden hypoxia caused by insufficient cardiac output or pulmonary ventilation dysfunction. If blood oxygen saturation decreases continuously or significantly, the load can be reduced to ensure basic oxygen delivery safety; body temperature indirectly reflects metabolic rate and heat production. Under long-time or high-intensity load, abnormal increase of core body temperature may indicate impaired heat dissipation mechanism or excessive metabolic state, increasing cardiovascular pressure. By monitoring skin temperature or core body temperature trend, if abnormal warming occurs, the load intensity can be appropriately reduced or the rest time can be prolonged to avoid the risk of heat stress superimposed on cardiac load; blood glucose and electrolyte levels. For diabetic patients or heart failure patients, blood glucose level significantly affects myocardial energy metabolism, and electrolyte balance such as potassium and calcium is directly related to myocardial excitability and contractility. Continuous blood glucose monitoring data or recent electrolyte test results can be accessed. When the data indicates the presence of hypoglycemia, hyperglycemia or electrolyte critical disorder, more conservative parameters can be used in the load scheme generation stage, or stricter safety margins can be introduced in the adjustment; body position information. Changes in body position will directly affect cardiac volume, blood pressure and cardiac preload through gravity effect. The built-in sensor identifies body position changes and dynamically adjusts the physiological baseline values and reaction thresholds for load regulation, for example, when changing from a lying position to a standing position, a set of higher heart rate safety thresholds and more sensitive blood pressure adjustment coefficients can be automatically enabled to adapt to the physiological changes caused by body position changes, ensuring the adaptability of load adjustment.

[0069] Based on this, the application discloses a process for acupuncture stimulation load dynamic adjustment, such asFigure 5 As shown, the initial acupuncture stimulation is started to be applied, the multi-channel cardiac magnetic signals and physiological data (such as ECG, blood pressure, blood oxygen saturation, etc.) are collected in real time, the physiological data is fused (such as weighted fusion, feature level fusion), and then the fused physiological signal set and the cardiac magnetic signals are further fused to generate a load adjustment instruction, and then the acupuncture stimulation parameters (intensity, frequency, duration) are dynamically adjusted according to the instruction, and finally it is judged whether the target load is reached or the termination condition is met; if yes, it is ended; if no, it is returned to continue the cycle, and the termination conditions include: heart rate / blood pressure exceeding the safety threshold, evaluation completion, user manual stop, etc.

[0070] The present application can more accurately distinguish physiological reactions and pathological stress by integrating multiple physiological parameters such as electrocardiogram, respiratory rate, blood oxygen saturation, body temperature, biochemical indicators and body position information, so as to more fully tap the treatment potential of individuals and maximize the efficacy under the premise of absolute safety.

[0071] 205, cardiac magnetic signal processing.

[0072] Specifically, the multi-channel cardiac magnetic signals are input into a multi-channel deep fusion network, the multi-channel cardiac magnetic signals are time-synchronized, and the time-synchronized multi-channel cardiac magnetic signals are weighted fused based on a preset weight, and a spatial dimension unified multi-dimensional cardiac magnetic signal is output; the multi-dimensional cardiac magnetic signal is input into a multi-level denoising neural network, and through convolution and recursive time series modeling of the multi-dimensional cardiac magnetic signal, layer-by-layer filtering of low-frequency environmental interference and high-frequency random noise is realized, and denoising and enhancement processing of the multi-dimensional cardiac magnetic signal is completed, wherein the multi-level denoising neural network includes a convolutional neural network and a recurrent neural network connected in series.

[0073] In the present embodiment, the collected multi-channel cardiac magnetic signals are input into a multi-channel deep fusion network, and a synchronization technology based on deep learning is specifically adopted to time-synchronize the multi-channel cardiac magnetic signals, and this process can be formalized as:

[0074] In the formula, is the denoised signal, is the different scales of the input signal, denotes a convolutional neural network function.

[0075] After completing high-precision time synchronization, the network further weights and fuses the time-synchronized multi-channel magnetocardiogram signals based on preset weights, and finally outputs a multi-dimensional magnetocardiogram signal with unified spatial dimensions. In this step, not only is the data aligned, but the spatial consistency and information density of the signal are also improved; then the system inputs the obtained multi-dimensional magnetocardiogram signal into a multi-level denoising neural network for deep purification. The multi-level denoising neural network includes a convolutional neural network and a recurrent neural network connected in series. Specifically, the network architecture is composed of multiple layers of convolutional neural networks and recurrent neural networks stacked together. When the network is working, the multi-dimensional magnetocardiogram signal is convolved and recursively modeled in time series. The CNN layer is responsible for extracting features and filtering out specific frequency noise within the local receptive field, while the RNN layer models the time series of the signal and captures its long-range dependencies. This series structure allows the network to process the signal at multiple scales, filtering out low-frequency noise to high-frequency noise layer by layer, thereby maximizing the denoising effect and completing the denoising and enhancement of the multi-dimensional magnetocardiogram signal. The process can be described as:

[0076] wherein, represents the mutual information between channels (i) and (j), and the goal is to achieve multi-channel signal synchronization denoising by maximizing mutual information.

[0077] In summary, by integrating the synchronous weighted fusion of the multi-channel deep fusion network and the convolutional recursive modeling of the multi-level denoising neural network in series, high-precision time alignment, spatial fusion, and adaptive multi-scale denoising and enhancement of the original magnetocardiogram signal are achieved, providing high signal-to-noise ratio input signals for subsequent feature extraction and functional evaluation.

[0078] 206. Magnetocardiogram signal feature extraction.

[0079] Specifically, the processed multi-dimensional magnetocardiogram signal is input into a deep learning feature extraction network to extract feature information related to myocardial function. The deep learning feature extraction network combines a convolutional neural network and a long short-term memory network. The convolutional neural network is used to extract the local frequency domain and morphological features of the processed multi-dimensional magnetocardiogram signal, and the long short-term memory network is used to capture the long-range temporal dependencies of the processed multi-dimensional magnetocardiogram signal.

[0080] In the embodiment, the processed high-quality multi-dimensional magnetocardiogram signal is input into a deep learning feature extraction network to extract feature information related to myocardial function. The deep learning feature extraction network aims to overcome the limitations of traditional artificial design features and realize automatic and intelligent learning of features. Specifically, a convolutional neural network (CNN) and a long short-term memory network (LSTM) are combined in the structure. The convolutional neural network is used to extract the local frequency domain and morphological features of the processed multi-dimensional magnetocardiogram signal. Through convolution operation in the time-space dimension of the signal, the local time-frequency pattern and waveform morphological change are effectively captured. At the same time, the long short-term memory network is used to capture the long-range temporal dependence of the processed multi-dimensional magnetocardiogram signal, model the time series of the signal, and understand its dynamic evolution law and correlation. This combination of CNN and LSTM realizes the comprehensive mining of time and frequency information of the magnetocardiogram signal. The process can be formally represented as:

[0081] In the formula, is the time series magnetocardiogram signal, CNN extracts time-frequency features, LSTM captures time series features, and the final output is the automatically extracted high-level feature.

[0082] After successfully extracting the high-level feature Further adaptive cardiac electrical activity state classification is performed. Specifically, a deep support vector machine or another deep neural network is used as a classifier to analyze the extracted features in real time, automatically identify and classify the electrical activity state of the heart, such as evaluating the real-time stress state and load level of the heart. The classification decision process can be represented as:

[0083] In the formula, is the evaluated cardiac state, is the signal feature extracted by the deep neural network, is the different cardiac state category. Through the deep learning-based classification method, the load and health status of the heart can be accurately and objectively reflected without relying on fixed thresholds set by humans, significantly improving the accuracy and reliability of state evaluation.

[0084] In addition, the application also integrates a high-order adaptive signal optimization and dynamic feedback mechanism. According to the real-time evaluation of the cardiac load response, a reinforcement learning algorithm is used to dynamically adjust the signal processing strategy. The optimization goal is guided by the loss function of reinforcement learning, which is defined as:

[0085] In the formula, represents the mathematical expectation, which reflects the average performance of the strategy in an uncertain environment. to optimize the loss, for the reward of each moment (determined by the heart response), is a discount factor, for example, when the system detects that the heart load is too high, the algorithm can automatically adjust the parameters, so that the feature extraction network pays more attention to the stability features of the signal; on the contrary, when the load is low, the network is guided to enhance the extraction of high-frequency response components. This feedback-based closed-loop optimization realizes the online self-adjustment of the signal processing strategy, thereby continuously improving the quality of signal analysis and the real-time performance of the overall system.

[0086] It should be noted that the advanced feature extraction and classification function is built on the basis of the aforementioned multi-channel signal fusion and advanced synchronous analysis, wherein the fusion algorithm of the multi-channel deep fusion neural network can be represented as:

[0087] In the formula, is the final fusion signal, is the CNN output of the i-th channel signal, is the weight of each channel signal, by performing independent CNN preliminary feature extraction on each channel signal and then performing weighted fusion, not only high-precision multi-sensor data synchronization is achieved, but also deep temporal and spatial information fusion is completed, generating a high-quality, spatial resolution-enhanced fusion signal, which provides an important input for subsequent deep feature extraction.

[0088] Based on this, the application discloses an organization architecture for realizing the processing and feature extraction process of the magnetocardiogram signal, as shown in Figure 6 The main modules include: a time alignment module, which performs high-precision time synchronization on the original signals from the multi-channel magnetocardiogram sensor and eliminates the sampling delay between channels; a spatial fusion module, which performs weighted fusion on the spatial dimension of each channel signal on the basis of time alignment to generate a unified multi-dimensional magnetocardiogram signal; a multi-dimensional magnetocardiogram signal output, which outputs the high-quality magnetocardiogram signal after time and spatial fusion as the input for subsequent feature extraction; and a deep learning feature extraction network, which uses a hybrid architecture CNN-LSTM combining convolutional neural network CNN and long short-term memory network LSTM to perform deep feature extraction on the multi-dimensional magnetocardiogram signal and output magnetocardiogram feature information related to myocardial function.

[0089] 207, data input and multi-source data fusion.

[0090] Specifically, the feature information and the plurality of physiological data are input into the multi-modal fusion neural network, and the multi-modal fusion neural network is used to perform temporal and spatial alignment and feature-level fusion on the feature information and the plurality of physiological data, extract and output fusion features, wherein the fusion features are used to reflect the details of the cardiac electrical activity and the physiological state of the target object.

[0091] In the present embodiment, no subsequent myocardial function evaluation is performed, and the specific input data comes from two main sources, one is the feature information collected from the magnetocardiograph sensor and processed by the aforementioned deep network, and the other is the various physiological data obtained in real time through the physiological monitoring device. The system first synchronously processes these heterogeneous data sources in real time to ensure that they are strictly aligned on the time axis, thereby forming a multi-dimensional signal input. This overall input can be formally represented as:

[0092] In the formula, is the feature information of the magnetocardiograph signal, is the physiological data.

[0093] Subsequently, the feature information and the various physiological data are input into a multi-modal fusion neural network, which is mainly used to perform data fusion. Specifically, the data from different sensors are temporally and spatially aligned to ensure that the timestamps are consistent with the phases of potential physiological events. Then, feature-level fusion is performed in the deep layers of the network. The network automatically mines and integrates the complex nonlinear relationships and temporal-spatial dependencies between the magnetocardiograph signal features and various physiological data through learning. This fusion process can be formally represented as:

[0094] In the formula, represents a multi-channel convolutional neural network, is the fused feature representation.

[0095] Through the above processing, the network finally extracts and outputs a unified, high-level fusion feature This fusion feature integrates the details of cardiac electrical activity from the magnetocardiograph signal and the macro physiological state information obtained from real-time monitoring, forming a joint representation that can reflect both the microscopic electrical physiology and the overall functional state of the heart. This enables the subsequent evaluation model to make decisions based on more comprehensive and discriminative information, thereby laying an important data foundation for improving the overall accuracy and reliability of cardiac health evaluation.

[0096] 208, cardiac load capacity evaluation.

[0097] Specifically, the real-time heart rate value, the real-time blood pressure value and the real-time oxygen saturation value in the fusion feature are acquired, and the magnetocardiogram signal feature in the fusion feature is extracted; the product of the real-time heart rate value and the heart rate influence coefficient is calculated to obtain a first load contribution value; the product of the real-time blood pressure value and the blood pressure influence coefficient is calculated to obtain a second load contribution value; the product of the real-time oxygen saturation value and the oxygen saturation influence coefficient is calculated to obtain a third load contribution value; based on the magnetocardiogram signal feature, the QRS wave length value and the ST segment change value are acquired, and the product of the sum of the QRS wave length value and the ST segment change value and the electrocardiogram feature influence coefficient is calculated to obtain a fourth load contribution value; the first load contribution value, the second load contribution value, the third load contribution value and the fourth load contribution value are weighted and summed based on a preset weight to obtain a cardiac load capacity index for representing the cardiac load capacity.

[0098] In the embodiment, the key parameters for load evaluation need to be extracted from the unified fusion feature, specifically, the real-time heart rate value, the real-time blood pressure value and the real-time oxygen saturation value in the fusion feature are acquired, and the magnetocardiogram signal feature in the fusion feature is extracted; and the core of the cardiac load capacity evaluation process is to calculate a comprehensive cardiac load capacity index, which is obtained by weighted fusion of the contribution values in multiple dimensions. First, the linear contribution values of the physiological parameters are calculated respectively, wherein the expression of the first load contribution value related to the real-time heart rate value is:

[0099] In the expression, the real-time heart rate value is the heart rate influence coefficient is and the first load contribution value is

[0100] The expression of the second load contribution value related to the real-time blood pressure value is:

[0101] In the expression, the real-time blood pressure value is the blood pressure influence coefficient is and the second load contribution value is

[0102] The expression of the third load contribution value related to the real-time oxygen saturation value is:

[0103] In the expression, the real-time oxygen saturation value is the real-time oxygen saturation influence coefficient is and the third load contribution value is

[0104] Meanwhile, the QRS wave duration value and the ST segment change value in the magnetocardiogram signal characteristics are summed up, and the product of the electrocardiogram feature influence coefficient is calculated to obtain a fourth load contribution value, expressed as:

[0105] In the formula, QRS wave duration value, ST segment change, fourth load contribution value.

[0106] After obtaining the first load contribution value, the second load contribution value, the third load contribution value, and the fourth load contribution value, the cardiac load capacity index is calculated by a weighted data fusion algorithm, expressed as:

[0107] In the formula, , which represents the influence function of each physiological signal (such as HR, BP, SpO2, etc.) on the load, i.e., the first load contribution value, the second load contribution value, the third load contribution value, and the fourth load contribution value, is the weighted coefficient of each physiological data on the final load assessment.

[0108] It should be noted that when evaluating the cardiac load, a key function, i.e., the electrocardiogram influence function , is introduced and defined, which is specifically used to quantify the contribution of the electrocardiogram (ECG) signal to the cardiac load. By accurately analyzing two key waveform parameters in the ECG signal closely related to the electrical activity of the heart, the QRS wave duration and the ST segment change, the electrical physiological state of the heart under load is described, and its calculation formula is defined as:

[0109] In the formula, QRSduration represents the duration of the QRS wave, represents the change amount of the ST segment, is a coefficient representing the influence degree of the ECG signal on the load, which can usually be adjusted according to the electrical activity characteristics and clinical data of different patients. Therefore, through the fine analysis of the QRS wave and the ST segment change, a deep description of the electrical activity state of the heart is provided, making the evaluation of the cardiac load more detailed and accurate, especially for patients with abnormal electrocardiogram conduction or potential myocardial ischemia, which can more effectively identify excessive cardiac load or abnormal electrical physiological response.

[0110] Further, the QRS wave duration refers to the time experienced from the beginning to the end of the QRS wave in the electrocardiogram, directly reflecting the speed and synchronicity of the ventricular muscle cell depolarization process. Abnormal prolongation of the QRS wave duration is often associated with ventricular hypertrophy, bundle branch block, or increased cardiac load, and its calculation formula is:

[0111] wherein, is the end time of the QRS wave, is the start time of the QRS wave, which is accurately identified and located by real-time ECG signal processing algorithms, such as wavelet transform-based time-frequency analysis or adaptive peak detection algorithms, to calculate the duration of the QRS wave in each cardiac cycle.

[0112] The ST segment is an isoelectric segment on the ECG between the end point of the QRS wave and the start point of the T wave, reflecting the repolarization process of the ventricular muscle. The elevation or depression of the ST segment relative to the baseline (i.e. ΔST) is a key indicator for diagnosing myocardial ischemia, damage, and evaluating the stress state of the heart, and the calculation formula is:

[0113] wherein, is the final value of the ST segment, is the baseline value of the ST segment, usually referring to the initial ST segment level after the end of the QRS wave, which is accurately defined by ECG signal processing technology after identifying the QRS wave and T wave, and measuring the potential level of the ST segment interval.

[0114] Further, although QRS wave and ST segment parameters are traditionally extracted from ECG signals, the high-resolution magnetocardiogram signals used by the present application can also accurately reflect these electrical activity characteristics. In the magnetocardiogram signal, QRS wave and ST segment information can be extracted by similar waveform recognition algorithms, and the calculation formula for the duration is:

[0115] wherein, T endmag and T startmag represent the end time and start time of the QRS wave in the magnetocardiogram signal, respectively.

[0116] 209、heart pumping efficiency evaluation.

[0117] Specifically, the magnetocardiogram signal features in the fusion features are extracted, and the absolute value of the QRS wave, the time interval of the QRS wave, and the electrocardio conduction velocity are extracted in the magnetocardiogram signal features; the sum of the absolute values of all QRS waves is calculated to obtain a first sum value; the sum of the time intervals of all QRS waves is calculated to obtain a second sum value, and the product of the second sum value and the electrocardio conduction velocity is calculated to obtain a product value; the ratio of the first sum value to the product value is calculated to obtain a heart pumping efficiency index for representing the heart pumping efficiency.

[0118] In the embodiment, the evaluation of the heart pumping efficiency is based on the accurate extraction and comprehensive analysis of the key parameters of the QRS wave in the magnetocardiogram signal. The required parameters are obtained by extracting the magnetocardiogram signal features in the fused features. Specifically, the system first extracts the absolute value of the amplitude of the QRS wave , the time interval of the QRS wave , and the electrocardio conduction velocity from the magnetocardiogram signal features. It should be noted that the electrocardio conduction velocity directly affects the propagation time of the electrical activity in the myocardium, and is closely related to the coordination and efficiency of mechanical contraction. After the parameters are obtained, the system performs a series of calculations to quantify the pumping efficiency. First, the sum of the absolute values of the amplitudes of all QRS waves is calculated to obtain a first sum value. Second, the sum of the time intervals of all QRS waves is calculated to obtain a second sum value. Then, the product of the second sum value and the electrocardio conduction velocity is calculated to obtain a product value. Finally, the ratio of the first sum value to the product value is calculated to obtain a heart pumping efficiency index representing the heart pumping efficiency. The calculation formula is as follows:

[0119] In the formula, E is the heart pumping efficiency index.

[0120] Further, by using the magnetocardiogram signal as the parameter extraction source, the present application can obtain higher precision and clearer QRS wave amplitude and time interval data compared to the traditional electrocardiogram, thereby improving the accuracy of the pumping efficiency calculation. Moreover, the electrocardio conduction velocity is introduced into the calculation model of the pumping efficiency, breaking the limitation of traditional methods relying on QRS wave amplitude. The electrocardio conduction velocity reflects the propagation efficiency of the electrical activity of the heart, and the product term of the time interval better represents the overall time cost of the conversion, making the efficiency evaluation more comprehensive. The entire calculation process is based on the magnetocardiogram signal features after deep processing and fusion, and combines the key variable of the electrocardio conduction velocity, ensuring the accuracy, comprehensiveness, and reliability of the evaluation of the heart pumping capacity of different individuals and under different load conditions, avoiding the one-sidedness of traditional evaluation methods, and realizing the complete technical chain from high-quality signal acquisition to key parameter extraction to comprehensive physical model calculation of the heart pumping efficiency.

[0121] Based on this, the present application discloses a calculation flowchart of the heart pumping efficiency index, as shown in Figure 7 ​The input is based on the magnetocardiogram signal in the fusion feature, and the key waveform information such as QRS wave, P wave, and T wave is extracted. Then, processing and feature extraction are performed to extract the absolute value of the QRS wave amplitude |AQRS|, the time interval ΔT, and the electrocardio conduction velocity CV from the magnetocardiogram signal. Then, intermediate calculation is performed to calculate the sum of the absolute values of all QRS wave amplitudes:∑|AQRS|, and the sum of all QRS wave time intervals is calculated and multiplied by the electrocardio conduction velocity: (∑ΔT)×CV. Finally, the cardiac pumping efficiency index is calculated according to the formula The cardiac pumping efficiency index reflects the amplitude, time, and conduction efficiency of cardiac electrical activity, and can be used as an important quantitative index for evaluating the pumping function of the heart.

[0122] 210, cardiac stress response evaluation.

[0123] In this embodiment, in addition to the cardiac load capacity and the cardiac pumping efficiency, myocardial stress response analysis can also be performed, aiming to quantitatively evaluate the dynamic stress level of the heart under the intervention of acupuncture and other loads, combining traditional physiological index calculation and advanced deep learning modeling. The specific implementation steps include: First, data preprocessing is performed. The original physiological data of the target object collected in real time, especially the heart rate and blood pressure, will be denoised and standardized through a signal processing algorithm, and converted into time series data that can be used for analysis. The process can be represented as:

[0124] In the formula, and respectively represent the heart rate and blood pressure data after preprocessing, and are the original data.

[0125] Then, the load and cardiac electrical activity are trained through an LSTM model. Specifically, a long short-term memory network model is used to learn the complex time sequence correlation between the load application process and the cardiac electrophysiological response. The LSTM model receives the preprocessed heart rate, blood pressure time series data and other related features as input, and its internal state captures and encodes the time sequence pattern related to cardiac stress at each time point t. This process can be formalized as:

[0126] In the formula, represents the output of the LSTM model, and represents the time The instantaneous heart electrical activity state (including HRV, QRS wave, etc. characteristics), the model can identify the potential pattern of heart response under different load conditions through training, on the basis of model learning, the system performs heart stress response evaluation, specifically for stress index calculation, calculation based on heart rate and blood pressure relative to its individualized baseline value change, the system calculates at each sampling time t in the whole observation period T of load application, first calculate the ratio of heart rate and baseline heart rate at this time, and the ratio of blood pressure and baseline blood pressure, then calculate the absolute value of the difference between each ratio and 1, these two values respectively quantify the absolute change amplitude of heart rate and blood pressure at this time relative to the resting baseline, avoid the interference of change direction to intensity evaluation, finally, sum all the absolute values of these two at all time points in the whole observation period to obtain the comprehensive stress index Stress, the calculation formula is:

[0127] In the formula, is the baseline heart rate, which represents the heart rate under normal resting state; is the baseline blood pressure, which represents the blood pressure under normal resting state, the stress index represents the overall stress response intensity of the heart by accumulating the overall deviation of physiological parameters during the load period.

[0128] Further, to improve the accuracy of the model, the system also includes a model optimization and feedback mechanism. In the training phase, by comparing the stress value predicted by the LSTM model with the value calculated according to the above formula, a loss function is constructed:

[0129] In the formula, is the prediction result of the LSTM model, is the actual calculated stress response value; by continuously optimizing the parameters of the LSTM model through the back propagation algorithm, the prediction of the LSTM model is more consistent with the real physiological response.

[0130] 211、Heart function evaluation and early warning.

[0131] Specifically, the system acquires and integrates diverse individual characteristic information of the target subjects, and constructs a comprehensive decision feature vector based on this information. This diverse individual characteristic information includes static health records, genetic and family history, dynamic lifestyle data, and assessment results. The comprehensive decision feature vector is input into a pre-defined ensemble learning model, which outputs a cardiac risk score. This cardiac risk score is then compared with multi-level risk thresholds, including high-risk and medium-risk thresholds. When the cardiac risk score exceeds the high-risk threshold, an emergency warning is triggered, and an adjustment plan is generated, including medical intervention and medication adjustment recommendations. When the cardiac risk score is between the medium-risk and high-risk thresholds, a routine warning is triggered, and a management plan is generated, including lifestyle enhancement interventions and rehabilitation exercise recommendations. When the cardiac risk score is below the medium-risk threshold, a preventative guidance plan is generated, including health maintenance and optimization recommendations.

[0132] In this implementation, the diverse individual characteristic information of the target object is first acquired and integrated. This diverse individual characteristic information is systematically organized into the following categories: static health records, i.e., basic medical history, including diagnosed disease history, such as hypertension, diabetes, and heart disease; genetic and family history, referring to genetic factors, such as the history of heart disease in immediate family members; dynamic lifestyle data, i.e., lifestyle and behavioral data, covering diet, exercise habits, and smoking and drinking history; and assessment context data, which not only includes physiological data reflecting real-time status, such as heart rate, blood pressure, and blood oxygen saturation, but also integrates the assessment results generated in the current assessment cycle, such as the calculated cardiac pumping capacity and cardiac load capacity index, which are key dynamic inputs for risk scoring. By integrating the above information, a comprehensive decision feature vector is constructed based on the diverse individual characteristic information. This vector is essentially a data set that integrates static attributes, dynamic indicators, and the core findings of this assessment.

[0133] Subsequently, the comprehensive decision feature vector is input into a preset ensemble learning model, such as a random forest or a gradient boosting decision tree, the input vector is analyzed by learning the complex mapping relationship between the features and the risk in a large amount of historical data, and finally a quantitative heart risk score is output. It needs to be clear that the decrease of heart pumping capacity and the overloading of heart are the key factors that the model focuses on when assessing risk and directly affect the value; after obtaining the risk score, the system executes a detailed definition of the grading response mechanism, that is, the heart risk score is compared with the multi-level risk threshold. The system presets at least two key critical points of high risk threshold and medium risk threshold. When the heart risk score exceeds the high risk threshold, it indicates that there is an emergency risk, the system immediately triggers an emergency warning, and generates an adjustment scheme, the core content of which is medical intervention and drug adjustment suggestion, aiming to perform emergency disposal to alleviate the high risk state; when the heart risk score is between the medium risk threshold and the high risk threshold, it indicates that there is a risk that needs to be actively managed, the system triggers a regular warning, and generates a management scheme, focusing on lifestyle reinforcement intervention and rehabilitation exercise suggestion, guiding patients to control risk through behavior change; when the heart risk score is lower than the medium risk threshold, it indicates that the current risk is low, and the system generates a preventive guidance scheme, the content of which is mainly health maintenance and optimization suggestion, to consolidate and improve the heart health level; the generation of the above adjustment scheme is based on the result of the heart risk score, the automatically generated personalized intervention measures, the specific content of which is different according to the risk level, can cover multiple dimensions such as drug treatment, for example, adjusting antihypertensive drugs, lipid-lowering drugs, exercise suggestion, for example, developing or adjusting rehabilitation training plan, diet adjustment, for example, controlling salt and fat intake, and lifestyle adjustment, for example, quitting smoking and limiting alcohol, and the ensemble learning model for scheme generation can comprehensively consider historical data, patient physiological data and individual characteristics, automatically optimize and recommend the most suitable intervention measure combination for the current risk state.

[0134] As shown in Figure 8 , the system analyzes and evaluates the diversified individual feature information of the target object through the ensemble learning model, the specific process includes: the input layer obtains and integrates the static health records including demographic information, medical history, examination results, genetic and family history data, activity, sleep, diet and other dynamic lifestyle data, and diversified features such as heart pumping efficiency index; fusion and modeling, the comprehensive decision feature vector [F1, F2, …, Fn] is constructed through the feature fusion engine, and is input into the ensemble learning model (such as random forest, XGBoost) for training and prediction; risk score output, the model outputs a quantitative heart risk score; threshold comparison and layered warning, compare the risk score with the preset multi-level risk threshold, and divide it into high risk, medium risk and low risk levels.

[0135] 212、report generation and feedback.

[0136] In this embodiment, after the myocardial function evaluation process is completed, the system automatically integrates all the key data and conclusions generated during the analysis to generate a comprehensive myocardial function evaluation report, which includes but is not limited to: a summary of the acquisition process of the cardiac multi-channel magnetocardiogram signal and physiological data, quality indicators after signal processing, dynamic trend chart of the cardiac electrical activity under acupuncture load, quantitative impact analysis of load application on key physiological parameters, and the final calculated myocardial health score. All the contents are displayed in the graphical user interface through clear charts, curves and statistical data, so that the doctor can intuitively and quickly grasp the cardiac function status and the whole evaluation process of the patient. To meet the diversified clinical needs, the system also provides a customized report function. The doctor can selectively check the parameters or analysis results that need to be highlighted in the system interface according to the specific diagnosis focus or research needs, for example, a brief report showing only the load-response curve and risk score can be customized, or a detailed technical report containing all the original signal processing intermediate steps can be generated. This function enhances the flexibility of the system and makes the report output more suitable for individualized diagnosis and treatment decision-making process, providing great convenience for the doctor to conduct in-depth analysis and develop treatment plans. In view of the sensitive medical and health data processed by the system, the present application attaches great importance to data security and patient privacy protection. At the data storage level, all patient personal information, original physiological data, magnetocardiogram signals and generated evaluation reports are stored with strong encryption using the advanced encryption standard (AES encryption algorithm), ensuring that even if the data is accessed illegally, it cannot be directly read. At the data transmission level, whether it is between the device and the server or the network communication when the doctor terminal retrieves the report, an encrypted channel is established through the secure transmission layer protocol (TLS), effectively preventing data from being eavesdropped or tampered with during transmission, and ensuring the confidentiality and integrity of the data from acquisition, processing to final review.

[0137] In summary, as shown in Figure 9 The system architecture of the present application mainly includes the following modules: an acupuncture stimulation control unit for outputting adjustable load parameters (frequency, intensity, duration); a multi-channel magnetocardiogram signal acquisition unit for acquiring cardiac magnetic field signals; a plurality of physiological data acquisition units for acquiring physiological waveforms such as ECG, blood pressure and blood oxygen saturation; a physiological data fusion module and a magnetocardiogram signal fusion module for fusion processing of physiological data and magnetocardiogram signals respectively to generate a physiological signal set; a multi-channel deep fusion network and a feature extraction network for time alignment, denoising, spatial fusion and feature extraction of the magnetocardiogram signal; and a cardiac load evaluation model for outputting the cardiac load capacity and the cardiac pumping efficiency based on the fusion features.

[0138] Further, as a specific implementation of the Figure 1 method, the present application embodiment provides a myocardial function evaluation device, as shown in Figure 10As shown, the device comprises: a multi-source acquisition module 301, a load adjustment module 302, a signal processing module 303, and a feature evaluation module 304.

[0139] The multi-source acquisition module 301 is configured to acquire multi-channel cardiac magnetic signals of a target object by using multiple cardiac magnetic sensors during the process of applying acupuncture stimulation to the target object to form a cardiac load, and simultaneously acquire multiple physiological data of the target object by using a physiological monitoring device; The load adjustment module 302 is configured to perform standardization and fusion processing on the multiple physiological data, and then perform secondary fusion on the fused physiological data and the multi-channel cardiac magnetic signals to generate a physiological signal set, and dynamically adjust the load parameters of the acupuncture stimulation based on the physiological signal set and the basic physiological data of the target object; The signal processing module 303 is configured to perform time alignment and spatial fusion on the multi-channel cardiac magnetic signals by using a multi-channel deep fusion network to obtain multi-dimensional cardiac magnetic signals, and perform denoising and enhancement processing on the multi-dimensional cardiac magnetic signals, and then extract feature information related to myocardial function by using a deep learning feature extraction network; The feature evaluation module 304 is configured to perform real-time synchronization and fusion processing on the feature information and the multiple physiological data to obtain fused features, and evaluate the myocardial function state of the target object based on the fused features and by using a preset cardiac load evaluation model to generate an evaluation result, wherein the evaluation result includes a cardiac load capacity and / or a cardiac pumping efficiency.

[0140] In a specific application scenario, the load adjustment module 302 is specifically configured to calculate the load parameters of the acupuncture stimulation by using a preset optimization algorithm based on the basic physiological data, individual feature information, and historical medical data of the target object, wherein the load parameters include the intensity, frequency, and stimulation time of the acupuncture stimulation; in the process of applying the acupuncture stimulation to the target object, calculate the heart rate change value and the blood pressure change value in real time according to the physiological signal set; when the heart rate change value exceeds a preset heart rate change safety threshold, and / or when the blood pressure change value exceeds a preset blood pressure change safety threshold, calculate a first difference value between the heart rate change value and the heart rate change safety threshold, and / or a second difference value between the blood pressure change value and the blood pressure change safety threshold; calculate an acupuncture load adjustment amount according to the first difference value and / or the second difference value, and dynamically adjust the load parameters according to the acupuncture load adjustment amount; and optimize the adjusted load parameters by using a preset machine learning algorithm based on the load response data of the target object in the historical acupuncture stimulation process.

[0141] In a specific application scenario, the signal processing module 303 is specifically configured to input the multi-channel magnetocardiogram signal into a multi-channel deep fusion network, time-synchronize the multi-channel magnetocardiogram signal, and perform weighted fusion on the time-synchronized multi-channel magnetocardiogram signal based on a preset weight, to output a multi-dimensional magnetocardiogram signal with unified spatial dimensions; input the multi-dimensional magnetocardiogram signal into a multi-level denoising neural network, perform convolution and recursive time series modeling on the multi-dimensional magnetocardiogram signal, realize layer-by-layer filtering of low-frequency environmental interference and high-frequency random noise, and complete denoising and enhancement processing of the multi-dimensional magnetocardiogram signal, wherein the multi-level denoising neural network includes a convolutional neural network and a recurrent neural network connected in series; input the processed multi-dimensional magnetocardiogram signal into a deep learning feature extraction network, and extract feature information related to myocardial function, wherein the deep learning feature extraction network combines a convolutional neural network and a long short-term memory network, the convolutional neural network is used to extract local frequency domain and morphological features of the processed multi-dimensional magnetocardiogram signal, and the long short-term memory network is used to capture long-range time series dependence of the processed multi-dimensional magnetocardiogram signal.

[0142] In a specific application scenario, the feature evaluation module 304 is specifically configured to input the feature information and the plurality of physiological data into a multi-modal fusion neural network, perform spatio-temporal alignment and feature-level fusion on the feature information and the plurality of physiological data by using the multi-modal fusion neural network, and extract and output fusion features, wherein the fusion features are used to reflect details of cardiac electrical activity and physiological state of the target object.

[0143] In a specific application scenario, the evaluation result is a cardiac load capacity; the feature evaluation module 304 is further configured to obtain a real-time heart rate value, a real-time blood pressure value, and a real-time oxygen saturation value in the fusion features, and extract a magnetocardiogram signal feature in the fusion features; calculate a product of the real-time heart rate value and a heart rate influence coefficient to obtain a first load contribution value; calculate a product of the real-time blood pressure value and a blood pressure influence coefficient to obtain a second load contribution value; calculate a product of the real-time oxygen saturation value and an oxygen saturation influence coefficient to obtain a third load contribution value; based on the magnetocardiogram signal feature, obtain a QRS wave duration value and an ST segment change value, sum the QRS wave duration value and the ST segment change value, and calculate a product of the sum and an electrocardiogram feature influence coefficient to obtain a fourth load contribution value; and based on a preset weight, perform weighted summation on the first load contribution value, the second load contribution value, the third load contribution value, and the fourth load contribution value to obtain a cardiac load capacity index for representing the cardiac load capacity.

[0144] In a specific application scenario, the evaluation result is the heart pumping efficiency; the feature evaluation module 304 is further configured to extract a magnetocardiogram signal feature in the fused feature, extract an absolute value of a QRS wave, a time interval of the QRS wave, and a heart electrical conduction velocity in the magnetocardiogram signal feature; calculate a sum of the absolute values of all the QRS waves to obtain a first sum value; calculate a sum of the time intervals of all the QRS waves to obtain a second sum value, and calculate a product of the second sum value and the heart electrical conduction velocity to obtain a product value; and calculate a ratio of the first sum value to the product value to obtain a heart pumping efficiency index for representing the heart pumping efficiency.

[0145] In a specific application scenario, as shown in Figure 11 The device further includes a decision suggestion module 305, which is specifically configured to obtain and integrate diversified individual feature information of the target object, and construct a comprehensive decision feature vector based on the diversified individual feature information, wherein the diversified individual feature information includes a static health record, genetic and family history, dynamic lifestyle data, and an evaluation result; input the comprehensive decision feature vector into a preset integrated learning model, output a heart risk score, and compare the heart risk score with a multi-level risk threshold, wherein the multi-level risk threshold includes a high risk threshold and a medium risk threshold; when the heart risk score exceeds the high risk threshold, trigger an emergency warning and generate an adjustment scheme, wherein the adjustment scheme includes a medical intervention and a drug adjustment suggestion; when the heart risk score is between the medium risk threshold and the high risk threshold, trigger a regular warning and generate a management scheme, wherein the management scheme includes a lifestyle reinforcement intervention and a rehabilitation exercise suggestion; and when the heart risk score is lower than the medium risk threshold, generate a preventive guidance scheme, wherein the preventive guidance scheme includes a health maintenance and optimization suggestion.

[0146] It should be noted that other corresponding descriptions of the functions of the heart muscle function evaluation device provided in this embodiment can be referred to the corresponding descriptions in Figure 1 and Figure 3 , which will not be described here in detail.

[0147] Based on the above method as shown in Figure 1 and Figure 3 , accordingly, the present embodiment also provides a storage medium having a computer program stored thereon, which is executed by a processor to implement the above-mentioned heart muscle function evaluation method.

[0148] Based on such understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.), including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the heart muscle function evaluation method of each embodiment scenario of the present application.

[0149] Based on the above, Figure 1 and Figure 3 The method shown, and Figure 10 and Figure 11 The illustrated embodiment of the myocardial function assessment device, in order to achieve the above objectives, such as... Figure 12 As shown, this embodiment also provides a physical device for myocardial function assessment. This device includes a communication bus, a processor, a memory, and a communication interface. It may also include input / output interfaces and a display device. The various functional units can communicate with each other via the bus. The memory stores a computer program, and the processor executes the program stored in the memory to perform the myocardial function assessment method described in the above embodiment.

[0150] Optionally, the physical device may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0151] Those skilled in the art will understand that the structure of the physical device for assessing myocardial function provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0152] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs to be identified. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0153] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms, or it can be implemented by hardware. By applying the technical solution of this application, acupuncture stimulation is used as a controllable and non-invasive load application method. During the entire load application process, cardiac magnetic signals and various physiological data are continuously and synchronously collected. Compared with traditional data acquisition methods such as electrocardiogram and echocardiography, the real-time data stream can continuously depict the complete dynamic response curve of the heart from rest to load and then to recovery, which is convenient for real-time grasp of the evolution trend of myocardial function and improves the timeliness and continuity of assessment. Multi-channel cardiac magnetic sensing technology is introduced, combined with multi-channel deep fusion network and adaptive noise reduction and enhancement algorithm, to perform super-synchronous acquisition and high-fidelity processing of weak cardiac magnetic signals. Among them, the cardiac magnetic signal is extremely sensitive to changes in cardiac electrical activity and can reflect deep electrophysiological details that are difficult to capture by traditional electrocardiogram, thereby accurately identifying and quantifying cardiac electrical activity under acupuncture load. This application addresses subtle and rapid changes in myocardial pumping function. Instead of relying on a single signal, it employs a dual-fusion architecture—a secondary fusion of physiological data and real-time synchronous fusion of feature information—to deeply integrate cardiac magnetic signals with macroscopic physiological indicators. Utilizing a deep learning feature extraction network and a cardiac load assessment model, it extracts effective features and performs intelligent evaluation. This comprehensive assessment of myocardial function from multiple levels yields accurate results, providing a reliable basis for clinical decision-making. Finally, this application constructs an intelligent closed-loop system that not only passively monitors but also dynamically adjusts acupuncture load parameters based on real-time generated physiological signal sets, ensuring the load remains within the safe and effective range for the target individual. The evaluation results also provide feedback to optimize signal processing and load adjustment strategies. The entire evaluation process is safe and reliable, providing direct guidance for precision treatment. In summary, this method combines acupuncture load with cardiac magnetic technology, integrating signal processing, data fusion, and artificial intelligence algorithms to dynamically, continuously, and accurately assess myocardial function, providing a reliable basis for clinical decision-making.

[0154] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or can be modified to be located in one or more apparatuses different from this embodiment. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0155] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of any particular implementation scenario. The above disclosures are merely a few specific implementation scenarios of this application; however, this application is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A method for assessing myocardial function, characterized in that, include: During the process of applying acupuncture stimulation to the target object to create cardiac load, multiple magnetic sensors are used to collect multi-channel magnetic signals of the target object, and physiological monitoring equipment is used to collect various physiological data of the target object. The various physiological data are standardized and fused, and then the fused physiological data is fused with the multi-channel magnetic-cardiogram signal to generate a physiological signal set. Based on the physiological signal set and the basic physiological data of the target object, the load parameters of the acupuncture stimulation are dynamically adjusted. The multi-channel magnetic field signals are time-aligned and spatially fused using a multi-channel deep fusion network to obtain multi-dimensional magnetic field signals. After denoising and enhancing the multi-dimensional magnetic field signals, feature information related to myocardial function is extracted using a deep learning feature extraction network. The feature information and the various physiological data are synchronized and fused in real time to obtain fused features. Based on the fused features and using a preset cardiac load assessment model, the myocardial function status of the target object is assessed to generate assessment results, wherein the assessment results include cardiac load capacity and / or cardiac pumping efficiency.

2. The method according to claim 1, characterized in that, The dynamic adjustment of the acupuncture stimulation load parameters based on the physiological signal set and the basic physiological data of the target object includes: Based on the target object's basic physiological data, individual characteristic information, and historical medical data, the load parameters of the acupuncture stimulation are calculated using a preset optimization algorithm. The load parameters include the intensity, frequency, and duration of the acupuncture stimulation. During the acupuncture stimulation applied to the target object, the heart rate change value and blood pressure change value are calculated in real time based on the physiological signal set. When the heart rate change value exceeds a preset heart rate change safety threshold, and / or when the blood pressure change value exceeds a preset blood pressure change safety threshold, calculate a first difference between the heart rate change value and the heart rate change safety threshold, and / or calculate a second difference between the blood pressure change value and the blood pressure change safety threshold; The acupuncture load adjustment amount is calculated based on the first difference and / or the second difference, and the load parameters are dynamically adjusted based on the acupuncture load adjustment amount. Based on the load response data of the target object during historical acupuncture stimulation, the adjusted load parameters are optimized using a preset machine learning algorithm.

3. The method according to claim 1, characterized in that, The process involves using a multi-channel deep fusion network to perform time-alignment and spatial fusion of the multi-channel magnetic resonance imaging (MRCI) signals to obtain multi-dimensional MRCI signals. After denoising and enhancing the multi-dimensional MRCI signals, a deep learning feature extraction network is used to extract feature information related to myocardial function, including: The multi-channel magnetic field signals are input into a multi-channel deep fusion network to synchronize the multi-channel magnetic field signals in time. The time-synchronized multi-channel magnetic field signals are then weighted and fused based on preset weights to output a multi-dimensional magnetic field signal with unified spatial dimensions. The multidimensional magnetic field signal is input into a multi-level denoising neural network. By performing convolution and recursive time-series modeling on the multidimensional magnetic field signal, the low-frequency environmental interference and high-frequency random noise are filtered out layer by layer, thus completing the denoising and enhancement processing of the multidimensional magnetic field signal. The multi-level denoising neural network includes a concatenated convolutional neural network and a recursive neural network. The processed multidimensional magnetic field signal is input into a deep learning feature extraction network to extract feature information related to myocardial function. The deep learning feature extraction network combines a convolutional neural network and a long short-term memory network. The convolutional neural network is used to extract the local frequency domain and morphological features of the processed multidimensional magnetic field signal, and the long short-term memory network is used to capture the long-range temporal dependencies of the processed multidimensional magnetic field signal.

4. The method according to claim 1, characterized in that, The real-time synchronization and fusion processing of the feature information and the various physiological data to obtain fused features includes: The feature information and the various physiological data are input into a multimodal fusion neural network. The multimodal fusion neural network is used to perform spatiotemporal alignment and feature-level fusion on the feature information and the various physiological data, and to extract and output fused features. The fused features are used to reflect the details of the cardiac electrical activity and physiological state of the target object.

5. The method according to claim 1, characterized in that, The assessment result is the cardiac workload capacity; the assessment of the myocardial function status of the target object based on the fusion features and using a preset cardiac workload assessment model, generating assessment results, includes: The real-time heart rate, real-time blood pressure, and real-time oxygen saturation values ​​are obtained from the fusion features, and the magnetic cardiomyocyte signal features are extracted from the fusion features. The first load contribution value is obtained by multiplying the real-time heart rate value by the heart rate influence coefficient. The second load contribution value is obtained by multiplying the real-time blood pressure value by the blood pressure influence coefficient. The product of the real-time oxygen saturation value and the oxygen saturation influence coefficient is calculated to obtain the third load contribution value; Based on the aforementioned magnetic field signal characteristics, the QRS duration value and ST segment variation value are obtained. After summing the QRS duration value and the ST segment variation value, the product with the electrocardiogram characteristic influence coefficient is calculated to obtain the fourth load contribution value. The first load contribution value, the second load contribution value, the third load contribution value, and the fourth load contribution value are weighted and summed based on preset weights to obtain a cardiac load capacity index used to characterize the cardiac load capacity.

6. The method according to claim 1, characterized in that, The evaluation result is the cardiac pumping efficiency; the evaluation of the myocardial function status of the target object based on the fusion features using a preset cardiac load assessment model, generating evaluation results, includes: Extract the magnetic-cardiogram signal features from the fusion features, and extract the absolute value of the QRS wave amplitude, the time interval of the QRS wave, and the electrocardiographic conduction velocity from the magnetic-cardiogram signal features; Calculate the sum of the absolute values ​​of the amplitudes of all QRS waves to obtain the first sum; Calculate the sum of the time intervals of all QRS waves to obtain the second sum, and calculate the product of the second sum and the ECG conduction velocity to obtain the product value; The ratio of the first sum to the product is calculated to obtain the cardiac pumping efficiency index, which characterizes the cardiac pumping efficiency.

7. The method according to claim 1, characterized in that, After generating the evaluation results, the method further includes: The diverse individual characteristic information of the target object is acquired and integrated, and a comprehensive decision feature vector is constructed based on the diverse individual characteristic information. The diverse individual characteristic information includes static health records, genetic and family history, dynamic lifestyle data, and the assessment results. The comprehensive decision feature vector is input into a preset ensemble learning model to output a cardiac risk score, and the cardiac risk score is compared with a multi-level risk threshold, wherein the multi-level risk threshold includes a high-risk threshold and a medium-risk threshold. When the cardiac risk score exceeds the high-risk threshold, an emergency warning is triggered and an adjustment plan is generated, which includes medical intervention and medication adjustment recommendations. When the cardiac risk score is between the medium-risk threshold and the high-risk threshold, a routine warning is triggered and a management plan is generated, wherein the management plan includes lifestyle enhancement interventions and rehabilitation exercise recommendations; When the cardiac risk score is below the medium-risk threshold, a preventive guidance plan is generated, wherein the preventive guidance plan includes health maintenance and optimization recommendations.

8. A myocardial function assessment device, characterized in that, include: The multi-source acquisition module is used to acquire multi-channel magnetic signals of the target object using multiple magnetic sensors during the process of applying acupuncture stimulation to the target object to create cardiac load, and at the same time, to acquire various physiological data of the target object using physiological monitoring equipment. The load adjustment module is used to standardize and fuse the various physiological data, then fuse the fused physiological data with the multi-channel magnetic-cardiogram signal to generate a physiological signal set, and dynamically adjust the load parameters of the acupuncture stimulation based on the physiological signal set and the basic physiological data of the target object. The signal processing module is used to perform time alignment and spatial fusion on the multi-channel magnetic field signals using a multi-channel deep fusion network to obtain multi-dimensional magnetic field signals. After denoising and enhancing the multi-dimensional magnetic field signals, a deep learning feature extraction network is used to extract feature information related to myocardial function. The feature evaluation module is used to perform real-time synchronization and fusion processing on the feature information and the various physiological data to obtain fused features. Based on the fused features and using a preset cardiac load assessment model, the myocardial function status of the target object is evaluated, and an evaluation result is generated. The evaluation result includes cardiac load capacity and / or cardiac pumping efficiency.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images