Integrated monitoring system integrating 12-lead electrocardiogram and pulse oxyhemoglobin saturation

Abstract

The invention relates to the technical field of medical instrument diagnosis, discloses a comprehensive monitoring system fusing 12-lead electrocardiogram and pulse blood oxygen saturation, and aims to solve the problem that in the prior art, due to hypothesis of peripheral perfusion uniformity, regional hemodynamic anomalies cannot be recognized. According to the system, 12-lead electrocardiosignals and four-channel pulse blood oxygen and photoelectric volume pulse wave signals of limbs are synchronously collected and preprocessed, then electrocardio and perfusion characteristic parameters are extracted in parallel, and a limb perfusion difference index matrix is constructed; further, through an electrocardio-perfusion space-time correlation fusion diagnosis engine, recognition and early warning of an electricity-hemodynamics coupling abnormal mode are achieved. According to the application, the relevance between the cardiac electrical activity and the peripheral perfusion state can be dynamically and comprehensively evaluated, and the early diagnosis and early warning capability of diseases such as aortic dissection and congenital heart disease can be remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of medical devices, specifically relating to a comprehensive monitoring system that integrates 12-lead electrocardiogram and pulse oxygen saturation. Background Technology

[0002] Cardiovascular and respiratory diseases are leading causes of death and disability worldwide, making early identification and dynamic monitoring crucial for clinical intervention. In modern healthcare, electrocardiography (ECG) and pulse oximetry (SpO2) monitoring have become two core non-invasive methods for assessing cardiac electrophysiological status and peripheral oxygenation function. A 12-lead ECG simultaneously acquires multi-angle potential changes in the heart using 10 electrodes on the body surface (4 limb electrodes and 6 chest leads), comprehensively reflecting key pathological information such as arrhythmias, myocardial ischemia, and conduction abnormalities, and has long been considered the gold standard for cardiovascular diagnosis. Meanwhile, pulse oximetry monitoring utilizes the difference in absorption of red and infrared light in oxygenated and deoxygenated hemoglobin, non-invasively acquiring arterial oxygen levels and pulse waveforms through peripheral optical sensors. It is widely used for hypoxemia screening, oxygen therapy guidance, and circulatory status assessment. Currently, these two parameters are typically integrated into multi-parameter monitors, achieving simultaneous display of cardiac electrical activity and single-point oxygenation status, forming the basic framework for routine clinical monitoring.

[0003] Pulse oxygen saturation monitoring commonly employs a single-site measurement mode (e.g., one finger on one side). Its underlying logic is based on the crucial assumption that peripheral perfusion in the human body is highly homogeneous in spatial distribution, and a single-point reading is sufficient to represent the overall oxygenation and circulatory status. However, this assumption does not hold true in many critical or structural cardiovascular pathological conditions. For example, in acute aortic dissection, the intimal tear can selectively block arterial branches supplying specific limbs, resulting in significantly lower oxygen saturation on one upper or lower limb compared to the contralateral side. In newborns with ductus arteriosus-dependent congenital heart disease, a patent ductus arteriosus can cause a systemic difference in oxygenation levels between the upper and lower limbs, manifesting as typical "differential cyanosis." Furthermore, in various shock states, although systemic perfusion generally declines, the degree of damage and recovery time of peripheral microcirculation in different limbs may exhibit significant heterogeneity. In these situations, relying solely on single-point SpO2 monitoring not only fails to reveal regional perfusion abnormalities but may also obscure crucial diagnostic clues due to incomplete information, delaying the assessment of the true nature of the condition. Although clinicians can manually attach multiple pulse oximeters for comparison when there is a high suspicion, this operation relies on subjective judgment, lacks standardized procedures, and is difficult to achieve continuous and synchronous multi-point dynamic monitoring, which cannot meet the urgent needs of precision medicine for early warning and quantitative assessment.

[0004] While existing technologies can provide high-precision 12-lead electrocardiograms and single-point oxygen saturation data, they suffer from structural deficiencies in integrating multi-dimensional hemodynamic information. The fundamental contradiction lies in this: on the one hand, 12-lead ECGs provide a panoramic view of cardiac electrical activity; on the other hand, SpO2 monitoring remains limited to a single peripheral site, failing to achieve a synergistic spatial mapping with ECG information. This results in a loss of the ability to identify high-risk conditions such as aortic lesions, congenital shunts, or regional hypoperfusion through differences in oxygenation across multiple limbs. This technological fragmentation—comprehensive electrophysiological assessment but one-sided perfusion evaluation—makes it difficult for existing integrated monitoring systems to provide targeted early warnings in complex clinical scenarios requiring combined analysis of cardiac function and peripheral circulation distribution. Therefore, there is an urgent need for an integrated monitoring system capable of simultaneously acquiring 12-lead ECGs and limb pulse oxygen saturation data. This system, through real-time comparison and fusion analysis of multi-point perfusion information, can overcome the inherent limitations of single-point monitoring, providing a novel technological approach for identifying asymmetric blood flow distribution and related critical illnesses. Summary of the Invention

[0005] In view of this, the problem to be solved by the present invention is to provide a comprehensive monitoring system that integrates 12-lead electrocardiogram and pulse oxygen saturation.

[0006] The technical problem to be solved by the present invention is to overcome the limitations of the existing scheme that combines 12-lead electrocardiogram monitoring with single-point pulse oximetry monitoring. Due to the inherent assumption of uniformity of systemic peripheral perfusion, it is unable to effectively identify and quantify regional and asymmetric hemodynamic abnormalities caused by pathological conditions such as aortic dissection, specific congenital heart disease, or severe shock, resulting in missing diagnostic information, insufficient early warning capabilities, and delayed intervention.

[0007] To address the aforementioned technical problems, this invention provides a comprehensive monitoring system integrating 12-lead electrocardiogram (ECG) and pulse oximetry. It constructs a multi-channel, highly synchronous physiological signal acquisition front-end to acquire, in real-time and in parallel, a 12-lead ECG data stream reflecting the overall electrophysiological activity of the heart, and a 4-channel pulse oximetry and photoplethysmography (PPO) wave data stream accurately characterizing the local perfusion and oxygenation status of the extremities. Based on this, the invention introduces a multimodal data processing and fusion analysis architecture. First, it performs rigorous timestamp alignment and signal quality optimization preprocessing on the signals from each channel. Then, it extracts key electrophysiological feature parameters from the ECG signal and blood oxygenation, perfusion, and waveform morphology feature parameters from the pulse wave signal in parallel. The core of this invention is the establishment of a limb perfusion difference quantification analysis unit. By calculating and tracking the differences and ratios of blood oxygen saturation, perfusion index, and pulse wave conduction time among the limbs in real time, it generates a set of dynamic index matrices characterizing the symmetry of peripheral circulation. Finally, through a heart... The electro-perfusion spatiotemporal correlation fusion diagnostic engine performs deep spatiotemporal correlation analysis on the cardiac events revealed by electrocardiograms and the limb perfusion difference matrix, identifying abnormal electro-hemodynamic coupling patterns with specific clinical significance. This generates highly specific diagnostic inferences and graded early warning information for regional perfusion disorders, achieving a comprehensive, dynamic, and correlated assessment of cardiac electrophysiology and peripheral hemodynamic status. Furthermore, to enhance the system's early warning capabilities for major adverse cardiovascular events and MACE, a category association learning-based prediction algorithm—CSP (CardioSync Predictor)—is embedded. This algorithm can dynamically learn the temporal dependencies between electrocardiogram and perfusion parameters, achieving an upgrade from "rule matching" to "data-driven prediction."

[0008] According to one aspect of the present invention, a comprehensive monitoring system integrating 12-lead electrocardiogram and pulse oximetry is provided, the system comprising: A multi-channel physiological signal synchronous acquisition module is configured to simultaneously acquire 12-lead ECG signal timing data and 4-channel pulse oxygen saturation and photoplethysmography pulse wave signal timing data through ten ECG electrodes deployed on the patient's body surface and four optical sensors fixed to the patient's extremities, and assign a unified high-precision timestamp to all acquired data streams. An ECG-blood oxygen data preprocessing unit is connected to the output of the multi-channel physiological signal synchronous acquisition module. It is used to perform baseline drift correction, power frequency interference filtering and electromyographic noise suppression on the received ECG signal time series data, and to perform motion artifact identification and removal, and DC component and AC component separation on the photoplethysmography pulse wave signal time series data. A multidimensional feature parallel extraction unit is connected to the output of the ECG-oxygen data preprocessing unit. It is used to extract heart rate, arrhythmia events, QRS complex width, ST segment offset, and QT interval dispersion in real time from the preprocessed 12-lead ECG signal. At the same time, it independently extracts pulse oxygen saturation value, pulse rate, perfusion index, pulse wave amplitude, rise time, and tidal wave and dicrotic wave morphological parameters for each limb channel from the preprocessed 4-channel photoplethysmography pulse wave signal. A limb perfusion difference quantitative analysis unit is connected to the output of the multidimensional feature parallel extraction unit. It is used to receive perfusion-related feature parameters of the limb channels and calculate and generate a limb perfusion difference index matrix in real time based on preset mathematical operation rules. The matrix includes the blood oxygen saturation difference between any two limbs, the perfusion index ratio, and the pulse arrival time difference based on the electrocardiogram R wave. An electrocardiogram-perfusion spatiotemporal correlation fusion diagnostic engine has its input terminals connected to the output terminals of the multidimensional feature parallel extraction unit and the limb perfusion differential quantitative analysis unit, respectively. The engine has a set of clinical diagnostic rule knowledge base embedded in it, which is used to perform spatiotemporal correlation matching between the change pattern of electrocardiogram feature parameters and the dynamic evolution of the limb perfusion differential index matrix, so as to identify predefined electro-hemodynamic coupling abnormal patterns pointing to specific pathological states. A comprehensive physiological state assessment and early warning output module is connected to the output of the ECG-perfusion spatiotemporal correlation fusion diagnostic engine. It is used to process the input through the CSP prediction algorithm submodule based on the coupled abnormal pattern recognition results output by the engine, fuse the prediction results with the rule base output, generate the final diagnosis through weighted voting (the weights are dynamically adjusted based on the confidence level), and finally generate structured early warning information containing specific pathological state inferences, confidence assessments, and risk level classifications, which is then visualized through the user interface.

[0009] In one embodiment of the present invention, the multi-channel physiological signal synchronous acquisition module specifically includes an electrocardiogram (ECG) signal acquisition submodule and a 4-channel pulse oximetry acquisition submodule. The ECG signal acquisition submodule comprises ten silver-silver chloride patch electrodes, connected via shielded leads to a preamplifier with a built-in 24-bit analog-to-digital converter. The preamplifier differentially amplifies the weak voltage signals from the electrodes by at least 1000 times and digitizes them at a sampling frequency of 1000Hz. The 4-channel pulse oximetry acquisition submodule comprises four identical finger-clip or strap-type optical probes. Each probe integrates a red light-emitting diode with a center wavelength of 660nm and an infrared light-emitting diode with a center wavelength of 940nm, as well as a high-sensitivity silicon photodiode receiver. The submodule alternately drives the light-emitting diodes in the four probes via a time-division multiplexing driving circuit and synchronously acquires the output photocurrent signals of the corresponding photodiodes, digitizing them via independent transimpedance amplifiers and a 16-bit analog-to-digital converter at a sampling frequency of 200Hz. The multi-channel physiological signal synchronous acquisition module is equipped with a master clock generator driven by a temperature-compensated crystal oscillator. The synchronous clock signal generated by the generator is distributed to the analog-to-digital converters of the electrocardiogram signal and the pulse oxygenation signal to ensure that all data sampling points are strictly aligned on the time base.

[0010] Furthermore, the ECG-oxygen data preprocessing unit employs a wavelet transform-based denoising method when processing the ECG signal. This method involves multi-scale wavelet decomposition of the original ECG signal to identify and suppress high-frequency electromyographic noise and low-frequency baseline drift at specific scale coefficients. A notch filter is used to precisely filter out 50Hz or 60Hz power frequency interference. For the photoplethysmography (PPG) signal, the unit uses an adaptive filtering algorithm based on a triaxial accelerometer signal to eliminate motion artifacts. The DC component of the signal is extracted using a moving average method, and the AC component representing pulse oscillations is extracted using a bandpass filter.

[0011] Furthermore, the multidimensional feature parallel extraction unit applies the Pan-Tompkins algorithm for real-time detection and localization of the QRS complex when extracting ECG features. Using this as a reference point, it defines the start and end points of the P and T waves, and then calculates standard parameters such as heart rate, PR interval, QRS duration, and QT interval. The unit identifies arrhythmic events such as atrial fibrillation and premature ventricular contractions through morphological matching and rhythm analysis algorithms. For the calculation of ST segment offset, the PR segment is used as the isoelectric line reference, and the voltage offset value at 60 ms after point J is measured. When extracting pulse wave features, the perfusion index is calculated by multiplying the AC component amplitude to the DC component amplitude of the photoplethysmography (PPG) signal by 100. The pulse arrival time is defined as the time interval from the peak of the R wave on the ECG to the peak of the main wave of the PPG signal corresponding to the cardiac cycle.

[0012] As a core implementation of this invention, the limb perfusion difference quantitative analysis unit generates a limb perfusion difference index matrix, specifically a 4x4 symmetric matrix. The off-diagonal elements include: the difference in blood oxygen saturation between the right and left upper limbs, the difference in blood oxygen saturation between the right and left lower limbs, the difference in blood oxygen saturation between the right and right lower limbs, the difference in blood oxygen saturation between the left and left lower limbs, the difference in blood oxygen saturation between the left and right upper limbs, and the difference in blood oxygen saturation between the left and right lower limbs. The matrix also includes the perfusion index ratio and pulse arrival time difference between corresponding limb pairs. The unit continuously updates this matrix and compares each element value with a preset clinical normal fluctuation threshold. When the absolute value of any element exceeds the threshold, it is marked as abnormal.

[0013] Furthermore, the clinical diagnostic rule knowledge base internally embedded in the ECG-perfusion spatiotemporal correlation fusion diagnostic engine contains a set of "condition-action" rules. A typical rule example is: when the engine detects ECG characteristics of sinus tachycardia without clear evidence of myocardial ischemia, and the limb perfusion difference index matrix shows that the absolute value of the difference in blood oxygen saturation between the right and left upper limbs is greater than 0.03, and the perfusion index of the right upper limb is less than 0.5 of the perfusion index of the left upper limb, and this state lasts for more than 5 minutes, the engine determines it as a highly suspected coupling abnormality pattern of type A aortic dissection involving the brachiocephalic trunk. Another rule example is: when the monitored subject is a newborn and the ECG characteristics are normal, but the limb perfusion difference index matrix shows that the difference in blood oxygen saturation between the right upper limb and any lower limb is greater than 0.05, it is determined as a suspected coupling abnormality pattern of ductus arteriosus-dependent pulmonary or systemic circulation disorder.

[0014] Furthermore, the comprehensive physiological state assessment and early warning output module processes data using the CSP prediction algorithm. The specific steps of this algorithm are as follows: First, the input layer receives output from the multi-dimensional feature parallel extraction unit, including 12-lead ECG features (such as heart rate, ST segment shift, and QT interval dispersion) and limb blood oxygenation features (such as blood oxygen saturation difference and perfusion index ratio). The input data is converted into time series tokens, each token corresponding to a variable's sequence within a time window. Second, a category-related attention mechanism is used, assigning dynamic weights to different variables through variable-level self-attention and cross-attention calculations. For example, the association weight between ECG variables (such as ST segment shift) and blood oxygenation variables (such as the blood oxygen difference between the right upper limb and left lower limb) is quantified through attention scores. This mechanism can automatically focus on key risk factors (such as perfusion differences related to aortic dissection). Then, the time-series fusion module uses an encoder-decoder structure to process long-term data (such as follow-up records spanning several months), preserving temporal sequence information through location encoding, and outputting the probability of case events occurring within the next 3 months. Finally, the output layer generates prediction results (such as risk scores and confidence levels) and merges them with the original rule base. The final diagnosis is generated through weighted voting, and the training uses the Focal Loss loss function to solve the data imbalance problem.

[0015] According to another aspect of the present invention, a comprehensive monitoring method integrating 12-lead electrocardiogram and pulse oximetry is provided, the method comprising the following steps: A multi-channel physiological signal synchronous acquisition module is used to synchronously acquire the patient's 12-lead electrocardiogram signal time-series data and the 4-channel pulse oxygen saturation and photoplethysmography pulse wave signal time-series data of the extremities, and assign a unified timestamp to all data streams. An ECG-blood oxygenation data preprocessing unit is used to perform digital filtering, noise reduction, and artifact removal on the acquired raw ECG signals and photoplethysmography pulse wave signals. A multidimensional feature parallel extraction unit is used to extract the set of electrocardiographic feature parameters from the preprocessed electrocardiogram signal and from the 4-channel photoplethysmography pulse wave signal. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0017] In the attached diagram: Figure 1 This is a schematic diagram of the overall technical architecture of a comprehensive monitoring system that integrates 12-lead electrocardiogram and pulse oximetry proposed in this invention. Figure 2This is a schematic diagram of the core principle framework of the limb perfusion differential quantitative analysis unit in this invention; Figure 3 This is a flowchart illustrating the logical process of synchronous acquisition of multi-channel physiological signals and preprocessing of ECG-blood oxygenation data in this invention. Figure 4 This is a flowchart illustrating the logical process of parallel extraction of multidimensional features and spatiotemporal correlation fusion diagnosis of electrocardiogram-perfusion in this invention. Figure 5 This is a schematic diagram of the multi-level interaction relationship and data flow between the terminal and the cloud in this invention; Detailed Implementation The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be noted that when a component is described as "fixed to" another component, it can be directly on the other component or may have a component in between. When a component is considered "connected to" another component, it can be directly connected to the other component or may have a component in between. When a component is considered "set on" another component, it can be directly set on the other component or may have a component in between. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] The method begins with step S101: A multi-channel physiological signal synchronous acquisition module simultaneously acquires the patient's 12-lead ECG signal timing data and the timing data of 4-channel pulse oximetry and photoplethysmography (PPO) wave signals from the extremities, assigning a unified timestamp to all data streams. This step relies on a highly integrated hardware front-end. The multi-channel physiological signal synchronous acquisition module includes an ECG signal acquisition submodule and a 4-channel pulse oximetry acquisition submodule. The ECG signal acquisition submodule deploys ten silver-silver chloride patch electrodes, fixed to the patient's standard 12-lead positions: four limb electrodes (right arm, left arm, right leg, and left leg) and six chest electrodes (V1 to V6). These electrodes are connected to a preamplifier via shielded lead wires. This preamplifier has a built-in 24-bit analog-to-digital converter that differentially amplifies the weak voltage signals from the electrodes by at least 1000 times and digitizes them at a sampling frequency of 2000Hz. The 4-channel pulse oximetry acquisition submodule comprises four identical finger-clip optical probes, fixed to the distal ends of the patient's right hand, left hand, right foot, and left foot, respectively. Each probe integrates a red LED with a center wavelength of 660nm, an infrared LED with a center wavelength of 940nm, and a high-sensitivity silicon photodiode receiver. This submodule employs a time-division multiplexing drive circuit, alternately driving the LEDs of the four probes according to a preset timing sequence, ensuring that only one probe is active at any given time, thus avoiding optical signal crosstalk. The photocurrent signal output from the corresponding photodiode is converted into a voltage signal by an independent transimpedance amplifier and digitized at a sampling frequency of 200Hz by a 16-bit precision analog-to-digital converter. To ensure strict temporal alignment of all physiological signals, the module incorporates a master clock generator driven by a temperature-compensated crystal oscillator. Its highly stable synchronous clock signal is simultaneously distributed to all analog-to-digital converters for both ECG and pulse oximetry signals, ensuring that each sampling point is assigned a precise timestamp based on the same time reference, thus laying the foundation for subsequent multimodal data fusion.

[0021] Next, step S102 is executed: using an ECG-oxygenation data preprocessing unit, digital filtering, denoising, and artifact removal are performed on the acquired raw ECG and photoplethysmography (PPG) signals. This unit receives the raw data stream from the multi-channel physiological signal synchronous acquisition module and executes two independent but structured preprocessing procedures in parallel. For ECG time-series data, the preprocessing procedure first applies a wavelet transform-based denoising algorithm. This algorithm uses the dB6 wavelet basis function to perform a 6-level wavelet decomposition on the raw signal, obtaining detail coefficients and approximation coefficients at different scales. For high-frequency electromyographic noise, soft thresholding is applied to the detail coefficients of the first to third levels by setting a threshold; for low-frequency baseline drift, the approximation coefficients of the sixth level are reconstructed and subtracted from the raw signal to achieve baseline correction. Subsequently, the signal is passed through a second-order IIR notch filter with a center frequency of 50Hz or 60Hz (automatically switched according to the power grid standard) to accurately filter out power frequency interference. For time-series photoplethysmography (PPG) signals, the preprocessing workflow first uses motion signals acquired by a triaxial accelerometer integrated with the optical probe as a reference input to construct an adaptive filter. This filter employs a minimum mean square error algorithm to estimate and remove artifact components caused by limb movement from the original PPG signal in real time. After artifact removal, the signal is fed into a bandpass filter with a cutoff frequency of 0.05Hz to 10Hz to separate the AC component characterizing pulse fluctuations, and the DC component is extracted using a moving average filter with a window length of 5s. This step ensures that the signal upon which subsequent feature extraction depends has a high signal-to-noise ratio and good morphological fidelity.

[0022] The following step, S103, is executed: A multi-dimensional feature parallel extraction unit extracts a set of electrocardiographic feature parameters from the preprocessed ECG signal and independently extracts perfusion and oxygenation feature parameters for each limb channel from the 4-channel photoplethysmography (PPG) signal. This unit employs a parallel computing architecture, simultaneously processing data from 12 ECG leads and 4 PPG channels. For ECG feature extraction, the Pan-Tompkins algorithm is first applied to perform real-time QRS complex detection on all lead signals. This algorithm generates an enhanced QRS feature wave by performing bandpass filtering, differentiation, square sum integration, and moving window integration on the signal, and then accurately locates the R-wave peak position using a dynamic thresholding method. A cardiac cycle window is defined by retrospectively elongating the R-wave peak by 150 ms and extending it backward by 300 ms. Within this window, the P-wave initiation, P-wave end, QRS initiation, QRS end, T-wave initiation, and T-wave end are identified using a combination of morphological template matching and slope analysis. Based on these key points, standard parameters such as heart rate, PR interval, QRS duration, QT interval, and ST segment offset were calculated. The ST segment offset was calculated using the midpoint of the PR segment as the isoelectric reference, measuring the voltage deviation at point J 60ms relative to this reference. Simultaneously, arrhythmia events such as atrial fibrillation, premature ventricular contractions, and atrioventricular block were identified through rhythm regularity analysis and morphological abnormality detection algorithms. Regarding pulse wave feature extraction, for each limb channel signal, the pulse oxygen saturation value was first calculated based on the logarithmic transformation of the ratio of the red and infrared AC component amplitudes, mapped to the final SpO2 percentage value through a pre-stored calibration curve. The pulse rate was determined by the reciprocal of the time interval between adjacent pulse wave peaks. The perfusion index was defined as the ratio of the AC component amplitude to the DC component amplitude multiplied by 100. Furthermore, morphological parameters such as pulse wave amplitude, rise time (time from the waveform's inception to the peak), and the relative height and time interval between tidal waves and dicrotic waves were extracted. All extracted feature parameters are assigned the same timestamp as the original signal, forming a structured feature data stream.

[0023] Next, step S104 is executed: the perfusion and oxygenation characteristic parameter set of the four limb channels is input into a limb perfusion difference quantitative analysis unit, which calculates and generates a limb perfusion difference index matrix in real time. The core function of this unit is to transform the data of the four independent limb channels into a set of quantitative indicators that can characterize the symmetry of peripheral circulation. The index matrix is ​​constructed as a 4x4 symmetric matrix, with its rows and columns corresponding to the right upper limb, left upper limb, right lower limb, and left lower limb, respectively. The diagonal elements of the matrix are set to zero, and the off-diagonal elements are filled with the difference index calculated from the corresponding limb pairs. Specifically, for any two limbs i and j, the blood oxygen saturation difference ΔSpO2(i,j) = SpO2(i) - SpO2(j), the perfusion index ratio PI_ratio(i,j) = PI(i) / PI(j), and the pulse arrival time difference ΔPAT(i,j) = PAT(i) - PAT(j) are calculated. The pulse arrival time (PAT) is defined as the time interval from the peak of the R wave on the electrocardiogram (ECG) to the peak of the main wave of the corresponding limb photoplethysmogram (PPT). This timestamp alignment relies on the unified time reference established in step S101. This unit continuously updates this matrix with a period of 1 second and compares each element value in the matrix with a preset clinical normal fluctuation threshold in real time. For example, the normal threshold for the blood oxygen saturation difference is set to A, the normal range for the perfusion index ratio is set to 0.5 to 2.0, and the normal threshold for the pulse arrival time difference is set to 50 ms. When the absolute value of any element exceeds its corresponding threshold and continues to exceed the preset confirmation time window, the element is marked as abnormal, and a local abnormality flag is triggered.

[0024] Finally, step S105 is executed: the set of electrocardiographic feature parameters and the limb perfusion difference index matrix are input into an electrocardiogram-perfusion spatiotemporal correlation fusion diagnostic engine for deep spatiotemporal correlation analysis to identify predefined abnormal electro-hemodynamic coupling patterns and generate structured early warning information. This engine internally contains a clinical diagnostic rule knowledge base, which consists of a set of "condition-action" rules, each encoding a typical multimodal manifestation under a specific pathological state. The engine's workflow involves matching the real-time input electrocardiographic features and perfusion difference matrix with all rule conditions in the knowledge base. The matching process considers not only the static values ​​of the parameters but also their dynamic evolution patterns, such as the duration and rate of change of the abnormal state. A typical rule condition is: electrocardiographic features show sinus rhythm, heart rate greater than 100 beats per minute, and no ST segment elevation or depression; simultaneously, the limb perfusion difference index matrix shows that the absolute value of ΔSpO2 (right upper limb, left upper limb) is greater than A, PI_ratio (right upper limb, left upper limb) is less than B, and the above state lasts for more than 30 seconds. When this condition is met, the engine determines a highly suspected type A aortic dissection involving the brachiocephalic trunk coupling anomaly and generates structured information including this inference, confidence assessment (calculated based on the degree and duration of parameter deviation from the threshold), and risk level (e.g., Level 1 warning). Another rule applies to newborns: normal ECG characteristics; ΔSpO2 (right upper limb, either lower limb) greater than 0.05; the engine determines a suspected ductus arteriosus-dependent congenital heart disease. The engine's output is sent to the comprehensive physiological state assessment and warning output module, which transforms the structured information generated by the CSP prediction algorithm into user-visual content, including text alerts, graphical difference matrix heatmaps, and suggested clinical interventions.

[0025] During the execution of the above method, the internal data flow and processing logic constitute a closed-loop, highly synchronous monitoring and diagnostic process. From the raw signal acquisition in S101 to the final warning output in S105, each step depends on the accurate output of the previous step, and a unified timestamp mechanism ensures strict alignment of multimodal data in the spatiotemporal dimensions. This design fundamentally solves the problem of regional perfusion information loss caused by single-point blood oxygen monitoring in existing technologies. By quantifying the perfusion differences among the limbs and correlating them with cardiac electrical activity, it achieves early and accurate identification of complex critical illnesses.

[0026] At the system level, to support the execution of the above methods, the integrated monitoring system of this invention includes the following necessary components: A multi-channel physiological signal synchronous acquisition module, whose hardware configuration is as described above, serves as the data entry point for the entire system. An ECG-oxygen data preprocessing unit, implemented as an embedded digital signal processor, internally equipped with the aforementioned wavelet denoising, notch filtering, and adaptive motion artifact elimination algorithms. A multi-dimensional feature parallel extraction unit, implemented as a multi-core microcontroller, with each core responsible for handling feature extraction tasks for specific leads or channels, ensuring real-time performance. A limb perfusion differential quantitative analysis unit, implemented as a dedicated computing module, responsible for performing matrix operations and threshold comparisons. An ECG-perfusion spatiotemporal correlation fusion diagnostic engine, implemented as a rule-based inference engine, with its knowledge base stored in non-volatile memory, which can be maintained and expanded through software updates. A comprehensive physiological state assessment and early warning output module, implemented as a human-computer interaction interface, including a high-resolution display screen and an audio alarm device, used to present the final diagnostic and early warning results to medical personnel. A CSP prediction algorithm submodule, a deep learning model based on the Transformer architecture, is a submodule of the integrated physiological state assessment and early warning output module. Designed specifically for multivariate time-series medical data, its core function is to capture the interaction relationships between different physiological variables (such as ECG characteristics and blood oxygenation parameters) through category association learning, thereby predicting the risk of future cardiovascular events (such as major adverse cardiovascular events, MACE). All these modules are interconnected via a high-speed internal bus, sharing a unified clock source and power management unit, collectively forming a fully functional and reliable integrated monitoring system. The system's design ensures the precise execution of methodological steps and provides clinicians with a novel diagnostic tool that transcends the limitations of traditional single-point monitoring.

[0027] The embodiments of the present invention have been described in detail above, but the content described is only a preferred embodiment of the present invention and should not be considered as limiting the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of this patent.

Claims

1. A comprehensive monitoring method integrating 12-lead electrocardiogram and pulse oximetry, characterized in that, include: A multi-channel physiological signal synchronous acquisition module is used to synchronously acquire the patient's 12-lead electrocardiogram signal timing data and the 4-channel pulse oxygen saturation and photoplethysmography pulse wave signal timing data acquired by 4 optical sensors fixed to the patient's extremities, and assign a unified high-precision timestamp to all acquired data streams. Using an ECG-blood oxygenation data preprocessing unit, baseline drift correction, power frequency interference filtering, and electromyographic noise suppression are performed on the 12-lead ECG signal time series data, and motion artifact identification and removal, and DC and AC component separation are performed on the 4-channel photoplethysmography pulse wave signal time series data. Using a multidimensional feature parallel extraction unit, heart rate, arrhythmia events, QRS complex width, ST segment offset and QT interval dispersion are extracted in real time from the pre-processed 12-lead ECG signal. Pulse oxygen saturation, pulse rate, perfusion index, pulse wave amplitude, rise time and tidal wave and diabetic wave morphological parameters are extracted independently for each limb channel from the pre-processed 4-channel photoplethysmography pulse wave signal. The perfusion and oxygenation characteristic parameter set of the limb channels is input into a limb perfusion difference quantitative analysis unit. Based on the preset mathematical operation rules, a limb perfusion difference index matrix is ​​calculated and generated in real time. The matrix includes the blood oxygen saturation difference between any two limbs, the perfusion index ratio, and the pulse arrival time difference based on the electrocardiogram R wave. The set of electrocardiographic feature parameters and the limb perfusion difference index matrix are input into an electrocardiogram-perfusion spatiotemporal correlation fusion diagnostic engine. The engine has a set of clinical diagnostic rule knowledge base embedded in it, which is used to perform spatiotemporal correlation matching between the change pattern of electrocardiographic feature parameters and the dynamic evolution of the limb perfusion difference index matrix, so as to identify predefined electro-hemodynamic coupling abnormal patterns pointing to specific pathological states. Based on the identification results of the abnormal electro-hemodynamic coupling pattern, structured early warning information containing specific pathological state inferences, confidence assessments, and risk level classifications is generated and presented visually through a user interface.

2. The integrated monitoring method for combining 12-lead electrocardiogram and pulse oximetry according to claim 1, characterized in that, The method involves simultaneously acquiring 12-lead ECG signal timing data and 4-channel pulse oximetry and photoplethysmography pulse wave signal timing data of the patient through a multi-channel physiological signal synchronous acquisition module, including: Ten silver-silver chloride patch electrodes are deployed in the patient's standard 12-lead position and connected to a preamplifier with a built-in 24-bit precision analog-to-digital converter to amplify the ECG signal differentially by at least 1,000 times and digitize it at a sampling frequency of 1,000 Hz. Four finger clip-on optical probes are fixed to the fingernails of the patient's right hand, left hand, right foot, and left foot respectively. Each probe integrates a red light-emitting diode with a center wavelength of 660nm, an infrared light-emitting diode with a center wavelength of 940nm, and a high-sensitivity silicon photodiode receiver. A time-division multiplexing drive circuit is used to drive each probe alternately, and the photocurrent signal is digitized at a sampling frequency of 200Hz. A master clock generator driven by a temperature-compensated crystal oscillator generates a synchronous clock signal, which is distributed to all analog-to-digital converters to ensure that all data sampling points have a unified time reference.

3. The integrated monitoring method for combining 12-lead electrocardiogram and pulse oximetry according to claim 2, characterized in that, The process of using an ECG-oxygenation data preprocessing unit to process ECG signals and photoplethysmography (PPG) pulse wave signals includes: The ECG signal is decomposed into six layers of wavelet based on dB6 wavelet basis function. The first to third layer detail coefficients are processed by soft thresholding to suppress electromyographic noise. The sixth layer approximation coefficients are reconstructed and subtracted from the original signal to correct baseline drift. Power frequency interference is filtered out by a second-order IIR notch filter with a center frequency of 50Hz or 60Hz. For the photoplethysmography (PPG) signal, an adaptive filter is constructed using fused triaxial accelerometer signals to remove motion artifacts. The AC component is extracted using a 0.5Hz to 10Hz bandpass filter, and the DC component is extracted using a 5s window moving average filter.

4. The integrated monitoring method for combining 12-lead electrocardiogram and pulse oximetry according to claim 3, characterized in that, The extraction of electrocardiographic feature parameter sets through a multi-dimensional feature parallel extraction unit includes: The Pan-Tompkins algorithm was used to detect QRS complexes in 12-lead ECG signals. The cardiac cycle window was defined with the R wave peak as the reference, and the key start and end points of the P wave, QRS complex and T wave were identified by morphological template matching and slope analysis. Based on the key start and end points, heart rate, PR interval, QRS duration, QT interval and ST segment offset are calculated. The ST segment offset is measured with the midpoint of the PR segment as the isopotential line reference, and the voltage offset value is measured 60ms after point J. Atrial fibrillation, premature ventricular contractions, and atrioventricular block events are identified through rhythm regularity analysis and morphological abnormality detection algorithms.

5. The integrated monitoring method for combining 12-lead electrocardiogram and pulse oximetry according to claim 4, characterized in that, The process involves independently extracting perfusion and oxygenation feature parameter sets for each limb channel using a multi-dimensional feature parallel extraction unit, including: The pulse oxygen saturation value is obtained by logarithmic transformation of the amplitude ratio of red light and infrared light AC components and mapping through a pre-stored calibration curve. The pulse rate is determined by the reciprocal of the time interval between the main peaks of adjacent pulse waves; The injection index is obtained by weighting the ratio of the AC component amplitude to the DC component amplitude by 100. Pulse wave amplitude, rise time, relative height and time interval between tidal wave and dicrotic wave were extracted as morphological parameters.

6. The integrated monitoring method for combining 12-lead electrocardiogram and pulse oximetry according to claim 5, characterized in that, The limb perfusion differential analysis unit generates a 4x4 symmetric matrix of limb perfusion differential indicators, whose off-diagonal elements include: Differences in blood oxygen saturation between the right upper limb and the left upper limb, the right upper limb and the left lower limb, the right upper limb and the right lower limb, the left upper limb and the left lower limb, the left upper limb and the right lower limb, and the left lower limb and the right lower limb; The ratio of perfusion index between corresponding limb pairs; Using the peak time of the R wave in the electrocardiogram as a benchmark, the time difference between the peak times of the main wave of the corresponding limb photoplethysmography pulse wave is taken as the pulse arrival time difference.

7. The integrated monitoring method for combining 12-lead electrocardiogram and pulse oximetry according to claim 6, characterized in that, The limb perfusion differential quantitative analysis unit continuously updates the limb perfusion differential index matrix and compares the value of each element in the matrix with a preset clinical normal fluctuation threshold. When the absolute value of any element exceeds the threshold, it is marked as abnormal.

8. The integrated monitoring method for combining 12-lead electrocardiogram and pulse oximetry according to claim 7, characterized in that, The ECG-perfusion spatiotemporal correlation fusion diagnostic engine's internally embedded clinical diagnostic rule knowledge base contains the following rules: When the electrocardiogram features sinus tachycardia without ST segment elevation or depression, and the limb perfusion difference index matrix shows that the absolute value of the difference in blood oxygen saturation between the right and left upper limbs is greater than A, the perfusion index of the right upper limb is lower than that of the left upper limb by B, and this state lasts for more than 30 seconds, it is judged as a highly suspected type A aortic dissection involving the brachiocephalic trunk coupling abnormality pattern. When the monitored subject is a newborn with normal electrocardiographic characteristics, and the difference between the blood oxygen saturation of the right upper limb and the blood oxygen saturation of either lower limb is greater than Kid C, it is judged as a suspected coupling abnormality pattern of ductus arteriosus-dependent congenital heart disease.

9. The integrated monitoring method for combining 12-lead electrocardiogram and pulse oximetry according to claim 8, characterized in that, The generated structured early warning information includes: The specific pathological state is inferred based on the results of coupling anomaly pattern recognition. The confidence score is calculated based on the degree and duration of parameter deviation from the threshold. The risk level is determined based on the clinical risk level classification standard, and level one to three warning signals are generated.

10. A comprehensive monitoring system integrating 12-lead electrocardiogram and pulse oximetry, characterized in that, include: A multi-channel physiological signal synchronous acquisition module is used to synchronously acquire 12-lead ECG signal timing data and 4-channel pulse oxygen saturation and photoplethysmography pulse wave signal timing data through 10 ECG electrodes deployed on the patient's body surface and 4 optical sensors fixed to the patient's extremities, and assign a unified high-precision timestamp to all acquired data streams. An ECG-blood oxygen data preprocessing unit is connected to the output of the multi-channel physiological signal synchronous acquisition module. It is used to perform baseline drift correction, power frequency interference filtering and electromyographic noise suppression on the received ECG signal time series data, and to perform motion artifact identification and removal, and DC component and AC component separation on the photoplethysmography pulse wave signal time series data. A multidimensional feature parallel extraction unit is connected to the output of the ECG-oxygen data preprocessing unit. It is used to extract heart rate, arrhythmia events, QRS complex width, ST segment offset and QT interval dispersion in real time from the preprocessed 12-lead ECG signal. It also independently extracts pulse oxygen saturation value, pulse rate, perfusion index, pulse wave amplitude, rise time and tidal wave and dicrotic wave morphological parameters for each limb channel from the preprocessed 4-channel photoplethysmography pulse wave signal. A limb perfusion difference quantitative analysis unit is connected to the output of the multidimensional feature parallel extraction unit. It is used to receive perfusion-related feature parameters of the limb channels and calculate and generate a limb perfusion difference index matrix in real time based on preset mathematical operation rules. The matrix includes the blood oxygen saturation difference between any two limbs, the perfusion index ratio, and the pulse arrival time difference based on the electrocardiogram R wave. An electrocardiogram-perfusion spatiotemporal correlation fusion diagnostic engine has its input terminals connected to the output terminals of the multidimensional feature parallel extraction unit and the limb perfusion differential quantitative analysis unit, respectively. The engine has a set of clinical diagnostic rule knowledge base embedded in it, which is used to perform spatiotemporal correlation matching between the change pattern of electrocardiogram feature parameters and the dynamic evolution of the limb perfusion differential index matrix, so as to identify predefined electro-hemodynamic coupling abnormal patterns pointing to specific pathological states. A comprehensive physiological state assessment and early warning output module is connected to the output of the ECG-perfusion spatiotemporal correlation fusion diagnostic engine. It is used to process the input through the CSP prediction algorithm submodule based on the coupled abnormal pattern recognition results output by the engine, generate structured early warning information containing specific pathological state inference, confidence assessment and risk level classification, and present it visually through the user interface. The processing logic of the CSP prediction algorithm submodule includes: Step 1: The input layer receives the output from the multi-dimensional feature parallel extraction unit, including 12-lead ECG features. The input data is converted into time series tokens, and each token corresponds to the sequence of a variable within a time window. Step 2: Using a category-related attention mechanism, dynamic weights are assigned to different variables through variable-level self-attention and cross-attention calculations; Step 3: Use an encoder-decoder structure to process long-term data, preserve temporal sequence information through positional encoding, and output the probability of case events occurring within the predicted time period; Step 4: The output layer generates prediction results and merges them with the original rule base. The final diagnosis is generated through weighted voting. The training uses the Focal Loss loss function to solve the data imbalance problem. The multi-channel physiological signal synchronous acquisition module includes: An ECG signal acquisition submodule includes ten silver-silver chloride patch electrodes, shielded leads, a preamplifier, and a 24-bit precision analog-to-digital converter, used to differentially amplify and digitize ECG signals at a sampling frequency of 2000Hz; A 4-channel pulse oximetry acquisition submodule includes 4 finger clip optical probes, a time-division multiplexing drive circuit, a transimpedance amplifier, and a 16-bit precision analog-to-digital converter, used to digitize the photocurrent signal at a sampling frequency of 200Hz; A master clock generator driven by a temperature-compensated crystal oscillator is used to generate a synchronous clock signal and distribute it to all analog-to-digital converters.

Images