Systems and methods for non-invasive cardiovascular health assessment

Abstract

A computer-implemented method for non-invasive cardiovascular health assessment includes receiving, at a processor, physiological signals of a subject, where the physiological signals include at least a cardiac ECG signal (electrical signal) and a PPG signal (photoplethysmography signal); processing the ECG signal and the PPG signal through an AI model (artificial intelligence model) executed by the processor; and estimating, by the AI model, one or more demographic and anthropometric parameters based on the processing of the ECG and PPG signals.

Description

Technical Field

[0001] This disclosure relates to a system and method for noninvasive cardiovascular health assessment. This disclosure may also relate to a system and method for noninvasive cardiovascular health assessment based on ECG and PPG signals. Background Technology

[0002] Cardiovascular health remains a central challenge in contemporary medicine, as it directly impacts life expectancy and quality of life. Accurate estimation of demographic and anthropometric characteristics such as age, sex, height, weight, and body mass index (BMI) is crucial not only for effective clinical surveillance but also for early detection of cardiovascular disease (CVD) risk, individualized prevention strategies, and customized treatment plans. Despite the importance of these parameters, current methods for obtaining them are limited. Manual anthropometry is susceptible to human error, operator variability, and equipment-specific inconsistencies, making it unreliable in large-scale or telemedicine settings. Invasive or lab-based measurements, while accurate, are impractical for routine monitoring due to their associated discomfort, risks, and time constraints.

[0003] Even when electrocardiograms (ECG) and photoplethysmograms (PPG) are used in practice, their applications are primarily limited to blood pressure (BP) estimation, arrhythmia detection, or pulse analysis, lacking an established framework for demographic or anthropometric predictions. Furthermore, many existing artificial intelligence (AI) systems in healthcare suffer from bias and fairness issues, leading to uneven predictive quality across different sexes, ages, or population subgroups. Therefore, an improved system is needed to address the shortcomings of conventional methods used to determine demographic and anthropometric parameters. Summary of the Invention

[0004] This disclosure aims to provide a system and method for noninvasive cardiovascular health assessment that will overcome or at least substantially improve some of the shortcomings of the prior art, or at least provide an alternative.

[0005] According to a first aspect of this disclosure, a computer-implemented method for noninvasive cardiovascular health assessment is provided. The method includes: At the processor, physiological signals from the subject are received, wherein the physiological signals include at least ECG signals and PPG signals; The AI ​​model executed by the processor processes the ECG signal and the PPG signal; and Using the AI ​​model, one or more demographic and anthropometric parameters are estimated based on the processing of the ECG and PPG signals.

[0006] In one example, the processing steps include: The received ECG and PPG signals are preprocessed. Feature extraction is performed on the preprocessed ECG and PPG signals; and Select one or more optimal features; and The demographic and anthropometry parameters are estimated based on one or more of the optimal features.

[0007] In one example, the processing steps further include performing time processing on the preprocessed ECG and PPG signals.

[0008] In one example, the time processing includes capturing the temporal correlation and spatial relationship between the one or more optimal features.

[0009] In one example, the demographic parameters include age and sex, and the one or more anthropometric characteristics include weight, height, and BMI.

[0010] In one example, the method further includes: Using the AI ​​model, based on the estimated demographic and anthropometric parameters, the subject's cardiovascular age is calculated; and Output the cardiovascular age, and / or compare the calculated cardiovascular age with the subject's physiological age to identify the risk of vascular aging.

[0011] In one example, the method further includes: tracking changes in the estimated BMI or weight over time to detect abnormal trends, and presenting the changes in the estimated BMI or weight on a user interface.

[0012] In one example, the method further includes: Generate visual indicators to display estimates of the demographic and anthropometric parameters; and Present the visual indicator on the user interface; and The visual indicators include one or more of the following: saliency maps, heatmaps, or attention heatmaps.

[0013] In one example, the AI ​​model includes a recurrent neural network (RNN) model, wherein the RNN model is trained to perform steps of processing received ECG and PPG signals and output estimates of one or more demographic and anthropometric parameters.

[0014] In one example, the method may optionally apply an ensemble approach to combine predictions from different model architectures. This can improve robustness and reliability.

[0015] According to another aspect of this disclosure, a system for noninvasive cardiovascular health monitoring is provided. The system includes: A computing device including a processor and a storage unit operatively coupled; A user interface that is operatively coupled to or integrated into the computing device; The storage unit is used to store an AI model, which can be executed by the processor. The storage unit also includes instructions that, when executed by the processor, cause the processor to perform the following operations: Receive physiological signals from the subject, wherein the physiological signals include at least ECG signals and PPG signals; The AI ​​model is used to process the ECG signal and the PPG signal. Using the AI ​​model, based on the processing of the ECG and PPG signals, one or more demographic and anthropometric parameters are estimated; and The estimated demographic and anthropometric parameters are displayed on the user interface.

[0016] In one example, the user interface may be a screen, such as an LED or LCD screen. Alternatively, the screen may be a touchscreen.

[0017] In one example, the processor is programmed to: The received ECG and PPG signals are preprocessed. Feature extraction is performed on the preprocessed ECG and PPG signals; and Select one or more optimal features; and The demographic and anthropometry parameters are estimated based on one or more of the optimal features.

[0018] In one example, the processor is programmed to perform timing processing on the preprocessed ECG and PPG signals.

[0019] In one example, during time processing, the processor is programmed to capture the temporal correlations and spatial relationships between the one or more optimal features.

[0020] In one example, the demographic parameters include age and sex, and the one or more anthropometric characteristics include weight, height, and BMI.

[0021] In one example, the processor is programmed to: By applying the AI ​​model, the cardiovascular age of the subject is calculated based on the estimated demographic and anthropometric parameters; and Output the cardiovascular age, and / or compare the calculated cardiovascular age with the subject's physiological age to identify the risk of vascular aging.

[0022] In one example, the processor is programmed to track changes in the estimated BMI or weight over time to detect abnormal trends and to present the changes in the estimated BMI or weight on a user interface.

[0023] In one example, the processor is also programmed to: Generate visual indicators (i.e., visual markers) to demonstrate estimates of the demographic and anthropometric parameters; The visual indicator (i.e., visual marker) is presented on the user interface; and The visual indicators include one or more of the following: saliency maps, heatmaps, or attention heatmaps.

[0024] In one example, the visual indicator may be displayed on the user interface.

[0025] In one example, the AI ​​model includes an RNN model, wherein the RNN model is trained to perform steps of processing received ECG and PPG signals and output estimates of one or more demographic and anthropometric parameters.

[0026] In one example, the AI ​​model is pre-trained on a dataset of unlabeled ECG signals, PPG signals, or both using a self-supervised learning framework, wherein the self-supervised learning framework is a masked self-supervised learning framework, segments of the unlabeled signals are masked, and the AI ​​model is trained to reconstruct the masked segments.

[0027] In one example, the AI ​​model may further include: An encoder, trained to extract features from the ECG and / or PPG signals; and One or more task-specific classification heads are used to perform regression to estimate anthropometric parameters and / or to perform classification to estimate demographic parameters; The encoder is coupled to one or more task-specific classification heads such that the output of the encoder is the input of each classification head, and the encoder is located upstream of each classification head.

[0028] In one example, the encoder can be pre-trained. In another example, the encoder can be a PatchTST encoder.

[0029] In one example, the AI ​​model can be pre-trained using a self-supervised learning method. This pre-training provides general feature representations that can be transferred to multiple downstream tasks without requiring complete retraining.

[0030] In one example, the AI ​​model may optionally include an integrated architecture that combines multiple predictions from different model architectures, wherein the multiple model architectures can define the AI ​​model. This can improve robustness and reliability.

[0031] According to another aspect of this disclosure, a wearable device for non-invasive cardiovascular health monitoring is provided. The wearable device includes: A sensor array for continuously acquiring ECG and PPG signals from the user; and A computing device including a processor and a storage unit operatively coupled; The computing device is operatively coupled to the sensor array; The storage unit is used to store AI models, wherein the AI ​​models are executable by the processor. The storage unit also includes instructions that, when executed by the processor, cause the processor to perform the following operations: Receive physiological signals from the subject, wherein the physiological signals include at least ECG signals and PPG signals; The AI ​​model is used to process the ECG signal and the PPG signal; and Using the AI ​​model, one or more demographic and anthropometric parameters are estimated based on the processing of the ECG and PPG signals.

[0032] In one example, the wearable device may include a user interface, or the wearable device may be used to communicate with a user interface. The one or more estimated demographic parameters and the one or more estimated anthropometric parameters may be presented on the user interface.

[0033] According to another aspect of this disclosure, an AI model for noninvasive cardiovascular health assessment is provided, which is used in the method described in the first aspect. The AI ​​model includes: Machine learning (ML) models or deep learning (DL) models used for feature extraction and selection of optimal features; and A recurrent neural network or transformer for time processing of the optimal features. The AI ​​model is trained to process ECG and PPG signals; and The AI ​​model is trained to estimate one or more demographic and anthropometric parameters based on the processing of the ECG and PPG signals.

[0034] In one example, the AI ​​model includes an integration of multiple deep learning models with different architectures, wherein the estimates of the demographic and anthropometry parameters are generated by combining the outputs of the multiple deep learning models.

[0035] In one example, the AI ​​model is trained on a dataset of unlabeled ECG and PPG signals using a self-supervised learning framework, wherein the self-supervised learning framework is a masked self-supervised learning framework, segments of the unlabeled signals are masked, and the AI ​​model is trained to reconstruct the masked segments.

[0036] In one example, the AI ​​model includes: An encoder, trained to extract features from the ECG and / or PPG signals; and One or more task-specific classification heads are used to perform regression to estimate anthropometric parameters and / or to perform classification to estimate demographic parameters; The encoder is coupled to one or more task-specific classification heads such that the output of the encoder is the input of each classification head, and the encoder is located upstream of each classification head.

[0037] In one example, the encoder can be pre-trained. In another example, the encoder can be a PatchTST encoder.

[0038] According to another aspect of this disclosure, a data processing device for non-invasive cardiovascular health monitoring is provided. The data processing device includes means for performing the methods described herein.

[0039] According to another aspect of this disclosure, a computer program is provided. The computer program includes instructions, wherein, when the program is executed by a computer, the instructions cause the computer to perform the method according to the first aspect (i.e., the method described in the first aspect).

[0040] According to another aspect of this disclosure, a computer-readable medium is provided. The computer-readable medium includes instructions, wherein, when executed by a computer, the instructions cause the computer to perform the method according to the first aspect.

[0041] According to another aspect of this disclosure, a system for noninvasive cardiovascular health assessment is provided. The system includes: One or more processors; and A memory storing instructions that, when executed by the one or more processors, cause the system to perform the following operations: Receive physiological signals from the subject, wherein the physiological signals include at least one ECG signal and a PPG signal; The received ECG and PPG signals are preprocessed to reduce noise and normalize the signals; The preprocessed signal is input into an AI model, the AI ​​model including a deep learning architecture; and The AI ​​model generates estimates of multiple demographic and anthropometric parameters of the subject, including age, sex, height, weight, and BMI, based on a detailed analysis of the temporal and spatial patterns within the preprocessed signal.

[0042] In one example, the deep learning architecture is a recurrent neural network (RNN).

[0043] In one example, the deep learning architecture is based on the Transformer architecture.

[0044] In one example, the AI ​​model is pre-trained on a dataset of unlabeled ECG signals, PPG signals, or both using a self-supervised learning framework.

[0045] In one example, the self-supervised learning framework is a masked self-supervised learning framework, wherein segments of the unlabeled signal are masked, and the AI ​​model is trained to reconstruct the masked segments.

[0046] In one example, the AI ​​model includes an integration of multiple deep learning models with different architectures, and the estimate is generated by combining the outputs of the multiple deep learning models.

[0047] In one example, the instructions also cause the system to integrate additional sensor data with the ECG and PPG signals, the additional sensor data being selected from a group consisting of respiratory signals, oxygen saturation (SpO2) signals, and accelerometer signals.

[0048] In one example, the instructions also cause the system to implement an Explainable AI (XAI) module to provide an interpretable basis for the generated estimates, wherein the XAI module generates at least one of a saliency map or an attention heatmap.

[0049] In one example, the system is integrated into a wearable device, and the instructions also cause the system to perform real-time on-device inference to generate the estimate.

[0050] In one example, the AI ​​model is optimized using one or more compression techniques selected from the group consisting of model quantization, pruning, and bias-only fine-tuning for deployment on the wearable device.

[0051] In one example, the instructions also cause the system to perform longitudinal tracking by repeatedly generating the estimates over time to monitor changes in at least one of the subject's estimated weight, BMI, or calculated cardiovascular age.

[0052] According to another aspect of this disclosure, a method for noninvasive cardiovascular health assessment is provided. The method includes: Physiological signals of a subject are received at one or more processors, the physiological signals including at least one ECG signal and a PPG signal; The received ECG and PPG signals are preprocessed by the one or more processors to ensure signal quality and consistency. The temporal and spatial features extracted from the preprocessed signal are analyzed using an AI model executed on one or more processors; and The AI ​​model estimates multiple demographic and anthropometric parameters of the subject, including age, sex, height, weight, and BMI.

[0053] In one example, the method further includes: pre-training the AI ​​model on a dataset of unlabeled physiological signals using a self-supervised learning framework prior to the analysis step.

[0054] In one example, the AI ​​model includes a Transformer-based architecture for capturing time-dependent patterns in the preprocessed signal.

[0055] In one example, the method further includes training and updating the AI ​​model using a federated learning framework to protect data privacy across multiple data sources.

[0056] In one example, the method further includes applying an ethical bias check mechanism to the AI ​​model to promote fairness and mitigate prediction discrepancies across different population subgroups.

[0057] In one example, the method further includes: Based on the estimated demographic and anthropometric parameters, the cardiovascular age of the subject was calculated; and The calculated cardiovascular age is compared with the subject's physiological age to identify the risk of vascular aging.

[0058] In one example, the method can be performed by a telemedicine platform to facilitate remote patient monitoring.

[0059] In one example, the method includes generating personalized healthcare recommendations for the subject based on the estimated parameters.

[0060] According to another aspect of this disclosure, a system for noninvasive cardiovascular health assessment is provided. The system includes: A method for collecting pre-processed ECG and PPG signals from subjects; Artificial intelligence-based methods that integrate machine learning / deep learning (ML / DL) with recurrent neural network (RNN) or Transformer architecture modules for feature extraction and temporal learning; and The deep learning model predicts the subject's age, gender, height, weight, and BMI based on the preprocessed signal, wherein the architecture of the deep learning model is capable of detailed analysis of temporal and spatial signal patterns.

[0061] In one example, the system helps maintain the quality, alignment, and consistency of ECG and PPG signals acquired from available online CVD datasets. Furthermore, the hybrid machine learning / deep learning (ML / DL) model can extract features from these signals.

[0062] In one example, the DL / ML model includes an RNN component or a Transformer to capture temporal and spatial relationships within the signal. This combination enables the system to make thorough estimates of cardiovascular parameters.

[0063] In one example, the system also includes an ensemble learning mechanism in which multiple deep AI models with different architectures are combined to improve the accuracy and robustness of age, gender, height, weight, and BMI estimates.

[0064] In one example, the system can be enhanced with an ethical bias checking mechanism to promote fairness and ensure clinically reliable parameter estimation. Furthermore, the system addresses potential discrepancies in predictions.

[0065] In one example, the system can be integrated into existing healthcare systems, telemedicine platforms, and wearable health devices for real-time cardiovascular health assessment and remote patient monitoring, thereby promoting proactive healthcare management.

[0066] In one example, the system could demonstrate its effectiveness through clinical trials, obtain regulatory approval as a non-invasive cardiovascular health assessment technology, thereby advancing medical diagnostics and improving patient outcomes.

[0067] According to another aspect of this disclosure, a system for noninvasive cardiovascular health assessment is provided. The system includes: A method for collecting pre-processed ECG and PPG signals from subjects; AI-based methods that combine ML / DL approaches with RNN or transformer architectures for feature extraction and temporal learning; and The DL model predicts the subject's age, gender, height, weight, and BMI based on the preprocessed signal, wherein the architecture of the DL model is capable of detailed analysis of temporal and spatial signal patterns.

[0068] In one example, the quality, alignment, and consistency of ECG and PPG signals obtained from an available online CVD dataset are ensured, and the hybrid ML / DL model extracts features from the signals.

[0069] In one example, the DL / ML model includes RNN components or transformers to capture temporal correlations and spatial relationships within the signal. This combination enables the system to make thorough estimates of cardiovascular parameters.

[0070] In one example, the system includes an ensemble learning mechanism in which multiple deep AI models with different architectures are combined to improve the accuracy and robustness of age, gender, height, weight, and BMI estimates.

[0071] In one example, the system can be enhanced with an ethical bias checking mechanism to promote fairness and ensure clinically reliable parameter estimation. Furthermore, the system addresses potential discrepancies in predictions.

[0072] In one example, the system can be integrated into existing healthcare systems, telemedicine platforms, and wearable health devices for real-time cardiovascular health assessment and remote patient monitoring, thereby promoting proactive healthcare management.

[0073] In one example, the AI ​​model could be validated through clinical trials to obtain regulatory approval as a non-invasive cardiovascular health assessment technology, thereby advancing medical diagnostics and improving patient outcomes.

[0074] In one example, the AI ​​model is pre-trained on unlabeled ECG and / or PPG signals using a self-supervised learning framework, and then evaluated for demographic parameter estimation.

[0075] In one example, the AI ​​model includes an XAI module, comprising saliency maps, attention visualizations, or equivalent interpretability methods to provide clinicians with interpretable reasoning for each estimate.

[0076] One example includes a longitudinal tracking mechanism to monitor changes in estimated weight, BMI, or cardiovascular age over time, enabling early detection of abnormal health trends.

[0077] In one example, the model is optimized for deployment on wearable devices using compression techniques, including model quantization, pruning, or bias-only fine-tuning, enabling real-time on-device inference.

[0078] In one example, a federated learning framework is used to perform model training and updates, thereby protecting data privacy while aggregating knowledge across multiple devices or institutions.

[0079] In one example, additional sensor modalities such as breathing, SpO2, or accelerometer signals are integrated with ECG and PPG to improve the robustness of demographic estimation in dynamic or wearable settings.

[0080] The term “include” (and its grammatical variations) as used in this article is used in the open sense of “having” or “containing”, rather than in the closed sense of “consisting of only”.

[0081] It should be understood that if any prior art information is cited herein, such citation does not imply an admission that such information is part of common general knowledge in the field in Australia or any other country. Attached Figure Description

[0082] Specific embodiments of this disclosure will now be described by way of example with reference to the accompanying drawings.

[0083] Figure 1A schematic diagram of an exemplary embodiment of a system for noninvasive cardiovascular health assessment is shown.

[0084] Figure 2 A flowchart illustrating an exemplary embodiment of a method for noninvasive cardiovascular health assessment is shown.

[0085] Figure 3 A flowchart of another exemplary embodiment of a method for noninvasive cardiovascular health assessment is shown.

[0086] Figure 4 A schematic diagram of an exemplary wearable device for non-invasive cardiovascular health assessment or cardiovascular monitoring is shown.

[0087] Figure 5 An exemplary architecture of an AI model used in noninvasive cardiovascular health assessment is shown.

[0088] Figure 6 An exemplary architecture of the PatchTST encoder in an AI model is shown.

[0089] Figure 7 The training and validation process of an AI model for non-invasive cardiovascular health assessment is shown.

[0090] Figure 8 Further alternative embodiments of the non-invasive cardiovascular health assessment method are shown. Detailed Implementation

[0091] Cardiovascular health is the cornerstone of overall health, and accurate estimation of demographic and anthropometric parameters such as age, sex, height, weight, and BMI plays a crucial role in effective healthcare management. Conventional methods for obtaining these parameters are often invasive, time-consuming, or error-prone, highlighting the need for non-invasive alternatives.

[0092] The rapid development of wearable technologies such as smartwatches, chest straps, and fingertip PPG sensors has made continuous and non-invasive cardiovascular monitoring both feasible and scalable. Accurate demographic estimation using ECG and PPG has significant clinical implications. Age is directly associated with atherosclerosis and cardiovascular disease; abnormal weight, BMI, or height patterns may predict an increased risk of hypertension, obesity, and metabolic diseases; and gender differences are known to affect cardiovascular outcomes. The increasing availability of large, publicly available ECG and PPG datasets enables robust data-driven modeling.

[0093] This disclosure relates to a system and method for noninvasive cardiovascular health monitoring. This disclosure may also relate to a wearable device. The device includes an integrated computing device capable of enabling noninvasive cardiovascular health monitoring. This disclosure relates to a system, method, and / or wearable device that incorporates machine learning or other AI methods to perform cardiovascular health assessment by estimating demographic and / or anthropometric characteristics from signals measured noninvasively.

[0094] Figure 1 An example of a system 100 for noninvasive cardiovascular health monitoring is shown. (See reference...) Figure 1 A system 100 for noninvasive cardiovascular health monitoring includes: a computing device 200 comprising a processor 202 and storage units 204, 206 operably coupled to each other; and a user interface 212 operably coupled to or integrated into the computing device 200. Storage units 204, 206 are used to store an AI model 300, wherein the AI ​​model is executable by the processor 202. Storage units 204, 206 also include instructions that, when executed by the processor 202, cause the processor to perform the following operations: receive physiological signals from a subject, wherein the physiological signals include at least ECG and PPG signals; process the ECG and PPG signals using the AI ​​model 300; estimate one or more demographic and anthropometric parameters using the AI ​​model based on the processing of the ECG and PPG signals; and display the estimated demographic and anthropometric parameters on the user interface 212.

[0095] System 100 combines ECG and PPG signals with advanced AI technology to predict key demographic characteristics with high accuracy. This approach employs signal preprocessing for noise reduction and normalization, followed by machine learning and deep learning architectures applied in the AI ​​model. These architectures may include recurrent neural networks and transformer models for automated feature extraction and temporal learning.

[0096] Age is directly related to the occurrence of arteriosclerosis and cardiovascular disease. Demographic and anthropometric parameters estimated by AI model 300 provide indicators of the subject's cardiovascular health. Abnormal weight, BMI, or height patterns may predict an increased risk of cardiovascular disease, hypertension, or other conditions, and generally indicate poor cardiovascular health. This system can accurately, in real-time, and non-invasively predict these characteristics, thus enabling non-invasive assessment of the subject's cardiovascular health.

[0097] System 100 may include one or more sensors. For example... Figure 1As shown in the example configuration, system 100 may include an ECG sensor 102 and a PPG sensor 104. Sensors 102 and 104 are electrically communicatively connected to computing device 200. Computing device 200 may receive ECG signals and PPG signals from sensors 102 and 104, respectively. The ECG sensor may include a chest strap or multiple electrodes. The PPG sensor may be a fingertip PPG sensor or a wrist-worn sensor, such as a pulse oximeter.

[0098] The computing device 200 can be implemented using a computer with a suitable user interface. The computing device 200 can be implemented using any computing architecture, including: a portable computer, tablet computer, personal computer (PC), smart device, Internet of Things (IoT) device, edge computing device, client / server architecture, "dumb" terminal / host architecture, cloud computing-based architecture, or any other suitable architecture. The computing device 200 can be appropriately programmed to implement a method for non-invasive cardiovascular health monitoring as described herein.

[0099] The computing device 200 may be implemented as a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic component, discrete gate or transistor logic, discrete hardware component, or any combination thereof intended to perform the functions described herein. The computing device 200 may be programmed to implement a method for non-invasive cardiovascular health monitoring as described herein.

[0100] A general-purpose processor can be a microprocessor, but alternatively, it can be any conventional processor, controller, microcontroller, circuit, and / or state machine. A processor can also be implemented as a combination of computing components, such as a combination of a DSP and a microprocessor, several microprocessors, one or more microprocessors combined with a DSP core, or any other such configuration. In one example, computing device 200 can be implemented as a microprocessor.

[0101] The computing device 200 can be programmed to estimate one or more demographic parameters and / or anthropometry parameters. In one example, the demographic parameters include age and sex, and the one or more anthropometry characteristics include weight, height, and BMI.

[0102] like Figure 1The diagram illustrates a system 100 for noninvasive cardiovascular health assessment or monitoring. The system includes or can be implemented by a computing device 200. Figure 1 A schematic diagram of the components of the computing device 200 is further shown.

[0103] In the illustrated example, computing device 200 includes suitable components necessary for receiving, storing, and executing corresponding computer instructions. These components may include processor 202 (i.e., processing unit 202). These components may include one or more of a central processing unit (CPU), a math co-processing unit (MathProcessor), a graphics processing unit (GPU), or a tensor processing unit (TPU) for tensor or multidimensional array computation or manipulation operations. Computing device 200 also includes read-only memory (ROM) 204, random access memory (RAM) 206, and input / output devices such as disk drive 208, and input devices 210 such as Ethernet ports, USB ports, etc.

[0104] The computing device 200 may also include a user interface 212. The user interface 212 may be a display 212. The display 212 may be a liquid crystal display, a light-emitting display, or any other suitable display. The user interface 212 may be a screen, such as an LED or LCD screen. Optionally, the screen may be a touchscreen. The computing device 200 may also include a communication link 114. The user interface 212 may be integrated into the computing device or may be a separate remote interface. In the case of remote configuration, the processor 202 may be operatively coupled to the user interface 212 at least via a wired or wireless connection to enable data communication.

[0105] The computing device 200 may include instructions that can be stored in ROM 204, RAM 206, or disk drive 208 and executed by processor 202. Multiple communication links 214 may be provided, which may be connected differently to one or more computing devices, such as servers, personal computers, terminals, wireless or handheld computing devices, IoT devices, smart devices, and edge computing devices. At least one of the multiple communication links 214 may be connected to an external computing network via a telephone line or other types of communication link.

[0106] The computing device 200 may include storage devices such as disk drive 208, which may include solid-state drives, hard disk drives, optical drives, tape drives, or remote or cloud-based storage devices. The computing device 200 may use a single disk drive or multiple disk drives, or remote storage services. The computing device 200 may also have a suitable operating system residing in its storage units.

[0107] The computing device 200 may also include one or more databases for storing one or more types of data. For example, the computing device 200 may include a training data database 220, such as ECG and PPG signals with corresponding demographic and anthropometric labels (e.g., age, sex, height, weight, and BMI). These datasets may be publicly available cardiovascular health assessment datasets containing ECG and PPG signals. These datasets can provide a baseline truth for the AI ​​model during training to learn the correlation between biophysical signals and individual characteristics. The training datasets can be periodically updated with additional data to improve model performance.

[0108] The computing device 200 can also provide the computing power necessary to operate or interface with the AI ​​model 300. The AI ​​model 300 can be a machine learning network, such as a neural network, to provide various functions and outputs. The AI ​​model 300 can be implemented locally, or it can be accessed or partially accessed via a server or cloud-based service. The AI ​​model 300 can also be untrained, partially trained, or fully trained, and / or can be retrained, modified, or updated over time. Training can be updated using new training datasets.

[0109] Computing device 200 may include one or more GPUs operatively coupled to a CPU (i.e., processor). Computing device 200 may also include further hardware elements operatively connected to the CPU and / or GPU to provide the computing device components required to implement AI models such as machine learning networks or machine learning models. AI model 300 may be stored in a memory unit (e.g., ROM). Details regarding AI model 300 are described below.

[0110] Processor 202 is programmed to preprocess the received ECG and PPG signals, extract features from the preprocessed ECG and PPG signals, and select one or more optimal features. Processor 202 is also programmed to estimate demographic and anthropometric parameters based on one or more optimal features. Processor 202 and computing devices are used to estimate demographic and anthropometric parameters based on these optimal features.

[0111] In one example, demographic parameters include age and sex, and one or more anthropometric features include weight, height, and BMI. Processor 202 can execute AI model 300 to estimate demographic and anthropometric parameters. AI model 300 can be trained to estimate other demographic and anthropometric parameters.

[0112] Processor 202 can also be programmed to perform time processing on the preprocessed ECG and PPG signals. AI model 300 can be trained to perform time processing on the ECG and PPG signals. In one example, during time processing, the processor is programmed to capture the temporal correlations and spatial relationships between one or more optimal features.

[0113] The processor 202 can also be programmed to calculate the cardiovascular age for the subject by applying an AI model based on estimated demographic and anthropometric parameters. The processor 202 can also be programmed to output the cardiovascular age and / or compare the calculated cardiovascular age with the subject's chronological age to identify the risk of vascular aging.

[0114] In one example, processor 200 can be programmed to track changes in estimated BMI or weight over time to detect abnormal trends and display these changes on a user interface. Furthermore, the processor can be further programmed to generate visual indicators (i.e., visual markers) to illustrate the estimates of demographic and anthropometric parameters. Processor 202 can also generate additional visual indicators to display changes in demographic and anthropometric parameters.

[0115] The processor 202 can also be programmed to provide one or more of the following functions: longitudinally tracking health parameters to monitor aging processes or BMI fluctuations; comparing cardiovascular age with physiological age to enable early identification of vascular aging and disease risk; or early detection of comorbidities, including hypertension, obesity, and CVD susceptibility, based on demographic trends. The AI ​​model 300 can be trained to output one or more of the described functions.

[0116] Processor 202 can also be programmed to display these functions as visual indicators. Processor 202 can be programmed to render this visual indicator (i.e., visual marker) on the user interface. Visual indicators include one or more of the following: saliency maps, heatmaps, or attention heatmaps. One or more visual indicators can be displayed on the user interface.

[0117] System 100 can be integrated into real-world healthcare environments, including, for example, wearable or mobile devices, telemedicine platforms, or cloud-based or federated learning frameworks. Wearable and mobile devices can include, for example, smartwatches, chest straps, or fingertip PPG devices for continuous monitoring. Integrating the system as part of a telemedicine platform enables remote patient assessment and proactive care services. Integrating System 100 into a privacy-preserving, HIPAA-compliant cloud-based or federated learning framework can achieve population-level improvements without compromising patient data.

[0118] Figure 2 An example of a computer-implemented method 400 for noninvasive cardiovascular health assessment is shown. Method 400 may begin at step 402. Step 402 includes: receiving physiological signals from subject 10 at a processor. The physiological signals include at least ECG and PPG signals. Step 404 includes: processing the ECG and PPG signals by an AI model 300 executed by processor 202. Step 406 includes: estimating one or more demographic and anthropometric parameters based on the processing of the ECG and PPG signals using the AI ​​model. Method 300 may be executed by processor 202 of computing device 200. Computing device 200 may receive ECG and PPG signals from EEG sensor 102 and PPG sensor 104.

[0119] Figure 3 A further example of a computer-implemented method 500 for noninvasive cardiovascular health assessment is shown. Method 500 begins at step 502. Step 502 includes receiving an ECG signal and a PPG signal. Step 504 includes preprocessing the received ECG and PPG signals. Signal preprocessing may include noise reduction, alignment of the ECG and PPG signals (i.e., waveforms), and normalization to ensure cross-device consistency of sensor signals from various sensor types. Signal preprocessing can enhance the quality of the received signals.

[0120] Preprocessing may include: segmenting the raw ECG and / or PPG signals into fixed-length windows, resampling to achieve consistent frequencies across different sources, and normalizing to mitigate device-to-device variability.

[0121] Step 506 includes feature extraction from the preprocessed ECG and PPG signals. Feature extraction can be performed by AI model 300. In one example, the AI ​​model can apply ML and DL algorithms to identify waveform features related to demographic characteristics.

[0122] Step 508 includes selecting one or more optimal features from the feature extraction. In one example, the AI ​​model 300 may be programmed to select one or more optimal features. The AI ​​model 300 may employ machine learning or deep learning methods to automatically extract meaningful features from the input signal.

[0123] Step 510 includes applying an AI model (e.g., machine learning) to the optimal features. In one example, temporal learning may be applied by AI model 300. In this example, AI model 300 may use one or more RNN or converter architectures to capture dynamic, time-dependent patterns in the signal. Optionally, an ensemble approach may be performed by AI model 300 to combine predictions from different model architectures, thereby improving robustness and reliability. In one example, temporal processing may include capturing the temporal correlations and spatial relationships between one or more optimal features.

[0124] Step 512 includes estimating demographic and anthropomorphic parameters based on optimal features. The AI ​​model 300 can be trained to estimate these parameters. The estimated demographic and anthropomorphic parameters can be displayed on the user interface 212. The estimated parameters are a good indicator of the subject's cardiovascular health.

[0125] Method 500 can also be used for other clinical applications beyond demographic and anthropometric parameter estimation. Optionally, the method may include additional steps 514 and 516. Step 514 may include: calculating the cardiovascular age for the subject based on the estimated demographic and anthropometric parameters using an AI model. Step 516 may include: outputting the cardiovascular age and / or comparing the calculated cardiovascular age with the subject's chronological age to identify the risk of vascular aging. Method 500 may also include, in step 518: longitudinal tracking of health parameters to monitor aging progress or BMI fluctuations.

[0126] In step 516, the method may include comparing cardiovascular age with physiological age. Steps 516 and / or 518 can identify vascular aging and disease risk at an early stage. Method 500 can also detect comorbidities, including hypertension, obesity, and CVD susceptibility, at an early stage based on demographic trends. Steps 514-518 may be optional steps.

[0127] In one example, method 500 may include step 520. Step 520 may include: tracking changes in the estimated BMI or weight over time to detect abnormal trends and presenting the changes in the estimated BMI or weight on a user interface.

[0128] Optionally, method 500 may include step 22: generating a visual indicator (i.e., a visual marker) to display estimates of demographic and anthropometric parameters, and presenting the visual indicator on user interface 212. The visual indicator may include one or more of the following: a saliency map, a heatmap, or an attention heatmap.

[0129] Optionally, computing device 200 can be used to implement one or more XAI mechanisms or XAI models. XAI models can be incorporated into AI model 300. XAI mechanisms can provide the ability to generate visual indicators (e.g., saliency maps or attention heatmaps). These visual indicators provide clinicians with interpretable insights into model decisions.

[0130] System 100 and its components can be configured or implemented as wearable devices to enable continuous patient (i.e., subject) monitoring, remote patient assessment, and proactive care services. Figure 4 A schematic diagram of an exemplary wearable device 600 for non-invasive cardiovascular health assessment or cardiovascular monitoring is shown. Figure 4 An exemplary embodiment of wearable device 600 is illustrated. Wearable device 600 may include sensor array 602 for continuously acquiring ECG and PPG signals from a user. Sensor array 602 may include at least two types of sensors 102, 104. Sensors 102, 104 may be used to sense ECG and PPG signals, respectively. ECG sensor 102 may be a chest strap or electrodes integrated into sensor array 602. Sensor array 602 may include a frame or mounting structure, and electrodes may be mounted on the frame. The frame may be aligned appropriately on the subject's body (i.e., the patient's body) such that ECG sensor 102 is in an operable position. PPG sensor 104 may be a pulse oximeter or other PPG sensor. PPG sensor 104 may be a fingertip-mounted pulse oximeter or a wrist-worn pulse oximeter.

[0131] Wearable device 600 includes computing device 604. Computing device 604 may include all the same components as computing device 200 described above. The user interface, i.e., the display, may be a remote display for the wearable device. Computing device 602 includes processor 606 and storage unit 608, wherein processor 606 and storage unit 608 are operatively coupled to each other. Computing device 604 may be operatively coupled to sensor array 602 such that measured sensor signals can be received at the computing device. The storage unit may be used to store an AI model. The AI ​​model used in wearable device 600 may be AI model 300 described herein. AI model 300 may be executed by processor 606, and storage unit 608 also includes instructions that, when executed by processor 606, cause the processor to perform the following operations: receive physiological signals of a subject, wherein the physiological signals include at least ECG signals and PPG signals; process the ECG signals and PPG signals through the AI ​​model; and estimate one or more demographic and anthropometric parameters through the AI ​​model based on the processing of ECG signals and PPG signals. Demographic and anthropometric parameters can be displayed on the user interface.

[0132] In one exemplary implementation, wearable device 600 may perform method 400 or 500 for noninvasive cardiovascular health monitoring as described herein.

[0133] The AI ​​model 300 used in system 100 and wearable device 600 will be described in more detail below. In one example, the AI ​​model includes an RNN model, which is trained to perform steps of processing received ECG and PPG signals and output estimates of one or more demographic and anthropometric parameters.

[0134] Figure 5 An exemplary architecture of an AI model 300 is shown. The AI ​​model may include an encoder 302 and one or more task-specific classification heads 304, 306. The encoder 302 may be trained to extract features from ECG and / or PPG signals. The task-specific classification heads 304, 306 may be trained to apply regression estimation of anthropometric parameters and / or apply classification estimation of demographic parameters. The encoder 302 is coupled to one or more task-specific classification heads 304, 306 such that the output of the encoder is the input of each classification head, and the encoder is upstream of each classification head. In one example, the encoder may be pre-trained. In one example, the encoder may be a PatchTST encoder. Model 300 may include a regression head 304 and a classification head 306. Each head leads to an output block. The output block may output estimates of demographic and anthropometric parameters.

[0135] In one example, regression head 304 can be used to estimate anthropometric parameters such as age, weight, BMI, height, etc. Classification head 306 can be used to estimate demographic parameters such as sex.

[0136] Figure 6 An exemplary architecture of the PatchTST encoder 302 is illustrated. The encoder 302 may include a multi-head attention block 310 coupled to a residual connection and normalization block 312. The output of the multi-head attention block 310 is fed into the residual connection and normalization block 312. Additionally, the input may also be fed forward into block 312. The residual connection and normalization block 312 may add the input to the output of the multi-head attention block and normalize the added data. The encoder 302 may further include a feedforward block 314 operatively coupled to the residual connection and normalization block 312. The encoder may include a second residual connection and normalization block 316 adapted to add the output of feedforward block 314 to the input of the feedforward block. The output of the encoder 302 may be features extracted from ECG and PPG signals. Optionally, the encoder 302 may also perform the preprocessing steps described herein.

[0137] In one example, the AI ​​model may optionally include an ensemble architecture that combines multiple predictions from different model architectures, where multiple model architectures can define the AI ​​model. This can improve robustness and reliability. Ensemble learning can be introduced to further improve robustness, and performance can be enhanced by optional extensions such as self-supervised pre-training, interpretable AI mechanisms, or the integration of additional modalities (e.g., breathing or SpO2).

[0138] AI Model 300 can be pre-trained using a self-supervised learning method. Pre-training provides general feature representations that can be transferred to multiple downstream tasks without requiring complete retraining. AI Model 300 can be pre-trained on large training datasets.

[0139] The training datasets include publicly available cardiovascular health assessment datasets containing synchronized ECG and PPG signals, along with corresponding demographic and anthropometric labels such as age, sex, height, weight, and BMI. These datasets provide the necessary baseline truth for learning the correlation between biophysical signals and individual characteristics through AI-driven analytics.

[0140] To ensure broad applicability and robust model training, the dataset employed encompasses a diverse population, covering multiple age groups, genders, and a wide range of anthropometric values ​​(BMI, weight, and height). This diversity helps support generalization across different patient groups and reduces potential bias in downstream predictions.

[0141] Optionally, the dataset can be extended with additional wearable sensor modalities, such as respiratory waveforms, oxygen saturation (SpO2), or accelerometer data. These optional inputs can further enhance the robustness of the system, especially in mobile or dynamic monitoring scenarios.

[0142] As part of training AI Model 300, preliminary results were obtained using the PPG-BP dataset (a publicly available database containing photoplethysmography (PPG) signals and corresponding demographic and biometric information). This dataset was also used to evaluate the feasibility of directly predicting demographic and anthropometric parameters from PPG signals.

[0143] The feasibility of AI Model 300 was evaluated using the PPG-BP dataset, a publicly available database of PPG signals with accompanying demographic and biometric information. Figure 7 The training and validation process of AI model 300 is shown.

[0144] Reference Figure 7 The training and validation process 700 includes a pre-training phase 702. During pre-training, the system employs masked self-supervised learning (SSL) on the training dataset. This enables robust feature learning from large-scale unlabeled data. A PatchTST-based architecture is used, where the input time-series segments are divided into blocks, and a portion of these blocks is randomly masked. Model 300 is trained to reconstruct the masked portions, forcing encoder 302 to capture temporal dynamics, waveform morphology, and underlying physiological features. This pre-training strategy provides general feature representations that can be transferred to multiple downstream tasks without requiring complete retraining. The pre-training phase is used to train encoder 302. Pre-trained weights can be generated, for example, weights for the encoder.

[0145] During the pre-training phase 702, the labeled data can be preprocessed using a blockization module. The masked portion can have, for example, a masking ratio of 0.4. The projection module can segment the time series into blocks. Linear layers can be used to reconstruct these blocks.

[0146] Step 704 involves training the AI ​​model 300 using a training dataset. The training phase can be used to train classification heads 304 and 306, and the remainder of model 300. During the training phase, model weights are determined so that the AI ​​model 300 can estimate demographic and anthropometric parameters. The model weights can be stored and applied to the model.

[0147] Step 706 includes evaluation, i.e., validation of the AI ​​model 300. Following pre-training, model 300 is directly evaluated on a labeled downstream task using demographic and anthropometric parameters as the target output. At this stage, the pre-trained encoder 302 extracts features from the input PPG signal, and task-specific prediction heads 304 and 306 are applied to regression (age, height, weight, BMI) or classification (gender).

[0148] Performance was evaluated using standard regression and classification metrics. Mean absolute error (MAE) and root mean square error (RMSE) were used to evaluate performance on the regression task. F1 score and area under the receiver operating characteristic curve (AUROC) were used to evaluate performance on the classification task. Cross-validation and statistical analysis ensured the robustness of the reported results, resulting in more robust AI models.

[0149] As part of validating the feasibility of the AI ​​model, the inventors conducted preliminary experiments on the PPG-BP dataset. A pre-trained AI model 300, including the PatchTST encoder, was applied to the PPG signals for demographic parameter estimation. The results are shown in Table 1 below.

[0150]

[0151] Table 1: Preliminary results of the proposed model on the PPG-BP dataset These results demonstrate the potential of Model 300 to accurately estimate demographic and anthropometric parameters using only PPG signals. The findings of the tests provide strong preliminary evidence supporting the applicability and accuracy of the described AI-based model approach. These results also demonstrate the robustness of the AI ​​model in estimating demographic and anthropometric parameters.

[0152] Figure 8 Further alternative implementations of a non-invasive cardiovascular health assessment method are illustrated. Method 800 can be performed by a system or wearable device. Method 800 provides an AI-based approach for automatic feature extraction, followed by feature learning and prediction based on the learned features. Method 800 includes collecting and processing ECG and PPG signals, optimizing data quality through sophisticated preprocessing techniques. Machine learning models, including neural networks, then analyze the signals to estimate age, sex, height, weight, and BMI with unparalleled accuracy.

[0153] Method 800 utilizes publicly available CVD databases containing ECG and PPG signals, as well as demographic data. The data can be partitioned into training, validation, and test sets, and deep learning can be applied to automatically extract highly salient features from the input signals, or machine learning can be applied for specific feature extraction. Subsequently, AI algorithms such as recurrent neural networks can be used to train and optimize each model to estimate age, sex, height, weight, and BMI. The overall model can be composed of a hybrid approach, using ML / DL techniques for feature extraction and recurrent neural networks to facilitate temporal learning of time-series waveforms. Rigorous evaluation and cross-validation techniques ensure robust and reliable performance. The trained model can then be validated using benchmark truth data, addressing ethical concerns, and its potential for real-world integration will be explored, such as in telemedicine applications and clinical trials. Method 800 is an optional alternative.

[0154] The systems, methods, and wearable devices described herein have the potential to significantly transform the healthcare industry through their ability to facilitate remote monitoring, streamline clinical diagnosis, and enhance preventative healthcare strategies. Furthermore, the seamless integration of this technology with telemedicine and wearable health devices makes it a holistic solution within the modern healthcare environment.

[0155] By enabling accurate, real-time, and non-invasive predictions of these characteristics, this system supports early risk identification, proactive intervention, and remote monitoring, and also provides broader applications in telemedicine, personalized medicine, and even health insurance risk assessment.

[0156] The systems 100 and methods 400 and 500 described in this paper fill this gap by introducing an AI-enabled, non-invasive system that integrates ECG and PPG signals to automatically estimate demographic characteristics. System 100 is advantageous because it uses a non-invasive method to acquire ECG and PPG signals, making the acquisition of these signals for subjects both simple and comfortable. System 100 also additionally uses a trained AI model to estimate demographic and anthropometry parameters. These demographic and anthropometry parameters can be used to accurately determine cardiovascular health status.

[0157] This system and method are advantageous because they employ a trained AI model, providing a transformative approach to estimating age, sex, height, weight, and BMI using ECG and PPG signals. The system and method utilize AI to improve the accuracy of estimating age, sex, height, weight, and BMI from non-invasive ECG and PPG signals. This represents an advancement in cardiovascular health engineering as it eliminates the need for invasive procedures or less accurate methods. Unlike conventional methods, this AI-based innovation provides a comprehensive overview of an individual's cardiovascular health in a single assessment. This holistic approach accelerates diagnosis and provides a comprehensive assessment of a patient's cardiovascular health.

[0158] This system and approach are also advantageous because they facilitate the early detection of cardiovascular problems, enabling healthcare practitioners to take proactive strategies that may prevent the progression of chronic heart disease and improve patient health. The AI-enabled continuous monitoring capabilities provide real-time insights into a patient's cardiovascular health status, facilitating timely intervention and adjustments to treatment plans. The system integrates effectively with existing healthcare systems and electronic health records, allowing healthcare professionals rapid access to critical patient data. Therefore, this integration enhances the overall quality of care provided.

[0159] The system and method for noninvasive cardiovascular health assessment using AI models described in this article have advantages including: real-time operation, clinical interpretability, scalability across devices and populations, and applicability in proactive healthcare management.

[0160] While not strictly necessary, the embodiments described with reference to the accompanying drawings can be implemented as an Application Programming Interface (API) or a set of libraries used by the developer, or can be included in another software application, such as a terminal or personal computer operating system or a portable computing device operating system. Typically, since program modules include routines, programs, objects, components, and data files that help perform specific functions, those skilled in the art will understand that the functionality of a software application can be distributed among multiple routines, objects, or components to achieve the same functionality required herein.

[0161] It should also be understood that any suitable computing system architecture can be used where the methods and systems of this disclosure are implemented wholly or partially by a computing system. This will include stand-alone computers, network computers, and dedicated hardware devices. When using the terms "computing system" and "computing device," these terms are intended to include any suitable configuration of computer hardware capable of implementing the described functions.

[0162] Those skilled in the art will understand that various changes and / or modifications can be made to this disclosure as illustrated in the specific embodiments without departing from the spirit or scope of this disclosure as broadly described. Therefore, this embodiment should be considered illustrative rather than restrictive in all respects.

[0163] Furthermore, it should be noted that the implementation can be described as a process, depicted as a flowchart, block diagram, structural diagram, or block diagram. Although a flowchart can describe operations as a sequential process, many operations can be performed in parallel or simultaneously. Moreover, the order of operations can be rearranged. A process terminates when its operations are completed. A process can correspond to a method, function, procedure, subroutine, subroutine, etc., in a computer program. When a process corresponds to a function, its termination corresponds to the function returning to the calling function or the main function.

[0164] The various aspects of the systems and methods described above are operable on any type of general-purpose computer system or computing device, including but not limited to desktop computers, laptops, tablets, smart TVs, game consoles, or mobile devices. The term "mobile device" includes, but is not limited to: wireless devices, mobile phones, smartphones, mobile communication devices, user communication devices, personal digital assistants, mobile handheld computers, laptops, wearable electronic devices such as smartwatches and head-mounted devices, e-book readers and reading devices capable of reading electronic content, and / or other types of mobile devices that are typically carried by an individual and / or have some form of communication capability (e.g., wireless, infrared, short-range radio, cellular, etc.).

[0165] In all its aspects, embodiments of this disclosure may be embodied as computer-implemented processes, machines (such as electronic devices, or general-purpose computers or other devices that provide a platform on which computer programs can be executed), processes performed by such machines, or articles of art.

Claims

1. A computer-implemented method for non-invasive cardiovascular health assessment, characterized in that, include: At the processor, physiological signals from the subject are received; The physiological signals include at least cardiac ECG signals (electrical signals) and PPG signals (photoplethysmography signals). The processor executes an AI model (artificial intelligence model) to process the ECG signal and the PPG signal. as well as Using the AI ​​model, one or more demographic and anthropometric parameters are estimated based on the processing of the ECG and PPG signals.

2. The method according to claim 1, characterized in that, in, The processing steps include: The received ECG and PPG signals are preprocessed. Feature extraction is performed on the preprocessed ECG and PPG signals; and Select one or more optimal features; and The demographic and anthropometry parameters are estimated based on one or more of the optimal features.

3. The method according to claim 2, characterized in that, in, The processing steps also include: performing time processing on the preprocessed ECG signal and PPG signal.

4. The method according to claim 3, characterized in that, in, The time processing includes capturing the temporal correlation and spatial relationship between the one or more optimal features.

5. The method according to claim 1, characterized in that, in, The demographic parameters include age and sex, and the one or more anthropometric characteristics include weight, height, and body mass index (BMI).

6. The method according to claim 1, characterized in that, Also includes: The subject's cardiovascular age is calculated using the AI ​​model based on the estimated demographic and anthropometric parameters. as well as Output the cardiovascular age, and / or compare the calculated cardiovascular age with the subject's physiological age to identify the risk of vascular aging.

7. The method according to claim 6, characterized in that, Also includes: Track changes in estimated BMI or weight over time to detect abnormal trends and present the changes in estimated BMI or weight on the user interface.

8. The method according to claim 1, characterized in that, Also includes: Generate visual indicators to display estimates of the demographic and anthropometric parameters; as well as The visual indicator is displayed on the user interface; as well as The visual indicators include one or more of the following: saliency maps, heatmaps, or attention heatmaps.

9. The method according to claim 3, characterized in that, in, The AI ​​model includes an RNN (Recurrent Neural Network) model, wherein the RNN model is trained to perform steps of processing received ECG and PPG signals and output estimates of one or more demographic and anthropometric parameters.

10. A system for non-invasive cardiovascular health monitoring, characterized in that, include: A computing device including a processor and a storage unit operatively coupled; A user interface that is operatively coupled to or integrated into the computing device; The storage unit is used to store an AI model (artificial intelligence model), wherein the AI ​​model can be executed by the processor. The storage unit also includes instructions, which, when executed by the processor, cause the processor to perform the following operations: Receive physiological signals from the subject. The physiological signals include at least cardiac ECG signals (electrical signals) and PPG signals (photoplethysmography signals). The AI ​​model is used to process the ECG signal and the PPG signal. Using the AI ​​model, based on the processing of the ECG and PPG signals, one or more demographic and anthropometric parameters are estimated; and The estimated demographic and anthropometric parameters are displayed on the user interface.

11. The system according to claim 10, characterized in that, in, The processor is programmed to: The received ECG and PPG signals are preprocessed. Feature extraction is performed on the preprocessed ECG and PPG signals; as well as Select one or more optimal features; as well as The demographic and anthropometry parameters are estimated based on one or more of the optimal features.

12. The system according to claim 11, characterized in that, in, The processor is programmed to perform timing processing on the preprocessed ECG and PPG signals.

13. The system according to claim 12, characterized in that, in, During time processing, the processor is programmed to capture the temporal correlations and spatial relationships between the one or more optimal features.

14. The system according to claim 10, characterized in that, in, The demographic parameters include age and sex, and the one or more anthropometric characteristics include weight, height, and body mass index (BMI).

15. The system according to claim 10, characterized in that, in, The processor is programmed to: By applying the AI ​​model, the cardiovascular age of the subject is calculated based on the estimated demographic and anthropometric parameters; as well as Output the cardiovascular age, and / or compare the calculated cardiovascular age with the subject's physiological age to identify the risk of vascular aging.

16. The system according to claim 15, characterized in that, in, The processor is programmed to track changes in the estimated BMI or weight over time to detect abnormal trends and to present the changes in the estimated BMI or weight on the user interface.

17. The system according to claim 10, characterized in that, in, The processor is also programmed to: Generate visual indicators to display estimates of the demographic and anthropometric parameters; The visual indicator is presented on the user interface; and The visual indicators include one or more of the following: saliency maps, heatmaps, or attention heatmaps.

18. The system according to claim 11, characterized in that, in, The AI ​​model includes an RNN (Recurrent Neural Network) model, wherein the RNN model is trained to perform steps of processing received ECG and PPG signals and output estimates of one or more demographic and anthropometric parameters.

19. The system according to claim 18, characterized in that, in, The AI ​​model is pre-trained on a dataset of unlabeled ECG signals, PPG signals, or both using a self-supervised learning framework, wherein the self-supervised learning framework is a masked self-supervised learning framework, segments of the unlabeled signals are masked, and the AI ​​model is trained to reconstruct the masked segments.

20. A wearable device for non-invasive cardiovascular health monitoring, characterized in that, include: A sensor array used to continuously acquire ECG and PPG signals from the user; A computing device including a processor and a storage unit operatively coupled; The computing device is operatively coupled to the sensor array; The storage unit is used to store an AI model (artificial intelligence model), wherein the AI ​​model can be executed by the processor. The storage unit also includes instructions that, when executed by the processor, cause the processor to perform the following operations: The subject receives physiological signals, including at least cardiac ECG signals (electrical signals) and PPG signals (photoplethysmography signals). The AI ​​model is used to process the ECG signal and the PPG signal; and Using the AI ​​model, one or more demographic and anthropometric parameters are estimated based on the processing of the ECG and PPG signals.

Images