Description:[001] FIELD OF THE INVENTION
[002] The present invention relates to health monitoring of patients, and particularly, to a method and system for continuous, non-invasive estimation of a plurality of physiological parameters’ values associated with a subject patient.

[003] BACKGROUND OF THE INVENTION
[004] Cardio-pulmonary diseases represent a significant global health challenge, leading to millions of cases of morbidity and mortality each year. Early diagnosis and treatment of these conditions can substantially mitigate their impact. Continuous monitoring of vital signs such as respiration rate (RR), heart rate (HR), and blood pressure (BP), are crucial for detecting the onset and progression of these diseases, as well as for assessing the overall health status of patients. While HR and BP are well-established indicators of cardiovascular risk, abnormal RR serves as a critical and independent marker for conditions including cardiac arrest and chronic pulmonary diseases, particularly in high-risk patients.
[005] Despite enormous clinical importance, the respiration rate is recorded significantly less than other vital signs, mostly due to time constraints and a lack of respiration monitoring devices. Conventionally, the respiration rate is monitored using capnography, inductance plethysmography, impedance pneumography, and oronasal pressure transducers. However, these methods are often obtrusive, costly, require cumbersome set-up with multiple attachments to the body, and cause intrusion in natural breathing. The traditional methods for heart rate (electrocardiography) and blood pressure (sphygmomanometer and oscillometry) estimation involve several electrodes and an inflatable cuff attached to the body that causes obtrusion and, hence, is unsuitable for continuous and ambulatory monitoring. Further, the blood oxygen saturation (SpO2) is generally measured with pulse oximeter, which is an unobtrusive and non-invasive method. Moreover, no unobtrusive device can measure all these vital parameters together using only the PPG signal.
[006] Therefore, there is a need for improved technique for unobtrusive, accurate, and reliable estimation of RR, HR, BP, and SpO2.

[007] OBJECTS OF THE INVENTION:
[008] Some of the objects of the present disclosure, which at least one embodiment herein satisfies, are listed herein below.
[009] The primary objective of the present invention is to provide a comprehensive, non-invasive, and unobtrusive solution for continuous monitoring of key vital signs such as heart rate (HR), respiration rate (RR), systolic blood pressure (SBP), diastolic blood pressure (DBP), and blood oxygen saturation (SpO2) using a single photoplethysmography (PPG) signal.
[0010] These and other objects and advantages will become more apparent when reference is made to the following description and accompanying drawings.

[0011] SUMMARY OF THE INVENTION
[0012] This summary is provided to introduce concepts related to continuous, non-invasive estimation of a plurality of physiological parameters’ values associated with a subject patient. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.
[0013] In an aspect of the present invention, a system for continuous, non-invasive estimation of a plurality of physiological parameters’ values associated with a subject is described. The system includes a receiving module, a preprocessing module, and a machine-learning model. The receiving module is configured to receive a single photoplethysmography, PPG, signal from a sensing device positioned on the subject. The preprocessing module is configured to preprocess at least one PPG data obtained from the received PPG signal to normalize the at least one PPG data to ensure consistent scaling across one or more variables of the at least one PPG data. the preprocessing module is further configured to segment the normalized at least one PPG data into a plurality of windows. Each of the windows is of a second predetermined period of time, and two consecutive windows comprise an overlap of a third predetermined period of time between them.
[0014] The machine-learning model is configured to extract a plurality of features from the segmented PPG data to identify one or more patterns relevant to the plurality of physiological parameters. The machine learning model is further configured to capture one or more temporal dependencies in the extracted plurality of features. The machine learning module is further configured to estimate the plurality of physiological parameters’ values by aggregating the captured plurality of temporal dependent features into a fixed-size representation. The preprocessing module is configured to normalize the estimated plurality of physiological parameters’ values. The machine learning model is trained with the normalized PPG data and the normalized plurality of physiological parameters’ values.
[0015] In an embodiment of the present invention, the plurality of physiological parameters’ values include Heart Rate, HR, Respiration Rate, RR, Systolic Blood Pressure, SBP, Diastolic Blood Pressure, DBP, and Blood Oxygen Saturation, SpO2, values.
[0016] In another embodiment of the present invention, the second predetermined period of time is 5 seconds.
[0017] In another embodiment of the present invention, the third predetermined period of time is 3 seconds.
[0018] In another embodiment of the present invention, the normalization is performed by Z-squared technique.
[0019] In another embodiment of the present invention, the machine-learning model includes a Convolutional Neural Networks, CNN, a Recurrent Neural Networks, RNNs, and a Gated Recurrent Units, GRU. The CNN includes a plurality of Conv1D layers with increasing filter sizes, a plurality of Residual Blocks, RB, and a Global Average Pooling, GAP.
[0020] In another embodiment of the present invention, the CNN is configured to extract the plurality of features from the segmented PPG data.
[0021] In another embodiment of the present invention, the plurality of Conv1D layers include five Conv1D layers with increasing filter sizes for the SBP, DBP, and SpO2 estimation.
[0022] In another embodiment of the present invention, the RB enhances the feature extraction capabilities for the SBP and DBP estimation. The GAP is applied after the CNN and RNN to aggregate the temporal features into the fixed-size representation for the estimation of plurality of physiological parameters’ values.
[0023] In another embodiment of the present invention, the RNN includes two Long Short-Term Memory, LSTM, layers for the SBP and DBP estimation, and a single LSTM layer for the SpO2 estimation. The LSTM layers are configured to capture one or more temporal dependencies in the plurality of features.
[0024] In another embodiment of the present invention, the machine learning model is trained based on Adam optimizer technique. Further, the machine learning model is evaluated based on Mean Squared Error, MSE, as a loss function.
[0025] In another aspect of the present invention, a method of continuous, non-invasive estimation of a plurality of physiological parameters’ values associated with a subject is described. The method includes the step of receiving, from a sensing device positioned on the subject, a single photoplethysmography, PPG, signal. The method further includes the step of preprocessing at least one PPG data obtained from the received PPG signal. The step of preprocessing includes the sub-step of normalizing the at least one PPG data to ensure consistent scaling across one or more variables of the at least one PPG data. The step of pre-processing includes the sub-step of segmenting the normalized at least one PPG data into a plurality of windows. Each of the windows is of a second predetermined period of time. Further, two consecutive windows include an overlap of a third predetermined period of time between them.
[0026] The method further includes the step of extracting, by a machine-learning model, a plurality of features from the segmented PPG data to identify one or more patterns relevant to the plurality of physiological parameters. The method further includes the step of capturing, by the machine-learning model, one or more temporal dependencies in the extracted plurality of features. The method further includes the step of estimating the plurality of physiological parameters’ values by aggregating the captured plurality of temporal dependent features into a fixed-size representation. The estimated plurality of physiological parameters’ values are normalized to ensure consistent scaling across one or more variables of the PPG data. The machine learning model is trained with the normalized PPG data and the normalized plurality of physiological parameters’ values.
[0027] Various objects, features, aspects, and advantages of the inventive subject matter will become more apparent from the following detailed description of preferred embodiments, along with the accompanying drawing figures in which like numerals represent like components.

[0028] BRIEF DESCRIPTION OF DRAWINGS:
[0029] The illustrated embodiments of the subject matter will be understood by reference to the drawings, wherein like parts are designated by like numerals throughout. The following description is intended only by way of example and simply illustrates certain selected embodiments of devices, apparatus, systems, and methods that are consistent with the subject matter as claimed herein, wherein:
[0030] FIG. 1 illustrates a block diagram depicting a system configured to continuous, non-invasive estimate a plurality of physiological parameters’ values associated with a subject patient, in accordance with an exemplary embodiment of the present disclosure;
[0031] FIG. 2 illustrates a schematic block diagram of the system of Fig. 1, in accordance with an exemplary embodiment of the present disclosure;
[0032] FIG. 3 illustrates a schematic block diagram of a machine learning model of the system of Fig. 2, in accordance with an exemplary embodiment of the present disclosure;
[0033] FIG. 4 illustrates an architectural diagram depicting a system configured to continuous, non-invasive estimate a plurality of physiological parameters’ values associated with a subject patient, in accordance with an exemplary embodiment of the present disclosure; and
[0034] FIG. 5 illustrates a schematic block diagram depicting a method for continuous, non-invasive estimation a plurality of physiological parameters’ values associated with a subject patient, in accordance with an exemplary embodiment of the present disclosure.
[0035] The figures depict embodiments of the present subject matter for the purposes of illustration only. A person skilled in the art will easily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles of the disclosure described herein.

[0036] DESCRIPTION OF THE INVENTION:
[0037] The following is a detailed description of embodiments of the disclosure depicted in the accompanying drawings. The embodiments are in such detail as to clearly communicate the disclosure. However, the amount of detail offered is not intended to limit the anticipated variations of embodiments; on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the present disclosure as defined by the appended claims.
[0038] While the embodiments of the disclosure are subject to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the figures and will be described below. It should be understood, however, that it is not intended to limit the disclosure to the particular forms disclosed, but on the contrary, the disclosure is to cover all modifications, equivalents, and alternatives falling within the scope of the disclosure.
[0039] The terms “comprises”, “comprising”, or any other variations thereof used in the disclosure, are intended to cover a non-exclusive inclusion, such that a device, system, or assembly that comprises a list of components does not include only those components but may include other components not expressly listed or inherent to such system, or assembly, or device. In other words, one or more elements in a system or device proceeded by “comprises… a” does not, without more constraints, preclude the existence of other elements or additional elements in the system or device.
[0040] The present invention relates to method and system for continuous, non-invasive estimation of a plurality of physiological parameters’ values associated with a subject patient. The present invention describes a deep learning-based model designed to measure vital health parameters including heart rate (HR), respiration rate (RR), systolic blood pressure (SBP), diastolic blood pressure (DBP), and blood oxygen saturation (SpO2) simultaneously using only photoplethysmography (PPG) signal. This model continuously monitors and analyzes the PPG signal to provide real-time measurements of HR, RR, SBP, DBP, and SpO2. By incorporating advanced deep learning techniques, all five vital parameters can be simultaneously estimated with high accuracy.
[0041] The cardio-pulmonary diseases are a major global health concern, causing millions of morbidity and mortality cases annually. Early diagnosis and continuous monitoring of vital parameters such as HR, RR, and BP are essential for detecting and managing these diseases, improving patient outcomes, and reducing healthcare costs. Traditional methods for monitoring these vital signs are often obtrusive, costly, and require cumbersome setups, making continuous monitoring challenging. The present invention addresses these issues by providing a non-invasive, unobtrusive, and cost-effective solution using a single PPG signal. This deep learning-based model offers comprehensive and simultaneous assessment of HR, RR, SBP, DBP, and SpO2, enhancing patient comfort, enabling early detection, and improving health outcomes for individuals at risk of cardiopulmonary diseases. Utilizing photoplethysmography (PPG) signals, the model overcomes the limitations of traditional methods by offering a comprehensive, unobtrusive solution that enhances patient comfort and compliance.
[0042] The present invention leverages a deep learning model including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRU) to process the PPG signals and estimate the aforementioned physiological parameters. The model is meticulously designed to capture and analyze the intricate patterns in the PPG signals, ensuring high accuracy in the simultaneous estimation of all five vital parameters. The deep learning model has been trained using the Beth Israel Deaconess Medical Center (BIDMC) PPG and Respiration dataset which is part of the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC II) database. The dataset includes data collected from 53 individual patients, providing a range of physiological data across different subjects. The dataset used to train the model is aligned or matched with other physiological waveforms in the MIMIC II database. Further, the PPG signals are sampled at a frequency of 125 Hertz (Hz) and other physiological parameters (e.g., heart rate, blood pressure) are recorded at a lower frequency of 1 Hertz.
[0043] Now coming to the methodology of the present invention, at least one PPG data from the PPG signal received, from a sensing device positioned on a subject patient’s body is normalized, and thereafter segmented into a plurality of windows, to capture vital cardiac details. Each of the windows is of a second predetermined period of time, e.g. 5-second, and two consecutive windows include an overlap of a third predetermined period of time between them, e.g. 3-second. A Z-squared normalization is applied to the PPG data, HR, RR, and SpO2 values to ensure consistent scaling across variables. The CNN is used for feature extraction from the PPG signal. A plurality of Conv1D layers with increasing filter sizes are employed for SBP, DBP, and SpO2 prediction. The residual blocks are used within the CNN enhance the model's representational power for SBP and DBP prediction. Two LSTM layers are used for SBP and DBP prediction, and a single LSTM layer for SpO2 estimation. These layers capture temporal dependencies in the PPG data, crucial for accurate vital sign estimation. Global Average Pooling is applied after CNN and RNN layers to aggregate temporal features into fixed-size representations which enhances the model's ability to capture the most relevant information for each physiological parameter. The Adam optimizer is used for training the model. The Mean Squared Error (MSE) is employed as the loss function to evaluate the model performance.
[0044] For better understanding, one or more embodiments of the present invention shall be described with respect to the earlier-mentioned drawings.
[0045] FIG. 1 illustrates a block diagram (100) depicting a system (102) configured to continuous, non-invasive estimate a plurality of physiological parameters’ values associated with a subject patient, in accordance with an exemplary embodiment of the present disclosure. FIG. 2 illustrates a schematic block diagram (200) of the system (102) of Fig. 1, in accordance with an exemplary embodiment of the present disclosure. FIG. 3 illustrates a schematic block diagram (300) of a machine learning model (206) of the system (102) of Fig. 2, in accordance with an exemplary embodiment of the present disclosure. FIG. 4 illustrates an architectural diagram (400) depicting a system (102) configured to continuous, non-invasive estimate a plurality of physiological parameters’ values associated with a subject patient, in accordance with an exemplary embodiment of the present disclosure.
[0046] As illustrated, the system (102) includes a receiving module (202), a preprocessing module (204), and a machine-learning model (206). The receiving module (202) is configured to receive a single photoplethysmography (PPG) signal (104, 402) from a sensing device (not shown in figures) positioned on the subject. In one or more embodiment, the receiving module (202) may receive the PPG signals (104, 402) from the sensing device over a first network (404) which may be wired or wireless. The sensing device may be an electrode adapted to position on the subject patient’s body, e.g., finger, or a wearable device which can be attached to other part of the body of the subject patient. The sensing device may be capable of recording the PPG signal of the subject patient.
[0047] The preprocessing module (204) is configured to preprocess at least one PPG data derived from the received PPG signal (104, 402) to normalize the at least one PPG data to ensure a consistent scaling across one or more variables of the at least one PPG data.
[0048] The PPG signal (104, 402) includes a plurality of PPG data at different data points. In an example, the received PPG signal is of 125Hz sampling frequency which means 125Hz in 1 second so there may be 625 PPG data points in 5 seconds. During training (as described in the subsequent paragraphs), the PPG data is segmented into 5-second windows due to the inference model's requirement for 5-second PPG data. To ensure consistency, the machine learning model must also be trained on 5-second segments of PPG data. The normalization of the PPG signal ensures consistent scaling across the entire PPG signal, addressing variations in amplitude or baseline shifts caused by factors like sensor placement or lighting conditions. By normalizing the PPG signal itself, all data points are rescaled to a standard range, making it easier for the model to detect patterns related to physiological parameters without being influenced by signal magnitude fluctuations or noise. Thus, the one or more variables may refer to amplitude (magnitude of the waveform), baseline (shifts in the signal level), and/or pulse shape (heights of peaks and troughs).
[0049] The preprocessing module (204) is further configured to segment the normalized PPG data into a plurality of windows. Each of the windows is of a second predetermined period of time, and two consecutive windows include an overlap of a third predetermined period of time between them. In an embodiment, the second predetermined period of time is 5 seconds, and the third predetermined period of time is 3 seconds. The overlapping of two consecutive windows has been done to overcome the problem associated with less data for training. Basically, two windows are taken and a new window is formed by overlapping them which results in an increase in the number of windows.
[0050] The machine-learning model (206) is configured to extract a plurality of features from the segmented PPG data to identify one or more patterns relevant to the plurality of physiological parameters (106, 406). The machine learning model (206) is further configured to capture one or more temporal dependencies in the extracted plurality of features. The technical feature "capture one or more temporal dependencies in the extracted plurality of features" refers to the model's ability to understand how certain patterns or information in the input signal (PPG data in this case) change or evolve over time. In time-series data like PPG, which is collected continuously over time, the current value of the signal is often influenced by past values. Thus, capturing these temporal dependencies helps the model understand the relationships between past, present, and future data points to make better predictions.
[0051] The machine learning module (206) is further configured to estimate the plurality of physiological parameters’ values by aggregating the captured plurality of temporal dependent features into a fixed-size representation. The preprocessing module (204) is configured to normalize the estimated plurality of physiological parameters’ values (106, 406). The machine learning model (206) is trained with the normalized PPG data and the normalized plurality of physiological parameters’ values (106, 406).
[0052] In an embodiment of the present invention, the plurality of physiological parameters’ values (106, 406) include Heart Rate (HR), Respiration Rate (RR), Systolic Blood Pressure (SBP), Diastolic Blood Pressure (DBP), and Blood Oxygen Saturation (SpO2) values.
[0053] In another embodiment of the present invention, the normalization of PPG data and physiological parameters’ values is performed by Z-squared technique.
[0054] In another embodiment of the present invention, the machine-learning model (206) includes a Convolutional Neural Networks (CNN) (302), a Recurrent Neural Networks (RNNs) (304), and a Gated Recurrent Units (GRU) (306). The CNN (302) includes a plurality of Conv1D layers with increasing filter sizes, a plurality of Residual Blocks (RB) (318), and a Global Average Pooling (GAP) (320). Each of the layers are described in details in the subsequent paragraphs of the present disclosure.
[0055] In another embodiment of the present invention, the CNN (302) is configured to extract the plurality of features from the segmented PPG data.
[0056] In another embodiment of the present invention, the plurality of Conv1D layers include five Conv1D layers (308, 310, 312, 314, 316) with increasing filter sizes for the SBP, DBP, and SpO2 estimation.
[0057] In another embodiment of the present invention, the RB (318) enhances the feature extraction capabilities for the SBP and DBP estimation. The GAP (320) is applied after the CNN (302) and RNN (304) to aggregate the temporal features into the fixed-size representation for the estimation of plurality of physiological parameters’ values (106, 406).
[0058] In another embodiment of the present invention, the RNN (304) includes two Long Short-Term Memory (LSTM) layers (322, 324) for the SBP and DBP estimation, and a single LSTM layer (326) for the SpO2 estimation. The LSTM layers (322, 324, 326) are configured to capture one or more temporal dependencies in the plurality of features.
[0059] In another embodiment of the present invention, the machine learning model (206) is trained based on Adam optimizer technique. Further, the machine learning model (206) is evaluated based on Mean Squared Error, MSE, as a loss function.
[0060] The machine-learning model (206) is specifically configured to extract both spatial and temporal features from the PPG signals using a combination of CNNs (302) and RNNs (304) (like LSTMs and GRUs). The CNN layers (302) are strategically placed to capture spatial features, such as localized waveform patterns and peak-to-trough variations within the segmented PPG data. Meanwhile, the RNN layers (304) are designed to capture temporal dependencies, allowing the machine learning model to learn how the PPG signal evolves over time and identify trends in physiological parameters. Additionally, the model’s architecture, including the arrangement of layers, hyperparameters, optimizers, and loss functions, is finely tuned to enhance learning and ensure optimal feature extraction and prediction accuracy. This comprehensive configuration allows the model to effectively learn both spatial details and temporal relationships within the PPG data, ultimately improving the prediction of physiological parameters such as heart rate, SpO2, SBP, and DBP.
[0061] CNN (Convolutional Neural Network) for Feature Extraction
[0062] The CNN consists of 1D Convolutional layers (Conv1D), which operate on 1D data like time-series signals (PPG in this case). Each Conv1D layer applies a set of filters to the input signal. Each filter convolves over the input sequence, computing local features from the PPG signal. The convolution is performed using a sliding window approach where each filter slides across the input signal and computes a weighted sum of the inputs within the filter's receptive field.
[0063] Increasing Filter Sizes:
[0064] The five Conv1D layers are designed with increasing filter sizes, meaning that the size of the convolutional window (receptive field) grows as the network depth increases.
[0065] First Layers: Capture low-level, fine-grained features such as small-scale oscillations in the PPG signal, which could correspond to local variations in heart rate.
[0066] Deeper Layers: Capture more abstract, high-level patterns, such as complex relationships between signal peaks and troughs that might correlate with SBP, DBP, and SpO2 variations.
[0067] Activation Functions: After each convolution, a non-linear activation function (e.g., ReLU) is applied to introduce non-linearity into the machine learning model, enabling the model to learn complex, non-linear mappings between the input and output.
[0068] Residual Block (RB)
[0069] The RB is employed to enhance feature extraction. In traditional deep networks, as the depth increases, information can degrade due to the vanishing gradient problem, leading to diminished model performance. To combat this, the RB introduces skip connections that allow the input of the block to bypass the convolutional layers and be added directly to the output. This allows for the preservation of both the original and transformed information, ensuring that feature gradients propagate effectively through the network.
[0070] Mathematically, for a residual block 𝐻(𝑥), the function learned by the block is 𝐻(𝑥)=𝐹(𝑥)+𝑥, where 𝐹(𝑥) represents the transformation learned by the convolutional layers and x is the input that bypasses the layers.
[0071] Global Average Pooling (GAP) layer
[0072] The output of the final Conv1D layers goes through a Global Average Pooling (GAP) layer. The GAP layer replaces traditional fully connected layers by computing the average of each feature map, creating a compact and fixed-length vector.
[0073] The GAP layer serves two functions:
[0074] The GAP layer drastically reduces the dimensionality of the feature maps without losing important information. It makes the model more robust to overfitting by not relying on dense layers with numerous parameters.
[0075] The GAP layer outputs a feature vector that represents the aggregated spatial information from the PPG signal. This fixed-size representation is then passed to the RNN layers for further temporal modeling.
[0076] Temporal Feature Learning Using RNN (Recurrent Neural Network)
[0077] Once spatial features are extracted by the CNN, the model incorporates Recurrent Neural Networks (RNN) to capture the temporal dependencies within the PPG signal. The PPG is inherently a time-series signal, where current values depend on previous signal values.
[0078] Long Short-Term Memory (LSTM) is a special kind of RNN that can learn long-term dependencies by controlling the flow of information using gates. Each LSTM layer contains:
● Forget Gate: Decides what information to discard from the cell state.
● Input Gate: Determines what new information to store in the cell state.
● Output Gate: Decides what information to output.
[0079] The machine learning model contains two stacked LSTM layers to predict SBP and DBP. Stacking LSTM layers deepens the model's temporal understanding, allowing it to model complex sequences of heartbeats, pressure variations, and their long-term correlations with SBP and DBP. The stacked architecture enhances the model's capacity to learn patterns across both short-term and long-term time steps.
[0080] GRU (Gated Recurrent Unit):
[0081] The GRU layers can be used as an alternative to LSTM for capturing temporal dependencies. The GRUs are a simpler version of LSTMs, containing fewer gates (no output gate) and are computationally cheaper while maintaining comparable performance for sequence-based tasks. After the LSTM layers have processed the PPG features over time, another Global Average Pooling (GAP) layer is applied to reduce the temporal features into a compact fixed-size vector.
[0082] The GRU layer performs temporal aggregation, similar to the previous spatial GAP layer in the CNN. It takes all the temporal information modeled by the LSTM and creates a final fixed-size representation that summarizes the entire time sequence. This compact feature vector is then used to predict the physiological parameters.
[0083] The output of the GAP layer is passed through a final regression layer for the estimation of SBP, DBP, and SpO2 values. This layer computes the predicted values by applying a linear transformation on the fixed-size feature vectors produced by the GAP.
[0084] Regression Output:
[0085] The final output is continuous-valued predictions (since this is a regression problem) for SBP, DBP, and SpO2 based on the extracted spatial and temporal features.
[0086] Loss Function: Mean Squared Error (MSE)
[0087] The loss function used is Mean Squared Error (MSE), which is common in regression tasks. The loss is computed as the average of the squared differences between the predicted and actual values.
[0088] The machine learning model (206) is trained using the Adam optimizer, an adaptive optimization technique that computes adaptive learning rates for each parameter. The Adam optimizer technique combines the advantages of two other popular optimizers: AdaGrad and RMSProp, making it suitable for training deep neural networks by dynamically adjusting the learning rate based on the first and second moments of the gradients.
[0089] In one or more embodiments, the system (102) may be part of a larger computer system (410) and/or maybe operatively coupled to a network (e.g., a second network 408) with the aid of a communication interface to facilitate the transmission of and sharing data and predictive results. The computer network may be a local area network (LAN), an intranet and/or extranet, an intranet and/or extranet that is in communication with the Internet, or the Internet. The network in some cases is a telecommunication and/or a data network, and may include one or more computer servers. In an example, the communication network includes, but not limited to, 2G network, 3G network, 4G network, LTE network, 5G network, 6G network, and so forth. The network, in some cases with the aid of a computer system, may implement a peer-to-peer network, which may enable devices coupled to the computer system to behave as a client or a server. In other examples, the system, the database, and the server may be integrated network node or a single integrated unit.
[0090] The system (102) may communicate with one or more other systems by the interfaces (e.g., network adapters). The memory (210) or memory locations may be, e.g., random-access memory, read-only memory, flash memory. The system may also include at least one electronic storage units (e.g., hard disks), and peripheral devices, such as cache, other memory, data storage, and/or electronic display adapters.
[0091] The system (102) may also include one or more IO Managers as software instructions that may run on the one or more processors (208) and implement various communication protocols such as User Datagram Protocol (UDP), Modbus, MQ Telemetry Transport (MQTT), Open Platform Communications Unified Architecture (OPC UA), Semiconductor's equipment interface protocol for equipment-to-host data communications (SECS/GEM), Profinet, or any other protocol, to access data in real-time from disparate data sources via any communication network, such as Ethernet, Wi-Fi, Universal Serial Bus (USB), Zigbee, Cellular or 5G connectivity, etc., or indirectly through a device’s primary controller, through a Programmable Logic Controller (PLC) or through a Data Acquisition System (DAQ), or any other such mechanism.
[0092] Further, the CPU(s) may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, logic circuitries, and/or any devices that manipulate data based on operational instructions. Among other capabilities, the one or more processor(s) (208) are configured to fetch and execute computer-readable instructions stored in the memory (210) of the system (102).
[0093] Further, the memory (210) may store one or more computer-readable instructions or routines, which may be fetched and executed to create or share data units over a network service. The memory (210) may include any non-transitory storage device including, for example, volatile memory such as RAM, or non-volatile memory such as EPROM, flash memory, and the like.
[0094] Further, the processing devices(s) may be implemented as a combination of hardware and programming device(s) (for example, programmable instructions) to implement one or more functionalities of the processing device(s). In examples described herein, such combinations of hardware and programming may be implemented in several different ways. In one example, the programming for the processing device(s) may be processor executable instructions stored on a non-transitory machine-readable storage medium and the hardware for the processing device(s) may include a processing resource (for example, one or more processors (208)), to execute such instructions. In other examples, the processing devices(s) may be implemented by electronic circuitry.
[0095] FIG. 5 illustrates a schematic block diagram depicting a method (500) for continuous, non-invasive estimation a plurality of physiological parameters’ values associated with a subject patient, in accordance with an exemplary embodiment of the present disclosure.
[0096] As illustrated, the method (500) includes the step of receiving (502), from a sensing device positioned on the subject, a single photoplethysmography (PPG) signal. The method (500) further includes the step of preprocessing (504) at least one PPG data obtained from the received PPG signal. The step of preprocessing (504) includes the sub-step of normalizing the at least one PPG data to ensure a consistent scaling across one or more variables of the at least one PPG data. The step of pre-processing (504) includes the sub-step of segmenting the normalized at least one PPG data into a plurality of windows. Each of the windows is of a second predetermined period of time. Further, two consecutive windows include an overlap of a third predetermined period of time between them. The method (500) further includes the step of extracting (506), by a machine-learning model, a plurality of features from the segmented PPG data to identify one or more patterns relevant to the plurality of physiological parameters. The method (500) further includes the step of capturing (508), by the machine-learning model, one or more temporal dependencies in the extracted plurality of features. The method (500) further includes the step of estimating (510) the plurality of physiological parameters’ values by aggregating the captured plurality of temporal dependent features into a fixed-size representation. The estimated plurality of physiological parameters’ values are normalized to ensure consistent scaling across one or more variables of the PPG data. The machine learning model is trained with the normalized PPG data and the normalized plurality of physiological parameters’ values.
[0097] The results on standard BIDMC PPG and respiration dataset including signals from 53 patients with ground truth values of HR, RR, SpO2, SBP, and DBP, validates performance through rigorous evaluation metrics (RMSE, MAE, R2, NMAE, NRMSE) and correlation coefficients computed against actual values, and evaluates accuracy under specific tolerance levels, demonstrating robustness in predicting physiological changes with high precision.
[0098] The present invention achieves impressive accuracy across all vital signs: HR (RMSE=2.43 bpm, MAE=1.50 bpm), RR (RMSE=1.87 brpm, MAE=1.23 brpm), SpO2 (RMSE=0.80%, MAE=0.52%), SBP (RMSE=3.26 mmHg, MAE=0.95 mmHg), and DBP (RMSE=2.41 mmHg, MAE=0.79 mmHg).
[0099] The present invention demonstrates strong performance metrics (R2 > 0.71 for all vital signs), indicating the model's ability to explain variance in the data.
[00100] High correlation coefficients above 0.85 for RR, SpO2, and above 0.98 for HR, SBP, and DBP highlight strong linear relationships between predicted and actual values.
[00101] The present invention can be utilized in multiple ways/places. For example:
● 1. Hospital settings, e.g., ICU and general wards for non-invasive, continuous vital sign monitoring.
● 2. Home Healthcare: Remote monitoring of chronic conditions and elderly care.
● 3. Emergency Medical Services (EMS): Real-time vital sign assessment in pre-hospital care.
● 4. Sports and Fitness Monitoring: Athlete performance optimization and injury prevention.
● 5. Telemedicine and Telehealth: Integration into virtual consultations for remote patient monitoring.
● 6. Military and Defence Applications: Health monitoring for military personnel in remote locations.
● 7. Ambulatory monitoring: Unobtrusive measurement for ambulatory monitoring of vital parameters.
● 8. Sleep study and stress testing: Suitable for long term continuous monitoring that will be helpful for sleep study and stress testing.
[00102] A few of the major advantages of the present invention over the conventional solutions:
● The present invention simultaneously estimates heart rate (HR), respiration rate (RR), systolic blood pressure (SBP), diastolic blood pressure (DBP), and blood oxygen saturation (SpO2) from a single photoplethysmography (PPG) signal, as the traditional solutions typically require separate devices for each vital sign monitoring, leading to a cumbersome and uncomfortable experience for the patient. By leveraging deep learning techniques to process a single PPG signal, the present invention provides a comprehensive assessment of multiple vital parameters at once, enhancing both efficiency and patient comfort.
● The present invention uses a single PPG signal to monitor vital signs, eliminating the need for multiple sensors and lead attachments to the body of the patient. The traditional monitoring methods often involve uncomfortable and intrusive setups, such as multiple electrodes or inflatable cuffs to be worn in the body. The present invention offers a non-invasive and unobtrusive alternative using which vital parameters can be measured just from the fingertip or from the wrist, hence, making it suitable for continuous monitoring in various settings, from hospitals to home care.
● The present invention estimates multiple vital signs (HR, RR, SpO2, SBP, DBP) simultaneously from a single PPG signal, reducing the need for multiple devices and improving patient comfort.
● The present invention offers a non-invasive and unobtrusive monitoring solution, enhancing patient compliance and facilitating long-term use in various healthcare settings.
● The present invention utilizes existing PPG technology, eliminating the need for additional sensors or complex setups, thereby reducing costs associated with monitoring equipment.
● The present invention provides real-time assessment of vital signs, enabling timely clinical interventions and continuous monitoring of patient health status.
● The present invention supports healthcare providers with accurate and reliable data, aiding in better clinical decision-making and personalized patient care.
[00103] Those of skill in the art would appreciate that the various illustrative blocks, modules, elements, components, methods, and algorithms described herein may be implemented as electronic hardware, computer software, or combinations of both. To illustrate this interchangeability of hardware and software, various illustrative blocks, modules, elements, components, methods, and algorithms have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application. Various components and blocks may be arranged differently (e.g., arranged in a different order, or partitioned in a different way) all without departing from the scope of the subject technology.
[00104] It is understood that any specific order or hierarchy of blocks in the processes disclosed is an illustration of example approaches. Based upon design preferences, it is understood that the specific order or hierarchy of blocks in the processes may be rearranged, or that all illustrated blocks be performed. Any of the blocks may be performed simultaneously. In one or more implementations, multitasking and parallel processing may be advantageous. Moreover, the separation of various system components in the implementations described above should not be understood as requiring such separation in all implementations, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.
[00105] It should be noted that the description and figures merely illustrate the principles of the present subject matter. It should be appreciated by those skilled in the art that conception and specific embodiment disclosed may be readily utilized as a basis for modifying or designing other structures for carrying out the same purposes of the present subject matter. It should also be appreciated by those skilled in the art by devising various systems that, although not explicitly described or shown herein, embody the principles of the present subject matter and are included within its spirit and scope.
[00106] Furthermore, all examples recited herein are principally intended expressly to be for pedagogical purposes to aid the reader in understanding the principles of the present subject matter and the concepts contributed by the inventor(s) to further the art and are to be construed as being without limitation to such specifically recited examples and conditions. The novel features which are believed to be characteristic of the present subject matter, both as to its organization and method of operation, together with further objects and advantages will be better understood from the following description when considered in connection with the accompanying figures.
[00107] Although embodiments for the present subject matter have been described in language specific to package features, it is to be understood that the present subject matter is not necessarily limited to the specific features described. Rather, the specific features and methods are disclosed as embodiments for the present subject matter. Numerous modifications and adaptations of the system/device of the present invention will be apparent to those skilled in the art, and thus it is intended by the appended claims to cover all such modifications and adaptations which fall within the scope of the present subject matter.
, Claims:We claim:
1. A system for continuous, non-invasive estimation of a plurality of physiological parameters’ values associated with a subject, comprising:
- a receiving module configured to receive, from a sensing device positioned on the subject, a single photoplethysmography, PPG, signal;
- a preprocessing module configured to preprocess at least one PPG data obtained from the received PPG signal to:
o normalize the at least one PPG data to ensure consistent scaling across one or more variables of the at least one PPG data; and
o segment the normalized at least one PPG data into a plurality of windows, wherein each of the windows is of a second predetermined period of time, and wherein two consecutive windows comprise an overlap of a third predetermined period of time between them; and
- a machine-learning model configured to:
o extract a plurality of features from the segmented PPG data to identify one or more patterns relevant to the plurality of physiological parameters; and
o capture one or more temporal dependencies in the extracted plurality of features; and
o estimate the plurality of physiological parameters’ values by aggregating the captured plurality of temporal dependent features into a fixed-size representation,
wherein the preprocessing module is configured to normalize the estimated plurality of physiological parameters’ values, and wherein the machine learning model is trained with the normalized PPG data and the normalized plurality of physiological parameters’ values.
2. The system as claimed in claim 1, wherein the plurality of physiological parameters’ values comprise Heart Rate, HR, Respiration Rate, RR, Systolic Blood Pressure, SBP, Diastolic Blood Pressure, DBP, and Blood Oxygen Saturation, SpO2, values.
3. The system as claimed in claim 1, wherein the second predetermined period of time is 5 seconds.
4. The system as claimed in claim 1, wherein the third predetermined period of time is 3 seconds.
5. The system as claimed in claim 1, wherein the normalization is performed by Z-squared technique.
6. The system as claimed in claim 1, wherein the machine-learning model comprises a Convolutional Neural Networks, CNN, a Recurrent Neural Networks, RNNs, and a Gated Recurrent Units, GRU, and wherein the CNN comprises a plurality of Conv1D layers with increasing filter sizes, a plurality of Residual Blocks, RB, and a Global Average Pooling, GAP.
7. The system as claimed in claim 6, wherein the CNN is configured to extract the plurality of features from the segmented PPG data.
8. The system as claimed in claim 6, wherein the plurality of Conv1D layers comprise five Conv1D layers with increasing filter sizes for the SBP, DBP, and SpO2 estimation.
9. The system as claimed in claim 6, wherein the RB enhances the feature extraction capabilities for the SBP and DBP estimation, and wherein the GAP is applied after the CNN and RNN to aggregate the temporal features into the fixed-size representation for the estimation of plurality of physiological parameters’ values.
10. The system as claimed in claim 6, wherein the RNN comprises two Long Short-Term Memory, LSTM, layers for the SBP and DBP estimation, and a single LSTM layer for the SpO2 estimation, and wherein the LSTM layers are configured to capture one or more temporal dependencies in the plurality of features.
11. The system as claimed in claim 1, wherein the machine learning model is trained based on Adam optimizer technique and wherein the machine learning model is evaluated based on Mean Squared Error, MSE, as a loss function.
12. A method of continuous, non-invasive estimation of a plurality of physiological parameters’ values associated with a subject, comprising:
- receiving, from a sensing device positioned on the subject, a single photoplethysmography, PPG, signal;
- preprocessing at least one PPG data obtained from the received PPG signal for:
o normalizing the at least one PPG data to ensure consistent scaling across one or more variables of the at least one PPG data; and
o segmenting the normalized at least one PPG data into a plurality of windows, wherein each of the windows is of a second predetermined period of time, and wherein two consecutive windows comprise an overlap of a third predetermined period of time between them;
- extracting, by a machine-learning model, a plurality of features from the segmented PPG data to identify one or more patterns relevant to the plurality of physiological parameters;
- capturing, by the machine-learning model, one or more temporal dependencies in the extracted plurality of features; and
- estimating the plurality of physiological parameters’ values by aggregating the captured plurality of temporal dependent features into a fixed-size representation,
wherein the estimated plurality of physiological parameters’ values are normalized to ensure consistent scaling across one or more variables of the PPG data, and wherein the machine learning model is trained with the normalized PPG data and the normalized plurality of physiological parameters’ values.

Dated this 14th day of November, 2024

[SONAL MISHRA]
-DIGITALLY SIGNED-
IN/PA-3929
OF L.S. DAVAR & CO.
ATTORNEY FOR THE APPLICANT(S)