Wearable device for detecting atrial fibrillation using photoplethysmography (PPG) sensor

Abstract

A smart ring (104) for detecting atrial fibrillation using photoplethysmography is disclosed The smart ring (104) includes a PPG sensor for gathering PPG signals corresponding to the blood flow data of the user (102). A user device (106) receives the blood flow data and extracts a plurality of features from the blood flow data including time between heartbeats (RR intervals), variability in RR intervals, absence of P waves, presence of rapid and irregular heartbeat, SpO2, amplitude and width of the peaks, harmonic components of PPG, pulse wave velocity, etc. A weightage is assigned to each of the features. Further, the user device (106) implements a machine learning model (166) to detect a probability of occurrence of AFib based on the blood flow data. The user device (106) may further notify the user (102) about the probability of occurrence of AFib.

Description

WEARABLE DEVICE FOR DETECTING ATRIAL FIBRILLATION USING PHOTOPLETHYSMOGRAPHY (PPG) SENSORFIELD OF INVENTION

[0001] The present invention relates to a wearable device, and specifically relates to a device to detect abnormal heart rhythm using a photoplethysmography (PPG) sensor.BACKGROUND

[0002] The subject matter discussed in the background section should not be assumed to be prior art merely as a result of its mention in the background section. Similarly, a problem mentioned in the background section or associated with the subject matter of the background section should not be assumed to have been previously recognized in the prior art. The subject matter in the background section merely represents different approaches, which in and of themselves may also correspond to implementations of the claimed technology.

[0003] Regular health monitoring is important for accomplishing long-term wellness goals. Today, there are many wearable devices such as smart watches, available to collect a plurality of health parameters non-invasively. Wearable devices may collect health parameters using a plurality of sensors such as SpO2 sensor, ECG sensor, etc. The wearable devices also render the health parameters to a user via the screen or an external device like smartphone.

[0004] These wearable devices are not capable of interpreting the health parameters to determine a medical condition. However, such wearable devices are precise in monitoring heartbeat, SpO2 level or exercise mode of the user, but these lack in detecting critical health conditions such as heart attack, panic attack, stroke, etc. Further, wearable devices are also prone to miscalculations due to a lot of external factors such as sweat, motion, humidity, or other environmental factors. So, such devices are unable to provide detailed insights of health parameters precisely and reliably and detect critical health conditions.

[0005] Thus, there remains a need for a wearable device for detecting atrial fibrillation using a photoplethysmography sensor.SUMMARY OF THE INVENTION

[0001] This summary is provided to introduce aspects related to a wearable device for detecting atrial fibrillation using photoplethysmography sensor and the aspects are further described below in the detailed description. This summary is not intended to identify essential features of the claimed subject matter nor is it intended for use in determining or limiting the scope of the claimed subject matter.

[0002] In an embodiment, a method of detecting atrial fibrillation (AFib) is disclosed. The method comprises receiving, by a device, blood flow data of a user. Further, extracting, by a feature extraction module implementing on the device, a plurality of features from the blood flow data. The plurality of features include time between heartbeats (RR intervals), variability in RR intervals, absence of P waves, presence of rapid and irregular heartbeat, SpO2 level, amplitude and width of peaks, harmonic components of PPG signals, and pulse wave velocity. Furthermore, the method includes predicting, by a machine learning (ML) model, a probability of occurrence of AFib by processing the plurality of features; and providing, on the device, the probability of occurrence of AFib as a notification to the user.

[0003] In an aspect, the blood flow data is collected using a PPG sensor present in the smart ring.

[0004] In an aspect, extracting the plurality of features from the blood flow data comprises filtering, by a pre-processing layer of the feature extraction module, the blood flow data using a data signal processing (DSP) filter to remove one or more of motion artifacts, baseline drift, and noise. Further, segmenting, by a segmentation layer of the feature extraction module, the blood flow data into at least one of time-based and beat-wise frames, and detecting peaks and troughs in the blood flow data. Furthermore, extracting, by a time-domain feature layer of the feature extraction module, pulse amplitude, pulse width, pulse wave velocity, RR intervals, and variability in RR intervals using standard deviation of normal-to-normal intervals (SDNN) and root mean square of successive differences (RMSSD). Further, applying, by a frequencydomain feature layer of the feature extraction module, at least one of a Fourier and a wavelet transform on the blood flow data to obtain harmonic components and power spectral density. Further, identifying, by a clinical indicator layer of the feature extraction module, at least one of irregular and rapid heartbeats from the variability in RR intervals identified using the harmonic components and power spectral density.

[0005] In an aspect, the clinical indicator layer further identifies absence of P waves, and estimates SpCh level.

[0006] In an aspect, the ML model is a logistic regression model trained on a dataset of blood flow data corresponding to multiple users.

[0007] In an aspect, the device is at least one of a user device and a smart ring.

[0008] In an aspect, the ML model is implemted on the user device.

[0009] In an aspect, a lightweight executable version of the ML model is executed on the smart ring.

[0010] In an aspect, the method comprises transferring the plurality of features to a server and predicting, by a machine learning (ML) model implemented on the server, a probability of occurrence of AFib by processing the plurality of features.

[0011] In an aspect, predefined weightages are assigned to each feature of the plurality of features.

[0012] In an aspect, the notification is provided as at least one of a visual output, vibration, and an audio output.

[0013] In an aspect, the method further comprising determining, by the ML model, one or more heart conditions upon identifying AFib, wherein the one or more heart conditions comprise isolated extrasystoles, increased heart rate variability (HRV), frequent extrasystoles, trigeminy episode, trigeminy, bigeminy episode, bigeminy, tachycardia, tachycardia episode, bradycardia, bradycardia episode, atrial fibrillation, and atrial flutter.

[0014] In an aspect, the probability of occurrence of AFib and the plurality of features are transferred to the server for future reference.

[0015] In an embodiment, a device for detecting atrial fibrillation (AFib) is disclosed. The device includes one or processors and a memory coupled with the one or more processors, wherein the memory stores program instructions configured to receive a machine learning (ML) model pre-trained on a server, wherein the server is communicatively coupled with the device. Further, the memory stores program instructions configured to receive blood flow data of a user.

[0016] Furthermore, the memory stores program instructions configured to extract a plurality of features from the blood flow data using a feature extraction module implementing on the device. The plurality of features include time between heartbeats (RR intervals), variability in RR intervals, absence of P waves, presence of rapid and irregular heartbeat, SpO2 level, amplitude and width of peaks, harmonic components of PPG signals, and pulse wave velocity.

[0017] The memory stores program instructions configured to predict a probability of occurrence of AFib by processing the plurality of features by the ML model. Further, the memory stores program instructions configured to provide the probability of occurrence of AFib as a notification to the user.

[0018] In an aspect, the device is at least one of a user device and a smart ring.

[0019] Other aspects and advantages of the invention will become apparent from the following description, taken in conjunction with the accompanying drawings, illustrating by way of example the principles of the invention.BRIEF DESCRIPTION OF THE DRAWINGS

[0006] The accompanying drawings constitute a part of the description and are used to provide further understanding of the present invention. Such accompanying drawings illustrate the embodiments of the present invention which are used to describe the principles of the present invention. The embodiments are illustrated by way of example and not by way of limitation in the figures of the accompanying drawings in which like references indicate similar elements. It should be noted that references to “an” or “one” embodiment in this invention are not necessarily to the same embodiment, and they mean at least one. In the drawings:

[0007] Fig. 1 illustrates an environment diagram of a wearable device for detecting atrial fibrillation, in accordance with an embodiment of the present invention;

[0008] Fig. 2 illustrates a block diagram of a user device connected to the wearable device for detecting atrial fibrillation, in accordance with an embodiment of the present invention;

[0009] Figs. 3a to 3c depict photoplethysmograms of regular rhythm, isolated extrasystoles, and increased HRV respectively, in accordance with an embodiment of the present invention;

[0010] Figs. 4a to 4i depict photoplethysmograms of frequent extrasystoles, trigeminy episode, trigeminy, bigeminy episode, bigeminy, tachycardia, tachycardia episode, bradycardia, and bradycardia episode respectively, in accordance with an embodiment of the present invention;

[0011] Figs. 5a and 5b depict photoplethysmograms of atrial fibrillation and atrial flutter respectively, in accordance with an embodiment of the present invention; and

[0012] Fig. 6 illustrates a flow chart of a method of detecting AFib, in accordance with an embodiment of the present invention.DETAILED DESCRIPTION OF THE INVENTION

[0013] Exemplary embodiments now will be described with reference to the accompanying drawings. The disclosure may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey its scope to those skilled in the art. The terminology used in the detailed description of the particular exemplary embodiments illustrated in the accompanying drawings is not intended to be limiting. In the drawings, like numbers refer to like elements.

[0014] It is to be noted, however, that the reference numerals used herein illustrate only typical embodiments of the present subject matter, and are therefore, not to be considered for limiting its scope, for the subject matter may admit to other equally effective embodiments.

[0015] The detailed description includes specific details for the purpose of providing a thorough understanding of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced without these specific details.

[0016] The present invention relates to a method for detecting atrial fibrillation (AFib) using photoplethysmography (PPG) signals acquired from a wearable device. The method enables continuous, non-invasive monitoring of cardiac activity and identification of AFib episodes by processing PPG data collected via a wearable device such as, a smart ring, smartwatch, or smart band. Although the method is described herein with reference to implementation using a smart ring for illustrative purposes, it is understood that the same approach may be applied to any suitable wearable form factor.

[0017] The method includes steps for acquiring blood flow data, pre-processing the signal to remove artifacts, extracting time-domain and frequency-domain features, and applying a trained machine learning model, such as a logistic regression model, to identify patterns indicative of AFib. The method facilitates real-time prediction and user notification via a connected user device, thereby supporting early intervention and improved cardiac health monitoring.

[0018] Further, the present invention relates to a wearable device for monitoring heartbeat and rhythm and detecting atrial fibrillation using photoplethysmography (PPG). The wearable device may be a smart watch, smart band, or an electronic / smart ring. Although the details have been provided successively with reference to a smart ring merely for the sake of explanation, it must be understood that the invention could be fairly implemented in a similar manner using any other wearable device, such as the ones listed above.

[0019] Fig. 1 illustrates an environment diagram (100) of a wearable device, such as a smart ring (104) capable of detecting atrial fibrillation, in accordance with an embodiment of the present invention. The wearable device may include but is not limited to the smart ring (104), a wristband, a smart watch, and a standalone PPG sensor. The smart ring (104) may be made using a hypoallergenic material for allowing comfortable and continuous wear by a user (102).

[0020] The smart ring (104) may comprise an outer layer, a middle layer, and an inner layer. The outer layer may be made of a rigid and anti-rust material, such as titanium, transparent material, translucent material such as hardened glass, fibre, sapphire glass, or any other scratch-proof hard material. The middle layer positioned between the outer layer and the inner layer may be a flexible Printed Circuit Board (PCB). The flexible PCB (or rigid Flex) may house a plurality of sensors to capture health parameters of the user (102). One of the plurality of sensors is a PPG sensor configured to collect blood flow data of the user (102) using optical reflection of infrared light or visible light.

[0021] Further, a microcontroller may be mounted on the flexible PCB. All the sensors mounted on the flexible PCB may be connected to the microcontroller. The sensors may transmit values of the health parameters detected by them to the microcontroller, in real-time. The microcontroller may obtain values of the health parameters from the plurality of sensors based on some internal and external triggers associated with the plurality of sensors. Themicrocontroller may also store values of the health parameters (such as, heart parameters) in its own memory or a separate memory element mounted on the flexible PCB.

[0022] A wireless module may also be mounted on the flexible PCB to wirelessly communicate the health parameters to a user device (106), such as a smartphone or a laptop. The wireless module may work on one or more of Bluetooth, a Wi-Fi, a radio frequency, and Near Field Communication (NFC). Further, a battery may be used to power the plurality of sensors, the micro-controller, wireless module, and any other sensor used in the smart ring (104).

[0023] The inner layer of the smart ring (104) may come in contact with the user’s finger once the user (102) wears the smart ring (104). The inner layer may be made of a semitransparent, or completely transparent material. Materials such as glass, plastic, resin, or silicone may be used to fabricate the inner layer. Transparency of the inner layer would allow the sensors to obtain reading from the finger of the user. For example, the PPG sensor may be able to transmit light and obtain reflection of the light through the inner layer.

[0024] The PPG sensor collects signal corresponding to the change in volume of blood flow. The PPG sensor involves illuminating the skin with a light source, typically a Lightemitting diode (LED), and measuring the changes in light absorption or reflection caused by the pulsatile nature of blood flow. Further, the change in blood flow data of the user (102) may be stored in a memory of the smart ring (104). The memory may be a Random-access memory (RAM) or a Read-only memory (ROM). Further, the memory may store a set of computer readable instructions to perform various steps such as, transmitting user data to the user device (106), etc. Further, the smart ring (104) may transmit the user’s blood flow data gathered by the PPG sensor to the user device (106). The user device (106) may receive the blood flow data of the user (102) in real-time. The user device (106) may be, but not limited to, a smartphone, a tablet, a PC, or any other handheld device.

[0025] Further, the user device (106) is configured to extract a plurality of features from the blood flow data of the user (102). The blood flow data may include noise introduced by a plurality of external factors such as, motion of the user, sweating, etc. The user device (106) may extract the plurality of features, upon removing the noise from the blood flow data. The plurality of features of the blood flow data may include, but not limited to, time between heartbeats (RR intervals), variability in RR intervals, an absence of P waves, a presence ofrapid and irregular heartbeat, SpO2, amplitude and width of the peaks, harmonic components of PPG, pulse wave velocity.

[0026] Further, the RR interval indicates the time interval between two consecutive QRS complexes on the electrocardiogram. The RR interval may be measured from the peak or the beginning of an R wave to the peak or the beginning of the next one. Further, the P wave represents atrial depolarization, which results in atrial contraction, or atrial systole. SpO2 indicates peripheral oxygen saturation that measures the amount of oxygen bound to haemoglobin in user’s red blood cells. Further, the amplitude and peak indicate the maximum amplitude and peak of a Photoplethysmogram. The harmonic components of the PPG indicate the noise introduced in the blood flow data due to motion of the user (102), or sweat or dust on the skin of the user (102). Further, the pulse wave velocity indicates the velocity at which the blood pressure pulse propagates through the circulatory system.

[0027] Fig. 2 illustrates a block diagram of the user device (106) connected to the wearable device (smart ring (104), as shown in Fig. 1) capable of detecting atrial fibrillation, in accordance with an embodiment of the present invention. The user device (106) may include at least one processor (162) and a memory (164) to detect AFib. The processor (162) is communicatively coupled with the memory (164). Further, the user device (106) may include a ML model (166) trained to predict a probability of occurrence of atrial fibrillation (AFib) in a user by processing features of the blood flow data. The ML model (166) may be, but not limited to, a logistic regression model, linear regression model, predictive Al model etc.

[0028] The user device (106) may further include one or more interfaces (160). The one or more interfaces (160), the at least one processor (162), and the memory (164) may be interconnected by a system bus and a power supply (not shown in the Fig. 2). The one or more interfaces (160) may be used to provide input to or fetch output from, the user device (106). The one or more interfaces (160) may be implemented as a Command Line Interface (CLI) or a Graphical User Interface (GUI). Further, Application Programming Interfaces (APIs) may also be used for remotely interacting with edge systems and cloud servers.

[0029] Further, the ML model is trained on a server (108) using a labelled dataset corresponding to a plurality of users and the blood flow data. The server (108) is communicatively coupled to the user device (106) via a Wi-Fi module, internet, local areanetwork, etc. The server (108) may be, but not limited to, an online server, a supercomputer, a PC, or any other device capable of training the ML model (166).

[0030] The processor (162) may include one or more general purpose processors (e.g., INTEL® or Advanced Micro Devices® (AMD) microprocessors) and / or one or more special purpose processors (e.g., digital signal processors or Xilinx® System On Chip (SOC) Field Programmable Gate Array (FPGA) processor), MIPS / ARM-class processor, a microprocessor, a digital signal processor, an application specific integrated circuit, a microcontroller, a state machine, or any type of programmable logic array.

[0031] The memory (164) may include, but is not limited to, non-transitory machine- readable storage devices such as hard drives, magnetic tape, floppy diskettes, optical disks, Compact Disc Read-Only Memories (CD-ROMs), and magneto-optical disks, semiconductor memories, such as ROMs, Random Access Memories (RAMs), Programmable Read-Only Memories (PROMs), Erasable PROMs (EPROMs), Electrically Erasable PROMs (EEPROMs), flash memory, magnetic or optical cards, or other type of media / machine- readable medium suitable for storing electronic instructions.

[0032] The memory (164) may include a plurality of storage locations that are addressable by the processor (162) and one or more interfaces for storing software programs and other necessary information (program instructions and machine learning model (166), and feature extraction module (168)) associated with the embodiments described herein. The processor (162) detects AFib by executing program instructions stored in the memory (164). The processor (162) may comprise hardware elements or hardware logic adapted to execute the software programs and manipulate data structures.

[0033] It will be apparent to those skilled in the art that other processor and memory types, including various computer-readable media, may be used to store and execute program instructions pertaining to the techniques described herein. Also, while the description illustrates various processes, it is expressly contemplated that various processes may be embodied as modules configured to operate in accordance with the techniques herein (e.g., according to the functionality of a similar process). Further, while the processes have been shown separately, those skilled in the art will appreciate that processes may be routines or modules within other processes.

[0034] Details are now provided about development of the trained ML model. At first, blood flow data of the user (102) is collected using one or more wearable devices, such as the smart ring (104). Similarly, multiple wearable devices may be worn by multiple users for collecting diverse data. The blood flow data may be stored and relayed to a connected device, such as the server (108), upon receiving a request through Bluetooth. The wearable devices may be worn continuously by multiple users for several days to record blood flow data for a large number of day-to-day activities. Such process aims at encompassing all changes in blood flow volume of the users. Additionally, a number of instances of a user having abnormal heartbeat or rhythm are enacted multiple times by multiple persons of different physical characteristics. This represents the “AFib” instances. The blood flow data from the PPG sensor may be filtered using a Data Signal Processing (DSP) filter, such as a finite impulse response filter and an infinite impulse response filter, to remove noise from the data induced due to motion, sweat, etc. In an embodiment, the DSP filter may be a software implemented filter.

[0035] After a sufficient amount of the labelled dataset corresponding to a plurality of users and the blood flow data are collected, the labelled dataset is processed using a feature extraction implemented by the user device (106). The feature extraction process extracts a plurality of features of the labelled dataset to reduce data dimensionality by analysing pulse wave patterns of the blood flow data. The plurality of features may include, but not limited to, time between heartbeats (RR intervals), variability in RR intervals, an absence of P waves, a presence of rapid and irregular heartbeat, SpO2, amplitude and width of the peaks, harmonic components of PPG, pulse wave velocity.

[0036] The feature extraction process is carried out by a feature extraction module (168) configured to operate through a series of operations. Initially, a pre-processing layer filters the blood flow data using a DSP filter to remove motion artifacts, baseline drift, and background noise. The blood flow data is then passed to a segmentation layer that divides the blood flow data into time -based or beat-wise frames and detects peaks and troughs that correspond to heartbeat events. Subsequently, a time-domain feature layer extracts core features including pulse amplitude, pulse width, and pulse wave velocity, along with R-R intervals, which denote the time between consecutive heartbeats. The time-domain feature layer also calculates variability in R-R intervals using standard deviation of normal-to-normal intervals (SDNN) and root mean square of successive differences (RMSSD), which are well-established heart rate variability metrics used in clinical assessments.

[0037] A frequency-domain feature layer transforms the blood flow data using at least one of a Fourier transform or wavelet transform to extract harmonic components and analyse power spectral density of the blood flow data. Frequency-domain features may be useful in identifying irregularities in waveform morphology that are characteristic of AFib. Additionally, a clinical indicator layer identifies rapid and irregular heartbeat patterns using the variations in RR intervals identified from the harmonic components and power spectral density. Further, the clinical indicator layer detects the absence of P waves (where ECG data is available), and estimates the peripheral capillary oxygen saturation (SpCh). Layered architecture of the feature extraction module ensures that a diverse and diagnostically meaningful set of features is extracted to support accurate and robust prediction of atrial fibrillation.

[0038] Further, the ML model (166) may be executed on the blood flow data of the PPG sensor based on the extracted features and a weightage of each feature. The weightage of each feature may be a pre-defined value assigned to each feature or learned with the model as AFib occurs. In an implementation, the ML model (166) may be made by combining the ML models pre-trained using dataset of reflected light signals captured by the PPG sensor from different users. The AFib may be detected by execution of the ML model (166) on the reading of the PPG sensor. Further, the ML model (166) may be a logistic regression model that may output the probability of AFib, upon observing pattern of the blood flow data of the various users. The logistic regression model may correspond to a simplified model as expressed in equation 1.

[0039] p(A ib) = 1 / (1 + e-(P0 + pixl + P2x2 +.. + pnxn)) ... (1)

[0040] In above equation 1, B0, pi, P2. . .pn are weights learned from the model, and xl, x2...xn are features extracted from the PPG signals. In one embodiment, upon training, a lightweight executable version of the ML model (166) may be implemented in the smart ring (104). In one embodiment, the plurality of features may be transferred to the server (108). Further, a machine learning (ML) model (166) implemented on the server (108), a probability of occurrence of AFib by processing the plurality of features.

[0041] In an embodiment, the user device (106) may also render the blood flow data and the prediction of the occurrence of AFib to a user interface (UI) in real-time via an application installed in the user device (106). In an event of detection of AFib, the user device (106) may notify the user (102) about the AFib through the UI. In some implementations, the wearable devices, such as the smart ring (104), a smartwatch or a smartband may alert the user (102)about AFib by providing notification over display, vibrations, sound patterns, or illumination of LEDs.

[0042] In an embodiment, the smart ring (104) may detect one or more heart conditions based on blood flow data, extracted features of the blood flow data, the associated weightage, and the probability of occurrence of AFib. In one embodiment, the ML model (166) in the user device (106) may detect the one or more heart conditions. The ML model (166) may be trained using a plurality of AFib patterns corresponding to the one or more heart conditions. The one or more heart conditions may include isolated extrasystoles, increased HRV, frequent extrasystoles, trigeminy episode, trigeminy, bigeminy episode, bigeminy, tachycardia, tachycardia episode, bradycardia, bradycardia episode, atrial fibrillation, atrial flutter, etc.

[0043] Figs. 3, 4, and 5 depict photoplethysmograms of different heart conditions. Figs. 3a to 3c depict photoplethysmograms of regular rhythm, isolated extrasystoles, and increased HRV respectively. Figs. 4a to 4i depict photoplethysmograms of frequent extrasystoles, trigeminy episode, trigeminy, bigeminy episode, bigeminy, tachycardia, tachycardia episode, bradycardia, and bradycardia episode respectively, all belonging to non-AFib arrhythmia category. Figs. 5a and 5b depict photoplethysmograms of possible AFib category i.e. atrial fibrillation and atrial flutter respectively.

[0044] Fig. 6 illustrates a flow chart of a method (600) of detecting AFib, in accordance with an embodiment of the present invention. In this regard, each block may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the blocks may occur out of the order noted in the drawings.

[0045] For example, two blocks shown in succession in Fig. 6 may in fact be executed substantially concurrently or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. Any process descriptions or blocks in flow charts should be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps in the process, and alternate implementations are included within the scope of the example embodiments in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionalityinvolved. In addition, the process descriptions or blocks in flow charts should be understood as representing decisions made by a hardware structure such as a state machine.

[0046] The order in which the method is described is not intended to be construed as a limitation, and any number of the described method blocks can be combined in any order to implement the method. Additionally, individual blocks may be deleted from the methods without departing from the scope of the subject matter described herein.

[0047] Furthermore, the method can be implemented in any suitable hardware, software, firmware, or combination thereof. Furthermore, the above-mentioned methods may be implemented in suitable hardware, computer-readable instructions, or a combination thereof. The steps of such methods may be performed by either a system under the instruction of machine-executable instructions stored on a non-transitory computer-readable medium or by dedicated hardware circuits, microcontrollers, or logic circuits. The method may include the following steps.

[0048] At step 602, blood flow data of a user may be received at a device. The blood flow data may be sent from a wearable device such as, a smart ring connected to the device. In one embodiment, the device may be the wearable device. The wearable device includes a PPG sensor. Further, the wearable device may include but is not limited to the smart ring, a wristband, a smart watch, and a standalone PPG sensor. In one embodiment, the device may be a user device. The user device may receive the blood flow data of the user in real-time. The user device may be, but is not limited to, a smartphone, a tablet, a PC, or any other handheld device.

[0049] At step 604, a feature extraction module implementing on the device extracts a plurality of features from the blood flow data. The plurality of features may include time between heartbeats (RR intervals), variability in RR intervals, absence of P waves, presence of rapid and irregular heartbeat, SpO2 level, amplitude and width of peaks, harmonic components of PPG signals, and pulse wave velocity.

[0050] Further, the RR interval indicates the time interval between two consecutive QRS complexes on the electrocardiogram. The RR interval may be measured from the peak or the beginning of an R wave to the peak or the beginning of the next one. Further, the P wave represents atrial depolarization, which results in atrial contraction, or atrial systole. SpO2 indicates peripheral oxygen saturation that measures the amount of oxygen bound tohaemoglobin in user’s red blood cells. Further, the amplitude and peak indicate the maximum amplitude and peak of a Photoplethysmogram. The harmonic components of the PPG indicate the noise introduced in the blood flow data due to motion of the user, or sweat or dust on the skin of the user. Further, the pulse wave velocity indicates the velocity at which the blood pressure pulse propagates through the circulatory system.

[0051] The feature extraction process is carried out by the feature extraction module configured to operate through a series of operations. Initially, a pre-processing layer filters the blood flow data using a DSP filter to remove motion artifacts, baseline drift, and background noise. The blood flow data is then passed to a segmentation layer that divides the blood flow data into time -based or beat-wise frames and detects peaks and troughs that correspond to heartbeat events. Subsequently, a time-domain feature layer extracts core features including pulse amplitude, pulse width, and pulse wave velocity, along with R-R intervals, which denote the time between consecutive heartbeats. The time-domain feature layer also calculates variability in R-R intervals using standard deviation of normal-to-normal intervals (SDNN) and root mean square of successive differences (RMSSD), which are well-established heart rate variability metrics used in clinical assessments.

[0052] A frequency-domain feature layer transforms the blood flow data using at least one of a Fourier transform or wavelet transform to extract harmonic components and analyse power spectral density of the blood flow data. Frequency-domain features may be useful in identifying irregularities in waveform morphology that are characteristic of AFib. Additionally, a clinical indicator layer identifies rapid and irregular heartbeat patterns using the variations in RR intervals identified from the harmonic components and power spectral density. Further, the clinical indicator layer detects the absence of P waves (where ECG data is available), and estimates the peripheral capillary oxygen saturation (SpO?). Layered architecture of the feature extraction module ensures that a diverse and diagnostically meaningful set of features is extracted to support accurate and robust prediction of atrial fibrillation.

[0053] At step 606, a machine learning (ML) model implemented on the device may predict a probability of occurrence of AFib by processing the plurality of features. The ML model may be, but is not limited to, a logistic regression model, a linear regression model, and a predictive Al model.

[0054] Further, the ML model is pre-trained on a server using a labelled dataset corresponding to a plurality of users and the blood flow data. The server is communicatively coupled to the user device via a Wi-Fi module, internet, local area network, etc. The server may be, but is not limited to, an online server, a supercomputer, a PC, or any other device capable of training the ML model.

[0055] In one embodiment, the ML model may be implemented on the server, and the method (600) may include transferring the plurality of features to the server and predicting, by the machine learning model implemented on the server, a probability of occurrence of AFib by processing the plurality of features.

[0056] At step 608, the device provides the probability of occurrence of AFib as a notification to the user. In an embodiment, the notification may be provided on the smart ring. In an implementation, the notification may be provided by at least one of a visual output, audio output and vibrations.Technical Advancement and Economic Significance

[0057] The wearable device for detecting atrial fibrillation using photoplethysmography, disclosed in the present invention, may have the following advantages over conventional art:Early detection of AFib and other heart conditions is facilitated using machine learningbased probability prediction, which enables timely medical intervention and reduces the risk of complications such as stroke.Energy-efficient processing is supported by employing lightweight models such as logistic regression that can run on embedded systems within wearables or smartphones without requiring continuous cloud connectivity.Reduced false alarms are achieved through a feature extraction module that filters out motion artifacts, baseline drift, and signal noise, enhancing the reliability of alerts generated for AFib.

[0058] The specification may refer to “an”, “another”, “one”, or “some” embodiment s) in several locations.

[0059] This does not necessarily imply that each such reference is to the same embodiment(s), or that the feature only applies to a single embodiment. Single features of different embodiments may also be combined to provide other embodiments.

[0060] The terms “or” and “and / or” as used herein are to be interpreted as inclusive or meaning any one or any combination. Therefore, “A, B or C” or “A, B and / or C” mean “any of the following: A; B; C; A and B; A and C; B and C; A, B and C.” An exception to this definition will occur only when a combination of elements, functions, steps or acts are in some way inherently mutually exclusive.

[0061] As used herein, the singular forms “a”, “an”, and “the” are intended to include the plural forms as well, unless expressly stated otherwise. It will be further understood that the terms “includes”, “comprises”, “including” and / or “comprising” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It will be understood that when an element is referred to as being “connected” or “coupled” to another element, it can be directly connected or coupled to the other element or intervening elements may be present. Furthermore, “connected” or “coupled” as used herein may include operatively connected or coupled. As used herein, the term “and / or” includes any and all combinations and arrangements of one or more of the associated listed items.

[0062] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skills in the art to which this disclosure pertains. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0063] Although implementations of a wearable device for detecting atrial fibrillation using photoplethysmography have been described in language specific to structural features and / or methods, it is to be understood that the appended claims are not necessarily limited to the specific features or methods described. Rather, the specific features and methods are disclosed as examples of implementations of a wearable device for detecting atrial fibrillation using photoplethysmography.

[0064] The invention has been described above with reference to numerous embodiments and specific examples. Many variations will suggest themselves to those skilled in this art in light of the above-detailed description. All such obvious variations are within the full intended scope of the appended claims.

Claims

WE CLAIM:

1. A method (600) of detecting atrial fibrillation (AFib), comprising: receiving, by a device (104, 106), blood flow data of a user (102); extracting, by a feature extraction module (168) implementing on the device (104, 106), a plurality of features from the blood flow data, wherein the plurality of features include time between heartbeats (RR intervals), variability in RR intervals, absence of P waves, presence of rapid and irregular heartbeat, SpO2 level, amplitude and width of peaks, harmonic components of PPG signals, and pulse wave velocity; predicting, by a machine learning (ML) model (166), a probability of occurrence of AFib by processing the plurality of features; and providing, on the device (104, 106), the probability of occurrence of AFib as a notification to the user (102).

2. The method (600) as claimed in claim 1, wherein the blood flow data is collected using a PPG sensor present in the smart ring (104).

3. The method (600) as claimed in claim 1, wherein extracting the plurality of features from the blood flow data comprises: filtering, by a pre-processing layer of the feature extraction module (168), the blood flow data using a data signal processing (DSP) filter to remove one or more of motion artifacts, baseline drift, and noise; segmenting, by a segmentation layer of the feature extraction module (168), the blood flow data into at least one of time-based and beat-wise frames, and detecting peaks and troughs in the blood flow data; extracting, by a time-domain feature layer of the feature extraction module (168), pulse amplitude, pulse width, pulse wave velocity, RR intervals, and variability in RR intervals using standard deviation of normal-to-normal intervals (SDNN) and root mean square of successive differences (RMSSD);applying, by a frequency-domain feature layer of the feature extraction module (168), at least one of a Fourier and a wavelet transform on the blood flow data to obtain harmonic components and power spectral density; and identifying, by a clinical indicator layer of the feature extraction module (168), at least one of irregular and rapid heartbeats from the variability in RR intervals identified using the harmonic components and power spectral density.

4. The method (600) as claimed in claim 1, wherein the clinical indicator layer further identifies absence of P waves, and estimates SpCh level.

5. The method (600) as claimed in claim 1, wherein the ML model (166) is a logistic regression model trained on a dataset of blood flow data corresponding to multiple users.

6. The method (600) as claimed in claim 1, wherein the device (104, 106) is at least one of a user device (106) and a smart ring (104).

7. The method (600) as claimed in claim 6, wherein the ML model (166) is implemented on the user device (106).

8. The method (600) as claimed in claim 6, wherein a lightweight executable version of the ML model (166) is executed on the smart ring (104).

9. The method (600) as claimed in claim 1, comprising: transferring the plurality of features to a server (108); and predicting, by a machine learning (ML) model (166) implemented on the server (108), a probability of occurrence of AFib by processing the plurality of features.

10. The method (600) as claimed in claim 1, wherein predefined weightages are assigned to each feature of the plurality of features.

11. The method (600) as claimed in claim 1, wherein the notification is provided as at least one of a visual output, vibration, and an audio output.

12. The method (600) as claimed in claim 1, wherein the notification is provided on the smart ring (104).

13. The method (600) as claimed in claim 1, wherein the method (600) further comprises: determining, by the ML model, one or more heart conditions upon identifying AFib, wherein the one or more heart conditions comprise isolated extrasystoles, increased heart rate variability (HRV), frequent extrasystoles, trigeminy episode, trigeminy, bigeminy episode, bigeminy, tachycardia, tachycardia episode, bradycardia, bradycardia episode, atrial fibrillation, and atrial flutter.

14. The method (600) as claimed in claim 1, wherein the probability of occurrence of AFib and the plurality of features are transferred to the server (108) for future reference.

15. A device (104, 106) for detecting atrial fibrillation (AFib), comprises: one or processors (162) and a memory (164) coupled with the one or more processors (162), wherein the memory (164) stores program instructions configured to: receive a machine learning (ML) model (166) pre-trained on a server (108), wherein the server (108) is communicatively coupled with the device (104, 106); receive blood flow data of a user (102); extract a plurality of features from the blood flow data using a feature extraction module (168) implementing on the device (104, 106), wherein the plurality of features include time between heartbeats (RR intervals), variability in RR intervals, absence of P waves, presence of rapid and irregular heartbeat, SpO2 level, amplitude and width of peaks, harmonic components of PPG signals, and pulse wave velocity; predict a probability of occurrence of AFib by processing the plurality of features by the ML model (166); and provide the probability of occurrence of AFib as a notification to the user (102).

16. The device (104, 106) as claimed in claim 15, is at least one of a user device (106) and a smart ring (104).

Images