| CPC A61B 5/7267 (2013.01) [A61B 5/0205 (2013.01); A61B 5/352 (2021.01); G16H 40/67 (2018.01); G16H 50/20 (2018.01); G16H 50/30 (2018.01); A61B 5/02405 (2013.01); A61B 5/02416 (2013.01); A61B 5/361 (2021.01); A61B 2562/0219 (2013.01)] | 9 Claims |
|
1. A processor implemented method for detection of Atrial Fibrillation (AF) and cardiac disorder from biological signals for managing cardia health or disorders of one or more users during locomotion or movement or day-to-day activities, comprising:
acquiring a plurality of biological signals by sensory devices from the one or more users while the one or more users is performing one or more activities, wherein the sensory devices are implantable devices or pulse device connected to the one more users to capture the plurality of biological signals of the one or more users while performing the one or more activities, wherein each of the plurality of biological signals is unique and pertains to the one or more corresponding users, wherein the plurality of biological signals comprises an electrical signal acquired from a resistive sensor or a capacitive sensor, a cardiac pulse signal acquired from an optical sensor, and an arterial pulse signal acquired from a piezo sensor or doppler, wherein the electrical signal comprises an ECG signal;
pre-processing the obtained plurality of biological signals to obtain a plurality of filtered signals, wherein
if the filtered signal is identified as the ECG signal, then identifying a corresponding lead for the ECG signal and checking if negative R peak is present in the filtered ECG signal, and if yes, then correcting the filtered ECG signal by inverting the ECG signal, wherein correction of the ECG signal based on negative R-peak detection provides reliable HRV features extraction from the ECG signal;
if the filtered signal is identified as the ECG signal, and R Peak detected in the ECG signal, then R-R intervals were checked for missing R Peaks and noisy artefacts, and if yes, then the R-R intervals of the ECG signal are corrected to provide a corrected ECG signal for extraction of HRV features;
selecting, automatically one or more feature extraction modules from a plurality of feature extraction modules including a Heart Rate Variability (HRV) feature extraction module, a f-wave feature extraction module and a fiducial feature extraction module, comprised in a memory based on one or more features comprised in each of the plurality of filtered signals by optimizing system resources including the memory and a processor and utilize the system resources to perform feature extraction of specific features set, by,
automatically invoking, by the hardware processor, the HRV feature extraction module, the f-wave feature extraction module and the fiducial feature extraction module from the memory by the processor, when all fiducial points including PORST points are present in the biological signal and detects presence of f-wave in the biological signal,
automatically invoking, by the hardware processor, the HRV feature extraction module and the f-wave feature extraction module from the memory, when one of a fiducial point is absent in the biological signal and detects presence of the f-wave in the biological signal, and
automatically invoking, by the hardware processor, the HRV feature extraction module and the fiducial feature extraction module from the memory, when all the fiducial points are present in the biological signal and detects absence of the f-wave in the biological signal,
wherein if the filtered signal is identified as the ECG signal, and R Peak is detected in the ECG signal, then slope based fiducial feature extraction is performed to output the fiducial point features;
wherein if the filtered signal is identified as the ECG signal, and R Peak is detected in the filtered ECG signal, and the filtered ECG signal is fed as an input into the F-wave feature extraction module and when slope based fiducial feature extraction is performed to output P, Q, R, and T points wherein ECG signal lying between P offset or P end point and T offset or T end point is extracted and filtered between a range of x and y wherein value of x is 4 Hertz and y is 10 Hertz, then the output is a F-wave signal, and the F-wave signal is utilized for F-wave features extraction;
dynamically extracting features set comprising at least one of one or more HRV features, one or more f-wave features, and one or more fiducial point features comprising the PQRST points, using the selected feature extraction module, wherein the one or more HRV features refer to variability in beat to beat intervals of the heart rate, essentially reflecting changing effect of sympathetic and parasympathetic modulation of an autonomic nervous system, wherein the one or more fiducial point features are extracted when R Peak are detected from the biological signal and a slope of the biological signal within a window size of ‘n’ number of samples is evaluated to extract morphological features, wherein the slope includes both positive and negative values due to increasing and decreasing peaks in an ECG waveform, and the window size depends on the number of samples between a Q peak and the R peak in the ECG signal, wherein, for finding the window size, the R peak is found by differentiating the ECG signal and the Q wave is detected as the negative peak immediately prior to the detected R peak, and the window is placed at a 1st sample and the slope between the 1st sample and (n+1)th sample is found and stored, further the window is placed on a 2nd sample and the slope between the 2nd sample and (n+2)th is found, wherein the window is placed at all samples till the (N−n)th sample and the slope is found and stored as Sslope signal;
applying a feature reduction technique on the extracted features set to obtain a reduced feature set;
segmenting the reduced features set into test data and training data;
generating one or more machine learning models for biological signal classification using the training data;
automatically selecting at least one machine learning model for biological signal classification from the one or more generated machine learning models for biological signal classification based on the input signal and applying the at least one selected machine learning model on the test data to classify each of the plurality of biological signals as one of Atrial Fibrillation (AF), a non-AF, a cardiac arrythmia disorder, or ischemia;
obtaining a plurality of accelerometer features from an inbuilt accelerometer or an externally connected wearable device of the implantable device connected to the one or more users;
automatically detecting, using an accelerometer classification model (ACM), one or more activities of the one or more users based on the obtained plurality of accelerometer features, wherein the detected one or more activities comprises walking, jogging, sitting, standing; and
classifying with improvisation, using the weighted sum classification model, each of the plurality of biological signals as one of Atrial Fibrillation (AF), a non-AF, a cardiac arrythmia disorder, or ischemia based on the output of the machine learning model and the detected one or more activities to characterize a cardiac health disorder of the one or more users,
continuously monitoring and managing cardiac health of the one or more users via the one or more hardware processors and communication interface or via the wearable device by using the classified biological signal(s) during locomotion or movement or day-to-day activities.
|