image
image
user-login
Patent search/

MACHINE LEARNING-BASED SYSTEM FOR REAL-TIME DETECTION AND CLASSIFICATION OF CARDIAC ARRHYTHMIAS

search

Patent Search in India

  • tick

    Extensive patent search conducted by a registered patent agent

  • tick

    Patent search done by experts in under 48hrs

₹999

₹399

Talk to expert

MACHINE LEARNING-BASED SYSTEM FOR REAL-TIME DETECTION AND CLASSIFICATION OF CARDIAC ARRHYTHMIAS

ORDINARY APPLICATION

Published

date

Filed on 5 November 2024

Abstract

The real-time detection and classification of cardiac arrhythmias from electrocardiogram (ECG) data using a machine learning based system is presented by this invention. The integrated system is targeted at improving diagnostic accuracy and efficiency by integrating several modules that work in unison to uphold efficient cardiac health monitoring. ECG Data Acquisition Module receives signals from multiple sources such as wearable devices and standard ECG machines. The Data Preprocessing Module lowers the raw data and extracts them without noise or artefacts to ensure input of high quality for further analysis. To this, the arrival of Machine Learning throws an important role to this arrhythmia detection, and it utilizes Deep Learning Algorithms to classify the input data into normal and arrhythmic patterns. If an arrhythmia is found, Classification and Alert Module generates timely alerts to healthcare providers and patients so that they can take the appropriate action immediately. Furthermore, the system incorporates a Feedback Loop for Model Improvement which autonomously steers the machine learning model towards improvement with ever new data. This invention significantly advances the ability to monitor cardiac health by offering a scalable, reliable and adaptable solution to the monitoring of cardiac arrhythmias in clinical and remote settings.

Patent Information

Application ID202441084375
Invention FieldBIO-MEDICAL ENGINEERING
Date of Application05/11/2024
Publication Number46/2024

Inventors

NameAddressCountryNationality
D JyothirmaiDepartment of CSE, B V Raju Institute of Technology, Vishnupur, Narsapur, Medak, Telangana 502313IndiaIndia
Sainadh Singh KshatriDepartment of EEE, B V Raju Institute of Technology, Vishnupur, Narsapur, Medak, Telangana 502313IndiaIndia
R PitchaiDepartment of CSE, B V Raju Institute of Technology, Vishnupur, Narsapur, Medak, Telangana 502313IndiaIndia
P Krishna KishoreDepartment of Information Technology, BVRIT HYDERABAD College of Engineering for Women, HyderabadIndiaIndia

Applicants

NameAddressCountryNationality
B V RAJU INSTITUTE OF TECHNOLOGYDepartment of CSE, B V Raju Institute of Technology,Vishnupur, Narsapur, Medak, Telangana 502313IndiaIndia

Specification

Description:Field of the invention
[001] In particular, the present invention concerns the detection of cardiac arrhythmia, and a system and method that accurately detect, classify, and monitor cardiac arrhythmic events using machine learning techniques, based upon electrocardiogram (ECG) data.
Description of Related Art
[002] Cardiac arrhythmias are any abnormalities of the heart's rhythm, from minor to severe rhythm irregularities to dangerously severe cardiac arrhythmias. Arrhythmias include atrial fibrillation, ventricular tachycardias, and bradycardias; all of which can be detrimental if not detected and treated. These conditions are also linked to risks such as heart failure, stroke, as well as sudden cardiac arrest, and managing these conditions depends upon timely detection.
[003] Current approaches for arrhythmia detection are traditionally met on a practice of staff used electrocardiogram (ECG) analysis by trained professions in case of arrhythmia. But manual interpretation is the time consuming and they are subject to human errors, susceptible to delayed or wrong diagnoses. Cardiac monitoring requires high demand for precision & speed, in particular in critical care & remote health settings, making it highly desirable for more advanced, automated solutions.
[004] There is a promising avenue for automating arrhythmia detection with recent advances in machine learning. With the exception of certain algorithms which are based on deep learning methods, machine learning algorithms are good at processing large datasets and finding patterns (or learning) within the ECG data. These algorithms find equal or superior sensitivity for arrhythmia detection to even experienced clinicians.
[005] However, current approaches often lack generalization without access to large amounts of labeled data for training, are computationally inefficient, and may suffer from high false positive rates that can result in unnecessarily giving medical interventions to patients or patient distress. Additionally, current machine learning models can also not adapt to new data or train on individual variations of the patient over time, which is necessary for applications of continuous monitoring.
[006] In order to overcome the limitations presented above, this invention aims to develop a highly accurate, real time detection system that can be integrated with many kinds of ECG acquisition devices including a hospital level monitor and a wearable health device. A structured, multi layered machine learning approach is used to combine accuracy whilst minimizing the computational demand, which makes it suitable for hospital and remote monitoring environments.
[007] Furthermore, the system incorporates a feedback mechanism to continuously fine tune its performance as new data streams are introduced, and adapt and remain reliable over the long term.
[008] The improvements in this invention seek to bring substantial enhancement in the detection of cardiac arrhythmias, and help healthcare providers to diagnose and manage cardiac conditions better, and empower patients to monitor their cardiac abnormalities and better while having the information at their fingertips. This forms a significant step toward the design of refined, scalable and flexible systems for cardiac health management.
SUMMARY
[009] To increase accuracy, efficiency and adaptability in arrhythmia diagnostics, the proposed invention is a real time machine learning based system for detecting and classifying the cardiac arrhythmias with ECG data. This system solves important problems in cardiac monitoring, including providing an automated solution that can adapt to many settings, ranging from hospital to remote patient monitoring.
[0010] The invention utilizes a multi-layer machine learning approach to achieve high accuracy while minimizing computational burden necessary for continuous real time monitoring without a penalty on available computational resources. The invention includes a feedback mechanism whereby it is able to adapt to new data as well as refine its accuracy over time to build a highly adaptive and dynamic solution.
[0011] The system consists of a number of components working together to reliably detect and categorize arrhythmias. The ECG Data Acquisition Module acquires the ECG signals from the patient (acquired from wearable devices, mobile health technology, or standard ECG machines).
[0012] This allows the system to be used in different settings and with different devices, so that it can works in a clinical and home environment. ECG signals are then collected and processed by the Data Preprocessing Module, for example, removing noise and artifacts, which are very common in ECG data, and may get in the way of accurate analysis. This module uses advanced filtering and smoothing techniques to clean up ECG data, leaving the ECG data reliable for additional analysis, thereby enhancing reliability of the detection process.
[0013] The Feature Extraction Module performs the preprocessing to the ECG data, and then extracts significant features from the ECG data including RR intervals, QRS duration, P wave morphology and other attributes used in arrhythmia detection. This input then allows the machine learning model to identify subtle patterns of different types of arrhythmias. A large dataset of annotated ECG readings is provided to the machine learning model (a CNN or RNN, or else a hybrid architecture), that is trained on a variety of arrhythmic and non-arrhythmic classes. This model structure facilitates the recognition of spatial and temporal patterns in ECG data, and jointly enhances the ability of the model to distinguish normal from abnormal rhythms with excellent performance.
[0014] After processing, classifying and alerting of the ECG data, the Classification and Alert Module will alert you when an arrhythmia is detected. When healthcare providers, caregivers, or the patient themselves get this alert, it means that high risk arrhythmias can be immediately intervened. It is designed to identify and differentiate between arrhythmias categories, in a way that provides healthcare providers with information about what kind of arrhythmia was detected in order to help them make smart, urgent decisions. This alert mechanism is especially valuable for patients who are at high risk of cardiac event, in which they provide proactive cardiology care.
[0015] The system also includes a Feedback Loop for Model improvement that helps it continues to improve it's performance as new data gets introduced to it. The system receives new ECG data and retrains and refines itself by feeding the new ECG data back into the machine learning model and keeping high accuracy all the time and adapting to individual patient variations. More than just robustifying the system, this also allows us to leverage this feedback to adapt to changes in patient health conditions.
BRIEF DESCRIPTION OF THE DRAWINGS
[0016] FIG. 1 A schematic diagram of data flow for Detection of Cardiac Arrhythmia;
DETAILED DESCRIPTION
[0017] An all-encompassing machine learning based system invention is introduced which can successfully detect as well as classify cardiac arrhythmias with high accuracy and real time responsiveness. Here, the system consists of interconnected modules, which are designed to perform a particular stage of the arrhythmia detection process, but work in concert to produce a robust, dependable and efficient solution suitable for both the clinical and remote surveillance environments.
[0018] The ECG Data Acquisition Module first collects ECG signals from the patient directly, and then processes them. The types of ECG sources supported by this module include hospital grade ECG machines, wearable monitoring devices and mobile health applications. The invention maintains compatibility with a variety of devices to allow flexibility in data acquisition to continuously monitor patients in various environments. Real-time or at intervals depending on the patients need and device capabilities, this module captures data.
[0019] The Data Preprocessing Module then cleans the raw ECG data from noise and artifacts that are quite typical in ECG signals and that can otherwise contribute to a loss in the accuracy of the analysis. Baselines wander, powerline interference, and motion artifacts are all addressed using advanced filtering techniques in this module.
[0020] The metal detritus is then removed from the ECG signal by cleaning it and then segmenting the cleaned ECG signal into individual heartbeats or cycles so that each cycle is isolated for further analysis. The importance of this preprocessing step is evident since top quality features are extracted in subsequent stages through accurate segmentation of the data and noise reduction.
[0021] The Key Characteristics are extracted in the Feature Extraction Module from preprocessed ECG data. Specifically, it makes use of advanced analytical techniques, including wavelet transforms or principal component analysis (PCA), to extract critical features, i.e., RR intervals, P-wave morphology, QRS complex duration, T-wave morphology and attributes (e.g., amplitude variations). They serve to distinguish normal from arrhythmic heart patterns. With focus on these key ECG characteristics, the invention is able to maximize its ability to detect even subtle irregularities associated with arrhythmias and to achieve high sensitivity and specificity in diagnosis.
[0022] The system was built based on the Machine Learning Model, which takes advantage of deep learning algorithms to classify ECG data in normal and arrhythmic patterns. The model can use Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs) or a hybrid, so it can handle both the spatial and temporal patterns in the ECG data. A large and diverse dataset of Labeled ECG recordings are used to train the model, spanning a very broad range of arrhythmic, as well as non-arrhythmic conditions.
[0023] This extensive training trains the model to recognize many arrhythmias: arrhythmia atrial fibrillation, arrhythmia ventricular tachycardia, arrhythmia bradycardia, etc. The model's architecture is multi layered so it can be thrown in the deep learning feast for the heavy flavor and can still work with high accuracy, and should not hinder its ability to be computed efficiently on real time applications.
[0024] After the ECG data is processed and classified, Classification and Alert Module translates the output from the model. This module sends an alert if an arrhythmia is detected to healthcare providers, caregivers or directly to the patient. This alert system provides very specific types of arrhythmia detected, meaning that healthcare professionals can rapidly and accurately identify what treatment their patients need. This proactive alert mechanism is particularly valuable for identifying subjects at high risk of cardiac events, who can be alerted early and will see better outcomes as a result.
, Claims:1. I/We Claim: A method for detecting cardiac arrhythmias, comprising:
a. Acquiring ECG data from a patient using an ECG device;
b. Preprocessing the ECG data to remove noise and artifacts;
c. Extracting features relevant to arrhythmia detection;
d. Utilizing a machine learning model trained to classify ECG data as normal or arrhythmic; and
e. Generating an alert if an arrhythmia is detected.
2. I/We Claim: A system for real-time arrhythmia detection, comprising:
a. An ECG Data Acquisition Module configured to collect and transmit ECG data;
b. A Data Preprocessing Module configured to filter and segment the data;
c. A Feature Extraction Module for deriving attributes indicative of arrhythmic events;
3. I/We Claim: A Machine Learning Model for classifying ECG data; and
4. I/We Claim: An alert mechanism that notifies healthcare providers when an arrhythmic event is detected.
5. I/We Claim: A feedback mechanism within the machine learning model that enables retraining using newly acquired data to refine its predictive accuracy and adaptability.
6. I/We Claim: The use of CNNs, RNNs, or a hybrid architecture in the machine learning model to capture temporal and spatial features within ECG data for accurate arrhythmia classification

Documents

NameDate
202441084375-COMPLETE SPECIFICATION [05-11-2024(online)].pdf05/11/2024
202441084375-DECLARATION OF INVENTORSHIP (FORM 5) [05-11-2024(online)].pdf05/11/2024
202441084375-DRAWINGS [05-11-2024(online)].pdf05/11/2024
202441084375-FORM 1 [05-11-2024(online)].pdf05/11/2024
202441084375-REQUEST FOR EARLY PUBLICATION(FORM-9) [05-11-2024(online)].pdf05/11/2024

footer-service

By continuing past this page, you agree to our Terms of Service,Cookie PolicyPrivacy Policy  and  Refund Policy  © - Uber9 Business Process Services Private Limited. All rights reserved.

Uber9 Business Process Services Private Limited, CIN - U74900TN2014PTC098414, GSTIN - 33AABCU7650C1ZM, Registered Office Address - F-97, Newry Shreya Apartments Anna Nagar East, Chennai, Tamil Nadu 600102, India.

Please note that we are a facilitating platform enabling access to reliable professionals. We are not a law firm and do not provide legal services ourselves. The information on this website is for the purpose of knowledge only and should not be relied upon as legal advice or opinion.