This study gives efficient approach to classify ECG signals with high accuracy. Each heartbeat is a combination of action impulse waveforms produced by different specialized cardiac heart tissues. Heartbeats classification faces some difficulties because these waveforms differ from person to another, they are described by some features. These features are the inputs of machine learning algorithm.
Multiples classifiers are proposed for ECG classification, these classifiers are used mostly in Big Data and Machine Learning fields by the weighted voting principle. Each classifier influences the final decision according to its performance on the training data.
Parameters of each classifier are adjusted on the basis of an individual classifier’s performance on the training data by applying the pseudoinverse technique.
Introduction
1. Background:
Millions of people suffer from irregular heartbeats (arrhythmias), which can be life-threatening. Therefore, accurate and low-cost detection methods are essential. Electrocardiograms (ECGs), which are non-invasive and widely used in clinical practice, play a central role in diagnosing heart conditions.
2. Problem Statement:
Classifying ECG signals accurately is challenging due to variability in heartbeat waveforms between individuals. These waveforms must be analyzed and their features extracted for machine learning algorithms to classify heartbeats as normal or abnormal.
3. Objectives:
Develop a high-accuracy ECG classification system using machine learning.
Use multiple classifiers and combine their decisions via a weighted voting method.
Adjust classifier parameters using the pseudoinverse technique.
Validate the approach on a large ECG dataset (205,146 records from 51 patients).
Implement using Big Data tools (Spark–Scala) on a local PC.
4. Proposed System Overview:
Treats heart disease detection as a binary classification problem (disease/no disease).
Uses supervised machine learning to train models on labeled ECG data.
Once trained, models can predict whether a patient has heart disease based on new ECG data.
5. Literature Review:
Discusses the use of the Modified Early Warning Score (MEWS) in hospitals to identify patients at risk of cardiac arrest. A retrospective study found that higher MEWS scores correlated with lower survival rates after cardiac arrest, although the study had limitations like small sample size and being conducted in only one emergency department.
6. System Analysis – Advanced ML System:
Proposes an advanced real-time cardiac arrest detection system using deep learning models (CNNs, RNNs, attention mechanisms).
Integrates multi-modal physiological data (ECG, HRV, blood pressure).
Employs wearable devices and real-time monitoring.
Uses transfer learning for faster development and improved performance.
Provides intelligent alerts to healthcare providers for early intervention.
7. System Design:
Architecture: Involves data input, visualization, preprocessing, model building using 5 ML classifiers, and final result output.
Use Case Diagram: Shows user interaction with the system for input and result viewing; admin can access and manage the patient database.
Data Flow Diagram: Describes the data journey from raw input → preprocessing → train/test split → model training/classification.
Sequence Diagram: Illustrates the step-by-step flow from user input to final prediction output.
Key Technologies and Approaches Used:
ECG signal processing and feature extraction
Machine Learning and Big Data tools (Spark–Scala)
Classification algorithms with weighted voting
Real-time monitoring using wearable devices
Deep learning (CNN, RNN, attention mechanisms)
Transfer learning for model enhancement
Conclusion
In conclusion, the utilization of machine learning approaches for the early detection of cardiac arrest presents a promising avenue for improving patient outcomes and reducing the burden of cardiovascular diseases. The proposed system harnesses the power of advanced neural network architectures, multi-modal data integration, and real-time monitoring capabilities to enable timely detection, personalized risk assessment, and proactive intervention. By continuously analyzing cardiovascular data in real-time and providing early alerts to healthcare providers or patients, the system facilitates prompt clinical response and potentially prevents adverse cardiac events. Furthermore, the system prioritizes interpretability, reliability, and scalability, ensuring its suitability for diverse healthcare settings and patient populations. Through ongoing validation, refinement, and collaboration with medical experts and stakeholders, the proposed system has the potential to revolutionize cardiac care, ultimately saving lives and improving the quality of life for individuals at risk of cardiac arrest.
References
[1] An-Yi Wang 1,Cheng-Chung Fang 2, Shyr-Chyr Chen 2, Shin-Han Tsai 3, Wei-Fong Kao 4 Affiliations expand Epub 2015 Dec 24.
[2] Yun Gi Kim # 1, Kyongjin Min # 2, Joo Hee Jeong 1, Seung-Young Roh 3, Kyung-Do Han 4, Jaemin Shim 1, Jong-Il Choi 5, Young-Hoon Kim 1 2024 Jan 27;14(1):2289.doi: 10.1038/s41598-024-52859-x.
[3] Vahid HOUSHYARIFAR, Mehdi Chehel AMIRANI,Turk J Elec Eng & Comp Sci (2017) 25: 1541 – 1553 ?c TUB¨ ?ITAK doi:10.3906/elk-1509-14,https://journals.tubitak.gov.tr/elektrik.
[4] Justin Chan, Thomas Rea, Shyamnath Gollakota & Jacob E. Sunshine npj Digital Medicine volume 2, Article number: 52 (2019).
[5] Ryo Ueno,Liyuan Xu, Wataru Uegami,Hiroki Matsui, Jun Okui, Hiroshi Hayashi, Toru Miyajima, Yoshiro Hayashi,David Pilcher, Daryl Jones,Published online 2020 Jul 13.
[6] Joon-Myoung Kwon 1,Youngnam Lee 2, Yeha Lee 2, Seungwoo Lee 2, Jinsik Park 2018 Jun 26;7(13):e008678. doi: 10.1161/JAHA.118.008678.
[7] Apeksha Shah1 , Swati Ahirrao1 , Sharnil Pandya1 *, Ketan Kotecha2 and Suresh Rathod1,published: 22 October 2021 doi: 10.3389/fpubh.2021.762303.
[8] Minsu ChaeHyo-Wook GilNam-Jun ChoHwamin Lee,Mathematics 2022, 10(12), 2049; https://doi.org/10.3390/math10122049 Submission received: 2 May 2022 / Revised: 1 June 2022 / Accepted: 9 June 2022 / Published: 13 June 2022.