Classification of ECG Signal Using CNN Algorithm Md Tarique Anis Department of Electrical Engineering National Institute of Technology Hamirpur Hamirpur, India 20mee108@nith.ac.in Dr Veena Sharma Department of Electrical Engineering National Institute of Technology Hamirpur Hamirpur, India veena@nith.ac.in Abstract— It is widely known that heart disease is among the most significant cause of loss of life worldwide. Therefore, early detection of heart diseases is important to reduce the rising death rate. An electrocardiogram can detect many types of heart diseases including abnormal heart rhythms. We propose here a method for categorizing heart diseases entirely focused on ECG by using a machine learning method, known as Convolution Neural Networks (CNNs) into five categories as per the Association for Advancement of Medical Instrumentation (AAMI) standards. The outcomes conveys that the suggested methodology outperforms several previously available methods for ECG signal classification in respect of the classification accuracy and computing efficiency. Keywords— Convolution Neural Networks (CNNs), ECG, Electrocardiogram, AAMI Standard. I. INTRODUCTION According to many health organizations including the WHO the cardiovascular illnesses is one of the leading causes of mortality worldwide. An electrocardiogram (ECG) is one of the most commonly used method used for getting the electrical activity generated by the heart. It is commonly used because of its non- invasive method. It is commonly employed in the detection of cardiovascular diseases. The term Arrhythmia refers to irregularities in the cardiovascular functions such as the rhythm, rate or the conduction of electrical signals through the heart and is the most commonly referred as cardiac diseases [1]. And these cardiac abnormalities can affect cardiac electrical activity, which can be recognized by analyzing an ECG waveform, composed of various electrical signals that are linked with heart activity and can provide important information about a patient's heart condition. There are mainly three types of ECG Signals: x Resting ECG: This ECG is carried out when the subject is lying down comfortably. x Stress or Exercise ECG: This ECG is carried out when the subject is using a treadmill or doing exercise on an exercise bike. x Ambulatory ECG or Holter ECG: In this type of ECG a small portable device consisting of electrodes is wrapped around the waist which can monitor the heartbeat for one or more days To get the most accurate diagnosis, an ECG using 10 electrodes capturing 12 leads (signals) is performed. Each lead examines the electrical activity of the cardiovascular system in a different way. 12 leads must be used to record an accurate result. Each ECG consist of five waves, which correspond to different phases of the heart's activity: P, Q, R, S, and T. In ECG waveform the first positive deflection is known as P wave. Depolarization of the ventricles is represented by the QRS complex, but the fact is that QRS complex may not always display all three waves, and its normal duration is between 0.08 seconds to 0.10 seconds i.e 80 to 100 milliseconds. Often, for doctors it is difficult to examine the lengthy ECG recordings in a relatively short time frame and observing the alterations caused by the abnormalities of the heart and the cardiovascular system, hence there is a need for automatic classification of ECG signals. Automation of ECG signal categorization is a difficult task to solve for many points. Primary ECG patterns from different patients can have morphological and temporal differences in their waveforms. Distinct patients with different cardiac beats may have similar ECG waveforms, while the same patient may have different ECG waveforms recorded at different time interval. The other factor involved in ECG signal classification is heart beat variability. Heart rate is affected by conditions such as stress, excitement, and exercise, and can result in difference in the ECG signal features, also there is no optimal classification rule for ECG signals.[1] ECG arrhythmia classification also faces the challenge of developing the classifier capable of classifying arrhythmia in real time. Automatic heart rate classification has been previously reported by many researchers using different functions and different classification methods to represent ECG signal [2]. Many classification and detection algorithms have been developed based on different techniques like artificial neural networks [3], hidden Markov models [4] and many more. But the main problem associated with the above classification model is that they perform good on training data but when exposed to ECG signal of different patient performance deteriorates. The extraction of features is an important step in ECG categorization. Various preprocessing and feature extraction strategies for ECG classification have been published by a 185 978-1-6654-8385-8/22/$31.00 c 2022 IEEE 2022 International Conference on Electronic Systems and Intelligent Computing (ICESIC) | 978-1-6654-8385-8/22/$31.00 ©2022 IEEE | DOI: 10.1109/ICESIC53714.2022.9783598 Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY HAMIRPUR. Downloaded on June 03,2022 at 05:50:03 UTC from IEEE Xplore. Restrictions apply.