ECG signal analysis using continuous wavelet transformation and deep neural network

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 09 | Sep 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 773
1Professor Grade2, SENSE, Vellore Institute of Technology, Katpadi, Tamil Nadu-632014
2,3,4,5 Student, SENSE, Vellore Institute of Technology, Katpadi, Tamil Nadu-632014
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The objective of this research is to develop an
algorithm that can identify and classify various
electrocardiogram (ECG) data beat kinds. Using the Deep
The neural network and the Continuous Wavelet Transform
(CWT). The goal is to teach a CNN how to distinguish
between ARR, CHF, and NSR. Identification and treatmentof
arrhythmias can reduce the risk of mortality from
cardiovascular disease (CVD). The electrocardiogram (ECG)
is examined beat by beat in clinical practise to make the
diagnosis, but this is frequently time-consuming and
challenging. In this study, we describeanautomatedmethod
for classifying ECGsbasedon ContinuousWaveletTransform
(CWT) and Convolutional Neural Network (CNN) (CNN).
While CNN is used to extract featuresfromthe2D-scalogram
created from the time-frequency components stated above,
CWT is used to break down ECG signals into discrete time-
frequency components. Four RR interval characteristics are
collected, combined with CNN features, andtheninputintoa
fully connected layer for ECG classification because the
surrounding R peak interval, also known as the RR interval,
is crucial for identifying arrhythmia. In the MITBIH
arrhythmia database, our method achieves an overall
performance of 70.75%, 67.47%, 68.76%, and 98.74% for
positive predictive value, sensitivity,F1-score,andaccuracy,
respectively. In comparison to earlier methods, our
technique raises the overall F1- score by 4.7516.85%.
1. INTRODUCTION
An irregular heartbeat known as an arrhythmia is one of the
main reasons why people die from cardiovascular disease
(CVD). While the majority of arrhythmias are benign, some
can be deadly. For instance, atrial fibrillation can cause
cardiac arrest and strokes. It needs to be treated right away
because it is so dangerous. The World Health Organization
(WHO) estimates that 17.5 million deaths worldwide were
attributable to CVD in 2012. By 2030, 23 million fatalities
from CVD are expected to have occurred. Additionally,
treatments for CVD, including medical care,areprohibitively
expensive. Over US $3.8 trillion is expected to be spent in
low- and middle-income countries between 2011 and 2025.
For this goal, researchers have developed a system that
automatically categorises heartbeats in ECG measurements.
Most methods involve categorization and feature extraction.
RR interval characteristics and heartbeat morphology are
frequently used. For categorization, a variety of methods
were used, including continuous wavelet transformation,
deep neural networks,andartificial neural networks(ANNs).
Even though these methods have a high level of
effectiveness, different people have ECG waves with very
diverse morphologies, and even the same patient can have
different ECG waves at different times. The fixed features of
these methods are insufficient for consistently
differentiating arrhythmia in differentindividuals. The rapid
advancement of deep neural networks has led to a recent
rise in popularity for methods based on deep learning. Deep
learning can automatically derive discriminant properties
from training data as a representation learning strategy.
Numerous studies indicate that deep learning-based
methods for classifying ECGs may be able to extract more
abstract traits and eliminate patient-specific discrepancies.
Because the ECG signal contains so many different types of
frequencies, it will be challenging to categorise the different
signals of the ECG, which will be made even more
challenging if we utilise straightforward deep neural
networking for the extraction. Shifting the ECG signal to
time-frequency domain is one easily imaginable way to
lessen the effects of aliasing of different frequency
components. Two well-liked time frequency methods are
Wavelet Transform (WT) and Short-Temporal Fourier
Transform (STFT). Although STFT served as inspiration for
WT, WT has the ability to provide both high frequency
resolution and low time resolution at low frequencies, as
well as high time resolution and low frequency resolution at
high frequencies. WT typically performs better in time-
frequency domain analysis than STFT. Using the Continuous
Wavelet Transform (CWT), which is a WT with a continuous
wavelet function, and the Convolutional Neural Network
(CNN), we develop an automaticECGcategorizationmethod.
CNN is a deep learning tool that has been used for
categorising images and successfully mimics the human
visual system. The ECG heartbeat signal is converted to the
time-frequency domain using the CWT, and features are
extracted from the 2D scalogram created from the time-
frequency components using CNN. The method combines
CNN's visual feature extraction capabilities with CWT's
expertise in multidimensional signal processing. To fully
Key Words: Cardiovascular disease; Deep neural
network; ECG signal classification; ARR; CHF; NSR
ECG signal analysis using continuous wavelet transformation and deep
neural network
Malaya kumar hota1, Somalaraju chenchu babu2, B. Bhargav reddy3 , A. Varun chowdary 4 , R. Sai
sumanth 5

utilise all of the data for ECG classification, the RR interval
characteristics are also obtained and fused into our CNN.
2. LITERATURE REVIEW
Over the past 20 years, numerous automatic ECG
categorization techniques have been put out. I therefore opt
for the journal that uses the CNN neural network to
distinguish between the various types of ECG readings.
Figure 1: ECG classification flow diagram
2.1 ECG-SIGNAL CLASSIFICATION
Arrhythmia There are several forms of ECG signals, and in
this study, we have mostly used the followingECGsignalsfor
categorization.
ARR stands for arrhythmias.
CHF stands for Congestive Heart Failure.
NSR stands for Normal Sinus Rhythm.
2.1.1 Arrhythmias
A problem with the rate or rhythm of the heartbeat is
known as an arrhythmia. The heart will beat irregularly, too
rapidly, or too slowly as a result of arrhythmias. The disease
known as tachycardia causes the heart to beat too fast.
2.1.2 Congestive heart failure
Bradycardia is a condition where the heart beats too slowly.
Congestive heart failure can result in heart failure.Theheart
cannot efficiently fill or pump blood (systolic) (diastolic).
Symptoms include shortness of breath, fatigue, and swollen
legs. having a rapid heartbeat Less salt may be consumed as
a treatment.
2.1.3 Normal sinus rhythm
using prescription medication and limiting fluid intake In In
some circumstances, a pacemaker or defibrillator may be
placed. The term "normal sinus rhythm" (NSR) refers to this
beat. rhythm characterising the average heartbeat of a
healthy human being that comes from the sinus node. The
NSR % is typically constant, but it can change depending on
the autonomic inputs that the sinus node receives.
Therefore, we previously used all three ECGsignal kinds.We
previously utilised CWT to classifythemanddeterminetheir
correctness. using a deep neural network as a tool.
Using an electrocardiogram (ECG), arrhythmias can be
identified (ARR). It involves checking the pulse and the
attitude. The clinical illness known as congestive heart
failure (CHF) occurs when the heart is unable topumpblood
at the rate required by the body's using tissues or when the
heart can only do so with a height in filling weight. When
referring to a specific type of sinus rhythm, NSR used to
mean that all other ECG readings fell withinthetypical range
of the breaking point, as depicted in the diagram.
Figure 2: Arrhythmia, CHF, normal ECG signals
2.2 CONTINUOUS WAVELET TRANSFORM
The continuous wavelet transform is a signal processingand
mathematics technique that is frequently used for image
compression, denoising, and other similar tasks. It is also
used in many other fields, such as solving partial differential
equations, financial time series analysis, and biomedical
signal processing, which includes ECG and EEG analysis.The
input layer of the convolutional neural network receives
coefficients right away as a "image," creating a "Transfer
learning" scenario. For this task, we only used Mortlet
Wavelet.
2.3 DEEP NEURAL NETWORK
Many different applications have made use of deep neural
networks. Natural language processing and pattern
recognition are two examples. processing and computer-
based learning Machine learning has provided enormous
benefits for decades prior to now. examples of how this has
an impact on our daily lives include effective web search,
self-driving automobiles, computer vision, and others
Understanding optical characteristics Particularly deep
neural networks have developed into effective machine

learning tools. both machine learning and artificial
intelligence A multi-layer artificial neural network is
referred to as a large neural network (DNN)(ANN).between
the input and output layers are further layers.
The success of deep neural networks has led to significant
advances, such as a 30% reduction in word error rates in
speech recognition over conventional methods (the largest
gain in 20 years) or a radically lower error rate in an image
recognition competition since 2011 (from 26% to 3.5%,
compared to 5% for humans). In order to analyse photos of
varied signals, we therefore use the "ALEX" neural network
in this research.
2.4 ALEXA NEURLA NETWORK
One of the large-scale image netvisual recognition networks,
the Alex neural network, was utilised to recognise every
distinct picture with a range of frequencies. There are eight
layers of parameters in the Alexnet that can be taught. Five
layers make up the model; the first is a max pooling layer,
followed by three fully connected layers. All of these layers,
with the exception of the output layer, use Relu activation.
They found that the training process was roughly six times
faster when the relu was used as an activation function. To
prevent overfitting, they also used dropout layers in their
model. The Image Net dataset is used to train the model as
well. About 14 million images from a thousand distinct
classifications are included in the Image Net dataset. As a
result, we chose this variant of the Alex neural network to
analyse the various types of ECG data because it had all the
characteristics of the Alex neural network.
There are three types of neural networks that are often
employed in all of them:
A. Multi-Layer Perceptrons (MLP);
B. Convolutional Neural Networks (CNN)
C. Recurrent Neural Networks (RNN)
A. Multiple-Neural network
A A feed forward artificial neural network is called a
multilayer perceptron (MLP) (ANN). An MLP has an input
layer, a hidden layer, and an output layer as its minimum
number of node layers. All nodes—aside from the input
nodes—are neuronswithnonlinearactivationfunctions. The
bulk of developers utilise this neural network because it is
one of the best ones.
B. Convolutional neural network
An artificial neural network called a convolutional neural
network (CNN) is designed specificallytoanalysepixel input
during image recognitionandprocessing.CNNsarepowerful
artificial intelligence (AI) systems that recogniseimagesand
videos using deep learning in addition to recommender
systems and natural languageprocessing(NLP).Amultilayer
perceptron-like technique used by CNNs has been tuned for
reduced processing requirements.
Input, output, and a hidden layer with several convolutional,
pooling, fully connected, and normalising layers make up a
CNN's three layers. A significantly more efficient system is
produced by removing restrictions and increasing image
processing efficiency.
C. Recurrent Neural Networks
Recurrent neural networks (RNN) are the most advanced
technique for sequential data and are used by Google voice
search and Apple's Siri. It is the first algorithm that, because
of its internal memory, remembers its input, making it
perfect for sequential data machine learning applications.
One of the algorithms responsible for the phenomenal
advancements in deep learning over the past few years is
this one. In this article, we'll discuss the principles of
recurrent neural networks' operation as well as their main
drawbacks and solutions.
3. METHODOLOGY
This project's primary objective was to categorise the
various ECG signal types using continuous wavelet
transformation [CWT]. To do this, we first used to generate
the final waveform of all the different ECG signal types,
which was primarily used to indicate the "efficiency" and
"accuracy" of the various ECG signals.
ECG Signals to Image conversion using CWT
We transfer the ECG signal to the timefrequency domain to
facilitate feature extraction because it consists of discrete
frequency components. The most popular time-frequency
analysis tool is CWT, which employs a set of wavelet
functions to deconstruct a signal in the time-frequency
domain. Therefore, we used the CWT to get the 2D-
scalogram waveform made up of various ECG signals.
Therefore, in this instance, the CWT is primarily utilised to
generate the following characteristics, which are then used
to categorise the various kinds of ECG signals. The Wavelet
"Analytic Morlet (amor)" is what we primarilyuse.Wavelets
having one-sided spectra and complex time values are
known as analytical wavelets. When creating a
timefrequency-analysis with the CWT, these wavelets are a
great option. 12 wavelet band-pass filters are used by CWT
for each octave (12 voices per octave).

Then, utilising all of the CWT's properties, we created
scologram images of each ECG signal stored in the database.
As a result, each 1D signal from the ECG signals is converted
into a CWT scalogram using the CWT, and each scalogram is
represented by a color-map of a jet of 128 colours. After
conversion, we received 900 separate 2d-scalogram images
of various ECG signals, such as ARR, CHF, and NSR, whichwe
used to identify them by creating various folders for each
type of ECG signal.
a. ECG signal classification using neural network
A neural network having more than two layers and a certain
level of complexity is referred to as a deep neural network.
Deep neural networks use sophisticated mathematical
models to handle data in complex ways. As a result, we were
able to evaluate the accuracy of all the various neural
network types that were present in the ECG database using
the "ALEX" neural network. Alex Krizhevsky developed the
deep neural network known as Alex Net. It was developed to
classify images for the ImageNet LSVRC2010 competition,
and it produced ground-breaking outcomes. Additionally, it
worked with a variety of GPUs.
Compared to earlier CNNs used for computer vision tasks,
Alex Net was significantly larger. It has650,000 neuronsand
60 million parameters, and training on two GTX 580 3GB
GPUs requires five to six days. Today's faster GPUs can
execute even more complex CNNs extremely well, even on
very large datasets. Interesting visual properties are
extracted using multiple convolutional kernels. A single
convolutional layer often has many kernels of the same size.
Following the first two Convolutional layers, the next
Overlapping Max Pooling layers are added. Direct coupling
exists between the third, fourth, and fifth convolutional
layers. The Overlapping Max Pooling layer, whose output is
routed through a series of twofullyconnectedlayers,follows
the fifth convolutional layer. The 1000 class label Softmax
classifier receives input from the second fully connected
layer.
As a result, every operation thathasbeen explainedhasbeen
carried out using an Alex neural network. We used this form
of Alex net for ECG signals to determine the correctness of
the various types of ECG signals images that are included in
the database. In order to produce what is the certain
accuracy of all ECG signals present, the alex net is used to
receive photographs as input. As a result, it requires 900
photographs of various ECG signals. This helps people
determine whether or not the ECG signals are good.
4. RESULTS
FIGURE 4 .Accuracy and loss of the ECG signal
Figure 3: CWT of ECG signal

FIGURE.5 Confusion Matrix
4.1 ACCURACY MEASUREMENT
Using an electrocardiogram (ECG), arrhythmias can be
identified (ARR). It involves checking the pulse and the
attitude. The clinical illness known as congestive heart
failure (CHF) occurs when the heart is unable topumpblood
at the rate required by the body's using tissues or when the
heart can only do so with a height in filling weight. A specific
type of sinus rhythm known as NSR is one in which all other
ECG readings remain within predeterminedtypical breaking
limits.
4.2 TRAINING PROGRESS OF THE SIGNAL
Therefore, we used to predict them using the 900 different
types of ECG images that are used to provide this much
accuracy. In this, 750 images are primarily taken into
account for training and 150 images are taken into account
for testing. We also used to have a confusion matrix in which
we used to have a separate accuracy value for the various
ECG signals and we used to calculate the error.
5. CONCLUSION
We developed a special ECG classification technique based
on CWT and a deep neural network. The ECG heartbeat
signal is first converted intothetimefrequencydomainusing
CWT to prevent the effects of aliasing of separate frequency
components. Then, features are recovered from a
decomposed time-frequency scalogram using Alexnet. The
strategy completely takes advantage of CWT'sadvantages in
multidimensional signal processingandalexnet'sadvantages
in image recognition. It was put to the test on the MITBIH
arrhythmia database using the inter-patient paradigm. Due
to its extremely accurateECGcategorization,ourmethod has
the potential to be used as a clinical additional diagnostic
tool. In general, early detection of ARR, CHF, and NSR is
essential because they are key contributors to
cardiovascular disease. After a thorough early diagnosis,
effective therapy, such as vagal stimulation or medications,
can reduce arrhythmia and avoid cardiovascular disease.
However, there are someotherneural networksthatprovide
greater efficiency, so we may enhance with the help of the
other neural networks to provide even higher accuracy.
Despite the fact that our technique achieves high overall
performance, in this case, we have been using deep neural
networks like Alexnet to provide the most accuracy. In
general, this can be made better with more annotated ECG
data. But classifying ECG heartbeats is expensive and time-
consuming. Nowadays, there are many freely accessible
unlabeled ECG databases, and the use of unsupervised
learning techniques like auto encoder may help to further
improve the performance of the F class in a practical way.
We'll give it another shot later.
6. REFERENCES
1. Kora, P., Annavarapu, A., Yadlapalli, P., Krishna, K. S. R., &
Somalaraju, V. (2017). ECG based atrial fibrillationdetection
using sequency ordered complex Hadamard transform and
hybrid firefly algorithm. Engineering Science and
Technology, an International Journal, 20(3), 1084-1091.
2. Padmavathi, K., & Ramakrishna, K. S. (2015). Detection of
Atrial Fibrillation using Autoregressive modeling.
International Journal ofElectrical andComputerEngineering
(IJECE), 5(1), 64-70.
3. Padmavathi, K., & Krishna, K. S. R. (2014, November).
Myocardial infarction detection using magnitude squared
coherence and support vector machine. In 2014
International Conference on Medical Imaging, m-Healthand
EmergingCommunicationSystems(MedCom)(pp.382-385).
IEEE.
4. Padmavathi, K., & Ramakrishna, K. S. (2015). Detection of
atrial fibrillation using continuous wavelet transform and
wavelet coherence. International Journal ofSystems,Control
and Communications, 6(4), 292-304.
5. Kora, P., & Krishna, K. S. R. (2016). ECG based heart
arrhythmia detection using wavelet coherence and bat
algorithm. Sensing and Imaging, 17(1), 12.
6. Majnaric L, Sabanovic S. “Cardiovasculardisease research
by using data from electronic health records”.
Atherosclerosis, 252:e41- e41.2016.
7. D. Zhang, "WaveletapproachforECGbenchmark meander
adjustment and clamor decrease", in Proc. IEEE Int. Eng.
Prescription. Biol. Soc, pp. 1212-1215. 2005.
8. Kora, P., Meenakshi, K., & Swaraja, K. (2019). Detection of
Cardiac Arrhythmia Using Convolutional Neural Network. In
Soft Computing and Signal Processing (pp. 519-526).
Springer, Singapore

ECG signal analysis using continuous wavelet transformation and deep neural network

More Related Content

Similar to ECG signal analysis using continuous wavelet transformation and deep neural network (20)

More from IRJET Journal (20)

Recently uploaded (20)

ECG signal analysis using continuous wavelet transformation and deep neural network