Heart diseases detection by ECG classification
Due to the rapid development of technology and increased usage of portable monitoring devices, a signi?cant amount of biomedical data is recorded daily to monitor and observe physiological condition of human body. These biomedical signals quantify the physio-logical activities of differe
2025-06-28 16:32:52 - Adil Khan
Heart diseases detection by ECG classification
Project Area of Specialization Artificial IntelligenceProject SummaryDue to the rapid development of technology and increased usage of portable monitoring
devices, a signi?cant amount of biomedical data is recorded daily to monitor and observe
physiological condition of human body. These biomedical signals quantify the physio-logical
activities of different organs such as heart, brain, muscles, cornea, etc. They are usually obtained
by placing one or more electrodes on the organ of interest. Electro-cardiogram (ECG) and
Electroencephalogram (EEG) are the most common biomedical signals recorded from heart and
brain, respectively[1].
An electrocardiogram is used to record the electrical activity of your heart .It is a
common procedure used to monitor heart’s status and detect heart problem in many
situation.ECG is a medical signal which measure the electrical activity of heart and also an
expert can analyze the abnormalities of heart by examing it.ECG has standard shape for normal
person any change in heart activities is reflected in ECG waveform. By examing ECG waveform
sheet physicians can easily ensure that if there is a heart disease due to family history ,smoking
,high cholesterol or another reason.
But examing ECG signal is a challing task ,because there are many complexities while
processing and analyzing ECG signal.heart rate is changing from person to person because of
physiological and different mental condition like exercise ,stress,energy and other physical
exercise.Due to which variability occur in ECG feature. Since these signals contain huge amount
of data ,therefore visuall scaning is very time consuming and costly ;subject to human error and
maybe inaccurate and very complex process,that would benefit from automation if it would be
done reliably and accurately.The automation of analyzing and classifying ECG will decrease the
physicians’s burdon and also diagnose and classify the disease upto large extant of accuracy.
The objective of the project is to process electrocardiogram for classification of heart disease.
For this purpose, we consider supervised deep learning techniques and choose CNN the most
appropriate one for feature extraction and classification.
We build a model based on deep learning (convalotional neural network+fully connected
layer ) to classify ECG signal for detection of various type of abnormalaties of heart . we use
python with machine learning libraries for the purpose of implementation of our model.
Due to the wide area of medical knowledge it is difficult for the medical experts
and physicians to cover all areas and gain knowledge about everything and there are
chances of wrong analysis of the disease also. So this project can be applied in hospitals
for the accurate analysis of heart diseases detection. Also for the testing and gaining of
basic information about age,sex, height, bloodpressure, bloodgroup, temperature etc it
can be utilized.
It saves cost and human resource as it removes different employs like for
carrying files, employs getting blood pressure, taking height measurements etc.It takes
all these values through putting different sensors on a single table and forwarding all
the information to the physicians through internet.
For the prediction of heart abnormalities usage of manual ECG devices are best
option but in emergency cases it is necessary to diagnose the heart abnormality quickly
while manual ECG takes too much time so in such cases this project can be used which
diagnoses the disease in very much less time. This project can be specially implemented
in portable devices like smart phones for diagnoses of heart diseases in rural areas
where diagnoses of heart diseases are big issue.
In this projrect we classify ECG signals with 1D CNN.
ECG classification and anomaly detection. With certain modifications and adaptations
over the traditional 2DCNNs, the proposed system can be tuned to classify directly the
raw data of the heart beats in any sampling rate, therefore, voiding the need for any
manual feature extraction and pre- or post-processing.
With the proper training the convolutional layers of CNNs can learn to extract
patient-specific features over which the MLP layers can learn to produce the final class
vectors of each beat. With the same limited training data as proposed in [10],[15]-[18],
we shall demonstrate that simple CNNs with only 3hidden layers will suffice to achieve a
superior classification performance rather than the complex ones that are commonly
used for deep learning tasks.
This paper shall further show that the proposed 1D CNNs are easier to train with
only few dozens of back-propagation (BP) epochs and can thus perform the
classification task with an utmost speed (requiring only few hundreds of 1D
convolutions). The data used for training the individual patient's classifier consists of
two parts: global(common to each patient) and local (patient-specific) training patterns.
While patient-specific data contains the first 5 minute segment of each patient’s ECG
record and is used as part of the training data to perform patient adaptation, the global
data set contains a relatively small number of representative beats randomly chosen
from each class in the training files and helps the classifier learn other arrhythmia
patterns that are not included in the patient-specific data.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 54000 | |||
| AWS,gpu computer enable | Equipment | 6 | 9000 | 54000 |