Automated Diagnosis of Cardiac Arrhythmia by using ECG and Machine Learning

Project Title

Automated Diagnosis of Cardiac Arrhythmia by using ECG and Machine Learning

Project Area of Specialization Artificial IntelligenceProject Summary

This research project focuses on the analysis of ECG signal to detect cardiac arrhythmia in humans. Cardiac arrhythmia is associated with irregular heartbeats and it is one of the most common cardiac associated diseases. The automated interpretation of the ECG signal is helpful in saving clinical time and also reduce patients suffering. The proposed system consists of single lead ECG sensor that is commonly utilized in long term ECG recordings. The low cost ECG sensor will be interfaced with raspberry pi. The obtained single will be filtered for noise removal and feature extractions. The position and magnitude of P, Q, R, S and T waves will be calculated from the recorded signal. The data will be recorded from healthy humans as well as humans with cardiac arrhythmia. By using extracted feature points, database is developed which is used to train machine learning algorithms for the detection of cardiac arrhythmia. The proposed system will help significantly in reducing the number of doctors required to deal with such large amount of cardiac patients. This also facilitates common population to keep their life style healthy and affordable.

Project Objectives

1. To interfacing of single lead ECG sensor with raspberry pi and capture and record the data.
2. To perform noise removal and feature extraction for the development of ECG database.
3. To train and test machine learning classifiers for the automated diagnosis of cardiac arrhythmia.

Project Implementation Method

The proposed system is based on single lead ECG sensor interfaced with raspberry pi. The raspberry pi is the main processing device and all the operation including filtering, feature extraction, frequency and time domain analysis etc. will be done by using raspberry pi. The data will be recorded while the subject is lying stationary in order to avoid any other electrical activity within the muscles. After data recording, the data will be filtered out to remove any noise and baseline interference. Bandpass filter is normally used to clean ECG signal. The main information about heart conditions can be obtained by using bandpass filter of specification 0.5Hz to 100Hz. The feature points (i.e. RR-intervals, QRS complex, waves peaks, waves onset etc.) to represents the ECG signal can be obtained by using neurokit2 (package available online for python) which utilizes peak detection algorithm and other filtration techniques to extract relevant information. The obtained ECG data from different human subjects will be utilized to train the machine learning classifiers to differentiate between normal and abnormal heartbeats. The proposed system will be helpful in diagnosing the cardiac diseases easily, swiftly and cheaply.

Benefits of the Project

The main benefit of this project is to facilitate cardiac patients who are increasing day by day by providing efficient and cost effective means of diagnosis process. The automation can provide better assistance to diagnose diseases. The proposed system will provide automated diagnosis which can be helpful for both doctors and patients. It can also be used for long term ECG recordings. The system will give doctors the ability to give more time to those patients who are suffering from severe cardiac issue.

Technical Details of Final Deliverable

The proposed system consist of single lead ECG sensor that can be attached to the cardiac patients easily with less wires. The classification of diseases will be purely based on the data collected from multiple cardiac patients. The resulting trained ML classifiers will be able to interpret ECG data of any new subject and provides the classification results in terms of associated disease. The data collected from patients can be stored electronically which can be accessed remotely as well.

Final Deliverable of the Project Hardware SystemCore Industry HealthOther Industries Medical Core Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Good Health and Well-Being for PeopleRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	60100
Raspberry pi 4	Equipment	1	34000	34000
Heart sensor	Equipment	1	11000	11000
ADC IC	Equipment	2	1500	3000
Electrodes	Equipment	3	700	2100
Related Hardware and expenses	Miscellaneous	1	10000	10000