it's a graduation project aims to Diagnose cardiovascular diseases in real-time using machine learning through extracting features from ECG signal with accuracy of 85% to 100%

1. 2019Diagnostic
ECG
Supervisor : Dr/Ahmed El-Sherbeny
Made by : Nourhan Slem Salm

2. Diagnostic ECG
Diagnosing cardiovascular diseases
in real-time using machine learning
through extracting features from
ECG signal with accuracy of
85% 100%

3. Content
What ?
Why ?
How ?
Results
Future Work
Diagnostic ECG
Methodology & Materials
Accuracy
What next
Motivation & Reasons

4. Diagnostic
ECG
01
02
Diagnostic
ECG
• Myocardia
• Dysrhythmia
• Cardiomyopathy
• Bundle branch block
• Normal
Detecting cardiovascular diseases
The ECG is the recording of the heart’s
electrical activity that is generated by
depolarization and repolarization of the
atria and ventricles.
ECG is an important tool for the
primary diagnosis of heart disease. One
cardiac cycle in an ECG signal consists
of the P-QRS-T waves .
ECG(Electrocardiogram)

5. Cardiovascular
Diseases
Cardiomyopathy
disease of the heart muscle that
makes it harder for your heart to
pump blood to the rest of your
body. Cardiomyopathy can lead
to heart failure.
03
Bundle Branch Block
Sometimes part of the heart's conduction
system is "blocked“.
04
05
06
Myocardia
medical name for a heart attack. Reducing
blood flow to the heart. This is usually the
result of a blockage in one or more of the
coronary arteries.
01
Dysrhythmia
Abnormal heart beat, the rhythm may
be irregular in its pacing or the heart
rate may be low or high such as sinus
arrhythmia and Tachycardia .
02

11. 30 min
2-3 h
After 6 h
Restoring flow to the affected
artery can abort an infarction.
some preservation of
myocardial function is achieved .
little or no myocardial salvage .
Time to treatment
can be a matter of life and death

12. Methodology

13. Project
data acquisition (DAQ) device
to analyze and measure live
signals anytime, anywhere.
MyDAQ
Simple electronic circuit made
to measure ECG signal
Electronic circuit
Used as analog digital
converter(ADC).
Arduino Uno
1. Downloaded from physionet
2. Measured from volunteers
3. Measured from MyDAQ
Database
Using python as language
programming
Code
Using rainforest model
Machine learning
algorism
Software
HardwareHardware

14. Hardware

15. MyDAQ
data acquisition (DAQ)
device to analyze and
measure live signals
anytime, anywhere.

18. Electronic
Circuit
Highpass filter Amplifier lowpass filter
Instrumentation
amplifier

19. Arduino Uno
Used as analog
digital
converter(ADC).

20. Project Stages
Stage 1
Dataset
Stage 2
Features
extraction
Stage 4
Selecting
Classifier
Stage 3
Pre-processing
features
Real-time
diagnosis
Stage 5

21. 1
Stage
Dataset
01
02
03
From volunteers
From physionet
From MyDAQ

22. From physionet
Bundle branch
block
3%
Myocardia
74%
Healthy
Control
17%
Dysrhythmia
2%
The PTB Diagnostic ECG Database
Institute in Germany ,digitizes
ECGs for research or
teaching purposes
PTB
1000 Hz
Fs
4783 record
No. of records
Cardiomyopathy
3%

23. From volunteers
9
No. of volunteers
Healthy : 240 Hz
Bradycardia : 150 Hz
Fs
982 record
No. of records
Bradycardia
50%
Healthy
50%

24. From MyDAQ
myocardia
50%
Healthy
50%
240 Hz
Fs
100 record
No. of records

25. 2
Stage
Features
extraction

26. Features extraction
Heart Rate
Mean
Variance
Correlation
power spectral density
(PSD)

28. 3
Stage
Pre-processing
Features

29. Pre-processing Features
1
Encoding categorical data
Alphabetic Numerical
3
75% training set
25% test set
Splitting the dataset into training
set and test set
4
machine learning equations are based
on the Euclidean distance. With a huge
number dominating a smaller number,
eventually the smaller number doesn't
exist
Feature scaling
2
a scenario in which two or more
variables are highly correlated; in
simple terms one variable can be
predicted from the others.
Solving dummy variables
problems

30. 4
Stage
Selecting
Classifier

31. Model Evaluation
Precision
portion of correct
positive classifications
from cases that are
predicted as positive.
𝑝𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 =
𝑇𝑃
𝑇𝑃 + 𝐹𝑃
× 100
Area Under ROC curve
Recall
portion of correct
positive classifications
from cases that are
actually positive
𝑅𝑒𝑐𝑎𝑙𝑙 =
𝑇𝑃
𝑇𝑃 + 𝐹𝑁
× 100
Accuracy
portion of correct
classifications from
overall number of cases.
𝑇𝑃 + 𝑇𝑁
𝑇𝑃 + 𝑇𝑁 + 𝐹𝑁 + 𝐹𝑃
× 100

33. Machine
learning
regression Classification
Decision tree
Rainforest
KNN
supervised unsupervised
clustering

34. Selecting Classifier

35. Random Forest

36. 5
Stage
Diagnosing in
Real-time

37. Diagnosing in Real-time
Loading
saved model
Continuous
diagnosing
every second
Transferring
the signal
Extracting
features

38. Results

39. Models’ Accuracy
98.2%88.3%
Myocardia
79.5%
CVDs Volunteers
MyDAQ
85.7%
Myocardia
100%
Normal
PTB database

40. Future Work

43. References
[1] I. A. Lecturer, M. N. Hossain, and S. M. Yahea Mahbub, “Baseline Drift Removal and De-
Noising of the ECG Signal using Wavelet Transform,” Int. J. Comput. Appl., vol. 95, no. 16,
pp. 975–8887, 2014.
[2] N. Dey, S. Borra, A. S. Ashour, and F. Shi, Machine Learning in Bio-Signal Analysis and
Diagnostic Imaging. Elsevier Science, 2018.
[3] “WHO | World Heart Day,” WHO, 2018. [Online]. Available:
https://www.who.int/cardiovascular_diseases/world-heart-day/en/. [Accessed: 11-Feb-2019].
[4] G. Roth, “CVD Causes One-Third of Deaths Worldwide,” J. Am. Coll. Cardiol., 2017.

44. References
[5] World Health Organization, “The top 10 causes of death,” 2018. [Online]. Available:
https://www.who.int/en/news-room/fact-sheets/detail/the-top-10-causes-of-death. [Accessed: 11-Feb-
2019].
[6] David Schulthorpe, “Innovation Is Our Best Hope Against Cardiovascular Disease - Personal
Health News.” [Online]. Available: http://www.personalhealthnews.ca/research-and-
innovations/innovation-is-our-best-hope-against-cardiovascular-disease. [Accessed: 19-Feb-2019].
[7] H. A. DeVon, N. Hogan, A. L. Ochs, and M. Shapiro, “Time to treatment for acute coronary
syndromes: the cost of indecision.,” J. Cardiovasc. Nurs., vol. 25, no. 2, pp. 106–14, 2010.
[8] “The PTB Diagnostic ECG Database.” [Online]. Available:
https://physionet.org/physiobank/database/ptbdb/. [Accessed: 12-Feb-2019].

45. Thank you