1 of 36

CardioPulse: Cardiovascular Disease Prediction through an Integrated Machine Learning Framework

Student Name:

Ali Aoun 20101001-038

Muneeba Javed 20101001-022

Hammad 20101001-166

Fariha 20101001-150

​

Supervised by:

Mr. Attique Ur Rehman

​

2 of 36

Outline

  • Introduction
  • Existing System
  • Problem Statement
  • Goals and Objectives
  • Main Modules
  • System Workflow
  • Tools and Technologies
  • Experiments and Results
  • Conclusion and Future Work
  • References

​

*

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

2

3 of 36

Introduction

  • Modern busy schedules promote unhealthy lifestyles, leading to anxiety and depression. To cope, people often turn to excessive smoking, drinking, and drugs, which can result in various diseases, such as cardiovascular issues and cancer.

​

  • Cardiovascular diseases affect a significant portion of the global population the highest number of death rates, globally. Almost 31% of the world’s deaths are because of the CVDs

​

  • Accurate cardiovascular risk prediction is essential to address the global health concern of heart diseases.

​

​

​

*

3

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

4 of 36

Introduction

  • Cardiovascular Diseases (CVDs) is a term which is used to describe a condition that affects the heart or blood vessels.

Four main types of CVDs include

Coronary Heart Disease,

Stroke (known as MiniStroke),

Peripheral arterial disease

Aortic disease

  • Some risk factors are responsible for these diseases including high blood pressure, smoking, diabetes, body mass index (BMI), cholesterol, age, family history, etc. These factors are different for different people. Age, gender, Stress, and unhealthy lifestyle are also some of the major factors which are responsible for the CVDs.

​

​

​

*

4

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

5 of 36

Existing System

Table 1.1: Literature Review Comparison

*

5

Reference

Research Paper/App

Methodology

Tool/techniques

Results

Limitation

[1]

Tanvi et al. [1]

Decision Tree, Random Forest.

14 features, Cleveland dataset.

93.24% accuracy

Generalizabiliy may be restricted as the study primarily relies on the Cleveland dataset.

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

[2]

Singh et al. [2]

Logistic Regression, SVM, pruning decision tree.

11 selected features, Heart Disease dataset.

​

87.1% accuracy (Logistic Regression )

​

Manual steps may lead to accuracy loss. Automation improves outcomes.

​

6 of 36

Existing System

Table 1.1: Literature Review Comparison

*

6

Reference

Research Paper/App

Methodology

Tool/techniques

Results

Limitation

[3]

Amanda et al [3]

Decision Tree, Naïve Bayes, SVM.

10 features, South African heart disease dataset.

Naïve Bayes produced best results. 82% accuracy.

Limited data instances and class imbalance may affect model performance. Further research could enhance sensitivity and specificity.

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

[4]

Ketut et al. [4]

KNN (with and without parameter weighting), Naïve Bayes, SVM.

18 parameters, dataset from Harapan Kita Hospital

KNN: 75.11%, 74.0%; Naïve Bayes, SVM not significant

Parameter weightings didn't improve accuracy. Different doctors may give inconsistent diagnoses. Naïve Bayes and SVM showed lower accuracy than KNN

7 of 36

Existing System

Table 1.1: Literature Review Comparison

*

7

Reference

Research Paper/App

Methodology

Tool/techniques

Results

Limitation

[5]

Rubini et al. [5]

Random Forest, Logistic Regression, Naïve Bayes, SVM

Framingham dataset

84.81% accuracy (Random Forest).

Reliance on a single algorithm (Random Forest). Focus on specific correlation (diabetes-heart disease) may not capture all relevant factors. Limited generalizability to correlations with other diseases. Sensitivity to parameter selection may impact accuracy.

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

8 of 36

Existing System

Table 1.1: Literature Review Comparison

*

8

Reference

Research Paper/App

Methodology

Tool/techniques

Results

Limitation

[6]

Hoda et al. [6]

KNN, Random Forest.

KNN, Random Forest, Framingham dataset.

KNN: 66.7%; Random Forest: 63.4% accuracy.

KNN outperformed random forests, but accuracy still needs improvement. Limited sample size hinders accuracy for different risk classes. Future work includes enhancing KNN with clustering techniques and increasing sample size for better accuracy.

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

9 of 36

Existing System

Table 1.1: Literature Review Comparison

*

9

Reference

Research Paper/App

Methodology

Tool/techniques

Results

Limitation

[7]

Randa et al. [7]

Naive Bayes.

13 features, Collective Heart Disease dataset (CAD).

92.0% accuracy.

Limited diversity in datasets. Focus on specific algorithms may overlook others. Possible risk of overfitting. Varying attributes in datasets. Future challenge in unifying heart disease data.

[8]

M. O. Butt

KNN

12 Attributes, Heart Failure dataset

84.11% with KNN

Limited data instances and class imbalance may affect model performance.

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

10 of 36

Problem Statement

​

There is a need for a comprehensive and versatile framework to predict cardiovascular diseases, as existing research insufficiently addresses the management of missing values and imbalanced class distributions. Additionally, an effective feature selection technique and a robust, efficient classification algorithm are required to handle the complexities of cardiovascular disease prediction.

​

*

10

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

11 of 36

Project Scope

​

​

*

11

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

​

  • The goal is to create a simple tool that uses advanced technology to find early signs of heart issues, helping to treat them quickly.

​

  • The main objective is to detect early warning signs of heart disease, facilitating timely intervention and treatment, and contributing to improved public health outcomes.

​

  • Committed to preventing heart problems and improving general health, CardioPulse shares the goal of leveraging technology to bring positive changes to people's lives.

​

​

12 of 36

Goals and Objectives

Goal: Create an innovative, data-driven system for accurately predicting and managing cardiovascular health risks through the integration of advanced machine learning and data analysis techniques.

​

  • Optimize Data Processing: Utilize advanced data processing techniques to optimize the interpretation of health-related data, enhancing the precision of risk assessments and improving overall cardiovascular health management.

​

  • Develop Machine Learning Models: Build and test specialized machine learning models for analyzing cardiovascular health data, tailored to meet the unique demands of the CardioPulse project.

*

12

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

13 of 36

Goals and Objectives (Cont..)

  • Ensure Platform Compatibility: Ensure that the CardioPulse system operates seamlessly on the Android platform, maximizing its accessibility for a wide range of users.

​

  • Evaluate Model Performance: Assess the performance of the developed system to ensure accurate and reliable prediction and management of cardiovascular health risks.

​

  • Enhance User Experience: Improve the overall experience of users by prioritizing user-friendly interfaces and intuitive functionalities in the CardioPulse system.

*

13

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

14 of 36

Main Modules

User Registration and Login:

Make it easy for users to create accounts and set up health profiles securely and effortlessly.

​

Health Data Collection and Storage:

Implement a structured database for storing and managing a variety of health-related data, ensuring data integrity and accessibility for accurate risk assessments.

​

​

​

*

14

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

15 of 36

Main Modules (Cont..)

Cardiovascular Risk Prediction:

Utilize advanced machine learning techniques to comprehensively analyze health data, empowering healthcare professionals with precise predictions of cardiovascular health risks.

​

Progress Monitoring and Reporting:

Enable users to track their cardiovascular health progress over time and receive personalized recommendations, fostering a data-driven approach to cardiovascular health management.

*

15

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

16 of 36

System Workflow

*

16

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

System Workflow Diagram:

17 of 36

*

17

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

Use Case Diagram:

18 of 36

Tools and Technologies

Tools:

  • Integrated Development Environment (IDE): Android Studio
  • RapidMiner.

​

Technologies:

  • Programming Languages: Java (for Android app development), Python (for machine learning integration)

​

  • Database: Firebase Firestore Database (for cloud-based storage)

​

  • Machine Learning and AI Libraries: Scikit-learn, TensorFlow,

Keras, NumPy, Pandas, Matplotlib.

*

18

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

19 of 36

Experiments and Results

*

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

​

19

Author

Dataset

Data Imbalance

Feature Selection

Classifier

Accuracy

[9] A. Rahim

Heart Disease

SMOTE

Automatic

Ensemble (LR & KNN)

98.0%

[8] M. O. Butt

Heart Failure Dataset

SMOTE

Optimize

KNN

84.11%

Singh et al. [2]

Heart Disease Dataset

SMOTE

11 Features

Logistic Regression

87.1%

Randa et al. [7]

Heart Disease Dataset

SMOTE

13 Features

Naïve-Bayes

92.0%

20 of 36

Experiments and Results

*

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

​

​

20

Stroke Dataset (F Fold = 10)

Classifier

Result

Random Forest

82.0%

Naive Bayes

87.50%

KNN(k=3)

89.10%

Heart Failure Dataset (F Fold = 10)

Classifier

Result

Random Forest

85.51%

Decision Tree

82.98%

KNN(k=3)

86.13%

21 of 36

Experiments and Results

*

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

​

​

21

Heart Disease Dataset (F Fold = 10)

Classifier

Result

Random Forest

98.10%

Decision Tree

89.45%

KNN(k=3)

99.05%

22 of 36

Experiments and Results

*

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

​

22

Best Results with F Folds of 10

Dataset

Data Imbalance

Feature Selection

Classifier

Accuracy

Heart Disease Dataset

SMOTE

Optimized

KNN(K=3)

99.05%

Stroke Dataset

SMOTE

Optimized

KNN(K=3)

89.10%

Heart Failure Prediction

SMOTE

Optimized

KNN(K=3)

86.13%

23 of 36

Conclusion and Future Work

  • Our project CardioPulse represents a significant stride in cardiovascular health management, leveraging advanced machine learning techniques to accurately predict and manage health risks.

​

  • The project's user-friendly design and emphasis on data processing and analysis ensure precise risk assessments and recommendations, revolutionizing the way cardiovascular health is managed.

​

  • CardioPulse utilizes modern technology for early detection and effective treatment of heart issues, enhancing public health outcomes.

​

​

*

23

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

24 of 36

Data Set

  • Data set name: Heart Failure Prediction Dataset
  • No. of attributes: 12
  • Instances: 918

​

​

*

24

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

25 of 36

Data Set

  • Data set name: Stroke Dataset
  • No. of attributes: 11
  • Instances: 5110

​

​

*

25

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

26 of 36

Data Set

  • Data set name: Heart Disease Dataset
  • No. of attributes: 14
  • Instances: 1025

​

​

*

26

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

27 of 36

App Preview

*

27

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

28 of 36

App Preview

*

28

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

29 of 36

App Preview

*

29

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

30 of 36

App Preview

*

30

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

31 of 36

App Preview

*

31

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

32 of 36

App Promo Video

*

32

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

33 of 36

Demo Video

*

33

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

34 of 36

References

[1] T. Sharma, S. Verma, and Kavita, ‘‘Prediction of heart disease using cleveland dataset: A machine learning approach,’’ Int. J. Recent Res. Aspects, vol. 4, no. 3, pp. 17–21, 2017.

​

[2] P. S. Kohli and S. Arora, ‘‘Application of machine learning in disease prediction,’’ in Proc. 4th Int. Conf. Comput. Commun. Autom. (ICCCA), Dec. 2018, pp. 1–4.

​

[3] A. H. Gonsalves, F. Thabtah, R. M. A. Mohammad, and G. Singh, ‘‘Prediction of coronary heart disease using machine learning: An experimental analysis,’’ in Proc. 3rd Int. Conf. Deep Learn. Technol., 2019, pp. 51–56.

​

[4] I. K. A. Enriko, M. Suryanegara, and D. Gunawan, ‘‘Heart disease diagnosis system with K-nearest neighbors method using real clinical medical records,’’ in Proc. 4th Int. Conf. Frontiers Educ. Technol., 2018, pp. 127–131.

*

34

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

35 of 36

References(Cont..)

[5] E. B. Randa, ‘‘An ensemble model for Heart disease data sets: A generalized model,’’ in Proc. 10th Int. Conf. Inform. Syst., 2016, pp. 191–196.

​

[6] P. E. Rubini, C. A. Subasini, A. V. Katharine, V. Kumaresan, S. G. Kumar, and T. M. Nithya, ‘‘A cardiovascular disease prediction using machine learning algorithms,’’ Ann. Romanian Soc. Cell Biol., vol. 25, no. 2, pp. 904–912, 2021. [Online]. Available: https://www.annalsofrscb.ro/index.php/journal/article/view/1040

​

[7] H. A. G. Elsayed and L. Syed, ‘‘An automatic early risk classification of hard coronary heart diseases using Framingham scoring model,’’ in Proc. 2nd Int. Conf. Internet Things, Data Cloud Comput., Mar. 2017, pp. 1–8

​

[8] M. O. Butt, A. Ur Rehman, S. Javaid, T. M. Ali and A. Nawaz, "An Application of Artificial Intelligence for an Early and Effective Prediction of Heart Failure," 2022 Third International Conference on Latest trends in Electrical Engineering and Computing Technologies (INTELLECT), Karachi, Pakistan, 2022, pp. 1-6.

​

*

35

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework

36 of 36

References(Cont..)

[9] A. Rahim, Y. Rasheed, F. Azam, M. W. Anwar, M. A. Rahim and A. W. Muzaffar, "An Integrated Machine Learning Framework for Effective Prediction of Cardiovascular Diseases," in IEEE Access, vol. 9, pp. 106575-106588, 2021, doi: 10.1109/ACCESS.2021.3098688.

*

36

CardioPulse: Cardiovascular Risk Prediction through an Integrated Machine Learning Framework