Table 15.
Detecting cardiovascular and heart disease using machine learning techniques.
| Data set | Approach | Results |
|---|---|---|
| Cleveland, Hungarian, Long-beach VA, Switzerland (UCI) [168] |
SVM, Naive Bayes, Random Forest, Multilayer Perceptron |
Accuracy: 98% |
| Cleveland (UCI) [37] | Hybrid Random Forest with a Linear Model |
Accuracy: 95.87% |
| Hungarian HD (UCI) [169] | Deep Learning Modified Neural Network (DLMNN) |
Security: 95.87% |
| Statlog Cleveland (UCI) [170] | Density-Based Spatial Clustering of Application with Noise (DBSCAN)/SMOTE-ENN/XGBoost |
STATLOG dataset Accuracy: 95.90 ± 5.55, Precision: 97.14 ± 5.71, Sensitivity: 94.67 ± 11.08, Cleveland dataset Accuracy: 98.40 ± 3.21, Precision: 98.57 ± 4.29, Sensitivity: 98.33 ± 5.00 |
| Cleveland (UCI) [45] | Modified Salp-Swarm Optimization-Adaptative Neuro-Fuzzy Inference System (MSSO-ANFIS) |
Accuracy: 99.45%, Precision: 96.54% |
| PTB Diagnostic ECG, Fantasia Database, St. Petersburg Institute of Cardiological Technics 12-lead Arrhythmia Database, PTB Diagnostic ECG Database, BIDMC Congestive Heart Failure Database [46] |
Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) |
Accuracy: 98.51%, Specificity: 97.89%, Sensitivity: 99.30%, Positive predictive value: 97.33% |
| PhysioNet [171] | Particle Swarm Optimization (PSO), Twin Support Vector Machine (TSVM) |
Accuracy: 96.68% |
| Framingham and Hungarian Kaggle, Health Dataset USA Health site [172] |
Beetle Swarm Optimization-Adaptive Neuro-fuzzy inference system (BSO-ANFIS) |
BSO-ANFIS of heart disease classification Accuracy: 99.1%, Precision: 99.37%, Specificity: 99.4%, Sensitivity: 99.21% BSO-ANFIS of multi-disease identification Accuracy: 96.08%, Precision: 98.63% |
| MIT-BIH Arrhythmia, BIDMC Congestive Heart Failure, MIT-BIH Normal Sinus Rhythm |
Deep Neural Network | Accuracy: 99% |
| Cleveland, South Africa, Z-Alizadeh Sani, Framingham, Statlog [173] |
Deep Belief Network | Cleveland Accuracy: 89.2% South Africa Accuracy: 89.5% Z-Alizadeh Sani Accuracy: 89.7% Framingham Accuracy: 90.2% Statlog cardiac disease Accuracy: 91.2% |
| Computing in Cardiology Challenge [174] | Neural Network | Accuracy: 97% |
| UCI Machine Learning Repository [175] | Artificial Neural Network (ANN) | Accuracy: 95.78% Precision: 95.2% Recall: 95.2% Equal rate of error: 4.32% |
| Dalian Medical University and Northeastern University [176] |
Three Decision Tree-based multilabel learning methods |
F1-score: 86.73% AUC: 90.80% Accuracy: 92.72% |
| Statlog, Cleveland, Hungary [177] | Deep Bidirectional Long Short-Term Memory with Elliptic Curve Cryptography dependent Diffi-Huffman algorithm |
Accuracy: 97.53% Sensitivity: 97.93% Specificity: 97.52% F1-Score: 7.65% |
| Cleveland [178] | Big data processing Apache Spark Apache Kafka |
Accuracy: 92.05% Sensitivity: 88.10% Specificity: 95.65% |
| UCI public repository [179] | Sine Cosine Weighted k-NN | Accuracy: 92.13% Precision: 88.21% Recall: 93.27% F1-Score: 90.60% RMSE: 0.1115 |
| MAHNOB-HCI, MMSE-HR, UBFC-Rppg, VIPL-HR [180] |
Convolutional Neural Network (CNN) | VIPL-HR Accuracy: 90% MAE: 5.23 RMSE: 7.21 |
| Kaggle repository [181] | Recursive feature elimination based Gradient Boosting (RFE-GB) |
Accuracy: 88.8% Precision: 88.8% Recall: 85% F1-score: 83% AUC: 84% MSE: 0.20 RMSE: 0.44 |
| Cleveland [182] | Multi-layer Perceptron for Enhanced Brownian Motion based on Dragonfly Algorithm |
Accuracy: 97.47% Sensitivity: 98.92% F1-score: 96.45% Specificity: 96.47% Kappa: 96.75% |
| Chapman University and Shaoxing People’s Hospital, China Physiological Signal Challenge [183] |
Deep Learning System | F1-score: 97.18% Precision: 97.36% Recall: 97.03% Accuracy: 98.73% |
| UCI Heart Disease [184] | Stacking Classifiers Model | Accuracy: 91.8% Precision: 92.6% Sensitivity: 92.6% Specificity: 91% |
| Breast Cancer, Heart Disease, Pima of UCI repository [185] |
Multi-layer Perceptron | Heart Disease prediction with MLP-AOA-AE Accuracy: 83.90% Precision: 84.69% F1-score: 83.84% Recall: 83.90% |
| Pulsewatch, UMMC Simband Stanford University’s PPG, MIMIC-III [186] |
Ensemble based feature selection | MIMIC-III Accuracy: 99% Sensitivity: 83% Specificity: 98% UMMC Accuracy: 94% Sensitivity: 95% Specificity: 91% |
| CapnoBase, MIMIC-II [187] | Growing Multilayer Network | CapnoBase Sensitivity: 98.49% Precision: 98.60% F1-score: 98.55% MIMIC-II Sensitivity: 96.01% Precision: 98.35% F1-score: 97.17% |
| Mindray, MIMIC [188] |
Knowledge Distillation Strategies | Estimation error Systolic BP: mmHG Diastolic BP: mmHG |