Table 16.

Detecting myocardial infarction using machine learning techniques.

Data Set	Machine Learning Approach	Results
Physikalisch-Technische Bundesanstalt (PTB) diagnostic ECG, AF Classification [PhysioNet] [189]	Convolutional Neural Network–Long Short-term memory (CNN-LSTM)	Sensitivity: 92.4%, Specificity: 97.7%, PPV 1: 97.2%, F1-score: 94.6%
Physikalisch-Technische Bundesanstalt [190]	Convolutional Neural Network	Accuracy: 98.13%, Sensitivity: 98.19%, Specificity: 98.09%
MIT-BIH, Electrocardiogram Vigilance with Electronic data Warehouse (ECG-ViEW II) [191]	Convolutional Neural Network (CNN), Recurrent Neural Network, XGBoost	Accuracy: 89.8%, 84.6%, 97.5%, Sensitivity: 93.2%, 78%, 93.5%, Specificity: 88.1%, 87.8%, 99.4%, F1-score: 89%, 82.8%, 97.1%, AUROC: 90.7%, 82.9%, 96.5%
Medical records from the hospital information system [192]	Random Forest	AUC: 85%, Accuracy: 82%
MIT PhysioNet PTB diagnostic ECG [193]	k-NN	Accuracy: 99.96% Sensitivity: 99.96% Specificity: 99.95%
Physikalisch-Technische Bundesanstalt diagnostic ECG [194]	Long Short-Term Memory (LSTM)	Accuracy: 89.56% Recall: 91.88% Specificity: 80.81%

¹ Positive predictive value.