Study | Data set | Number of participants | Algorithm used | Classification accuracy | AUC | Ventricular arrhythmias |
Okada et al59 | CMR imaging | 122 | Substrate spatial complexity analysis | 81.0% | 0.72 | 40 |
Kotu et al32 | CMR imaging | 54 | MATLAB, SVM and k-NN | 94.4% to 92.6% | 0.96 | – |
Ebrahimzadeh et al60 | ECG | 70 (35 normal, 35 sudden cardiac death) | kNN, MLP | 84.0% to 99.7% | – | – |
Au-Yeung et al37 | ECG | 788 | RF, SVM | – | 0.81 to 0.88 | 3 in 10 patients |
Marzec et al61 | CIED | 235 | RF, k-NN, STATA IC | 55.3% to 76.6% | 0.5 | 49 |
Shandilya et al 36 | ECG+PetCO2 | 153 | MDI model | 78.8% | 0.832 | – |
Howe et al33 | ECG | 41 | SVM | 81.9% | 0.75 | 115 |
Shandilya et al34 | ECG | 57 cardiac arrests (90 signals) | SVM | Up to 83.3% | 0.85 to 0.93 | 57 |
AUC, area under the curve; CIED, cardiac implantable electronic devices; CMR, cardiac MRI; k-NN, k nearest neighbours algorithm; MDI, multidomain integrative; MLP, multilayer perceptron; RF, random forest; STATA-IC, statistical software package; SVM, support vector machines.