Table 2.
Characteristics of the included studies.
| First author | Analytic model | Sample | Indication | Imaging | Comparison | Database |
|---|---|---|---|---|---|---|
| Cardiac arrhythmias | ||||||
| Alickovic et al. (2016) | RF | 47 | Arrhythmia detection | ECG | Five ECG signal patterns from MIT-BIH (normal (N), Premature Ventricular Complex (PVC), Atrial Premature Contraction (APC), Right Bundle Branch Block (RBBB) and Left Bundle Branch Block (RBBB)) and four ECG patterns from St. -Petersburg Institute of Cardiological Technics 12lead Arrhythmia Database (N, APC, PVC and RBBB) | St. Petersburg and MIT-BIH database |
| Au-Yeung et al. (2018) | 5sRF, 10sRF, SVM | 788 | Ventricular arrhythmia | ICD Data | SCD-HeFT study | |
| Hill et al. (2018) | Logistic-linear regression, SVM, RF | 2,994,837 | Development of AF/flutter in gen pop | Clinical data | ML compared with conventional linear statistical methods | UK Clinical Practice Research Datalink (CPRD) between 01–01-2006 and 31–12-2016 was undertaken |
| Kotu et al. (2015) | k-NN, SVM, RF | 54 | arrhythmic risk stratification of post MI patients | Cardiac MRI | Low LVEF and Scar versus textural features of scar | Single center |
| Ming-Zher Poh et al. (2018) | CNN | 149,048 | AF | ECG | Several publicly accessible PPG repositories, including the MIMIC-III critical care database,11 the Vortal data set from healthy volunteers12 and the IEEE-TBME PPG Respiratory Rate Benchmark data set.1 | |
| Xiaoyan Xu et al. (2018) | CNN | 25 | AF | ECG | MIT-BIH Atrial Fibrillation database | MIT-BIH Atrial Fibrillation Database |
| Coronary artery disease | ||||||
| Araki et al. (2016) | SVM classifier with five different kernels sets | 15 | Plaque rupture prediction | IVUS | 40 MHz catheter utilizing iMap (Boston Scientific, Marlborough, MA, USA) with 2,865 frames per patient (42,975 frames) and (b) linear probe B-mode carotid ultrasound (Toshiba scCNNer, Japan) | Single center |
| Araki et al. (2016) | SVM combined with PCA | 19 | Coronary risk assessment | IVUS | Single center | |
| Arsanjani et al. (2013) | boosting algorithm | 1,181 | Perfusion SPECT in CAD | Perfusion SPECT | 2 experts, combined supine/prone TPD | Single center |
| BaumCNN et al. (2017) | Custom-built algorithm | 258 | ctFFR in detecting relevant lesions | Invasive FFR determination of relevant lesions | the MACHINE Registry | |
| Coenen (2018) | Custom-built algorithm | 351 | Invasive FFR / Computational flow dynamics based (CFD-FFR) | CT angiography | Invasive FFR / Computational flow dynamics based (CFD-FFR) | 5 centers in Europe, Asia, and the United States |
| Dey et al. (2015) | boosting algorithm | 37 | Coronary CTA in ischemic heart disease patients to predict impaired myocardial flow reserve | CCTA | Clinical stenosis grading | Single center |
| Eisenberg et al. (2018) | boosting algorithm | 1925 | MPI in CAD | SPECT | Human visual analysis | The ReFiNE registry |
| Freiman et al. (2017) | Custom-built algorithm | 115 | CCTA in coronary artery stenosis | CCTA | Cardiac image analysis | The MICCAI 2012 challenge |
| Guner et al. (2010) | CNN | 243 | Stable CAD | Myocardial perfusion SPECT (MPS) |
SPECT evaluation and human–computer interaction One expert reader who has 10 years of experience and six nuclear medicine residents who have two to four years of experience in nuclear cardiology took part in the study |
Single center |
| Hae et al. (2018) | Logistic-linear regression, SVM, RF, boosting algorithm | 1,132 | Prediction FFR in stable and unstable angina patients | FFR, CCTA | Single center | |
| Han et al. (2017) | Logistic-linear regression | 252 | Physiologically significant CAD | CCTA and invasive fractional flow reserve (FFR | The DeFACTO study | |
| Hu (Xiuhua) et al. (2018) | Custom-built algorithm | 105 | Intermediate coronary artery lesions | CCTA | CCTA-FFR vs Invasive angiography FFR | Single center |
| Hu et al. (2018) | Boosting algorithm | 1861 | MPI in CAD | SPECT | True early reperfusion | Multicenter REFINE SPECT registry |
| Wei et al. (2014) | Custom-built algorithm | 83 | Noncalcified plaques (NCPs) detection on CCTA | CCTA | Single center | |
| Kranthi et al. (2017) | Boosting algorithm | 85,945 | CCTA in CAD | CCTA | 66 available parameters (34 clinical parameters, 32 laboratory parameters) | Single center |
| Madan et al. (2013) | SVM | 407 | Urinary proteome in CAD | global proteomic profile analysis of urinary proteome | Indian Atherosclerosis Research Study | |
| Zellweger et al. (2018) | Custom-built algorithm | 987 | CAD evaluation | N/A | Framingham scores | The Ludwigshafen Risk and Cardiovascular Health Study (LURIC) |
| Moshrik Abd alamir et al. (2018) | Custom-built algorithm | 923 | ED patients with chest pain -CTA analysis | CT Angiography | Single center | |
| Nakajima et al. (2017) | CNN | 1,001 | Previous myocardial infarction and coronary revascularization | SPECT | Expert consensus interpretations | Japanese multicenter study |
| Song et al. (2014) | SVM | 208 | Risk prediction in ACS | N/A | Single center | |
| VanHouten et al. (2014) | Logistic-linear regression, RF | 20,078 | Risk prediction in ACS | N/A | Single center | |
| Xiao et al. (2018) | CNN | 15 | Ischemic ST change in ambulatory ECG | ECG | Long-Term ST Database (LTST database) from PhysioNet | |
| Yoneyama et al. (2017) | CNN | 59 | Detecting culprit coronary arteries | CCTA and myocardial perfusion SPECT | Single center | |
| Stroke | ||||||
| Abouzari et al. (2009) | CNN | 300 | SDH post-surgery outcome prediction | CT head | Single center | |
| Alexander Roederer et al.(2014) | Logistic-linear regression | 81 | SAH-Vasospasm prediction | Passively obtained clinical data | Single center | |
| Arslan et al. (2016) | Logistic-linear regression, SVM, boosting algorithm | 80 | Ischemic stroke | EMR | Single center | |
| Atanassova et al. (2008) | CNN | 54 | Major stroke | Diastolic BP | 2 CNNs compared | Single center |
| Barriera et al. (2018) | CNN | 284 | Stroke (ICH and ischemic stroke) | CT head | Stroke neurologists reading CT | Single center |
| Beecy et al. (2017) | CNN | 114 | Stroke | CT head | Expert consensus interpretations | Single center |
| Dharmasaroja et al. (2013) | CNN | 194 | Stroke/intracranial hemorrhage | CT head | Thrombolysis after ischemic stroke | Single center |
| Fodeh et al. (2018) | SVM | 1834 | Atraumatic ICH | EHR review | Single center | |
| Gottrup et al. (2005) | kNN, Custom-built algorithm | 14 | Acute ischemic stroke | MRI | Applicability of highly flexible instance-based methods | Single center |
| Ho et al. (2016) | SVM, RF, and GBRT models | 105 | Acute ischemic stroke | MRI | Classification models for the problem of unknown time-since-stroke | Single center |
| Knight-Greenfield et al. (2018) | CNN | 114 | Stroke | CT head | Expert consensus interpretations | Single center |
| Ramos et al. (2018) | SVM, RF, Logistic-linear regression, CNN | 317 | SAH | CT Head | Delayed cerebral ischemia in SAH detection | Single center |
| SÜt et al. (2012) | MLP neural networks | 584 | Stroke mortality | EMR data | Selected variables using univariate statistical analyses | N/A |
| Paula De Toledo et al. (2009) | Logistic-linear regression | 441 | SAH | CT Head | Algorithms used were C4.5, fast decision tree learner, partial decision trees, repeated incremental pruning to produce error reduction, nearest neighbor with generalization, and ripple down rule learner | Multicenter Register |
| Thorpe et al. (2018) | decision tree | 66 | Stroke | Transcranial Doppler | Velocity Curvature Index (VCI) vs Velocity Asymmetry Index (VAI) | Single center |
| Williamson et al. (2019) | BOOSTING algorithm, RF | 483 | Risk stratification in SAH | True poor outcomes | Single center | |
| Xie et al. (2019) | Boosting algorithm | 512 | Predict Patient Outcome in Acute Ischemic Stroke | CT Head and clinical parameters | Feature selections were performed using a greedy algorithm | Single center |
| Heart failure | ||||||
| Andjelkovic et al. (2014) | CNN | 193 | HF in congenital heart disease | Echocardiography | Single center | |
| Blecker et al. (2018) | Logistic-linear regression | 37,229 | ADHF | Early ID of patients at risk of readmission for ADHF | 4 algorithms tested | Single center |
| Gleeson et al. (2016) | Custom-built algorithm | 534 | HF | Echocardiography and ECG | Data mining was applied to discover novel ECG and echocardiographic markers of risk | Single center |
| Golas et al. (2018) | Logistic-linear regression, boosting algorithm, CNN | 11,510 | HF | EHR | Heat failure patients to predict 30 day readmissions | Several hospitals in the Partners Healthcare System |
| Mortazavi et al. (2016) | Random forests, boosting, combined algorithms or logistic regression | 1653 | HF | Surveys to hospital examinations | Tele-HF trial | |
| Frizzell et al | Random forest and gradient-boosted algorithms | 56,477 | HF | EHR | Traditional statistical methods | GWTG-HF registry |
| Kasper Rossing et al. (2016) | SVM | 33 | HFpEF | Urinary proteomic analysis | Heart failure clinic (Single center) | |
| Kiljanek et al. (2009) | RF | 1587 | HF | Clinical diagnosis | Development of congestive heart failure after NSTEMI | CRUSADE registry |
| Liu et al. (2016) | Boosting algorithm, Logistic-linear regression | 526 | HF | Medical data, blood test, and echocardiographic imaging | Predicting mortality in HF | Single center |
SVM support vector machine, RF random forest, CNN convolutional neural network, kNN k-nearest neighbors, PCA principal component analysis, GBRT gradient boosted regression trees, MLP multilayer perceptron, HER electronic health record, HF heart failure, HFpEF heart failure with preserved ejection fraction, ADHF acute decompensated heart failure, SAH subarachnoid hemorrhage, SDH subdural hematoma, ICH intracerebral hemorrhage, CAD coronary artery disease, ACS acute coronary syndrome, CCTA coronary computed tomography angiography, FFR fractional flow reserve, IVUS intravascular ultrasound, ICD implantable cardioverter-defibrillator, AF atrial fibrillation, ECG electrocardiogram.