Table 1.

Study characteristics

Study	Year	Aim	Model	Training strategy	Testing strategy
Cohen-Shelly	2021	ECG screening for moderatesevere AS	CNN	129,788 random subjects	102,926 random subjects
Elias	2022	ECG identification of moderate or severe AS, AR, and MR	CNN (ValveNet)	43,165 patients	21,048 patients
Hata	2020	ECG classification of AS	CNN	128 ECG data	44 ECG data
Kwon	2020	ECG detection of significant AS	MLP and CNN	39,371 ECGs from 25,733 patients	10,865 ECGs from 10,865 patients (external validation)
Kwon JM	2020	ECG detection of MR	CNN	56,670 ECGs from 24,202 patients	10,865 ECGs of 10,865 patients (external validation)
Lin	2021	ECG prediction of MVP	SVM, LR, MLP	1654 subjects	552 subjects
Sawano	2022	ECG diagnosis of significant AR	2D-CNN, FC-DNN	19,136 ECGs from 10,460 patients	6,036 ECGs from 3,269 patients
Tison	2019	ECG detection of MVP	CNN-HMM–heuristic filter	170 manually segmented ECGs	36,186 sinus rhythm ECGs
Ulloa-Cerna	2022	ECG prediction of moderate or severe valvular disease (AS, AR, MR, MS, TR)	CNN with classification pipeline (min–max scaling, mean imputation, XGBoost classifier, and calibration)	2,232,130 ECGs from 484,765 adults	276058 patients
Vaid	2023	ECG identification of AS and MR	MLP and CNN	617,338 ECG-Echo pairs from 123,096 patients (MR) 617,338 ECG-Echo pairs for 128,628 patients (AS)	617,338 ECG-Echo pairs from 123,096 patients (MR) 617,338 ECG-Echo pairs for 128,628 patients (AS)

2D-CNN = two-dimensional convolutional neural network; AR = aortic regurgitation; AS = aortic stenosis; CNN = convolutional neural network; ECG = electrocardiogram; FC-DNN = fully connected deep neural network; HMM = hidden Markov model; LR = logistic regression; MLP = multilayer perceptron; MR = mitral regurgitation; MS = mitral stenosis; MVP = mitral valve prolapse; SVM = support vector machine; TR = tricuspid regurgitation.