Table 2.
Studies of AI’s Application in Left Ventricular Systolic Function—GLS.
| Authors | Year | Task | Models | Dataset | Results |
|---|---|---|---|---|---|
| Kawakami H. et al. [34] | 2021 | GLS assessment | AutoStrain | 561 subjects | Automated vs. manual GLS: r = 0.685, bias = 0.99%. Semi-automated vs. manual GLS: r = 0.848, bias = −0.90%. Automated vs. semi-automated GLS: r = 0.775, bias = 1.89%. |
| Salte I.M. et al. [22] | 2021 | GLS assessment | EchoPWC-Net | 200 studies | EchoPWC-Net vs. EchoPAC: r = 0.93, MD 0.3 ± 0.3%. |
| Evain E. et al. [36] | 2022 | GLS assessment | PWC-Net | >60,000 images | Automated vs. Manual GLS: r = 0.77, MAE 2.5 ± 2.1%. |
| Narula S. et al. [25] | 2016 | Disease detection (ATH vs. HCM) |
Ensemble model (SVM, RF, ANN) |
77 ATH, 62 HCM patients |
Sensitivity 0.96; specificity 0.77. |
| Sengupta P.P. et al. [26] | 2016 | Disease detection (CP vs. RCM) |
AMC | 50 CP patients, 44 RCM patients, and 47 controls |
AUC 0.96. |
| Zhang J. et al. [27] | 2021 | Disease detection(CHD) | Two-step stacking | 217 CHD patients, 207 controls |
Sensitivity 0.903; specificity 0.843; AUC 0.904. |
| Loncaric F. et al. [37] | 2021 | Disease detection (HT) | ML | 189 HT patients, 97 controls |
HT is divided into 4 phenotypes. |
| Yahav A. et al. [38] | 2020 | Disease detection (strain curve classification) |
ML | 424 subjects | Strain curve is divided into physiological, non-physiological, and uncertain categories (accuracy 86.4%). |
| Pournazari P. et al. [39] | 2021 | Prognosis analysis (COVID-19) |
ML | 724 subjects | BC (AUC 0.79). BC + Laboratory data + Vital signs (AUC 0.86). BC + Laboratory data + Vital signs + Echos (AUC 0.92). |
| Przewlocka-Kosmala M. et al. [40] | 2019 | Prognosis analysis (HFpEF) | Clustering | 177 HFpEF patients, 51 asymptomatic controls |
HFpEF is divided into 3 prognostic phenotypes. |
GLS, global longitudinal strain; MD, mean difference; MAE, mean absolute error; ATH, athletes; HCM, hypertrophic cardiomyopathy; SVM, support vector machine; RF, random forest; ANN, artificial neural networks; CP, constrictive pericarditis; RCM, restrictive cardiomyopathy; AMC, associative memory classifier; AUC, area under the receiver operating characteristic curve; CHD, coronary heart disease; HT, hypertension; ML, machine learning; HFpEF, heart failure with preserved ejection fraction; BC, baseline characteristics; Echos, echocardiographic measurements.