Table 1.
Artificial intelligence for image analysis and quantification.
| Authors | Summary | Data | Performance | |||
|---|---|---|---|---|---|---|
| Acquisition | Datasets | Patients | Metric value | Compared against | ||
| Left ventricular function assessment and quantification | ||||||
| Asch et al. | Automated EF using ML to assess LV function and volumes | 2D | 1 | > 50.000 | r = 0.95 | EA |
| Cannesson et al. | Automated EF using AI to assess LV function and volumes | 2D | 1 | 218 | r = 0.96 | EA |
| Hubert et al. | Automated diastolic function assessment | 2D | 1 | 50 | AUC 0.91 | OVS |
| Knackstedt et al. | Automated EF and strain using ML to assess LV function | 2D | 4 | 255 | ICC 0.83 | EA |
| Lancaster et al. | Automated diastolic function assessment | 2D | 1 | 866 | Kappa 0.62 | OVS |
| Medvedofsky et al. | Automated EF using ML to assess LV function and volumes | 3D | 6 | 180 | r 0.94 | EA |
| Rahmouni et al. | Automated EF using AI to assess LV function and volumes | 2D | 1 | 92 | r = 0.64 | EA |
| Sabovik et al. | Automated diastolic function assessment | 2D | 1 | 1,407 | AUC 0.88 | OVS |
| Tsang et al. | Automated EF using ML to assess LV function and volumes | 3D | 1 | 159 | r 0.87–0.96 | EA |
| Disease classification | ||||||
| Calleja et al. | Automated quantification using ML to assess aortic stenosis and regurgitation | 3D | 1 | 40 | ICC 0.99 | OIM |
| Casaclang et al. | Automated ventricular response to AS using ML | 2D | 1 | 246 | p < 0.001 | EA |
| Diller et al. | Automated segmentation using DL to detect congenital heart disease | 2D | 2 | 239 | AUC 0.98 | EA |
| Ghesu et al. | Automated detection valve morphology using DL | 3D | X | 869 | CE 45.2% | CT |
| Jeganathan et al. | Evaluate valve morphology using AI in mitral valve analysis | 3D | 1 | 4 | P = 0.0083 | EA |
| Jin et al. | Automated localizing prolapse using ML to evaluate mitral insufficiency | 3D | 1 | 90 | AC 0.89 | EA |
| Madani et al. | Automated diagnosis ventricular hypertrophy using DL | 2D | 1 | 79.937 | AUC 91.2 | EA |
| Moghaddasi et al. | Automated quantification mitral regurgitation using ML | 2D | 1 | 102 | AUC 0.99 | EA |
| Narula et al. | Automated discrimination HCM or athlete heart using ML | 2D | 1 | 139 | S&S p = 0.04 | EA |
| Pereira et al. | Automated detection aortic coarctation using DL | 2D | 1 | 163 | ER 12.9 | EA |
| Sanchez et al. | Automated clustering using ML for group classification | 2D | 4 | 156 | κ, 72.6% | EA |
| Sengupta et al. | Automated discrimination pericarditis or RCM using ML | 2D | 2 | 94 | AUC 0.89 | OIM |
| Zhang et al. | Automated discrimination HCM, amyloidosis, or PAH using DL | 2D | 1 | 14.035 | AUC >0.84 | EA |
AC, accuracy; AUC, area under curve; DC, dice coefficient; EA, expert assessment; EF, ejection fraction; EV, echo vendor; HCM, hypertrophic cardiomyopathy; IG, information gain; LV, left ventricular; ICC, intraclass correlation coefficient; OIM, other image modality; OVS, other validated scores; RCM, restrictive cardiomyopathy; SE, segmentation error; X, not available.