Table 3.

Summary of studies highlighting the role of artificial intelligence (AI) enhanced echocardiography.

Author	Aim of Study	Sample Size	Method of AI	Results
Asch et al. 36	To evaluate AI algorithms for the detection of endocardial boundaries and measurement of LV volumes and function	99	ML algorithm	Auto-EF values showed high consistency (mean absolute deviation =2.9%) and excellent agreement with the reference values: r=0.95, bias=1.0%, with sensitivity 0.90 and specificity 0.92 for detection of EF ≤35%
Narang et al. 37	To test whether novice users could obtain 10-view TTE studies of diagnostic quality	240	DL algorithm	> 90% of diagnostic quality for key parameters such as LV size and function, RV size and function, and pericardial effusion
Schneider et al. 38	To test the algorithm by having 19 first-year medical students without previous ultrasound knowledge scan patients	57	ML model	Successfully acquired at least one of the three required views 91% of the time
Cheema et al. 39	To describe the use of AI technology that guides novice users to acquire high-quality cardiac ultrasound images	5	DL algorithm	Accurately interpreted ventricular dysfunctions and helped in guiding medical therapies
Leclerc et al. 40	To evaluate the DL models in assessing 2D echocardiographic images, i.e., segmenting cardiac structures and estimating clinical indices	500	DL algorithm	It provided more consistent data with less variability
Ouyang et al. 41	To evaluate a video-based DL algorithm for accurate assessment of cardiac function in echocardiogram videos	55	DL algorithm	Accurately and rapidly calculated the EF (absolute error 4.1%)
Jafari et al. 42	To evaluate computationally efficient mobile application with POCUS for accurate LVEF estimation	427	DL algorithm	Accurately calculated LVEF with a median absolute error of 6.2% compared to an expert cardiologist assessment
Asch et al 43	To test the accuracy of AI algorithms LV volume and function based on POCUS exams	166	ML algorithm	The agreement with the reference EF values was good (intraclass correlation, 0.86–0.95), with biases <2%. ML classification of LV function showed similar accuracy to that by physicians in most views.
Tromp et al 44	To compare a DL interpretation of 23 echocardiographic parameter with three repeated measurements by core lab sonographers	600	DL algorithm	The point estimates of individual equivalence coefficient were all <0 and the upper bound of the 95% confidence intervals below 0.25, indicating that the disagreement between the DL and human measures was lower than the disagreement among three core lab readers.
Akerman et al 45	To analyze a single apical 4-chamber transthoracic echocardiogram video clip to detect HFpEF.	6756	ML algorithm	Excellent discrimination (area under receiver-operating characteristic curve: 0.97 [95% CI: 0.96-0.97] and 0.95 [95% CI: 0.93-0.96] in training and validation, respectively)
Kelsey et al. 46	To evaluate the role of AI echo in assessing cardiac structure and function among rural populations	138	DL algorithm	Adequate image quality for visual EF determination in 97%, with LV dimensions measurable in 88% and LA diameter in 91%

AI, artificial intelligence; DL, deep learning; ML, machine learning; LA, left atrium; LV, left ventricle; RV, right ventricle; EF, ejection fraction; POCUS, point-of-care ultrasound; HFpEF, heart failure with preserved ejection fraction and TTE, transthoracic echocardiography