Skip to main content

Table 1 Main results in subgroups based on image quality

From: Machine learning for the real-time assessment of left ventricular ejection fraction in critically ill patients: a bedside evaluation by novices and experts in echocardiography

Image quality

Good

Good

Good and fair

Good and fair

Fair and poor

Fair and poor

Green-flagged

Green-flagged

Measurement

Manual

Manual

Manual

Manual

Manual

Manual

ML

ML

Operator

Expert

Novice

Expert

Novice

Expert

Novice

Expert

Novice

Patients, n

41

41

84

84

54

54

80

80

Reproducibility

5 ± 4%

6 ± 4%

5 ± 4%

6 ± 5%

5 ± 4%

5 ± 5%

5 ± 4%

6 ± 5%

Correlation coef.

r = 0.86

r = 0.78

r = 0.85

r = 0.81

r = 0.86

r = 0.83

r = 0.86

r = 0.82

Bias (mean ± SD)

0 ± 6%

0 ± 7%

0 ± 6%

0 ± 7%

 − 1 ± 6%

 − 1 ± 6%

0 ± 6%

0 ± 7%

Sensitivity

65%

73%

65%

69%

Specificity

100%

100%

100%

100%

  1. ML = Machine learning, coef. = coefficient. Sensitivity and specificity were not calculated for small subgroups