Skip to main content

Table 2 Comparison of methods—testing set

From: Machine learning for early prediction of acute myocardial infarction or death in acute chest pain patients using electrocardiogram and blood tests at presentation

 

Sens (95% CI)

NPV (95% CI)

Ruled out % (95% CI)

Ruled out (n)

Missed AMI or Death

Rule out

     

 ESC 0 h

98.9 (96.15–99.87)

99.8 (99.30–99.96)

47.2 (45.1–49.3)

1123

2

 LogReg

97.8 (94.56–99.41)

99.6 (98.85–99.83)

38.5 (36.5–40.4)

915

4

 ANN

99.5 (97.03–99.99)

99.9 (99.37–99.99)

46.6 (44.6–48.7)

1109

1

 CNN-MB

99.5 (97.03–99.99)

99.9 (99.46–99.99)

55 (53.1–57)

1309

1

 CNN-RAW

99.5 (97.03–99.99)

99.9 (99.42–99.99)

50.8 (48.8–52.9)

1208

1

 

Spec (95% CI)

PPV (95% CI)

Ruled in % (95% CI)

Ruled in (n)

Incorrectly ruled in

Rule in

     

 ESC 0 h

97.4 (96.65–98.03)

63.9 (57.06–70.27)

6.6 (5.7–7.8)

158

57

 LogReg

98.4 (97.74–98.85)

65.1 (56.09–73.05)

4.3 (3.5–5.2)

103

36

 ANN

98.2 (97.58–98.73)

69.8 (62.05–76.51)

5.4 (4.5–6.4)

129

39

 CNN-MB

98.5 (97.89–98.96)

73.6 (65.86–80.12)

5.3 (4.4–6.2)

125

33

 CNN-RAW

98.2 (97.58–98.73)

70.5 (62.87–77.06)

5.5 (4.6–6.5)

132

39

  1. Performance with respect to rule-out (sensitivity and NPV) and rule-in (Specificity and PPV)
  2. NPV, negative predictive value; PPV, positive predictive value; n, number; ESC 0 h, 0 h arm of the European Society of Cardiology algorithm; LogReg, logistic regression; ANN, artificial neural network, CNN-MB, convolutional neural network trained on median beat ECG data; CNN-RAW, convolutional neural network trained on raw ECG data