Table 8 Overall performance across six datasets (EHST vs. Baselines).

From: Explainable attention-based deep learning for classification and interpretation of heart murmurs using phonocardiograms

Dataset

EHST (Ours)

Mean of 10 Baselines

Best Baseline

Acc (%)

F1 (%)

AUC

Acc (%)

F1 (%)

AUC

Acc (%)

F1 (%)

AUC

HeartWave

96.7

95.5

0.97

92.4

90.3

0.93

94.8

93.1

0.95

CirCor DigiScope

94.1

92.6

0.95

89.5

87.1

0.91

91.6

89.9

0.93

PhysioNet CinC

90.3

88.9

0.92

85.7

82.4

0.88

88.2

86.1

0.89

Pascal (A+B)

89.2

88.1

0.91

84.6

81.9

0.87

86.8

85.2

0.89

GitHub Valvular

92.5

91.8

0.93

88.6

86.7

0.90

91.2

90.3

0.91

Shenzhen (HSS)

91.9

90.7

0.92

87.4

85.5

0.88

89.7

88.2

0.90