Skip to main content

Table 2 The levels of prediction for combined (men + women) model

From: Prediction of dysphagia aspiration through machine learning-based analysis of patients’ postprandial voices

Model

Pre-trained models

Non-pre-trained models

mn40_as

mn30_as

mn4.0

mn3.0

AUC (Area under the curve)

AUC average

(95% CI)

0.8275

(0.7643, 0.8908)

0.8361

(0.7667, 0.9056)

0.8039

(0.7378, 0.8700)

0.8177

(0.7601, 0.8753)

AUC max in 10 folds

0.9500

0.9541

0.9691

0.9561

Accuracy (%)

Accuracy average

(95% CI)

71.47

(66.73, 76.21)

77.98

(70.07, 85.89)

73.43

(68.23, 78.63)

74.98

(70.18, 79.77)

Accuracy max in 10 folds

84.91

92.45

86.90

88.68

mAP (mean average precision, %)

mAP average

(95% CI)

83.62

(77.74, 89.51)

84.54

(78.57, 90.52)

81.05

(75.10, 87.00)

83.07

(78.13, 88.02)

mAP max in 10 folds

95.47

95.46

97.23

95.10

Sensitivity (%)

Sensitivity average

(95% CI)

71.47

(66.73, 76.21)

77.80

(69.87, 85.74)

73.55

(68.34, 78.77)

74.85

(70.07, 79.63)

Sensitivity max

in 10 folds

84.91

92.45

86.90

88.68

Specificity (%)

Specificity average

(95% CI)

72.43

(67.26, 77.60)

77.52

(69.75, 85.28)

73.16

(67.67, 78.64)

74.73

(69.01, 80.45)

Specificity max in 10 folds

85.91

93.94

88.39

90.91

Precision (%)

Precision average

(95% CI)

71.47

(66.80, 76.15)

77.78

(70.14, 85.42)

72.90

(68.17, 77.64)

74.06

(69.08, 79.03)

Precision max in 10 folds

84.05

91.67

85.10

88.46

F1 Score

F1 Score average

(95% CI)

0.7173

(0.6697, 0.7648)

0.7777

(0.6994, 0.8560)

0.7350

(0.6811, 0.7889)

0.7492

(0.7004, 0.7980)

F1 Score max in 10 folds

0.8510

0.9255

0.8720

0.8885

Loss

Loss average

(95% CI)

0.9225

(0.6930, 1.1520)

0.8524

(0.5410, 1.1640)

1.6013

(1.0110, 2.1920)

1.3553

(0.9250, 1.7860)

Loss max in 10 folds

1.6120

1.4136

3.1602

2.3892

Train accuracy (%)

Train accuracy average

(95% CI)

99.97

(99.91, 100.02)

99.98

(99.95, 100.02)

99.98

(99.94, 100.02)

99.93

(99.85, 100.02)

Train accuracy max

in 10 folds

100.00

100.00

100.00

100.00

Train loss

Train loss average

(95% CI)

0.0017

(0.0004, 0.0031)

0.0022

(0.0014, 0.0031)

0.0010

(− 0.0001, 0.0021)

0.0052

(− 0.0024, 0.0129)

Train loss max in 10 folds

0.0070

0.0045

0.0055

0.0350

  1. *All metrics represent the predictive performance on the Test Data except Train accuracy, and Train loss. The results presented in this table are the average predictive performance (95% CI) across all folds of each model after performing tenfold cross-validation
Back to article page

Journal of NeuroEngineering and Rehabilitation

ISSN: 1743-0003

Contact us