Skip to main content

Table 2 Segmentation performances (Mean (SD)) of intersection over union (IoU), Dice similarity coefficient score (DSC), precision (PR), recall (RC), relative volume difference (RVD), and volume of error (VOE) by Dense121 U-Net, Attention U-Net, simple U-Net, and SegNet for the mandibular third molar and the mandibular canal

From: Automatic classification of 3D positional relationship between mandibular third molar and inferior alveolar canal using a distance-aware network

Mandibular third molar

Segmentation models

IoU

DSC

Precision

Recall

RVD

VOE

SegNet

0.841 ± 0.183

0.898 ± 0.156

0.939 ± 0.093

0.888 ± 0.176

0.049 ± 0.031

0.111 ± 0.027

simple U-Net

0.846 ± 0.181

0.902 ± 0.147

0.944 ± 0.092

0.889 ± 0.177

0.056 ± 0.034

0.105 ± 0.034

Attention U-Net

0.863 ± 0.161

0.916 ± 0.125

0.940 ± 0.086

0.911 ± 0.151

0.038 ± 0.021

0.093 ± 0.027

Dense121 U-Net

0.872 ± 0.161

0.920 ± 0.131

0.946 ± 0.091

0.918 ± 0.148

0.038 ± 0.025

0.088 ± 0.024

Mandibular canal

Segmentation models

IoU

DSC

Precision

Recall

RVD

VOE

SegNet

0.659 ± 0.212

0.770 ± 0.194

0.837 ± 0.171

0.741 ± 0.226

0.162 ± 0.104

0.357 ± 0.146

simple U-Net

0.721 ± 0.162

0.825 ± 0.133

0.899 ± 0.106

0.785 ± 0.176

0.192 ± 0.106

0.303 ± 0.114

Attention U-Net

0.770 ± 0.132

0.863 ± 0.102

0.898 ± 0.093

0.846 ± 0.143

0.131 ± 0.074

0.249 ± 0.077

Dense121 U-Net

0.766 ± 0.125

0.861 ± 0.096

0.911 ± 0.085

0.830 ± 0.136

0.135 ± 0.071

0.248 ± 0.075