Table 3 Accuracy assessment of automatic segmentation.

From: Convolutional neural network for automatic maxillary sinus segmentation on cone-beam computed tomographic images

Metric

Descriptive analysis

Automatic vs ground truth

Automatic vs refined

DSC

Mean

0.984

0.996

SD

0.004

0.004

Min

0.962

0.983

Max

0.991

0.999

IoU

Mean

0.968

0.992

SD

0.008

0.007

Min

0.926

0.967

Max

0.983

0.998

95% HD (mm)

Mean

0.232

0.109

SD

0.059

0.115

Min

0.200

0

Max

0.447

0.283

RMS (mm)

Mean

0.209

0.214

SD

0.072

0.123

Min

0.142

0.100

Max

0.445

0.372

  1. DSC dice similarity coefficient, IoU intersection over union, HD hausdorff distance, RMS root mean square, SD standard deviation, Min minimal value, Max maximal value.
Back to article page