Table 4 Comparison of sensitivity for orthopedic surgeons with and without CNN model assistance.

From: Automated fracture screening using an object detection algorithm on whole-body trauma computed tomography

Sensitivity (%)

Without CNN model assistance

With CNN model assistance

p value

Pelvic fracture

 Orthopedic surgeon 1

0.822

0.892

0.0158*

 Orthopedic surgeon 2

0.808

0.893

0.0060*

 Orthopedic surgeon 3

0.853

0.881

0.0184*

Rib fracture

 Orthopedic surgeon 1

0.580

0.747

0.0006*

 Orthopedic surgeon 2

0.596

0.729

0.0093*

 Orthopedic surgeon 3

0.651

0.719

0.1573

Spine fracture

 Orthopedic surgeon 1

0.649

0.814

0.0023*

 Orthopedic surgeon 2

0.676

0.795

< 0.0001*

 Orthopedic surgeon 3

0.762

0.838

0.0143*

Mean

 Orthopedic surgeon 1

0.697

0.829

< 0.0001*

 Orthopedic surgeon 2

0.676

0.818

< 0.0001*

 Orthopedic surgeon 3

0.764

0.826

0.0003*

  1. Orthopedic surgeon 1 and 2 had 3 years of experience. Orthopedic surgeon 3 had 8 years of experience. CNN, convolutional neural network.
  2. *p < 0.05.
Back to article page