Rib fracture detection in computed tomography images using deep convolutional neural networks

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Radiology

Background:

• Rib fractures are common injuries in thoracic trauma.
• Accurate detection of rib fractures is crucial for patient management.
• Current diagnostic methods rely on radiologist interpretation of CT images.

Purpose of the Study:

• To evaluate the performance of a deep convolutional neural network (DCNN) software in detecting rib fractures on computed tomography (CT) images.
• To compare the DCNN software's diagnostic accuracy against that of junior doctors.

Main Methods:

• CT images from 39 patients with thoracic injuries were analyzed.
• A gold standard for 256 rib fractures was established by two radiologists.
• Performance was compared using McNemar test and jackknife alternative free-response receiver operating characteristic (JAFROC) analysis.

Main Results:

• The DCNN software demonstrated significantly higher sensitivity (0.645) than Intern A (0.313) and Intern B (0.258).
• JAFROC analysis indicated the DCNN software was non-inferior to both interns, with figure-of-merits of 0.057 and 0.071, respectively.
• The DCNN software met the non-inferiority margin of -0.10.

Conclusions:

• The DCNN software is a viable alternative for rib fracture detection in CT images.
• This AI tool can assist clinicians, particularly those with less experience in interpreting imaging findings.
• The DCNN software shows potential to improve the accuracy and efficiency of rib fracture diagnosis.