Binomial Classification of Pediatric Elbow Fractures Using a Deep Learning Multiview Approach Emulating Radiologist...

Area of Science:

• Radiology
• Artificial Intelligence
• Deep Learning

Background:

• Pediatric elbow fractures are common injuries.
• Accurate diagnosis relies on expert interpretation of radiographic images.
• Deep learning offers potential for automated image analysis.

Purpose of the Study:

• To assess the feasibility of a deep learning model for classifying pediatric elbow radiographic abnormalities.
• To implement a multiview approach mimicking human radiologist review.
• To perform binomial classification of acute traumatic abnormalities.

Main Methods:

• Retrospective analysis of 21,456 pediatric elbow radiographic studies (58,817 images).
• A deep learning model combining convolutional neural network and recurrent neural network was developed.
• The model interpreted entire series of three radiographs for a multiview analysis.

Main Results:

• The model achieved an area under the receiver operating characteristic curve (AUC) of 0.95 and 88% accuracy.
• High sensitivity (91%) and specificity (84%) were observed for abnormality detection.
• The model demonstrated effectiveness in identifying various fractures and effusions.

Conclusions:

• Deep learning models can accurately classify acute pediatric elbow abnormalities on radiographs.
• The multiview recurrent neural network approach aids in fracture identification in pediatric patients.
• This technology shows promise for supporting radiologists in trauma settings.