A Neural Network Model for Intelligent Classification of Distal Radius Fractures Using Statistical Shape Model...

Area of Science:

• Orthopedic imaging analysis
• Artificial intelligence in medicine
• Biomechanical modeling

Background:

• Distal radius fractures are common, with misdiagnosis rates up to 29% in emergency settings.
• Existing AI fracture detection methods lack classification capabilities and require large datasets.
• Accurate classification of distal radius fractures is vital for effective treatment planning.

Purpose of the Study:

• To develop and validate an intelligent classifier for distal radius fractures.
• To combine a statistical shape model (SSM) with a neural network (NN) for fracture detection and classification.
• To utilize CT imaging data for an efficient and accurate diagnostic tool.

Main Methods:

• Collected 80 CT scans (43 normal, 37 fractures: Colles', Barton's, Smith's).
• Established distal radius SSM using mean values and PCA features, defining six morphological indicators.
• Trained an SSM+NN classifier with SSM features and validated using four-fold cross-validation.

Main Results:

• Successfully established SSMs for normal and fractured distal radii.
• Significant differences in morphological indicators (p < 0.001) across groups.
• Classifier achieved 97.5% accuracy on the test set with a mean AUC of 0.95.

Conclusions:

• The CT-based SSM+NN classifier accurately identifies and classifies distal radius fractures.
• This novel approach offers an efficient, automated tool for clinical diagnosis.
• Potential to enhance diagnostic efficiency and treatment planning in orthopedics.