A transfer learning-based approach for automated bone fracture classification in X-ray imaging

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Orthopedics

Background:

• Accurate bone fracture classification is crucial in medical imaging.
• Deep learning offers enhanced diagnostic precision and reduced human error.

Purpose of the Study:

• To propose EnsembleAttenBoneNet, an ensemble deep learning model for bone fracture classification.
• To improve the accuracy and robustness of fracture detection in X-ray images.

Main Methods:

• An ensemble model combining ResNet50 and EfficientNetB3 with a Squeeze-and-Excitation (SE) attention mechanism was developed.
• Preprocessing included resizing, normalization, and augmentation for improved generalization.
• Features were extracted, concatenated, and refined using the SE attention module.

Main Results:

• The EnsembleAttenBoneNet model achieved a classification accuracy of 99.48%.
• This performance surpassed individual models: EfficientNetB3 (98.56%) and ResNet50 (97.86%).

Conclusions:

• Integrating deep learning with attention mechanisms significantly enhances diagnostic accuracy for bone fractures.
• The model shows promise as a valuable tool for clinical fracture detection.
• Future work includes dataset expansion and real-world validation.