Automatic Fracture Detection Convolutional Neural Network with Multiple Attention Blocks Using Multi-Region X-Ray...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Accurate fracture detection in X-rays is crucial for timely medical intervention.
• Current methods can be time-consuming and may have limitations in accuracy.
• Deep learning offers potential for automated and enhanced diagnostic capabilities.

Purpose of the Study:

• To develop and evaluate an advanced combined attention Convolutional Neural Network (CNN) model for improved fracture detection in X-ray images.
• To assess the diagnostic efficacy of the AI model before and after optimization.
• To demonstrate the role of attention mechanisms in enhancing feature representation for medical image analysis.

Main Methods:

• Development of a combined attention CNN model incorporating squeeze blocks and convolutional block attention module (CBAM).
• Training and evaluation using a diverse dataset of fractured and non-fractured X-rays from multiple anatomical locations (hips, knees, lumbar, limbs).
• Assessment of diagnostic performance using computed tomography X-ray images, comparing pre- and post-optimization efficacy.

Main Results:

• The AI model achieved a high training accuracy of 99.98% and a validation accuracy of 96.72%.
• The attention-based CNN effectively focused on relevant features, improving fracture detection capabilities.
• The model demonstrated strong generalization across various anatomical locations.

Conclusions:

• The developed attention-based CNN model shows significant promise for accurate and automated fracture detection in medical imaging.
• Incorporating attention mechanisms enhances the model's ability to interpret complex features in X-rays.
• This AI approach can support clinicians, reduce examination time, and improve patient outcomes.