Scaphoid Fracture Detection by Using Convolutional Neural Network.

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Scaphoid fractures are common but often occult on initial X-rays.
• Traditional image processing for scaphoid fracture detection is time-consuming and requires manual input.
• Existing methods have limited effectiveness in detecting scaphoid fractures from X-ray images.

Purpose of the Study:

• To propose a novel two-stage convolutional neural network (CNN) for accurate scaphoid fracture detection.
• To enhance the diagnostic capabilities for occult scaphoid fractures using deep learning.
• To evaluate the performance of the proposed CNN model in detecting and classifying scaphoid fractures.

Main Methods:

• A two-stage CNN approach was developed, utilizing Faster R-CNN for scaphoid bone segmentation.
• The ResNet model, feature pyramid network, and convolutional block attention module were employed for fracture detection and classification.
• Performance was assessed using metrics including recall, precision, accuracy, sensitivity, specificity, and AUC.

Main Results:

• The scaphoid bone detection stage achieved 99.70% accuracy.
• Scaphoid fracture detection yielded a recall of 0.789, precision of 0.894, accuracy of 0.853, and AUC of 0.920.
• Scaphoid fracture classification demonstrated a recall of 0.735, precision of 0.898, accuracy of 0.829, and AUC of 0.917.

Conclusions:

• The proposed two-stage CNN effectively detects and classifies scaphoid fractures from X-ray images.
• This AI-driven method offers a promising solution for improving the diagnosis of scaphoid fractures.
• Future research should explore integrating multi-view X-ray images for enhanced CNN performance.