Comparative Analysis of Conventional and Focused Data Augmentation Methods in Rib Fracture Detection in CT Images.

Area of Science:

• Medical Image Processing
• Artificial Intelligence in Healthcare

Background:

• Rib fracture detection is crucial for patient diagnosis and treatment.
• Computed tomography (CT) imaging is a primary tool for identifying rib fractures.

Purpose of the Study:

• To compare the efficacy of traditional data augmentation versus focused data augmentation for rib fracture detection.
• To evaluate the performance of YOLOv8 models (YOLOv8n, YOLOv8s, YOLOv8m) with different augmentation strategies.

Main Methods:

• Utilized computed tomography (CT) images for rib fracture detection.
• Implemented and compared traditional data augmentation (Albumentations) with a novel focused data augmentation technique targeting fracture regions.
• Evaluated model performance using metrics including Precision, Recall, mAP@50, and mAP@50-95 on YOLOv8n, YOLOv8s, and YOLOv8m.

Main Results:

• Focused data augmentation demonstrated superior performance in specific metrics, notably enhancing recall for fracture detection.
• The YOLOv8s model with focused augmentation achieved a 5.70% increase in recall (0.8766) and a 2.18% increase in mAP@50 (0.9412).
• Traditional augmentation yielded better overall precision with the YOLOv8m model, indicating context-dependent benefits.

Conclusions:

• Focused data augmentation offers a promising strategy for improving the accuracy and reliability of rib fracture detection in medical imaging.
• The choice of data augmentation technique can significantly impact the performance of deep learning models in specialized medical applications.
• Further research into targeted augmentation methods could enhance diagnostic capabilities in radiology.