Enhancing Wrist Fracture Detection with YOLO

Diagnosing and treating abnormalities in the wrist, specifically distalradius, and ulna fractures, is a crucial concern among children, adolescents,and young adults, with a higher incidence rate during puberty. However, thescarcity of radiologists and the lack of specialized training among medicalprofessionals pose a significant risk to patient care. This problem is furtherexacerbated by the rising number of imaging studies and limited access tospecialist reporting in certain regions. This highlights the need forinnovative solutions to improve the diagnosis and treatment of wristabnormalities. Automated wrist fracture detection using object detection hasshown potential, but current studies mainly use two-stage detection methodswith limited evidence for single-stage effectiveness. This study employsstate-of-the-art single-stage deep neural network-based detection modelsYOLOv5, YOLOv6, YOLOv7, and YOLOv8 to detect wrist abnormalities. Throughextensive experimentation, we found that these YOLO models outperform thecommonly used two-stage detection algorithm, Faster R-CNN, in fracturedetection. Additionally, compound-scaled variants of each YOLO model werecompared, with YOLOv8m demonstrating a highest fracture detection sensitivityof 0.92 and mean average precision (mAP) of 0.95. On the other hand, YOLOv6machieved the highest sensitivity across all classes at 0.83. Meanwhile, YOLOv8xrecorded the highest mAP of 0.77 for all classes on the GRAZPEDWRI-DX pediatricwrist dataset, highlighting the potential of single-stage models for enhancingpediatric wrist imaging.