Abstract:
Femoral neck fractures are considered to be one of the most challenging orthopedic
conditions because of the technical difficulties in the management of these injuries and
possible complications, including nonunion and avascular necrosis. These fractures are
common in elderly patients and they mostly occur due to low energy trauma such as falls.
Early diagnosis and identification of femoral neck fractures are critical to proper clinical
management and in reducing complications. This study employs the YOLOv8 model,
which is a recent advancement in object detection, to detect and classify femoral neck
fractures in X-ray images. The YOLOv8 model shows impressive results, with the mean
Average Precision mAP50 of 97. 9%, a precision of 93. 5%, and a mAP50-95 of 62. 5%.
Our proposed system encompasses several stages: data collecting, data preprocessing,
model training, model validation, and model deployment. In the process of data
preprocessing several data augmentation methods was performed to improve the model’s
resilience. The YOLOv8 model was then trained with this dataset and further rigorous
testing conducted to determine the efficiency of the model. The results show that the
proposed model has great potential for the automatic detection and classification of
femoral neck fractures, which will be helpful for radiologists. When implemented in
clinical environments, this system may increase diagnostic accuracy, decrease the
workload, and consequently, benefit patients
