Deep learning for automated hip fracture detection and classification : achieving superior accuracy.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Orthopaedic Surgery

Background:

• Hip fractures are a common and serious injury, particularly in older adults.
• Accurate and timely diagnosis is crucial for effective treatment and patient outcomes.
• Current diagnostic methods can be subject to human error and variability.

Purpose of the Study:

• To develop and evaluate a deep learning-based model for the classification of hip fractures.
• To enhance the diagnostic accuracy of hip fracture detection.
• To assess the model's performance against orthopaedic surgeons' diagnoses.

Main Methods:

• A retrospective study utilizing 5,168 hip anteroposterior radiographs.
• A convolutional neural network (CNN)-based classification model was trained using DAMO-YOLO for data processing.
• The model was trained on four hip fracture types: Displaced, Valgus-impacted, Stable, and Unstable.

Main Results:

• The model achieved high accuracy, sensitivity, and specificity across all four fracture categories on the internal dataset.
• External validation demonstrated robust performance with high sensitivity and specificity for each fracture type.
• The model's performance metrics, including Intersection over Union (IoU) and Dice coefficient, were evaluated.

Conclusions:

• The developed deep learning model significantly improves hip fracture detection and classification accuracy.
• The AI model demonstrates substantial potential as a tool to aid clinicians in diagnosing hip fractures.
• This technology can assist in providing more accurate and consistent diagnoses, leading to better patient care.