Detecting intertrochanteric hip fractures with orthopedist-level accuracy using a deep convolutional neural network

Area of Science:

• Radiology
• Artificial Intelligence in Medicine
• Orthopedic Surgery

Background:

• Intertrochanteric hip fractures are a common injury, particularly in elderly populations.
• Accurate and timely diagnosis is crucial for effective patient management and outcomes.
• Radiographic interpretation remains the primary diagnostic modality.

Purpose of the Study:

• To compare the diagnostic performance of a convolutional neural network (CNN) against orthopedic surgeons for intertrochanteric hip fractures.
• To evaluate the potential of AI in improving fracture detection on proximal femoral radiographs.

Main Methods:

• A dataset of 3346 hip radiographs (1773 fractured, 1573 non-fractured) was analyzed.
• Images were cropped to focus on the proximal femur, excluding specific complex cases.
• Performance was compared between a CNN and five orthopedic surgeons.

Main Results:

• The CNN achieved higher accuracy (95.5%) than orthopedic surgeons (92.2%).
• CNN demonstrated superior sensitivity (93.9% vs. 88.3%) and comparable specificity (97.4% vs. 96.8%).
• Confidence intervals indicated statistically significant differences in accuracy and sensitivity.

Conclusions:

• The convolutional neural network outperformed orthopedic surgeons in diagnosing intertrochanteric hip fractures under specific conditions.
• AI holds significant potential as a screening tool for plain radiograph fracture detection.
• CNNs could be particularly valuable in emergency departments with limited access to orthopedic specialists.