Using Deep Learning to Suggest Treatment for Proximal Humerus Fractures

Area of Science:

• Orthopedic Surgery
• Medical Imaging
• Artificial Intelligence

Background:

• Proximal humeral fractures are common emergency department visits.
• Accurate diagnosis and treatment selection can be challenging and time-consuming.
• Consultation with orthopedic surgeons may cause delays.

Purpose of the Study:

• To develop a machine learning model for predicting treatment type for proximal humeral fractures.
• To assist emergency department practitioners in making timely therapeutic decisions.

Main Methods:

• A machine learning model was developed to analyze radiographic images of proximal humeral fractures.
• The model was trained to differentiate between nonoperative and operative treatment options.

Main Results:

• The model achieved 86% accuracy in predicting treatment type.
• It demonstrated an interobserver reliability (kappa) of 0.722 on the test dataset.
• This reliability surpasses the agreement typically seen between shoulder surgeons.

Conclusions:

• The machine learning model shows significant potential as a decision support system.
• It can expedite treatment decisions for proximal humeral fractures in emergency settings.
• The model may help reduce patient waiting times and improve efficiency of care.