Assessing the Influence of Screw Orientation on Fracture Fixation of the Proximal Humerus Using Finite Element...

Area of Science:

• Orthopedic surgery
• Biomechanical engineering
• Computational modeling

Background:

• Proximal humeral fractures present management challenges, with traditional fixation plates exhibiting high failure rates.
• Variable angle screws offer potential benefits, but comprehensive analysis of screw orientation effects is limited by computational expense of finite element (FE) studies.

Purpose of the Study:

• To develop adaptive neural network (ANN) models for predicting screw collision and humeral bone strain.
• To identify optimal screw orientations for improved fixation and reduced complication risk.
• To overcome the computational limitations of traditional FE analysis for exploring screw orientation parameter spaces.

Main Methods:

• Trained ANN models using data from FE simulations of variable angle screws in a proximal humerus fracture model.
• Validated ANN predictions against unseen FE data for screw collision and bone strain.
• Utilized trained ANNs to predict outcomes across a full factorial of screw orientations.

Main Results:

• ANN models achieved 84.4% accuracy in predicting screw collision.
• Accurate prediction of bone strain with high correlation (R²=0.99) and low error (RMSE=0.65%-5.49%).
• Screw orientation in the calcar region significantly impacts bone strain around all screws.

Conclusions:

• ANNs provide a computationally efficient method for analyzing screw orientation effects in fracture fixation.
• Optimizing screw placement, especially in the calcar region, is critical for minimizing bone strain and enhancing fixation stability.
• This approach facilitates identification of optimal and worst-case scenarios for screw fixation in proximal humerus fractures.