Novel radiographic stem version predictor from anterior-posterior radiographs

Area of Science:

• Orthopedic surgery
• Medical imaging analysis
• Computational geometry

Background:

• Implant orientation on radiographs impacts postoperative outcomes.
• Assessing 3D implant version from 2D radiographs is challenging due to projection complexities.
• Multiaxial rotations, including AP tilt and flexion, further complicate radiographic assessment.

Purpose of the Study:

• To introduce a novel computational method for evaluating radiographic stem version.
• To develop a technique robust to multiaxial rotations, specifically AP tilt and flexion.
• To enhance the accuracy and reliability of stem version assessment from radiographs.

Main Methods:

• Synthesized radiographic features using computational geometry and simulation, incorporating clinical error sources.
• Trained a Gaussian process regression model to predict radiographic stem version.
• Evaluated the impact of AP tilt on existing techniques (Weber) and assessed its feasibility for tilt measurement.

Main Results:

• The novel method achieved a high prediction accuracy (R²=0.98), significantly improving upon the Weber technique (R²=0.85).
• Similar high accuracy was confirmed in both in vitro and in silico datasets.
• AP tilt was shown to reduce the accuracy of the Weber technique and introduce ambiguity in assessment.

Conclusions:

• The developed feature-based method provides a reliable and robust measure of radiographic stem version, adaptable to multiaxial variations.
• This technique allows for accurate assessment of rotational changes in serial postoperative radiographs.
• A controlled radiographic view is necessary to ensure the measured version accurately reflects the implanted stem version.