Is it feasible to develop a supervised learning algorithm incorporating spinopelvic mobility to predict impingement...

Area of Science:

• Orthopedic Surgery
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Accurate implant positioning in total hip arthroplasty (THA) is crucial for stability, influenced by individual spinopelvic biomechanics.
• Predicting impingement in THA is challenging, with limited research using artificial intelligence (AI).

Purpose of the Study:

• To evaluate the feasibility of an AI algorithm for predicting impingement in THA based on patient phenotype and spinopelvic mechanics.
• To assess the accuracy of AI in identifying impingement presence, direction, and type.

Main Methods:

• A prospective, international, multicentre cohort study of 157 adults undergoing robotic arm-assisted THA.
• Utilized Light Gradient-Boosting Machine (LGBM) with tabular data to predict impingement; a secondary model integrated radiographs.
• Impingement identified using the robotic software's virtual range of motion (ROM) tool.

Main Results:

• The LGBM model achieved 70.2% accuracy in predicting impingement presence.
• LGBM estimated impingement direction with 85% accuracy; SVM determined impingement type with 72.9% accuracy.
• Integrating imaging data did not significantly improve prediction accuracy compared to tabular data alone.

Conclusions:

• This pilot study pioneers AI for impingement prediction in THA using real-world clinical data.
• The developed machine-learning algorithm shows potential for predicting impingement, its type, and direction.
• External validation and larger studies are necessary before clinical implementation of the AI algorithm.