Kinematic analysis of lumbar pedicle screw placement using an artificial intelligence framework

Area of Science:

• Neurosurgery
• Surgical Robotics
• Artificial Intelligence in Medicine

Background:

• Spine surgery increasingly incorporates robotics and artificial intelligence (AI).
• Assessing surgical skill, particularly trainee proficiency, lacks standardized objective metrics.
• AI-based hand motion detection offers a potential solution for skill evaluation.

Purpose of the Study:

• To apply AI-based motion analysis and machine learning (ML) to evaluate hand movements during lumbar pedicle screw placement.
• To generate objective metrics for assessing surgical skill and proficiency.
• To establish a framework for objective skill assessment in spine surgery.

Main Methods:

• AI-based motion tracking analyzed hand movements during freehand (FH) and robot-assisted (RB) pedicle screw placement on a sawbone model.
• Extracted kinematic metrics included distance, displacement, speed, velocity, acceleration, and jerk.
• Data augmentation expanded a limited dataset, and ML models classified data by training level and technique.

Main Results:

• Procedure time and movement distance decreased with experience, particularly in FH procedures.
• Kinematic analysis showed reduced speed, displacement, and jerk variability with increased training.
• Robot-assisted procedures demonstrated less movement variability; ML models accurately classified training levels and techniques.

Conclusions:

• A data processing pipeline using AI video analysis can quantify surgical proficiency in spine procedures.
• Specific motion metrics can differentiate between freehand and robot-assisted techniques and correlate with training levels.
• This study provides a foundation for a standardized, objective framework for spine surgery skill assessment.