Automatic Surgical Skill Rating Using Stylistic Behavior Components.

Area of Science:

• Robotics
• Surgical Skill Assessment
• Human-Computer Interaction

Background:

• Direct observation is the gold standard for surgical skill evaluation but is time-consuming.
• Existing automatic assessment methods often neglect user's stylistic behavior.
• Few studies focus on automatic rating of surgical skills using established Likert scales.

Purpose of the Study:

• To evaluate a previously developed stylistic behavior lexicon for automatic robotic surgical skill rating.
• To assess the lexicon's effectiveness in relation to the Global Evaluative Assessment of Robotic Skills (GEARS) domains.
• To compare automatic ratings derived from sensor data versus crowd-sourced lexicon ratings.

Main Methods:

• 14 subjects of varying skill levels performed tasks on a da Vinci surgical simulator.
• Limb kinematics and joint positions were recorded.
• Faculty experts rated performance using GEARS; videos were also rated via crowd-sourcing using the stylistic lexicon.
• Two regression models were trained: one with sensor data, one with crowd ratings.

Main Results:

• Both prediction models, using sensor measurements and crowd ratings, showed agreement with gold standard faculty ratings.
• The stylistic behavior lexicon demonstrated effectiveness in automatically rating robotic surgical skills.
• The study validates the potential of automatic assessment tools in surgical training.

Conclusions:

• The proposed stylistic behavior lexicon can effectively automate robotic surgical skill assessment.
• Automatic rating systems can complement traditional evaluation methods, reducing time and resource intensity.
• This approach offers a scalable solution for surgical skill evaluation in robotic surgery training.