Multi-modal Features Analysis and Performance Assessment for Endovascular Surgery Skills

Area of Science:

• Medical Simulation
• Surgical Skill Assessment
• Biomedical Engineering

Background:

• Objective assessment of interventional skills is crucial for vascular interventionalists.
• Traditional surgical assessment methods lack objectivity and accuracy.
• There is a need for advanced tools to evaluate endovascular procedural performance.

Purpose of the Study:

• To propose a simulator and analytical framework for evaluating endovascular procedural performance.
• To address the limitations of traditional surgical assessments.
• To enhance the objectivity and accuracy of interventional skill evaluation.

Main Methods:

• Developed a custom interventional operation simulator with kinematic and force sensing.
• Collected multi-modal feature data (morphological, interaction, kinematic, force) from 30 interventionalists and 10 novices performing guidewire tasks.
• Utilized long short-term memory (LSTM) networks for operational force prediction and support vector machines/Mahalanobis distance for skill assessment.

Main Results:

• The LSTM model accurately predicted operational forces, and forces/torques between instruments and the vascular model.
• Qualitative assessment achieved 84.17% accuracy in distinguishing novice and expert attempts.
• Quantitative assessment provided effective scoring for all procedural attempts.

Conclusions:

• The proposed approach enables prediction and assessment of interventional performance behaviors.
• The developed method effectively promotes the advancement of interventional skill assessment.
• This technology offers a pathway to more objective and reliable surgical training evaluations.