Learning-based endovascular navigation through the use of non-rigid registration for collaborative robotic...

Area of Science:

• Medical Robotics
• Endovascular Interventions
• Medical Imaging

Background:

• Endovascular interventions face challenges with 2D imaging and complex anatomies.
• Robotic catheter technology offers potential for reduced radiation and enhanced navigation.
• Integrating 3D imaging with robotic platforms can improve precision and safety.

Purpose of the Study:

• To develop a semiautonomous robotic catheter platform that addresses anatomical variability in aortic arches.
• To incorporate patient-specific anatomical information for optimizing catheter trajectories.
• To enhance the safety and precision of endovascular navigation.

Main Methods:

• Statistical modeling of expert catheter motions for trajectory encoding.
• Non-rigid registration to map catheter trajectories across anatomically diverse models.
• Generation of robot trajectories for collaborative cannulation under flow simulations.
• Cross-validation and performance assessment against manual approaches.

Main Results:

• Semiautomatic cannulation achieved success rates of 98.1% (dry) and 94.4% (flow simulations).
• The robotic approach resulted in smoother catheter paths compared to manual methods.
• Mean contact forces were reduced by 33.3%, with a 70.6% decrease in force variability.

Conclusions:

• The developed platform offers a method for catheter task planning adaptable to patient-specific anatomies.
• Insights gained can inform the design of ergonomic catheter navigation robots.
• The study demonstrates the potential of robotic assistance in improving endovascular procedure outcomes.