Motion Planning under Uncertainty for Medical Needle Steering Using Optimization in Belief Space

Area of Science:

• Robotics
• Medical Devices
• Computational Geometry

Background:

• Needle steering is complex due to nonholonomic systems and uncertain sensing.
• Ultrasound and X-ray imaging provide noisy, partial state information.
• Accurate needle placement is critical for effective medical interventions.

Purpose of the Study:

• To develop an optimization-based motion planner for steerable medical needles.
• To explicitly address motion and sensing uncertainties in 3D anatomy.
• To improve the accuracy and safety of image-guided needle interventions.

Main Methods:

• Formulated needle steering as a partially observable Markov decision process (POMDP).
• Approximated belief states as Gaussian distributions for planning in belief space.
• Computed locally optimal trajectories and controllers minimizing a cost function in belief space.

Main Results:

• The planner explicitly considered motion and sensing uncertainties.
• Optimization in belief space yielded higher quality plans than state-space planning.
• Simulated scenarios demonstrated the planner's effectiveness in complex anatomy.

Conclusions:

• Belief space optimization enhances motion planning for steerable needles.
• This approach improves trajectory planning under uncertainty for medical applications.
• The planner offers a more robust solution for image-guided robotic procedures.