Enhanced Real-Time Pose Estimation for Closed-Loop Robotic Manipulation of Magnetically Actuated Capsule Endoscopes

Area of Science:

• Robotics
• Medical Imaging
• Biomedical Engineering

Background:

• Magnetic actuation of capsule endoscopes enables automated endoscopic maneuvers.
• Current pose estimation methods struggle with magnetic field singularities and initialization accuracy.
• Singular regions degrade system accuracy, particularly with single magnetic field sources.

Purpose of the Study:

• To analyze the workspace limitations of existing magnetic pose estimation techniques.
• To develop a novel hybrid magnetic actuation system free from singularity regions.
• To validate the performance and robustness of the proposed system for capsule endoscopy.

Main Methods:

• Analysis of pose estimation workspace using a point-dipole magnetic field model.
• Development of a hybrid system utilizing static and time-varying magnetic field sources.
• Experimental validation of accuracy, workspace, update rate, and singularity robustness.

Main Results:

• Identified singular regions in existing pose estimation methods within typical capsule actuation areas.
• The proposed hybrid system demonstrated no regions of singularity.
• The system achieved comparable or superior performance to prior methods, without requiring precise initialization and showing robustness to singularities.

Conclusions:

• The novel hybrid magnetic actuation system overcomes limitations of existing pose estimation techniques.
• This system offers improved accuracy, robustness, and eliminates the need for exact initialization.
• The developed pose estimation system is suitable for closed-loop control and advances intelligent automation in magnetically actuated capsule endoscopy for clinical realization.