Traceable Particle Swarm Optimization for Electromagnetically Navigated Bronchoscopy

Abstract

This paper proposes a modified evolutionary algorithm called traceable particle swarm optimization (PSO) that boosts bronchoscope motion tracking during electromagnetically navigated bronchoscopy. Since electromagnetic (EM) tracking is usually deteriorated by uncertainties (e.g., patient respiratory motion or magnetic field distortion) that occur in interventions, we develop a traceable PSO framework by integrating EM sensor measurements and image intensity information into the standard PSO method. In particular, all evolutionary parameters in our PSO framework can be updated traceably or adaptively in accordance with spatial distance constraints and image similarity information, resulting in an advantageous performance in dynamic bronchoscope motion estimation. Experimental results based on dynamic phantom validation demonstrate that our proposed tracking scheme provides a more robust, accurate, and efficient approach for endoscope motion tracking than several current available methods. The average tracking accuracy of position and orientation was improved from (4.3 mm, 7.8°) to (3.3 mm, 6.5°) and the computational time was reduced from 1.0 to 0.8 seconds per frame without any acceleration devices or code optimization strategy.