Automated Intraprocedural Detection and Registration of Interventional Devices via Characteristic Fiducial Markers

Motivation: Image-guided interventions leverage MR-imaging to orient surgical tools relative to internal anatomy. Automated fiducial detection improves workflow efficiency and accuracy relative to manual selection. Goal(s): To develop and evaluate an autonomous approach for detecting and registering device fiducials during intraprocedural MRI. Approach: Twelve high-resolution volumetric scans of a trajectory guide with a characteristic fiducial design were acquired. Image processing methods determined the fiducial's location and orientation. Results: Across two cadaver and ten phantom scans, automated methods isolated the fiducial 10/12 times, and coregistered it to a template with a Dice coefficient of 0.82 in T2 acquisitions. Impact: Applied methods were able to leverage geometric information from characteristic fiducial markers to fully autonomously identify and orient a stereotactic trajectory guide, streamlining an often manual or multimodal step in many workflows for MRI-guided interventions.