Related Experiment Video
Updated: May 8, 2026
10:46
Non-fluoroscopic Catheter Tracking for Fluoroscopy Reduction in Interventional Electrophysiology
Published on: May 26, 2015
Catheter tip tracking for MR-guided interventions using discrete Kalman filter and mean shift localization.
Abubakr Eldirdiri1, Frédéric Courivaud, Rafael Palomar
1The Intervention Centre, Oslo University Hospital, Oslo, Norway, abubakr.eldiriri@rr-research.no.
International Journal of Computer Assisted Radiology and Surgery
|August 27, 2013
Summary
Stochastic algorithms accurately track MR-guided catheter tips using regional features. Mean shift and Kalman filters show robust performance even at lower resolutions.
Area of Science:
- Medical imaging and interventional procedures
- Computer vision and algorithm development
- Biomedical engineering
Background:
- MR-guided catheterization requires precise catheter tip localization.
- Manual tracking can be challenging and prone to error.
- Automated methods are needed for improved accuracy and efficiency.
Purpose of the Study:
- To develop and evaluate stochastic computer algorithms for automatic detection and tracking of marked catheter tips during MR-guided catheterization.
- To assess the performance of these algorithms using regional feature extraction and matching.
Main Methods:
- Developed algorithms utilize speeded-up robust features (SURF) from gradient images to create probability maps of catheter tip location.
- Stochastic localization frameworks, including mean shift (MS) and Kalman filter (KF), were employed to track the catheter tip using 2D MR image projections.
- The system was tested with an LC resonant circuit marker at various image resolutions (1-7 mm² pixel).
Main Results:
- Both MS and KF algorithms demonstrated robust tracking performance down to 3 mm image resolution, with localization errors around 1 mm.
- At 5 mm resolution, localization errors were approximately 2.2 mm for both algorithms.
- Tracking accuracy decreased at 7 mm resolution, with errors around 3.6-3.7 mm.
Conclusions:
- KF and MS algorithms provide comparable and accurate results across different MR image resolutions.
- The developed algorithms offer high robustness for catheter tip tracking at 3 mm resolution and acceptable reliability at 5 mm resolution with the tested resonant marker.
