Improved electromagnetic tracking for catheter path reconstruction with application in high-dose-rate brachytherapy
Elodie Lugez,Hossein Sadjadi,Chandra P Joshi,Selim G Akl,Gabor Fichtinger
DOI: https://doi.org/10.1007/s11548-017-1534-4
Abstract:Purpose: Electromagnetic (EM) catheter tracking has recently been introduced in order to enable prompt and uncomplicated reconstruction of catheter paths in various clinical interventions. However, EM tracking is prone to measurement errors which can compromise the outcome of the procedure. Minimizing catheter tracking errors is therefore paramount to improve the path reconstruction accuracy. Methods: An extended Kalman filter (EKF) was employed to combine the nonlinear kinematic model of an EM sensor inside the catheter, with both its position and orientation measurements. The formulation of the kinematic model was based on the nonholonomic motion constraints of the EM sensor inside the catheter. Experimental verification was carried out in a clinical HDR suite. Ten catheters were inserted with mean curvatures varying from 0 to [Formula: see text] in a phantom. A miniaturized Ascension (Burlington, Vermont, USA) trakSTAR EM sensor (model 55) was threaded within each catheter at various speeds ranging from 7.4 to [Formula: see text]. The nonholonomic EKF was applied on the tracking data in order to statistically improve the EM tracking accuracy. A sample reconstruction error was defined at each point as the Euclidean distance between the estimated EM measurement and its corresponding ground truth. A path reconstruction accuracy was defined as the root mean square of the sample reconstruction errors, while the path reconstruction precision was defined as the standard deviation of these sample reconstruction errors. The impacts of sensor velocity and path curvature on the nonholonomic EKF method were determined. Finally, the nonholonomic EKF catheter path reconstructions were compared with the reconstructions provided by the manufacturer's filters under default settings, namely the AC wide notch and the DC adaptive filter. Results: With a path reconstruction accuracy of 1.9 mm, the nonholonomic EKF surpassed the performance of the manufacturer's filters (2.4 mm) by 21% and the raw EM measurements (3.5 mm) by 46%. Similarly, with a path reconstruction precision of 0.8 mm, the nonholonomic EKF surpassed the performance of the manufacturer's filters (1.0 mm) by 20% and the raw EM measurements (1.7 mm) by 53%. Path reconstruction accuracies did not follow an apparent trend when varying the path curvature and sensor velocity; instead, reconstruction accuracies were predominantly impacted by the position of the EM field transmitter ([Formula: see text]). Conclusion: The advanced nonholonomic EKF is effective in reducing EM measurement errors when reconstructing catheter paths, is robust to path curvature and sensor speed, and runs in real time. Our approach is promising for a plurality of clinical procedures requiring catheter reconstructions, such as cardiovascular interventions, pulmonary applications (Bender et al. in medical image computing and computer-assisted intervention-MICCAI 99. Springer, Berlin, pp 981-989, 1999), and brachytherapy.