[An algorithm for passive marker localization in interventional MRI]

The localization of passive marker systems in interventional MRI is necessary to monitor the position and orientation of medical instruments that do not emit an MR signal. In this work an algorithm is presented that automatically detects a given marker system in an MR image with a precision better than one pixel. Therefore, a combination of a phase-only cross correlation algorithm with a subsequent center-of-mass analysis is utilized. The algorithm was evaluated in simulations and phantom experiments with respect to precision, noise sensitivity as well as the influence of unwanted signal amplitudes. Above a signal-to-noise ratio (SNR) of 4.5 a localization precision significantly better than the pixel dimension could be achieved. For SNR values of 6 and more the influence of unwanted signals on the localization could not be detected in the simulations. In phantom experiments the predicted precision of the marker localization could be realized which results for typical measurement parameters in a maximal deviation of the needle tip in an MR-guided needle injection of 0.6 mm.