Validation of MR thermometry: method for temperature probe sensor registration accuracy in head and neck phantoms.

PURPOSE: Magnetic resonance thermometry (MRT) is an attractive means to non-invasively monitor in vivo temperature during head and neck hyperthermia treatments because it can provide multi-dimensional temperature information with high spatial resolution over large regions of interest. However, validation of MRT measurements in a head and neck clinical set-up is crucial to ensure the temperature maps are accurate. Here we demonstrate a unique approach for temperature probe sensor localisation in head and neck hyperthermia test phantoms. METHODS: We characterise the proton resonance frequency shift temperature coefficient and validate MRT measurements in an oil-gel phantom by applying a combination of MR imaging and 3D spline fitting for accurate probe localisation. We also investigate how uncertainties in both the probe localisation and the proton resonance frequency shift (PRFS) thermal coefficient affect the registration of fibre-optic reference temperature probe and MRT readings. RESULTS: The method provides a two-fold advantage of sensor localisation and PRFS thermal coefficient calibration. We provide experimental data for two distinct head and neck phantoms showing the significance of this method as it mitigates temperature probe localisation errors and thereby increases accuracy of MRT validation results. CONCLUSIONS: The techniques presented here may be used to simplify calibration experiments that use an interstitial heating device, or any heating method that provides rapid and spatially localised heat distributions. Overall, the experimental verification of the data registration and PRFS thermal coefficient calibration technique provides a useful benchmarking method to maximise MRT accuracy in any similar context.

Motion robust magnetic susceptibility and field inhomogeneity estimation using regularized image restoration techniques for fMRI.

In functional MRI, head motion may cause dynamic nonlinear field-inhomogeneity changes, especially with large out-of-plane rotations. This may lead to dynamic geometric distortion or blurring in the time series, which may reduce activation detection accuracy. The use of image registration to estimate dynamic field inhomogeneity maps from a static field map is not sufficient in the presence of such rotations. This paper introduces a retrospective approach to estimate magnetic susceptibility induced field maps of an object in motion, given a static susceptibility induced field map and the associated object motion parameters. It estimates a susceptibility map from a static field map using regularized image restoration techniques, and applies rigid body motion to the former. The dynamic field map is then computed using susceptibility voxel convolution. The method addresses field map changes due to out-of-plane rotations during time series acquisition and does not involve real time field map acquisitions.