Methods and potentials of magnetic resonance imaging for monitoring radiofrequency hyperthermia in a hybrid system.

INTRODUCTION: Non-invasive thermometry (NIT) is a valuable and probably indispensable tool for further development of radiofrequency (RF) hyperthermia. A hybridization of an MRI scanner with a hyperthermia system is necessary for a real-time NIT. The selection of the best thermographic method is difficult, because many parameters and attributes have to be considered.
METHODS: In the hybrid system (Siemens Symphony/BSD-2000-3D) the standard methods for NIT were tested such as T1, diffusion (ADC: apparent diffusion coefficient) and proton-resonance-frequency shift (PFS) method. A series of three-dimensional datasets was acquired with different gradient-echo sequences, diffusion-weighted EPI spin-echo sequences and calculated MR-temperatures in the software platform AMIRA-HyperPlan. In particular for the PFS-method, corrective methods were developed and tested with respect to drift and other disturbances. Experiments were performed in phantoms and the results compared with direct temperature measurements. Then the procedures were transferred to clinical applications in patients with larger tumours of the lower extremity or the pelvis.
RESULTS: Heating experiments and MR-thermography in a homogeneous cylindrical phantom give an excellent survey over the potentials of the methods. Under clinical conditions all these methods have difficulties due to motion, physiological changes, inhomogeneous composition and susceptibility variations in human tissues. The PFS-method is most stable in patients yielding reasonable MR temperature distributions and time curves for pelvic and lower extremity tumours over realistic treatment times of 60-90 min. Pooled data exist for rectal tumour recurrencies and soft tissue sarcomas. The fat tissue can be used for drift correction in these patients. T1 and diffusion-dependent methods appear less suitable for these patients. The standard methods have different sensitivities with respect to the various error sources. The advantages and pitfalls of every method are discussed with respect to the literature and illustrated by the phantom and patient measurements.
CONCLUSIONS: MR-controlled RF hyperthermia in a hybrid system is well established in phantoms and already feasible for patients in the pelvic and lower extremity region. Under optimal conditions the temperature accuracy might be in the range of 0.5 degrees C. However a variety of developments, especially sequences and post-processing modules, are still required for the clinical routine.