Optimization of Single Voxel MR Spectroscopy Sequence Parameters and Data Analysis Methods for Thermometry in Deep Hyperthermia Treatments

Hartmann, J. MSc
Gellermann, J. MD
Brandt, T. MSc
Schmidt, M. PhD
Pyatykh, S. PhD
Hesser, J. PhD
Ott, O. MD
Fietkau, R. MD
Bert, C. PhD

¹Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
²Department of Radiation Oncology, University Hospital Tübingen, Tübingen, Germany
³Praxis/Zentrum für Strahlentherapie und Radioonkologie, Berlin, Germany
⁴Medical Faculty Mannheim, Experimental Radiation Oncology, Heidelberg University, Mannheim, Germany

C. Bert, Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Universitätsstraße 27, Erlangen 91054, Germany. Email: [email protected]

Received December 21, 2015

Received in revised form April 30, 2016

Accepted May 23, 2016

Technology in Cancer Research and Treatment 16(4):p 470-481, August 2017. | DOI: 10.1177/1533034616656310

Objective:

The difference in the resonance frequency of water and methylene moieties of lipids quantifies in magnetic resonance spectroscopy the absolute temperature using a predefined calibration curve. The purpose of this study was the investigation of peak evaluation methods and the magnetic resonance spectroscopy sequence (point-resolved spectroscopy) parameter optimization that enables thermometry during deep hyperthermia treatments.

Materials and Methods:

Different Lorentz peak-fitting methods and a peak finding method using singular value decomposition of a Hankel matrix were compared. Phantom measurements on organic substances (mayonnaise and pork) were performed inside the hyperthermia 1.5-T magnetic resonance imaging system for the parameter optimization study. Parameter settings such as voxel size, echo time, and flip angle were varied and investigated.

Results:

Usually all peak analyzing methods were applicable. Lorentz peak-fitting method in MATLAB proved to be the most stable regardless of the number of fitted peaks, yet the slowest method. The examinations yielded an optimal parameter combination of 8 cm³ voxel volume, 55 millisecond echo time, and a 90° excitation pulse flip angle.

Conclusion:

The Lorentz peak-fitting method in MATLAB was the most reliable peak analyzing method. Measurements in homogeneous and heterogeneous phantoms resulted in optimized parameters for the magnetic resonance spectroscopy sequence for thermometry.