Phase Coherent Averaging in Magnetic Resonance Spectroscopy Using Interleaved Navigator Scans: Compensation of Motion Artifacts and Magnetic Field Instabilities Thorsten Thiel, 1 * Michael Czisch, 2 Gregor K. Elbel, 2 and Juergen Hennig 1 The quality of spectra in 1 H magnetic resonance spectroscopy (MRS) is strongly affected by temporal signal instabilities during the acquisition. One reason for these instabilities are hardware imperfections, e.g., drifts of the main magnetic field in super- conducting magnets. This is of special concern in high-field systems where the specification of the field stability is close to the spectral linewidth. A second major potential source of arti- facts, particularly in clinical MRS, is patient motion. Using stan- dard acquisition schemes of phase-cycled averaging of the individual acquisitions, long-term effects (field drifts) as well as changes on a shorter time scale (motion) can severely reduce spectral quality. The new technique for volume-selective MRS presented here is based on the additional interleaved acquisition of a navigator signal during the recovery time of the metabolite acquisition. It corrects for temporal signal instabilities by means of a deconvolution of the metabolite and the navigator signal. This leads to phase-corrected individual metabolite scans and upon summation to a phase-coherent averaging scheme. The inter- leaved navigator acquisition does not require any user interaction or supervision, while sequence efficiency is maintained. Magn Reson Med 47:1077–1082, 2002. © 2002 Wiley-Liss, Inc. Key words: 1H MRS; phase coherent averaging; motion correc- tion; navigator acquisition Over the last years the spectral resolution in in vivo proton ( 1 H) NMR spectra has been improved significantly by the combination of high-field magnets and efficient computer- assisted shimming procedures (1). Singlet linewidths as low as 0.025 ppm were obtained in the rat brain at 9.4T. In combination with the use of short echo times, e.g., TE  10 ms, this leads to a substantial improvement in sensitiv- ity for the determination of previously unresolved signals, especially signal contributions from coupled spin systems (2). The increase of T 1 with higher main magnetic field strength B 0 and the use of smaller voxel sizes, necessary to maintain B 0 homogeneity in the presence of stronger sus- ceptibility effects, leads to longer repetition times and a higher number of averages. In high-field animal systems the volumes of interest are intrinsically smaller (about 4–64 l for mice), necessitating even more averages. Con- sequently, typical acquisition times for spectroscopic data- sets at high fields (4T) are often between 10 –30 min. Due to these rather long acquisition times, magnetic resonance spectra are often impaired by artifacts arising from tempo- ral variations of phase, frequency, and amplitude of the metabolite signals during the experiment. Frequency shifts cause a broadening of the spectral linewidth upon sum- mation, whereas phase artifacts may partly lead to destruc- tive interference of the individual averages and insuffi- cient artifact subtraction when phase cycling schemes are applied. In particular, at high fields these artifacts often arise from drifts of the main magnetic field. Technical specifications of the B 0 field stability for systems at 4T and above are between 0.05 ppm/hr and 0.1 ppm/hr and there- fore well within the limits of the linewidth that can be achieved in single-voxel magnetic resonance spectroscopy (MRS) using automated high-order shimming routines. This problem becomes even more pronounced in the case of prolonged in vivo MRS acquisitions, as used in dynamic pharmaceutical long-term studies or for applications em- ploying spectral editing techniques. Motion is a second major source of artifacts in magnetic resonance spectroscopy (3). The use of strong spoiling gradients, which are necessary to suppress unwanted co- herences from outside the voxel, amplifies this property. Uniform movements along one spatial direction lead to phase changes in individual acquisitions without affecting the signal amplitude. Rotations and more complex forms of macroscopic motion due to swallowing and breathing introduce both signal attenuation and random phase shifts of the FID. The latter produces destructive interference and, therefore, leads to additional signal losses through averaging (4). Gradient motion compensation uses a multiplicity of gradient pulses of suitable durations such that the net effect of the various time derivatives of position can be made equal to zero simultaneously. This leads, however, to prolongation of the echo time and is thus incompatible with the use of short echo times. Only coherent motion can be corrected but neither incoherent movements nor hard- ware instabilities are accounted for. The techniques proposed for the correction of motion artifacts in volume-selective MRS consist of retrospective rejections of single FIDs based on the measured displace- 1 Section of Medical Physics, Department of Diagnostic Radiology, University Hospital, Freiburg, Germany. 2 Max Planck Institute of Psychiatry, Mu ¨ nchen, Germany. Grant sponsor: Bruker Medical GmbH. *Correspondence to: Thorsten Thiel, Ph.D., Bruker BioSpin MRI, Rudolf- Plank-Str. 23, 76275 Ettlingen, Germany. E-mail: thorsten.thiel@medical.bruker.de Received 20 July 2001; revised 8 January 2002; accepted 4 February 2002. DOI 10.1002/mrm.10174 Published online in Wiley InterScience (www.interscience.wiley.com). 1077 © 2002 Wiley-Liss, Inc. Magnetic Resonance in Medicine 47:1077–1082 (2002) FULL PAPERS