Digital Signal Processing Reference
In-Depth Information
1
Basic tasks of signal processing in
spectroscopy
Magnetic Resonance Spectroscopy (MRS) and Magnetic Resonance Spectro
scopic Imaging (MRSI) are among the leading and most rapidly developing
noninvasive diagnostic modalities in medicine [1]-[7]. While MRS focuses
upon a single voxel, the combination of MRS and Magnetic Resonance Imag
ing (MRI) yields MRSI, which provides information from multiple voxels.
Thus, the hybrid nature of MRSI lies in its merging of the spectroscopic and
the morphological information of MRS and MRI, respectively. An auxiliary
terminology for MRSI is occasionally used via Chemical Shift Imaging (CSI).
However, this is not an appropriate alternative nomenclature, since it makes
merely an implicit reference to the most important, spectroscopic aspect of
this diagnostic modality.
Irrespective of whether using one or the other competing acronyms, MRSI
or CSI, the current practice with Magnetic Resonance (MR) in medicine re
duces the rich quantitative interpretation of the spectroscopic information
from MRSI/CSI mainly to the qualitative aspect of imaging. This is done
by overlaying a few metabolite concentration maps in different colors upon
anatomical images of the scanned tissue. Such presentations unnecessarily
simplify MRSI by conveying a misleading message that this spectroscopic
imaging can su
ciently be understood and used merely in terms of a color
coded MRI scan. Even when the conventional attempts are made to show the
spectroscopic information from MRSI, this is done by displaying exclusively
Fourier envelope spectra throughout a scanned tissue. Again this is only a
qualitative description. The colorcoded concentration maps superimposed
on MRI scans stem from fitting Fourier envelopes to determine the peak ar
eas. Since any such fitting is illconceived, via leastsquare adjustments of
free parameters to the hidden local components of the Fourier global shape
lines, it necessarily leads to biased and nonunique estimates of the sought
information. This could hardly constitute a clinically appealing avenue, since
instead of taking advantage of spectroscopic complementarity to imaging, the
currently implemented MRSI with its fitted Fourier envelopes offers to physi
cians new dilemmas, ambiguities and uncertainties. Such severe deficiencies
are inherited and imported directly from MRS. Therefore, if these hampering
drawbacks could be overcome first in MRS, this would dramatically improve
the prospect of MRSI for becoming a reliable and invaluable clinical modality
1
