Magnetic resonance in neuro-oncology: Achievements and challenges - Signal Processing in MRS with Biomedical Applications - page 269

8.2.3 Fitting estimates for concentrations of a small number

of metabolites

There is a limited number of metabolites identified with in vivo MRS pro
cessed via FFT. Danielsen and Ross [259] stated that in vivo MRS has intro
duced a “new neurochemistry”, providing the clinician with the possibility
of noninvasively accessing neurophysiologic and neurochemical information.
This, together with the exquisite detail of neuroanatomical imaging provided
by MRI, has been said to offer “an indispensable tool for stateoftheart
neurodiagnosis” (p. 23). However, the current Fourierbased applications of
MRS and MRSI in brain tumor diagnostics yield a relatively small number of
metabolites (low molecular weight, high concentration) that are observable on
clinical scanners. In other words, in fact, we are provided with an exceedingly
limited view of normal and abnormal brain chemistry.
Thus, we have seen for example that definitions of the presence or absence
of brain tumor are very frequently based solely upon the ratios of choline to
NAA and choline to creatine. Besides the abovediscussed problems of varying
cutpoints and numerous extraneous factors that could affect these ratios,
the fundamental issue is that MRS and MRSI have made great strides in
neurooncology by relying upon a mere handful of metabolites. This restricted
metabolite window stems directly from the limitations of conventional data
analysis based upon the FFT and accompanying postprocessing fitting and
other related phenomenological approaches.

8.2.4 Lack of component spectra of clinically important over-

lapping resonances for brain tumor diagnostics

As we have elaborated in detail earlier, the FFT is a nonparametric method
which provides only the shape spectrum, but not the parameters that de
fine the underlying components. Twodimensional correlated and in vitro in
MRS studies reveal that there is a much richer store of potentially informative
metabolites for brain tumor diagnostics than is currently extracted using in
vivo 1D MRS and MRSI. Many of the most diagnostically important compo
nents overlap [295]. For example, in the lipid - lactate region with localized
2D Jresolved MRS, Thomas et al. [89] were able to distinguish the lactate
peak from the overlapping lipid in a patient with glioblastoma. They con
clude: “in vivo Jresolved plots of human brain tumors indicate the exciting
potential of this technique in extracting additional information from the con
ventional MR spectrum” (p. 459). The content of the resonances at 0.9 ppm
and 1.3 ppm is further elucidated by in vitro MRS. The ratios between the
0.93 ppm and 1.3 ppm peaks have been highly variable with in vivo MRS
[298]. These highresolution studies reveal that the CH 3 signal at 0.9 ppm
arises from protein residues at 0.92 ppm and from lipids at 0.88 ppm. This
is due to a nonlipid contribution superimposed on the 0.9 ppm resonance.
The large lipid content of the normal human brain is not visible with in vivo