Recapitulation of Pad´e-optimized processing of biomedical time signals - Signal Processing in MRS with Biomedical Applications - page 386

physical resonances as encountered in normal breast versus fibroadenoma ver
sus malignant breast was thereby demonstrated. In particular, the FPT un
ambiguously delineated diagnostically important metabolites such as lactate,
as well as choline, phosphocholine and glycerophosphocholine that are very
closelylying and may represent MRretrievable molecular markers of breast
cancer. Within a very narrow spectral band, there were seven resonances,
including phosphocholine which was completely buried underneath a much
larger phosphoethanolamine peak.
Because of the socalled “glycerophosphocholine to phosphocholine switch”
associated with malignant transformation of the breast [307]-[309], it is vi
tal to identify and precisely quantify phosphocholine, as well as the other
resonances lying within this narrow frequency band. This di cult task was
achieved by the FPT. Moreover, due to this region of high spectral density,
there were approximately 80 times more spurious resonances than those which
were genuine. The FPT unequivocally separated out all of these spurious
peaks, so that the true spectral information that distinguished malignant
breast tissue from fibroadenoma and normal breast tissue could be evaluated
with full confidence.
Particular note should be made that for all three cases of breast tissue,
the total shape spectrum converged at a shorter signal length than did the
component shape spectra. The peak at 3.22 ppm on the total shape spectrum
appeared completely symmetrical without any hint that there were, in fact,
two components: a larger phosphoethanolamine and a smaller, underlying
phosphocholine peak at that position along the chemical shift axis. Because
the FPT exactly determines the number of resonances and the spectral pa
rameters of each of these, there is no speculation whatsoever.
Spectral overlap with multiplet resonances is recognized to be exceedingly
troublesome for MRS of the prostate [443]. Exact determination of the number
of genuine resonances becomes particularly vital in this setting. Our results
as presented in chapter 11 illustrate that the FPT can unequivocally resolve
and quantify a large number of overlapping resonances, including multiplets
of metabolites that distinguish normal glandular prostate, normal stromal
prostate and prostate cancer.
For all the presented applications of Padeoptimized MRS to problem areas
within oncology, the metabolite concentrations were exactly and unequivo
cally reconstructed. With the standard Fourier approach, metabolite concen
trations are estimated from the shape spectra by integrating the areas under
the peaks or fitting the peaks to a subjectively chosen number of Lorentzians
and/or Gaussians. Even for clearly delineated peaks, as noted, this procedure
of numerical quadrature is subjective due to the uncertainty about the lower
and upper integration limits.
With respect to breast cancer, notwithstanding the need to expand the
number of metabolites upon which the diagnosis is made, accurate quantifi
cation of total choline via the FPT would represent a major breakthrough
for early detection of this malignancy. This could also be important for ther