Multiplet resonances in MRS data from normal and cancerous prostate - Signal Processing in MRS with Biomedical Applications - page 339

ported that spectral overlap was a major problem in onedimensional HRMAS
studies of prostate cancer and other tissues. They noted, for example, that
phosphocholine at 3.23 ppm and glycerophosphocholine at 3.24 ppm could not
be completely resolved from each other; thus the sum of these two metabolites
concentrations was given.
Most recently this group of authors [444] used twodimensional HRMAS
and total correlation spectroscopy (TOCSY). Phosphocholine was reportedly
present in eleven of fifteen (73%) prostate cancer specimens whereas PCho
was only seen in 6 of 32 (19%) of the nonmalignant prostate samples. They
also found that PCho/GPC ratios were significantly higher in the cancerous
samples [444]. Another study performed by means of HRMAS [445] revealed
that lactate and alanine were significantly lower in benign glandular prostate
compared with malignant tissue.
It should be emphasized that a particular challenge within the spectra from
the prostate are the numerous multiplets. Within the regions of greatest spec
tral density, the di culties become even greater. In the study by Swanson
et al. [443] of the polyamine resonances, only the contribution from the pre
dominant spermine was included in fitting, whereas putrescine and spermi
dine could not be resolved. These di culties underscore the need for exact,
unequivocal quantification not only of spectra with completely distinct res
onances, but also for multiplets with overlapping resonances. As elaborated
earlier in this topic, the FPT would appear to be ideally suited to this task.

11.3 Performance of the fast Pade transform for MRS

data from prostate tissue

In the present spectral analysis, three FIDs were synthesized in the form
c n =

K
k=1 d k e inτ ω k from (3.1) with Im(ω k ) > 0, by using a sum of K = 27
damped complex exponentials exp (inτω k ) (1≤k≤27) [33]. The complex
amplitudes d k were timeindependent. As has been the case throughout this
book, ω k and d k are the fundamental angular frequencies and amplitudes
(ω k = 2πν k , where ν k is the linear frequency). We then quantified these time
signals using the FPT (−) , as per Ref. [10]. The bandwidth was 6000 Hz,
where the inverse of this bandwidth is the sampling time τ. The total signal
length was set as N = 1024.
The input data for the absolute values of amplitudes were generated accord
ing to the given mean metabolite concentrations, as well as the description of
multiplets and the total shape spectra for normal glandular prostate, normal
stromal prostate and prostate cancer as per Ref. [443]. The total metabolite
concentrations were split into multiplets to correspond to the spectra of Ref.
[443] and also to fill in what was missing.