Multiplet resonances in MRS data from normal and cancerous prostate - Signal Processing in MRS with Biomedical Applications - page 358

For the components of the myoinositol triplet (k = 23−25) the recon
structed chemical shifts, Im(ν − k ) and|d − k |were either exact or were correct

to 35 of the six decimal places (accounting for roundoff). For alanine, the
doublets of citrate doublets and creatine the chemical shifts and Im(ν − k ) were
exactly reconstructed. Here, the|d − k

|for these resonances were correct to 4
to 5 of the six decimal places. For polyamine (k = 9) and choline (k = 10)
the concentrations were correct to one decimal place.
The integer was correct for the computed concentration of phosphocholine
(k = 11) at N P = 500 and the spectral parameters were correct to 1−3 decimal
places. The concentration of glycerophosphocholine (k = 12) was twice higher
than it should be. The computed concentration of peak k = 14 (taurine
component) was underestimated by about 20% lower than the correct value
at N P = 500. The concentration of the scylloinositol (k = 16) was correct.
Peaks k = 15 and k = 17−19, i.e., the components of the two taurine triplets,
had the correct integer values for the concentrations at N P = 500. Taking into
account roundoff, the concentrations of the components of the myoinositol
triplet (k = 20−22) were correct to at least one of two decimal places.
Full convergence was attained at N P = 600 for all the reconstructed param
eters for all twentyseven resonances (lower panel (ii) of Table 11.6 ) . As was
true for the data from normal prostate, stability of convergence was main
tained at higher partial signal length and at the full signal length N for the
data corresponding to prostate cancer.
Figure 11.7 displays the Padereconstructed absorption component shape
spectra at N P = 54 and N P = 800 for the malignant prostate data. At
N P = 54 (upper panel (i)) eleven of the twentyseven peaks were resolved.
Again, it was at the two extremes of the spectrum that all the resonances
were resolved and with nearly correct heights for lactate and alanine at 1.33
ppm and 1.49 ppm, respectively, and for myoinositol and lactate at 4.07 ppm
and 4.12 ppm, respectively. In the denser spectral region from≈2.5 ppm to
3.65 ppm, seven of the 23 peaks were resolved.
Unlike what was seen for the normal prostate spectra at this signal length,
there were no dispersive modes. Singlets near 2.5 ppm and 2.75 ppm were
seen at N P = 54, although doublets of citrate should appear at each of those
regions.
The peaks near 3.55 ppm and 3.65 ppm corresponding to myoinositol are
seen as singlets, but should be triplets. There should have been ten peaks
between 3.0 ppm and 3.35 ppm, but only three peaks were seen. At N P = 800
as shown on the bottom panel (ii) of Fig. 11.7, all twentyseven resonances
were resolved with the correct heights, widths, positions as well as the phases
(zero).
Figure 11.8 shows the Padereconstructed total absorption shape spectra at
the two partial signal lengths N P = 54 (top panel (i)) and N P = 800 (bottom
panel (ii)) for the prostate cancer data. As was the case for the component
absorption spectrum for malignant prostate data, lactate and alanine at 1.33
ppm and 1.49 ppm and myoinositol and lactate at 4.07 ppm and 4.12 ppm