Recapitulation of Pad´e-optimized processing of biomedical time signals - Signal Processing in MRS with Biomedical Applications - page 382

12.3 Relevance of Pade-optimized MRS for diagnostics
in clinical oncology
Both MRI and in vivo MRS are becoming the noninvasive diagnostic methods
of choice for a wide range of medical applications, particularly in oncology.
Technological advances, most notably the use of very fast gradient variations
(echoplanar scanning) [1, 4] have allowed high quality imaging to be per
formed rapidly. Here, MRS yields biochemical information which often im
proves the specificity of MRI. Further, MRSI combines the chemical specificity
of MRS with the spatial localization techniques that have been developed for
MRI to obtain multiple MRS signals over a volume of tissue.
There are many advantages of MRbased modalities. These include multi
planar capabilities, excellent contrast among tissues, as well as the potentially
rich array of spectral information that can help distinguish malignant from
benign lesions, as well as from healthy tissue. The lack of exposure to ionizing
radiation renders MRbased modalities very attractive for early detection,
especially for screening surveillance of persons at high risk for specific cancers.
The combination of anatomic localization and insight into metabolic char
acteristics from spectral information can be decisive for accurate and timely
identification of cancers. This is especially true in the exceedingly di cult
and critically important differential diagnostic dilemmas such as distinguish
ing recurrent tumor from radiation necrosis or from postoperative changes.
These advantages have become particularly clear in neurooncology, where
MRS and MRSI are becoming a key modality for nearly all aspects of brain
tumor diagnostics, as reviewed in chapter 8. Another area in which MRI and
MRS have made an important impact is prostate cancer, as reviewed in chap
ter 11, where these methods in combination have provided diagnostic clarity
unmatched by literally any other noninvasive approach, and are critical in
improving the yield of biopsies. Further, MRS and MRSI have also helped
improve the diagnostic accuracy of a number of other malignancies, most
notably head and neck cancers, nonHodgkin's lymphoma as well as breast
cancer [7].
Notwithstanding these achievements, the full potential of MRI and MRS for
cancer diagnostics has not been realized. As we have intimated, in actual clin
ical practice, all too often the interpretation of automatically generated MR
spectra is shrouded by confusion and ambiguity. In order for MRS and MRSI
to become a standard diagnostic tool in the area where it is needed the most,
for clinical oncology, including especially cancer screening and surveillance, it
is necessary to go beyond technical (hardware) improvements, as important
as these are.
We will now briefly recapitulate why many of the major limitations of cur
rent applications of MRS and MRSI are due to the almost exclusive reliance
upon the conventional Fourierbased mode of data analysis, i.e., the FFT,