Magnetic resonance in neuro-oncology: Achievements and challenges - Signal Processing in MRS with Biomedical Applications - page 263

8.2.1 Poor resolution and SNR

Poor resolution and low SNR represent a critical limitation which has ham
pered wider use of MRS for clinical oncology. As stated, the FFT is a low
resolution estimator, lacking both extrapolation and interpolation capacities.
Moreover, with attempts to enhance resolution by increasing acquisition time,
SNR deteriorates. This is due to the fact that mainly noise is recorded later
in the time signal, particularly with 1.5T clinical scanners. Moreover, the
FFT is a linear transform and therefore imports noise from the time to the
frequency domain, further contributing to poor SNR.

8.2.1.1 Limited capability to detect small brain tumors

A number of the problems with current applications of MRS and MRSI in
brain tumor diagnostics are related to resolution and SNR issues. Particularly
troublesome in this regard is the limited possibility of MRS and MRSI to
detect very small brain tumors [211], at the very time at which therapeutic
interventions would have the best chance for success. As noted by Huang
et al. [289] in vivo MRS has low SNR, which severely limits capabilities to
determine critical characteristics such as the grade of brain tumors.
Attempts to improve the SNR have most often entailed either increasing
the acquisition time, or increasing the volume of brain tissue from which
data is acquired. The latter approach frequently results in a heterogeneous
voxel with a mixture of tissue types. There are special problems in achieving
adequate SNR using MRSI within scan times of reasonable length [4]. Because
of the vital importance of achieving volumetric coverage of brain tumors,
which are often heterogeneous, the SNR issues specifically related to MRSI
are of particular concern for neurooncology, as well as for cancer diagnostics
in general [290].

8.2.1.2 Attempts to improve resolution and SNR via stronger fields

and short echo times

Resolution and SNR for brain tumor diagnostics have been improved by using
MR scanners with higher magnetic field strength [291]. However, the detection
of small residual foci of brain tumor has been reported to be problematic even
when a 3T scanner was used [291]. Some recent data indicate that a multi
channeled phasedarray head coil may improve the possibility to predict the
spatial extent of brain tumors [113, 292]. Another strategy has been to use
short TE in an attempt to capture the larger signal intensities observed early
in the recording, and thereby also capture clinically important metabolites for
brain tumor diagnostics such as myoinositol, lipids and glutamine - glutamate
that decay rapidly and can therefore only be seen at short TE [293, 294].
However, at short TE the problems of relying upon fitting can be exacerbated
for certain metabolites. In addition, if problems with the baseline occur,
longer echo times are recommended [293]. Moreover, since T 2 relaxation times