Page 44 - GTM-2-3

Global Translational Medicine Deep learning by NMR-biochemical

establishes a prognosis, correlates with fatigue, and monitors The implementation of MRS in clinical use has been
the effectiveness of drug therapies . At our laboratory, a slow and remains unclear due to the high technical
[66]
training model in a semi‑supervised KNFST algorithm demands of spectroscopic methods and difficult
showed an optimized projection matrix, confidence band interpretations in the brain. However, MRI with MRS using
values, and class-wise MS lesion training data projections 1 H and P spectroscopic imaging provides information
31
into the null space. Semi‑supervised KNFST algorithm on demyelination, neuron loss, glial tissue formation,
iteratively selected samples in the learning phase from changes in glycogenolysis and lactate accumulation,
data. The classifier predicted new labels to accepted lesion energy metabolism and pH imbalance in brain diseases,
data samples, as shown in Figure 12 [64,67] . and inborn errors of metabolism. In the skeletal muscles,
dynamic P MRS is used to detect metabolic disorders,
31
4. New developments in MRSI multimodal approach while H MRI/MRS evaluates muscle degeneration with fat
1
and limitations accumulation. P MRS is inconclusive for liver metabolic
31
The potentials of 4D localized in vivo NMR image function, but H-MRS/MRI indicates metabolites, mobile
1
spectroscopy, slice selective N MRS, biopsy HRMAS lipids, and bone marrow characteristics. In the heart, MRSI
spectroscopy, and multimodal, multinuclear, and water- provides in vivo metabolites with global cardiomyopathy.
fat suppression NMR techniques have emerged for better The role of MRI/MRS in cancer differential diagnosis
tissue contrast with metabolomics on high magnetic fields and treatment is still unclear due to the lack of preclinical
of up to 11.4 tesla in whole human body imagers, as well as investigations on physiological and metabolic events on
an ultrahigh resolution at 21- and 36-tesla microimagers. NMR spectra and minimal clinical trials.
The localized tissue metabolism in selected diseased body From the technical point of view, coil sensitivity with
parts or biopsies was studied using modified surface a higher signal-to-noise ratio is crucial in enhanced
coils, multiple contrast media, modified radiofrequency metabolite quantification.
pulses with high resolutions, and dynamic gradient fields
for localized 2D spectral characterization or metabolic 5. MRSI diagnostic use in clinical trials and
screening for spectrally resolved metabolomics imaging healthcare: A physician’s guide
applications . Due to being non-invasive and non-
[20]
radioactive, NMR techniques have been proven safe, with The author considers MRS an evidence and fact-finding
no biological hazards associated with the current level method for following diseases with metabolite indications
of SAR energy exposures. In the next few years, NMR (shown in brackets) collected over three decades: i. focal
methods will emerge as a safer, cheaper, and cost-effective brain tissues for maturating brains in gestation (elevated
alternative modality compared to other diagnostic or NAA, glutamine + glutamate [Glx] peaks, low Cho,
therapeutic techniques using ionized radiation. taurine [Tau] peaks), recurrent brain tumor distinction
from radiation-induced necrosis (low NAA, Cr with high
MRI-digital histochemical correlation is now Cho peaks), assessment of prognosis in hypoxic-ischemic
established to exhibit malignancy-associated changes due encephalopathy (low NAA, low NAA/Cho), grading the
to NMR T1 variations where T1 relaxivities depend on low-grade and high-grade glioma (change in MI, gamma-
visible water contents, proton densities, environments in aminobutyric acid [GABA] peaks), and evaluation of
tissue, and pathology . indeterminate brain lesion to postpone resection/biopsy
[1]
MRI and biochemical correlation initially proposed (low NAA, low Cr, high Cho, lactate [Lac] peaks); ii.
in the nineties has been established in routine clinical diagnosis and monitoring of metabolic diseases, such as
[1]
assessment trials for tissue content chemical analysis, cerebral ischemia (high Lac peak), creatine deficiency
in vivo MRS, ex vivo MRS, and relaxivities using artificial (Cr peak missing), Canavan disease (increased NAA/tCr
metabolite solutions and MR database. MRI and ratio), non-ketotic hyperglycinemia (DWI, DTI [diffusion
2D-spectral characterization of small-size metabolites in tensor imaging]) and tractography, fractional anisotropy
diseased tissues is reported with correlated spectroscopy [FA], and diffusivity), maple syrup urine disease (edema,
COSY-MRI, nuclear Overhauser effect spectroscopy high Glx, tau, and alanine [Ala] peaks), MS periventricular
(NOESY)-MRI, and high-pressure liquid chromatography and parahippocampus hypointense lesions (low NAA/
(HPLC)-MRI with little success but greater potentials in total creatine [tCr] and Cho/tCr ratios), metachromatic
non-invasive molecular details in undefined molecular leukodystrophy (low NAA peak and high NAA in urine),
etiology of poorly understood diseases. The final outcome Parkinson’s disease substantia nigra (low NAA and
depends on significant information and cost-effective GABA peaks, high Lac/tCr ratio), Pelizaeus‑Merzbacher
constraints [20-61] . disease (high NAA+NAAG peaks and low Cho peaks),

Volume 2 Issue 3 (2023) 12 https://doi.org/10.36922/gtm.337