Automated whole-brain N-acetylaspartate proton MRS quantification

Area of Science:

• Neuroimaging
• Biophysics
• Medical Physics

Background:

• N-acetylaspartate (NAA) concentration is a key metric for neuronal health and density.
• Current NAA quantification relies on manual peak integration in proton magnetic resonance spectroscopy (MRS).
• Manual integration is susceptible to operator bias and baseline artifacts.

Purpose of the Study:

• To develop and validate a full spectral modeling approach for automatic whole-brain NAA (WBNAA) concentration estimation.
• To compare the reliability of the new spectral modeling method against manual integration.

Main Methods:

• Utilized whole-head proton (1H)-MRS data from 18 healthy adults at 3 Tesla.
• Employed a full spectral simulation for automatic NAA peak area estimation.
• Normalized NAA peak area by segmented brain volume from T1-weighted MRI to determine WBNAA.
• Compared WBNAA quantification using paired t-tests and likelihood ratio tests.

Main Results:

• Spectral modeling yielded WBNAA means comparable to manual integration but with significantly reduced between-subject variability (~50% reduction).
• Within-subject variability improved by 40% with spectral modeling (7.0% vs. 11.7%).
• High-quality modeling demonstrated by low Cramer-Rao lower bounds (<0.1%) and minimal residuals.

Conclusions:

• Full spectral modeling enhances WBNAA quantification reliability by minimizing variability and operator dependence.
• The improved technique offers better quality control and is more suitable for tracking neurological disorder progression and treatment response.
• This automated approach increases the utility of (1H)-MRS for clinical neuroscience research.