Can Magnetic Resonance Spectroscopy (MRS) detect Alzheimer's disease early?

NAA (a substance that indicates how healthy brain cells are) decreases in Alzheimer's disease, while myo-inositol (a marker of inflammation and cell loss) increases. A meta-analysis of dozens of studies confirms that these changes are already measurable in mild cognitive impairment (MCI), before the disease has fully developed. The hippocampus shows the strongest abnormalities, and the changes become more severe as the disease progresses.

In the posterior cingulate cortex, a brain region affected early in Alzheimer's disease, the ratio between NAA and myo-inositol is a useful measure. This ratio reliably distinguishes Alzheimer's disease from vascular dementia and from healthy brains. MRS can also pick up changes that are not yet visible on a conventional brain scan. Sensitivity is high, but specificity is moderate: not everyone with an abnormal measurement has Alzheimer's disease.

MRS also has potential for distinguishing two forms of dementia that are clinically very similar to each other. In a small study of 57 MCI patients, the NAA/creatine ratio in the posterior cingulate cortex distinguished Alzheimer's disease from dementia with Lewy bodies, with reasonably good diagnostic accuracy. However, the group was small (only 10 patients with Lewy bodies), so this result requires confirmation in larger studies.

MRS works best as part of a broader diagnostic workup. Combined with cerebrospinal fluid proteins (amyloid and tau), structural MRI and PET scanning, diagnostic certainty increases. Stand-alone use of MRS is insufficiently supported for clinical diagnosis. When a PET scan is added to the mix, the radiation dose also rises, which requires additional consideration in vulnerable patients.

AI models trained on MRS data achieved a classification accuracy of 95% in a small internal study of 111 patients. This sounds impressive, but external validation on independent patient groups is entirely absent. Such results are preliminary.