How do Uncompensated Micro-Shifts Systematically Confound Low-Concentration Metabolite Quantification in 1H-MRS Analysis?

PURPOSE: Linear combination model fitting is widely used for spectral analysis in 1H-MRS but relies on basis sets with fixed spin-system parameters, implicitly neglecting the sensitivity of chemical shifts to physiological and pathological tissue microenvironments. This study systematically evaluated the impact of frequency perturbations on spectral analysis and reassessed the validity of the Cramér-Rao lower bounds (CRLB) as a quality-control metric for low-concentration metabolites.

METHODS

The effects of frequency perturbations on spectral analysis were investigated using simulated datasets with different baseline models and lineshapes, followed by validation in a publicly available in vivo human brain 1H-MRS dataset (n = 31). Estimated metabolite concentrations, relative concentration changes, estimation errors, and CRLB values were analyzed across all conditions.

RESULTS

Frequency perturbations produced minimal quantification changes in high-concentration metabolites but substantially larger errors in several low-concentration metabolites. Across both simulations and in vivo validation, relative concentration changes generally remained below 3% for high-concentration metabolites but exceeded 10%-20% for several low-concentration metabolites, including Gln, GSH, Asp, and NAAG. Importantly, substantial estimation errors frequently occurred despite CRLB values remaining within conventionally acceptable ranges. These findings were consistently observed across different baseline models, lineshapes, and in vivo validation.

CONCLUSION

Physiologically relevant frequency perturbations can introduce substantial quantification bias that is not consistently reflected by CRLB. These findings identify environmentally induced resonance-frequency variations as an underrecognized source of quantification error in current linear-combination model fitting and suggest that CRLB alone is insufficient for assessing quantification reliability, particularly under conditions associated with altered brain temperature or tissue pH.