Robust Extrapolated Semi-Solid Magnetization Transfer Reference Fitting for Quantitative Amide Proton Transfer Imaging at 3 T: Application to Patients With Mild Cognitive Impairment and Mild Dementia.

Purpose: To refine the utility of the extrapolated semi-solid magnetization transfer reference (EMR) method for amide proton transfer (APT) quantification.

Methods: Twelve patients (7 mild cognitive impairment and 5 mild dementia; 66.3 ± 8.7 years) and 13 age-matched, cognitively normal volunteers (68.6 ± 8.0 years), as well as a 10% cross-linked bovine-serum-albumin phantom, were scanned at 3 T. A two-step, coarse-to-fine EMR fitting approach was developed, and a bias term was introduced to compensate for the discrepancy between the ideal model and practical data. The fitted model parameters and calculated APT^# and NOE^# (nuclear Overhauser effect) were compared across different EMR fitting methods and between the two groups. The normalized-root-mean-squared-error was used to measure the discrepancy between fitted curves and acquired Z-spectra.

Results: Phantom results confirmed that the proposed EMR method had lower fitting errors and closer-to-zero APT^# and NOE^# signals. Human results showed the specific APT^# signals peaked at 3.5 ppm. A noticeable APT^# increase in the hippocampus was seen in the mild cognitive impairment/mild dementia group (mean APT^# = 3.09%, compared to 2.41% in the cognitively normal group; p = 0.002). The APT^# signal in the hippocampus provided higher accuracy in differentiating between cognitively normal individuals and those with mild cognitive impairment/mild dementia than APT-weighted signal (an area-under-the-curve of 0.92 compared to 0.67).

Conclusion: The proposed EMR method enabled more accurate quantification of APT signals and could potentially facilitate the use of APT imaging in the diagnosis and staging of Alzheimer's disease and related dementias.