N. Schuff, D. Amend, M. El-Din, D. Norman, G. Fein, and M.W. Weiner
Neuropsychological studies have shown that elderly individuals with mild cognitive impairments (MCI) are at increased risk for future development of Alzheimers disease (AD) . MRI studies in MCI reported reductions of hippocampal volume similar to changes seen in AD [2,3]. However, it remains unknown if this hippocampal volume loss is a sensitive measure of pathological processes which presage AD, because hippocampal atrophy occurs during normal aging , does not distinguish AD from other dementias , and is attenuated by gliosis.
In AD, several 1H-MRS studies reported reduced levels of N-acetyl aspartate (NAA) in cortical and hippocampal regions, consistent with the known distribution of pathology in AD. In addition, we demonstrated earlier that the combination of hippocampal NAA and volume, measured by 1H-MRSI and by MRI respectively, discriminates AD from healthy elderly better than either measure alone . Therefore, we performed a study in MCI to test the hypothesis that: hippocampal NAA and volume together provide better discrimination between MCI and controls than either measure alone.
Seven MCI subjects (mean age (±SD) = 72.7 ± 10) with a clinical dementia rating =0.5 were studied by MRI and 1H-MRSI (1.5T Vision, Siemens Inc.) and compared to age and sex-matched healthy elderly (N= 17) and AD patients (N=12). Measurement of hippocampal volume (normalized to intracranial head size) was based on coronal T1-weighted MRls (3D MPRAGE, TR/TE=55/6 ms, 1.0 x 1.0mm² inplane resolution, 1.4mm thick partitions). Semi-automated tissue segmentation was performed using these T1 images together with axial spin density and T2-weighted images obtained with a double spin-echo sequence (TR/TE1/T2 = 400/20/80ms, 1.2 x 1.0 mm2 inplane resolution, 3mm thick contiguous slices).
1H-MRSI data was acquired from a 15 mm thick and typically 80 x 100 mm2 wide region of interest oriented parallel to the long axis of the hippocarnpus and with a nominal voxel size of about 1.2 ml (TR/TE= 1800/135 ms, 24 x 24 phase encoding). MR spectra were extracted from voxels located in head, body, and tail of right and left hippocampus. NAA concentration was calculated from fitting the NAA resonance line and referencing the line integral to phantom data with corrections for coil loading, B1 inhomogeneity, and T1,T2 relaxation rates. Atrophy correction for NAA (termed [NAA]) was performed by computing the tissue content of each voxel from coregistered segmented MRis with consideration of voxel “bleed”, chemical shift displacement, and B1 profiles. This computation also provided information about the tissue composition in each voxel. ANCOVA was used to test that [NAA) differences between groups were independent of atrophy and tissue composition. Statistical significance of MRI and MRSI findings were tested using ANOVA.
RESULTS AND DISCUSSION
In all groups, NAA and volume changes of right hippocampus were not significantly different from changes of left hippocampus. Therefore, the mean values of right and left hippocampi were averaged for further analysis. The table lists the results for elderly controls. MCI. and AD. Although hippocampal volume was smaller in MCI than in controls the difference of atrophy corrected [NAA] between the two groups was greater than the difference of volume. Changes of [NAAl were independent of atrophy. In AD and controls, [NAA] was also independent of tissue composition. This has not yet been determined in MCI. Furthermore, in MCI the NAA/Cr ratio (not listed) was lower by 9% and 7% when compared to controls and AD, respectively. Most importantly, [NAA] and hippocampal volumc together separated MCI from elderly controls better than either measure alone. When MCI is compared to AD, neither [NAA], nor volume, nor the combination of both measures separate MCI from AD.
Table: Volume, [NAA] and Volume x [NAA] from hippocampus in elderly controls (EC), MCI, and AD. (Mean value ± standard error; Diff. = difference).
Volume x [NAA]
|EC||3.00 ± 0.08||10.2 ± 0.2||30.4 ± 1.3|
|MCI||2.74 ± 0. 14||8.7 ± 0.5||23.5 ± 1.8|
|AD||2.4 2 ± 0. 12||8.5 ± 0.3||20.6 ± 1.6|
MCI vs. EC
|% Diff||– 8.7||– 14.7||– 22.7|
|Effect Size||0 .2||0 .4||0 .4|
MCI vs. AD
|% Diff||+ 11.7||+ 2.4||+ 12.3|
Given the increased risk of AD among subjects with MCI, the finding of reduced hippocampal NAA in MCI supports the possibility that 1H-MRSI detects metabolic changes in this region which presage AD. Although, variations in tissue composition could contribute to the NAA difference between MCI and healthy elderly, this seems unlikely since NAA changes in controls and AD did not show such correlation. The observation that both NAA and the combination of NAA and volume differences between MCI and controls exhibit larger effect sizes than volume alone suggests that measurement of NAA is beneficial for separating MCI from normal elderly and further supports the hypothesis of NAA being a more sensitive marker of hippocampal damage in MCI than atrophy. However, because it is unknown which MCI subjects in this population will ultimately develop AD and which will remain stable, it is undetermined to what extent these NAA changes in MCI reflect signs of early AD pathology. Consequently, the importance of the observation that NAA levels in MCI were not significantly different from those in AD cannot be determined until further longitudinal studies reveal the relation of hippocampal NAA changes and development of AD.
 Petersen, R. C. et.al., Neurology 43:A277, 1993.
 DeLeon M.J. et.al, Ann. New York Acad. Sciences. 777, 1-14, 1996.
 Parnetti, L. et.al., J Am Geriat. Soc 44: 133-8 1996.
 Fazekas, F.et.al., Jaumal of Nuclear Medicine 30:1607-15. /989.
 Laakso, M. P., et.al., Neurology, 46:678-81. 1996.
 Schuff, N., et.al., Proc. Society of Magnetic Resonance 308, 1996