Reproducibility of Brain Volume Changes in Longitudinal Voxel-Based Morphometry Between Non-Accelerated and Accelerated Magnetic Resonance Imaging.

BACKGROUND

Scan acceleration techniques, such as parallel imaging, can reduce scan times, but reliability is essential to implement these techniques in neuroimaging.

OBJECTIVE

To evaluate the reproducibility of the longitudinal changes in brain morphology determined by longitudinal voxel-based morphometry (VBM) between non-accelerated and accelerated magnetic resonance images (MRI) in normal aging, mild cognitive impairment (MCI), and Alzheimer's disease (AD).

METHODS

Using data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) 2 database, comprising subjects who underwent non-accelerated and accelerated structural T1-weighted MRI at screening and at a 2-year follow-up on 3.0 T Philips scanners, we examined the reproducibility of longitudinal gray matter volume changes determined by longitudinal VBM processing between non-accelerated and accelerated imaging in 50 healthy elderly subjects, 54 MCI patients, and eight AD patients.

RESULTS

The intraclass correlation coefficient (ICC) maps differed among the three groups. The mean ICC was 0.72 overall (healthy elderly, 0.63; MCI, 0.75; AD, 0.63), and the ICC was good to excellent (0.6-1.0) for 81.4%of voxels (healthy elderly, 64.8%; MCI, 85.0%; AD, 65.0%). The differences in image quality (head motion) were not significant (Kruskal-Wallis test, p = 0.18) and the within-subject standard deviations of longitudinal gray matter volume changes were similar among the groups.

CONCLUSION

The results indicate that the reproducibility of longitudinal gray matter volume changes determined by VBM between non-accelerated and accelerated MRI is good to excellent for many regions but may vary between diseases and regions.