Premium
P2–200: Self‐ versus informant‐based cognitive complaints: Relation of E‐Cog scores to imaging, biomarkers and clinical Status in ADNI‐2
Author(s) -
Risacher Shan,
Petersen Ronald,
Aisen Paul,
Jack Clifford,
Koeppe Robert,
Jagust William,
Farias Sarah Tomaszewski,
Mungas Dan,
Trojanowski John,
Shaw Leslie,
Weiner Michael,
Saykin Andrew
Publication year - 2013
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2013.05.845
Subject(s) - atrophy , neuroimaging , cog , cognition , psychology , effects of sleep deprivation on cognitive performance , medicine , cohort , voxel , audiology , neuroscience , radiology , artificial intelligence , computer science
segmentation. The current study hence aimed to determine whether choice of ADNI MPRAGE acquisition and RF coil impacts on the compatibility of common brain segmentation software packages.Methods: 3D T1-weighted images of 9 subjects were acquired on a 3T GE MR750 scanner using 8, 12 and 32-channel coils (ADNI-1 MPRAGE, ADNI-GO/2 MPRAGE, ADNIGO/2 accelerated MPRAGE, together with a FSPGR volume for comparison). Images were processed using SPM-8, FSL and FreeSurfer in order to determine grey matter (GM), white matter (WM) and cerebrospinal fluid (CSF) volumes. Volumes were analysed in SPSS using the Intra-class correlation co-efficient (ICC) as a reliability measure to examine how strongly volumes between each coil and sequencewere related.Results:On average, brain segmentation results using the ADNI-1MPRAGEwere the most comparable for GM using the 8(GM1⁄40.93), 12(GM1⁄40.88) and 32channel(GM1⁄40.87) coil, whereas the ADNI-GO-ACC sequence was most comparable for WM: 8(0.99), 12(0.99) and 32-channel(0.97) coil. No ADNI sequence was better than the others for segmenting CSF. Overall, the 8-channel coil produced the most consistent GM andWM segmentation: 8(GM1⁄40.91,WM1⁄40.97), 12(GM1⁄40.88,WM1⁄40.92) and 32-channel (GM1⁄40.76,WM1⁄40.91) coil. CSF segmentation using SPM-8(0.89) and FSL(0.89) was best using the 8-channel coil, but overall FreeSurfer segmentation using the 32-channel coil was the most comparable. Segmentation reproducibility on average across all sequences was consistently higher using FreeSurfer(GM1⁄40.87,WM1⁄40.95), compared to SPM8(GM1⁄40.84,WM1⁄40.92) and FSL(GM1⁄40.83,WM1⁄40.93). Conclusions: Segmentation comparability was dependent on tissue type and was most comparable in SPM-8, FSL and FreeSurfer in the following order: WM>GM>CSF. Comparability for FreeSurfer and the ADNI-1 sequence was best in both GM and WM. We have shown the extent to which the choice of coil, sequence and brain segmentation software impacts thevolumetric analysis of MRI data. This has implications for future study design and for initiatives aiming to combining multiple retrospective MRI studies.