Premium
P2‐028: Influence of the training library composition on a patch‐based label fusion method: Application to hippocampus segmentation on the ADNI dataset
Author(s) -
Coupe Pierrick,
Fonov Vladimir,
Eskildsen Simon,
Manjón José,
Arnold Douglas,
Collins Louis
Publication year - 2011
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.2011.05.918
Subject(s) - segmentation , atrophy , population , kappa , artificial intelligence , dice , computer science , temporal lobe , pattern recognition (psychology) , neuroimaging , hippocampus , alzheimer's disease , medicine , pathology , psychology , neuroscience , disease , mathematics , statistics , geometry , environmental health , epilepsy
(244 ApoE4 carriers and 231 ApoE4 noncarriers). Baseline EDTA plasma samples were collected and analyzed with a 190 analyte multiplex immunoassay panel based on the Luminex xMAP platform. Hippocampal segmentations were generated from 1.5T 3D T1-weighted brain MRI scans with a novel automated segmentation technique based on the AdaBoost machine learning method. Hippocampal thickness was analyzed with the radial distance technique. We applied linear regression models to relate hippocampal radial distance to the plasma concentrations of several proteins found to have different levels in the CSF in ApoE4 carriers and non-carriers. We adjusted for effects of age and gender in the pooled sample, and separately in ApoE4 carriers and non-carriers. For multiple comparisons correction, we used permutations with threshold p < 0.01. Results: High levels of interleukin 13 (IL13), a well-established inflammatory protein, showed a significant negative association with right hippocampal radial distance in the pooled sample (left ß1⁄4 0.21, p corrected1⁄4 0.44; right ß1⁄4 0.31, p corrected1⁄4 0.039). High levels of tissue inhibitor of metalloproteinases 1 (TIMP1), a protein that promotes cell proliferation and has an anti-apoptotic function, showed a significant negative association with hippocampal radial distance in the pooled sample on the left (ß 1⁄4 0.42, p corrected 1⁄4 0.027). Angiopoietin 2 (ANG2), a regulator of neuronal progenitor differentiation and migration, showed a significant positive association with hippocampal radial distance on the right in ApoE4 carriers (ß 1⁄4 0.53, p corrected 1⁄4 0.018), but a significant negative association on the left in ApoE4 non-carriers (ß 1⁄4 0.2, p corrected 1⁄4 0.026). Conclusions: Hippocampal atrophy in the ADNI cohort was associated with plasma levels of several proteins whose plasma levels were previously reported to differ between ApoE4 carriers and non-carriers in a FAD cohort. These plasma protein levels may inform pre-dementia diagnostic ascertainment in the future.