Premium
IC‐P‐088: ON THE LINK BETWEEN BLOOD‐BRAIN BARRIER LEAKAGE, WHITE MATTER HYPERINTENSITIES, NEURODEGENERATION, AND COGNITION
Author(s) -
Freeze Whitney M.,
Jacobs Heidi IL.,
Jong Joost J.,
Verheggen Inge CM.,
Gronenschild Ed,
Palm Miguel,
Hoff Erik I.,
Wardlaw Joanna M.,
Jansen Jacobus FA.,
Verhey Frans RJ.,
Backes Walter H.
Publication year - 2018
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.2018.06.2152
Subject(s) - hyperintensity , dementia , white matter , medicine , psychology , neuropsychology , magnetic resonance imaging , cognition , effects of sleep deprivation on cognitive performance , cardiology , neuropsychological assessment , neuroscience , disease , radiology
“norming” exist. We compared three methods head-to-head in the same data set. Methods: The test data set comprised anatomical MRIs from 102 amyloid-positive and 51 amyloid-negative ADNI2 subjects. Z-scores were derived using: 1) Local FreeSurfer v5.3 parcellations and the Potvin (2016, 2017) method, 2) NeuroQuant, a commercial, FDA-cleared, medical device software, and 3) Local FreeSurfer v5.3 parcellations and a Simple Linear model with age and ICV as independent variables built using 164 agematched cognitively normal, amyloid-negative ADNI2 subjects, also processed using FreeSurfer v5.3. The Potvin and Simple Linear methods are both based on FreeSurfer; the Potvin and NeuroQuant methods both employ non-linear fits to large reference data sets over the human lifespan. The following regions were evaluated: cortical gray matter, precentral, inferior temporal, inferior parietal and fusiform gyri, hippocampus and lateral ventricles. NeuroQuant output is capped at the 99 percentile (Z1⁄42.33) so any such regions were excluded. The relationship between Z-scores from the different methods was assessed using linear regression, Bland-Altman analyses and concordance between methods when dichotomized at a Z-score threshold of 61.65. Results:There was a bias towards higher Z-scores for NeuroQuant in many cortical regions. Smaller biases were observed for the hippocampus between Potvin and Simple Linear methods, and for the fusiform gyrus between Potvin and NeuroQuant methods. For the lateral ventricles, a non-linear relationship between the Simple Linear approach and both Potvin and NeuroQuant methods was observed, with greatest divergence at higher Z-scores. Concordance between methods ranged from 74.6% to 95.7%. Conclusions:Concordance between norming methods was high overall, but region-dependent biases and non-linear relationships exist between Z-scores resulting from these different vMRI norming methods. The norming method choice should be carefully considered in AD trials. Comparison with visual assessments of age-dependent abnormality should be considered.