Premium
Longitudinal network connectivity measurements in medial temporal lobe subregions discriminate preclinical Alzheimer’s from amyloid‐β negative controls
Author(s) -
Khandelwal Pulkit,
Xie Long,
Bassett Danielle S,
de Flores Robin,
Wolk David A,
Yushkevich Paul A,
Das Sandhitsu R
Publication year - 2020
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 - 10.1002/alz.047689
Subject(s) - entorhinal cortex , atrophy , temporal lobe , discriminative model , grey matter , neuroscience , lobe , psychology , nuclear medicine , hippocampus , magnetic resonance imaging , medicine , pathology , white matter , artificial intelligence , radiology , epilepsy , computer science
Background Finding sensitive outcome measures for disease progression in clinical trials of preclinical Alzheimer’s disease (AD) remains challenging. We hypothesize that longitudinal network connectivity measurements of the medial temporal lobe (MTL) subregions, sites of earliest tau pathology, are more sensitive to progression at early stages than annualized subregional atrophy rates, thus, increasing power to detect a significant effect in a shorter timeframe. Method Longitudinal T1‐weighted MRI scans of 229 amyloid‐β negative (Aβ‐) controls and preclinical AD (Aβ+ controls) from ADNI were included (Table 1). Anterior/posterior hippocampus, entorhinal cortex (ERC), Brodmann areas (BA) 35 and 36, and parahippocampal cortex (PHC) were segmented in baseline MRI (Figure 1, Xie et al . 2019). Deformation‐based morphometry was used to obtain follow‐up volume measurements, which were subsequently entered in a linear regression to estimate an annualized atrophy rate for each subregion. The network measurements of strength, clustering coefficient, and local efficiency (Soto et al . 2016) were obtained from an adjacency matrix (Figure 1) whose elements represented inter‐regional covariance in volume measurements over time. At least three time points were used for each subject. The preclinical AD group was compared to the Aβ‐ controls with ANCOVA and Bonferroni correction, using age as a nuisance covariate. We compared the discriminative power of the network measurements against the atrophy rates for follow‐up MRI scans within 4, 2 and 1.5 years from baseline. Result Discriminative ability of both atrophy rates and network measurements diminishes with decreasing number of follow‐up scans. However, at least one network measurement was still significantly stronger, in absolute terms, than atrophy rates for all three follow‐up scan time periods (Tables 2,3,4, Figure 2). Even with short follow‐up (<1.5 years), a number of network measures were significant while atrophy rate was not (Table 4). Conclusion The longitudinal network measurements of the MTL subregions are sensitive to disease progression in preclinical AD to a greater extent than longitudinal atrophy rates, perhaps because they account for inter‐region interactions in the network which may capture covariance patterns relevant to the neurodegenerative process. As a result, they may serve as efficient outcome measures in clinical trials, potentially allowing treatment effects to be detected earlier.