Premium
The heterogeneous brain: Mapping individualised patterns of atrophy in Alzheimer’s disease using spatial normative models
Author(s) -
Verdi Serena,
Kia Seyed Mostafa,
Marquand Andre F.,
Schott Jonathan M.,
Cole James H.
Publication year - 2021
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.057605
Subject(s) - neuroimaging , normative , psychology , outlier , population , alzheimer's disease neuroimaging initiative , disease , medicine , cognition , neuroscience , artificial intelligence , cognitive impairment , pathology , computer science , philosophy , environmental health , epistemology
Background Alzheimer’s Disease (AD) is highly heterogeneous, with marked individual differences in clinical presentation and neurobiology. Neuroimaging biomarkers have considerable utility in AD research, however, common statistical designs do not capture neuroanatomical heterogeneity, generally assuming the effects of AD on the brain will be the same in different patients. Spatial normative modelling is an emerging technique that can reveal individual patterns of neuroanatomy by quantifying deviations from normative ranges (Verdi et al., 2021). On multi‐site Alzheimer’s Disease Neuroimaging Initiative (ADNI) data, we applied a hierarchical Bayesian regression (HBR) spatial normative model (Kia et al., 2020). We compared patterns of cortical thickness heterogeneity in AD patients, people with Mild Cognitive Impairment (MCI), and Cognitively Normal (CN) people. Method Structural imaging data provided estimates of cortical thickness across 148 (ROIs), generated using FreeSurfer on n=563 T1‐weighted MRI scans acquired across 38 sites. We then applied the HBR spatial normative model, using a separate healthy reference dataset of n=33,073 to index population variability, predicting cortical thickness at each ROI using age and sex. Next, we applied transfer learning, to recalibrate the normative model to the CN participants from ADNI. This generated cortical thickness z‐scores across ROIs for each participant (Fig. 1). Z‐scores < ‐1.96 were identified as outliers. Result Linear regression revealed group differences of outliers summed across 148 ROIs: AD had significantly more ROI outliers than MCI and CN participants (β=10.99,95%CI=[5.55,16.42], p=8.28×10 ‐5 ) (Fig. 2). ANOVAs at each ROI demonstrated that this difference was within temporal regions (Fig. 3) , e.g., parahippocampal gyrus (F(2,23.33),CI=[ 0.29, 0.57],FDR p=3.60×10 ‐8 ). Calculating the percentage of outliers in each ROI within groups revealed that outlying ROIs in AD patients only partially overlap. The parahippocampal gyrus had the highest number of AD patients with outliers (n=27 out of n=59 (45.7%)); more than half of the AD group appeared normal in this region (Fig. 4). Conclusion Our novel quantitative estimate of spatial heterogeneity of cortical thickness in AD patients suggests that the impact of AD on the brain is not consistent between patients. Individualised patient neuroanatomical maps have the potential to be a marker of disease, and could be used to track an individual’s disease progression or treatment response.