AbstractBackgroundPrior studies have demonstrated that shape analysis of the hippocampus may provide a more accurate prognostic prediction of cognitive decline than simply volume. Here we deploy a novel automated pipeline for hippocampal morphometry to elucidate the surface regions sensitive to shape change along the AD spectrum.MethodsHippocampal volumes were segmented from 186 T1 images obtained from the Wake Forest Alzheimer’s Disease Research Center (ADRC) Clinical Core cohort including cognitively normal (CN, n = 62, mean age = 72.1, males = 29), mild cognitive impairment (MCI, n = 62, mean age = 73.2, males = 29), and dementia participants (DEM, n = 62, mean age = 73.4, males = 28). All chosen participants were matched for age, sex, and education. Correspondent surface meshes of each hippocampus were generated via spherical harmonics (SPHARM‐PDM) in SlicerSALT. Hippocampal mesh models were incorporated into a Multivariate Functional Shape Data Analysis (MFSDA), a linear model paradigm that builds associations between shape measurements, demographic information, and other clinical variables to determine if there exists a statistically significant morphological difference caused by variables of interest. Results were mapped onto a template hippocampus to visualize surface regions significantly associated with cognitive diagnosis. Adjudication of cognitive diagnosis by expert panel consensus occurred in accordance with NIA‐AA guidelines.ResultsFigure 1 displays surface regions of left hippocampus that significantly differed between DEM and CN participants. Figure 2 displays surface regions of left hippocampus that significantly differed between MCI and CN participants. Significant regions (p ≤ 0.05) are yellow‐red, while non‐significant regions (p > 0.05) are blue. Similar results were observed in right hippocampus (not shown).ConclusionShape differences between DEM and CN subjects are significant across the majority of the hippocampal surface, while shape differences between MCI and CN subjects appear significant primarily at the posterior portions of CA1/CA3 and prosubiculum subregions, suggesting these particular regions may be vulnerable to atrophic shape change during the early stages of AD, consistent with previous findings.
Read full abstract