Mapping Medial Temporal Lobe Longitudinal Change in Preclinical Alzheimer’s Disease

Abstract

AbstractBackgroundThe medial temporal lobe (MTL) is a site of early tau pathology and neurodegeneration in AD. In recent work [1], we reported that rates of change in MTL subregion volumes on longitudinal structural T1‐MRI differ significantly between cognitively unimpaired individuals with and without amyloid pathology (A+CU vs A‐CU), suggesting that these measures are effective biomarkers for tracking disease progression in preclinical AD. Here, we develop a specialized pipeline that quantifies pointwise longitudinal change across the MTL, revealing more informative spatial patterns of early neurodegeneration than summary volumetric longitudinal measures.MethodsOur pipeline, SkelDBM, combines skeletonization and deformation‐based morphometry to generate maps of annualized rate of atrophy on the skeleton of an MTL template (Figure 1). SkelDBM samples subjects’ atrophy maps along the skeleton of an MTL template, increasing robustness to subject‐to‐template registration errors. We evaluate SkelDBM using a subset of ADNI data studied in [1] (Table 1). Two preclinical AD groups (A+CU and tau‐positive A+CU) are compared to a control group (A‐CU) using general linear models (GLM) with age as nuisance covariate; permutation testing is used to derive statistically significant clusters.ResultsThe full A+CU vs A‐CU comparison shows significant differences in rates of change in right anterior hippocampus, right entorhinal cortex, right transentorhinal cortex, and left posterior hippocampus (Figure 2). Limiting the A+CU group to tau‐positive individuals yields more widespread differences, including most of the right anterior MTL cortex and both hippocampi. These patterns align with volumetric longitudinal data reported in [1] but provide a richer characterization of spatial atrophy patterns.ConclusionSkelDBM is sensitive to longitudinal atrophy occurring over relatively short periods (∼2 years) in preclinical AD and characterizes patterns of atrophy more comprehensively than summary volumetric measures. Summary measures are also subject to biases, since larger structures are typically associated with less measurement error and hence may show exaggerated statistical effects compared to smaller structures even when the underlying neurodegeneration rates are the same. Regional maps of longitudinal change may be more robust to this bias and may help discovery of “signature regions” associated with specific pathological drivers of neurodegeneration.[1] Xie et al., Human Brain Mapping, 2020.