Hippocampal subregional texture features associated with Alzheimer’s disease severity and cognition

Abstract Alzheimer’s disease (AD) exhibits selective vulnerability in hippocampal subfields, where microstructural alterations precede overt atrophy. Conventional morphometric measures, such as volume or thickness, are limited in their ability to capture subtle textural heterogeneity that reflects underlying cytoarchitectural disorganization, particularly perforant path fibers within the subiculum. The objective of this study is to determine whether hippocampal subregional texture-based radiomic features can serve as sensitive markers of disease severity and cognitive impairment along the AD continuum. This retrospective, multicenter, cross-sectional study used data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Participants who underwent coronal high-resolution T2-weighted “HighResHippo” magnetic resonance imaging (MRI) and cerebrospinal fluid (CSF) biomarker assessment within 1 month of each other between June 2011 and May 2025 were included. Automatic Segmentation of Hippocampal Subfields was used to delineate cornu ammonis 1–3, dentate gyrus, subiculum, and other medial temporal cortices, followed by quality control to exclude mislabeled or artifact-degraded segmentations. After bias-field correction, intensity normalization, and voxel-resampling, first-order, shape, and texture features—including gray-level run-length matrix (GLRLM) metrics—were extracted. Participants were categorized according to cognitive status and CSF amyloid-β (Aβ) positivity. Associations between radiomic features and Montreal Cognitive Assessment (MoCA) scores as well as CSF tau concentrations were evaluated using multiple linear regression, adjusting for age, sex, education, and APOE ε4 status. Among 1,264 screened participants, 241 met all inclusion criteria (mean age 73.4 ± 6.9 years; 54% women). GLRLM-based run entropy within the subiculum demonstrated a stepwise increase across the AD continuum (cognitively unimpaired-Aβ– cognitively unimpaired-Aβ+ mild cognitive impairment-Aβ+ AD dementia-Aβ+). Orientation-specific analyses revealed that the superior–inferior component—aligned with the dominant fiber orientation of the subicular efferent pathway—showed the strongest associations with clinical and biomarker indices. Higher SI run entropy correlated with lower MoCA performance (standardized β = –0.710 95% CI –1.416 to –0.038) and higher CSF tau levels (standardized β = 2.307 95% CI 0.767 to 3.848) compared with other directional components. Radiomic texture features derived from in vivo high-resolution T2-weighted MRI of hippocampal subregions—particularly subicular GLRLM run entropy aligned with superior–inferior tract orientation—track disease severity and cognitive impairment along the AD continuum. These findings support texture-based imaging biomarkers as sensitive indicators of early microstructural alterations in AD.