For decades, brain imaging in Alzheimer's research has relied on volume measurements — how much the hippocampus shrinks. That approach misses the earliest warning signs. New evidence suggests that subtle textural disorganization within hippocampal subfields, detectable by specialized MRI analysis, may track disease progression and cognitive decline more sensitively than conventional atrophy metrics, potentially reshaping how clinicians stage and monitor Alzheimer's disease.

This multicentre retrospective study drawn from the Alzheimer's Disease Neuroimaging Initiative used high-resolution coronal T2-weighted MRI — a protocol optimized for hippocampal subfield visualization — in participants who also underwent cerebrospinal fluid biomarker assessment within a one-month window. Researchers segmented distinct subfields including cornu ammonis regions 1–3, the dentate gyrus, and the subiculum, then extracted radiomic features beyond simple morphometry: first-order intensity statistics, shape descriptors, and grey-level run-length matrix (GLRLM) texture metrics. These textural signatures, particularly within the subiculum where perforant path fibers are selectively vulnerable early in Alzheimer's pathology, showed meaningful associations with both CSF amyloid-β positivity and cognitive status across the disease continuum.

This work sits at the intersection of radiomics and neurodegeneration, two rapidly converging fields. The subiculum finding is biologically compelling — this subfield is a primary target of tau pathology and perforant path disruption in Braak staging, and its microstructural disorganization logically precedes macroscopic shrinkage. GLRLM features capture the spatial non-uniformity of tissue texture, a computational proxy for cytoarchitectural breakdown that volume measurements simply cannot resolve. The key limitation here is the cross-sectional, retrospective design: associations cannot establish whether texture changes causally predict future decline or merely correlate with concurrent pathology. Generalizability also depends on the specific MRI acquisition protocol used, which is more demanding than standard clinical sequences. Still, if replicated prospectively, subfield texture radiomics could offer a non-invasive, scalable complement to CSF or PET biomarkers for early Alzheimer's staging — an incremental but clinically meaningful advance.