Characterizing the MRI signature of hippocampal sclerosis of aging

Hippocampal sclerosis of aging (HS-A), a common cause of dementia, can currently only be diagnosed at autopsy. We aimed to identify and evaluate MRI metrics to distinguish HS-A from Alzheimer's Disease neuropathologic change (ADNC) and cases with limited/no pathology.

HS-A (N = 5), ADNC (N = 10), and limited/no pathology (N = 12) cases were compared on postmortem MRI signatures: manually measured cornu ammonis (CA)1/subiculum thickness, grey matter signal intensity, and automated hippocampal subfield thickness metrics. Similar metrics were obtained in T₁-weighted antemortem MRI in an initial dataset (HS-A = 4, ADNC = 7, limited/no pathology = 25) for group differences and discrimination (HS-A vs ADNC). T₁-weighted metrics were then evaluated in a second dataset (HS-A = 6, ADNC = 18) and in a pooled post-hoc analysis combining HS-A and ADNC cases from both datasets (N_HS-A = 10, N_ADNC = 25).

Postmortem MRI showed hippocampal thinning and grey matter hypointensity in HS-A at the CA1-subiculum junction more severe than in ADNC and limited/no pathology cases. In antemortem MRI, differentiating HS-A and ADNC based on anterior/posterior manual measures of CA1/subiculum thickness and hippocampal volumes displayed good discrimination in dataset 1, but lower discriminative performance in dataset 2. In complementary analyses pooling both datasets and adjusting for age, manual thickness achieved good performance (area under the curve (AUC) = 0.80-0.87), while anterior, posterior, and whole hippocampal volumes showed excellent discrimination (AUC = 0.94-0.98).

The study included a relatively small and neuropathologically heterogeneous sample. The final antemortem analyses were exploratory, reflecting challenges in replicating findings across independent datasets. Classification was limited to HS-A and ADNC, leaving it uncertain how the metrics perform when comparing HS-A with other diagnostic groups. HS-A diagnoses were determined postmortem, often several years after MRI, and most measures relied on standard imaging sequences with limited resolution to assess fine-grained hippocampal subfields.

HS-A displays distinct changes in the hippocampus that are detectable through structural MRI. Associated quantifiable MRI metrics may serve as promising tools in aiding antemortem HS-A diagnosis but require further validation in larger cohorts and against other dementia-related diseases.