Abstract
INTRODUCTION: In contrast to Alzheimer's disease (AD), in which initial neurofibrillary tangles (NFT) are mainly limited to the (trans)entorhinal region (EC) and CA1/prosubiculum, primary age-related tauopathy (PART) has been suggested to exhibit an early predisposition to NFTs in the hippocampal CA2 subregion. METHODS: We created an artificial intelligence model that recognizes and quantifies NFTs of three different maturity levels in different hippocampal subfields. This model was applied to a population-based Vantaa 85+ cohort, including hippocampal tau-immunostained sections from 210 individuals aged ≥ 85 years. RESULTS: EC and CA1, but not CA2, had significantly higher NFT density in moderate AD compared to PART. CA2/CA1 and CA2/EC NFT ratios were inversely associated with amyloid beta (Aβ) deposition in PART versus AD. The maturation process was exacerbated by Aβ in AD versus PART. DISCUSSION: Compared to AD, a CA2 conspicuous pattern of hippocampal NFTs was more common in, but not exclusive to, PART. HIGHLIGHTS: Neurofibrillary tangles (NFTs) of varying maturity were studied by a new kind of artificial intelligence model in the oldest-old. Hippocampal NFT distribution differed between primary age-related tauopathy (PART) and Alzheimer's disease (AD). PART more often showed prominent NFT pathology in CA2 compared to AD. In PART, the NFT maturation process stood out in CA2 compared to other subfields. The NFT maturation process was aggravated by amyloid beta in AD compared to PART.