Developing Topics.

BACKGROUND: Previous research has identified heterogeneous patterns of atrophy progression in individuals with Alzheimer's disease, which are linked to interindividual differences in clinical presentation and cognitive decline. While such frameworks offer promise for individualized care and more targeted clinical trials, existing models usually require extensive pre-processing and high-field MRI. Here, we aim to improve accessibility of a previously proposed atrophy subtyping and staging model (Baumeister et al., 2024, Brain) by using accelerated pre-processing and relying only on standard T1-weighted images (1.5T or 3T).

METHOD: We included participants with available T1-weighted MRI scans from the DZNE Longitudinal Cognitive Impairment and Dementia Study (DELCODE; N = 813), the Alzheimer's Disease Neuroimaging Initiative (ADNI; phases 1, GO, 2, and 3; N = 2,117), as well as the Anti-Amyloid Treatment in Asymptomatic Alzheimer's Disease and Longitudinal Evaluation of Amyloid Risk and Neurodegeneration (A4/LEARN) studies (N = 1,241). Images were acquired at 3T across all cohorts, except for n = 715 ADNI participants who were scanned at 1.5T. Atrophy markers were extracted using a deep learning-based algorithm (FastSurfer; Henschel et al., 2020, NeuroImage), and the Subtype and Stage Inference (SuStaIn) algorithm (Young et al., 2018, Nat Commun) was used for atrophy progression modelling.

RESULT: We replicated our previous finding of two distinct limbic-predominant and hippocampal-sparing atrophy progression subtypes in DELCODE using the updated model. At the individual level, subtype and stage classifications showed strong agreement with those from the original model. When retrained de novo in the independent validation cohorts (ADNI-3T, ADNI-1.5T, and A4/LEARN), the model identified comparable patterns of brain atrophy, highlighting its generalizability across diverse datasets and commonly used MRI field strengths. While subtype and stage assignments were based solely on cross-sectional data, longitudinal MRI scans showed strong within-subject stability of subtype classifications and monotonously increasing in atrophy stages, aligning with model assumptions.

CONCLUSION: Assessing brain atrophy within the context of a global progression framework enables a more comprehensive evaluation than relying on single volume measures or isolated metrics. This work demonstrates that atrophy progression models can be adapted for real-world settings, including those without access to high-field MRI.

© 2025 The Alzheimer's Association. Alzheimer's & Dementia published by Wiley Periodicals LLC on behalf of Alzheimer's Association.

PMID: 41434039