Abstract
Neuropathological and neuroimaging studies have identified several (endo-)phenotypes of Alzheimer's disease (AD), suggesting a substantial heterogeneity in cerebral atrophy and tau spreading patterns. We included in our study a total of 320 participants, including healthy controls (N = 154) and patients across the AD spectrum (N = 166). We identified clusters of cerebral atrophy and tau PET uptake using a data-driven and similarity-based clustering approach, aiming to examine regional abnormality patterns in both modalities and differences in the clinical, cognitive, and biomarker characteristics among derived clusters. Abnormality patterns in tau PET and T1-weighted MRI within the same individuals revealed four distinct clusters for each imaging modality as surrogate markers of tau and neurodegeneration, respectively. The tau PET and atrophy clusters mainly showed substantial differences in their clustering allocations. While having the most severe biomarkers burden, the left temporal tau and diffuse atrophy clusters revealed the fastest clinical progression and steepest increase in tau PET uptake. Moreover, the diffuse atrophy cluster showed the fastest cortical volume loss, followed by the limbic-predominant atrophy cluster. Our results suggest characteristic differences between tau PET and atrophy clusters, especially for tau PET clusters, revealing more pronounced differences in cognitive profiles and disease biomarker trajectories than atrophy clusters.