Cognitive decline rates differ among individuals with the same diagnosis category. Fast decliners may differ from slow decliners in clinical, neuropsychological, and neuroimaging profiles. Identifying imaging markers associated with microstructural injury could improve early risk assessment and disease monitoring.

Participants with longitudinal cognitive testing and multimodal MRI were classified as cognitively normal (n=378), mild cognitive impairment (MCI; n=162), or demented (n=60). Individuals were classified as fast or slow decliners based on whether their slopes of executive function (EF) and episodic memory (EM) were above or below the group-specific median. Logistic regression models examined associations between free water (FW), fractional anisotropy (FA), peak-width of skeletonized mean diffusivity (PSMD), the index for diffusion along perivascular spaces (ALPS), and white-matter hyperintensity (WMH) burden with the likelihood of being a fast decliner. Models were adjusted for demographics, vascular risk factors, total cranial volume, and hippocampal volume. Numeric variables were standardized.

Among cognitively normal participants, higher FW (OR 1.4 [SE 0.2], p=0.046) and WMH (OR 1.4 [0.1], p=0.008) were associated with faster EF decline. In MCI, higher FW (OR 1.6 [0.2], p=0.045) was linked to faster EF decline. In demented participants, higher FW (OR 3.9 [0.5], p=0.006) and PSMD (OR 2.1 [0.3], p=0.032) were associated with faster EM decline.

Distinct neuroimaging patterns were found for each diagnostic group and cognitive domain. White-matter integrity biomarkers were more predictive of EF, which reflects early vascular and connectivity changes. In contrast, their predictive value for EM, which is more connected to advanced gray-matter degeneration, was evident only in demented participants. These stage-specific imaging features may help predict both the trajectory and rate of cognitive decline within each clinical group.