Disentangling global atrophy burden from local structural patterns reveals clinically relevant heterogeneity in mild cognitive impairment

Mild cognitive impairment (MCI) is a heterogeneous prodromal stage of Alzheimer's disease (AD). Imaging-based subtyping may conflate global neurodegenerative burden with local anatomical variation, limiting biological interpretability. We aimed to separate global atrophy burden from local structural patterns and assess their clinical relevance in MCI.

Baseline ADNI-1 data from 731 participants were analyzed, including 205 cognitively normal individuals, 351 with MCI, and 175 with AD. A global atrophy index (GAI) was defined as the z-standardized ratio of TIV-adjusted whole-brain cerebrospinal fluid volume to TIV-adjusted gray matter plus white matter volume. Candidate regions were screened in the full CN-MCI-AD sample. Within MCI, regional measures were residualized against age, sex, education, and GAI to derive local residual structural features. Principal component analysis and unsupervised clustering were applied to identify imaging-based subtypes. Clinical associations were assessed using multivariable regression, and nested models evaluated the incremental explanatory value of continuous residual features beyond subtype labels.

GAI showed a clear gradient across cognitively normal, MCI, and AD groups. After adjustment for GAI, two local residual structural patterns were identified within MCI: a thalamic-striatal lower-residual subtype and a frontal-paralimbic lower-residual subtype. Compared with the thalamic-striatal subtype, the frontal-paralimbic subtype showed greater cognitive impairment and poorer daily functioning, with higher ADAS-Cog13 and FAQ scores after covariate adjustment. Associations with neuropsychiatric symptoms were weak. Continuous PCA-derived residual features provided additional explanatory value for ADAS-Cog13 and FAQ beyond subtype labels.

MCI heterogeneity is not explained solely by global atrophy. Local residual structural patterns were associated with cognitive and functional differences, and continuous residual features captured additional clinical information beyond discrete subtype labels. Separating global atrophy burden from local anatomical organization may improve interpretation of imaging heterogeneity in prodromal AD.