Structural MRI and mild behavioral impairment as complementary predictors of conversion from amnestic MCI to...

Mild Behavioral Impairment (MBI) and amnestic Mild Cognitive Impairment (aMCI) are complementary early markers of Alzheimer's disease (AD), yet their combined neuroanatomical correlates and predictive value for conversion remain underexplored. In this study of 72 community-dwelling older adults (49 aMCI, 23 healthy controls), we retrospectively classified aMCI participants into non-converters (aMCI-NC, n = 31) and converters (aMCI-C, n = 18) based on longitudinal follow-up. Baseline structural MRI revealed that aMCI-C patients exhibited bilateral atrophy in the hippocampus, amygdala, and nucleus accumbens, alongside cortical thinning in temporal-limbic and parietal regions compared to both aMCI-NC and controls. Notably, in the non-converter group, MBI-Checklist scores correlated negatively with temporal-parietal cortical thickness, while Montreal Cognitive Assessment scores correlated positively with subcortical volumes. Cox proportional-hazards regression with supervised principal component analysis revealed that while MBI symptoms alone did not independently predict conversion, they demonstrated significant conditional predictive value when combined with structural MRI markers. Model robustness was confirmed via rigorous internal validation using nested leave-one-out cross-validation with bootstrap aggregation. Furthermore, time-varying analysis indicated that the protective effect of larger medial temporal-limbic subcortical volumes was strong initially but attenuated over follow-up, consistent with a depletable brain reserve. Given the limited conversion events, which constrained statistical power for smaller independent effects and restricted model complexity, replication in larger prospective cohorts is warranted. Despite this limitation, our results underscore the conditional utility of integrating MBI assessment with structural MRI to enhance early detection and risk stratification in memory clinic settings.