Machine learning‐based classification of Alzheimer's disease and its at‐risk states using personality traits, anxiety, and depression

Abstract

Abstract Background Alzheimer's disease (AD) is often preceded by stages of cognitive impairment, namely subjective cognitive decline (SCD) and mild cognitive impairment (MCI). While cerebrospinal fluid (CSF) biomarkers are established predictors of AD, other non‐invasive candidate predictors include personality traits, anxiety, and depression, among others. These predictors offer non‐invasive assessment and exhibit changes during AD development and preclinical stages. Methods In a cross‐sectional design, we comparatively evaluated the predictive value of personality traits (Big Five), geriatric anxiety and depression scores, resting‐state functional magnetic resonance imaging activity of the default mode network, apoliprotein E (ApoE) genotype, and CSF biomarkers (tTau, pTau181, Aβ42/40 ratio) in a multi‐class support vector machine classification. Participants included 189 healthy controls (HC), 338 individuals with SCD, 132 with amnestic MCI, and 74 with mild AD from the multicenter DZNE‐Longitudinal Cognitive Impairment and Dementia Study (DELCODE). Results Mean predictive accuracy across all participant groups was highest when utilizing a combination of personality, depression, and anxiety scores. HC were best predicted by a feature set comprised of depression and anxiety scores and participants with AD were best predicted by a feature set containing CSF biomarkers. Classification of participants with SCD or aMCI was near chance level for all assessed feature sets. Conclusion Our results demonstrate predictive value of personality trait and state scores for AD. Importantly, CSF biomarkers, personality, depression, anxiety, and ApoE genotype show complementary value for classification of AD and its at‐risk stages.

Key points Multi‐class support vector machine classification was used to compare the predictive value of well‐established and non‐invasive, easy‐to‐assess candidate variables for classifying participants with healthy cognition, subjective cognitive decline, amnestic mild cognitive impairment, and mild Alzheimer's disease. Personality traits, geriatric anxiety and depression scores, resting‐state functional magnetic resonance imaging activity of the default mode network, ApoE genotype, and CSF biomarkers were comparatively evaluated. A combination of personality, anxiety, and depression scores provided the highest predictive accuracy, comparable to CSF biomarkers, indicating complementary value. Established and candidate predictors had limited success in classifying SCD and aMCI, underscoring the heterogeneity of these cognitive states and emphasizing the need for standardizing terminology and diagnostic criteria.