Distinct cognitive profiles differentiate dementia with lewy bodies from Alzheimer's disease

Dementia with Lewy bodies (DLB) and Alzheimer's disease (AD) share overlapping cognitive and neuropsychiatric symptoms, complicating early differential diagnosis. This study aimed to compare multidimensional impairment patterns in AD and DLB and develop a simple, interpretable classification model based on clinical scales.

A total of 249 participants were included: 84 patients with AD, 82 with DLB, and 83 participants with normal cognition (NC). Participants completed assessments covering global cognition, six cognitive domains, neuropsychiatric and depressive symptoms. Missing values in cognitive scales were handled using multiple imputation, and results were pooled across all imputations. Then, Least Absolute Shrinkage and Selection Operator (LASSO), Support Vector Machine (SVM), and Random Forest (RF) were used to identify key variables. A final set of six scales was selected to build a logistic regression model distinguishing DLB from AD. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC) in the entire cohort, mild stage and dementia stage.

DLB patients showed greater deficits in attention, visuospatial processing, and neuropsychiatric symptoms; AD patients exhibited more pronounced memory impairment. At mild stage, DLB displayed more depressive symptoms and attention deficits but milder memory decline than AD. At dementia stage, DLB presented broader impairments in executive, visuospatial, attentional, with similar global cognition. The six-feature model achieved high diagnostic accuracy in the entire cohort (AUC=0.879, 95%CI: 0.802-0.957), mild stage (AUC=0.866, 95% CI: 0.788-0.943) and dementia stage (AUC=0.939, 95% CI: 0.839-0.999).

The study identified distinct cognitive profiles of DLB and AD, and developed a concise, clinically practical model with robust diagnostic utility across disease stages, supporting its use in outpatient and resource-limited settings.