Quantitative analysis of intersecting pentagon drawings for interpretable cognitive impairment screening

Background The intersecting pentagon drawing task in the Mini-Mental State Examination (MMSE), also referred to as the Pentagon Drawing Test (PDT), is a brief cognitive screening tool in Alzheimer's disease (AD) and other cognitive disorders. However, conventional manual scoring is coarse and may overlook clinically meaningful variation in drawing performance. Objective To evaluate whether quantitative PDT features can support interpretable screening for cognitive impairment. Methods We analyzed participants who completed the MMSE, Montreal Cognitive Assessment (MoCA), and PDT. Cognitive impairment was defined using education-adjusted MMSE and MoCA cutoffs. Eight quantitative drawing features characterized geometric integrity, spatial organization, and graphomotor characteristics. These features, together with demographic variables, were evaluated for prediction of a modified MMSE-derived cognitive score and classification of cognitive status using cross-validation. Results A total of 3478 participants were included, including 2463 with cognitive impairment and 1015 cognitively normal individuals. All eight quantitative drawing features were significantly lower in the cognitive impairment group. Shape regularity, vertices, alignment and intersection area proportion showed the strongest positive correlations with modified MMSE scores. The model achieved an R 2 of 0.42 for score prediction and showed a positive correlation between predicted and observed scores (Spearman ρ = 0.64). For cognitive impairment classification, the model achieved a ROC-AUC of 0.832, sensitivity of 82.5%, specificity of 67.4%, PPV of 86.0%, NPV of 61.4%, and F1-score of 0.842. Conclusions Quantitative features of PDT showed interpretable associations with cognitive screening outcomes. This approach may support scalable and objective cognitive screening and serve as a complement to clinical assessment.