A Novel Approach for Cognitive Clustering of Parkinsonisms through Affinity Propagation

Cluster analysis is widely applied in the neuropsychological field for exploring patterns cognitive profiles, but traditional hierarchical and non-hierarchical approaches could be often poorly effective or even inapplicable on certain type of data. Moreover, these need initial specification number clusters, based a priori knowledge not always owned. For this reason, we proposed novel method clustering through affinity propagation (AP) algorithm. In particular, AP regression residuals Mini Mental State Examination scores—a commonly used screening tool impairment—of cohort 49 Parkinson’s disease, 48 Progressive Supranuclear Palsy 44 healthy control participants. We found four where two clusters (68 30 participants) showed almost intact performance, one cluster had moderate impairment (34 participants), last more extensive deficit (8 participants). The findings showed, first time, an intra- inter-diagnostic heterogeneity profile Parkinsonisms patients. Our unsupervised learning represent reliable supporting neuropsychologists understanding natural structure performance neurodegenerative diseases.