Cognitive deficits in Parkinson's disease (PD) are heterogeneous entities, but a relationship between the heterogeneity of cognitive deficits and resting-state network (RSN) changes remains elusive. In this study, we examined five sub-domain scores according to Addenbrooke's Cognitive Examination-Revised (ACE-R) for the cognitive evaluation and classification of 72 non-demented patients with PD. Twenty-eight patients were classified as PD with normal cognition (PD-NC). The remaining 44 were subdivided into the following 2 groups using a hierarchical cluster analysis: 20 with a predominant decrease in memory (PD with amnestic cognitive deficits: PD-A) and 24 with good memory who exhibited a decrease in other sub-domains (PD with non-amnestic cognitive deficits: PD-NA). We used an independent component analysis of RS-fMRI data to investigate the inter-group differences of RSN. Compared to the controls, the PD-A showed lower FC within the ventral default mode network (vDMN) and the visuospatial network. On the other hand, the PD-NA showed lower FC within the visual networks and the cerebellum-brainstem network. Significant differences in the FC within the vDMN and cerebellum-brainstem network were observed between the PD-A and PD-NA, which provided a good discrimination between PD-A and PD-NA using a support vector machine. Distinct patterns of cognitive deficits correspond to different RSN changes.
Keywords: Cognitive deficits; Functional connectivity; Independent component analysis; Parkinson’s disease; Resting-state fMRI; Resting-state network.