Parkinson's disease (PD) is a chronic and progressive movement disorder of the central nervous system characterized by widespread alterations in several non-motor aspects such as mood, sleep, olfactory, and cognition in addition to motor dysfunctions. Advanced neuroimaging using functional connectivity reconstruction of the human brain has provided a vast knowledge on the pathophysiological mechanisms underlying this disorder, but this, however, does not cover the overall inter-/intra-individual variability of PD phenotypes. The present review is aimed at discussing to what extent the evidence provided by group-based neuroimaging analysis in this field of study (using seed-based, network-based, or graph theory approaches) may be generalized. In particular, we summarized the literature on the application of resting-state functional connectivity studies to explore different neural correlates of motor and non-motor symptoms of PD and the neural mechanisms involved in treatment effects: effects of levodopa or deep brain stimulation. The lesson learnt from one decade of studies provides consistent evidence on the role of the altered communication between the striato-frontal pathways as a marker of PD-related motor degeneration, whereas in the non-motor domain, several missing pieces of a complex puzzle are provided. However, the main target is to present a new era of intelligent neuroimaging applications, where automated multivariate analysis of functional connectivity data may be used for moving from group-level statistical results to personalized predictions in a clinical setting. Although in its relative infancy, the evidence gathered so far suggests a new era of clinical neuroimaging is starting.
Keywords: Functional magnetic resonance imaging; Machine learning; Parkinson’s disease; Resting-state functional connectivity; Seed-based approach.