Multimodal metabolic and functional network signatures for diagnosis of parkinson's disease: a PET/MR study.

Parkinson’s disease (PD) is a neurodegenerative disorder linked to altered cerebral glucose metabolism detectable via 18F-FDG-PET. This study integrated PET/MR and Jensen-Shannon similarity estimation (JSSE) to investigate metabolic/functional network dynamics in PD. 25 PD patients and 16 age/sex-matched normal controls (NCs) underwent visual analysis, standardized uptake value ratio (SUVR) comparisons, and JSSE-based metabolic/functional network construction. Topological properties, regional connectivity, and metabolic-functional correlations were analyzed. A support vector machine (SVM) model incorporating these features differentiated PD from NC. PD patients exhibited elevated SUVR in temporal and thalamic regions but reduced SUVR in parietal/precentral areas. Metabolic networks in PD showed increased assortativity yet decreased normalized clustering coefficients and disrupted small-worldness. Ten metabolic connections were impaired, with the left caudate-temporal pathway showing the strongest reduction. PD demonstrated significantly higher metabolic-functional network correlations than NCs. Machine learning identified metabolic network connectivity as the optimal classifier (AUC = 0.91). Our findings reveal PD-associated metabolic reorganization and validate PET/MR-derived network features as potential biomarkers. The multimodal framework advances pathophysiological understanding through network-level metabolic-functional interplay analysis, offering novel insights into PD diagnostics and mechanistic research.