Network-based diffusion analysis and functional connectivity in modelling atrophy progression for early detection of GBA-associated Parkinson’s disease

Parkinson’s disease (PD) is a prion-like neurodegenerative disorder driven by α-synuclein misfolding, spreading pathology across the brain, especially in cases with GBA mutations. Advances in MRI processing, such as graph analysis and connectomics, help map brain-wide functional connectivity and clarify neurodegenerative mechanisms. Network analysis reveals how diseases alter brain organization and hypothesizes abnormal protein spread. The Aggregation Network Diffusion (AND) model predicts future pathology based on baseline data, simulating misfolded protein spread and aggregation into neurotoxic forms. To account for individual variability in disease progression, AND must adapt to subject-specific dynamics for accurate atrophy modelling. Thirteen GBA-positive, 39 GBA-negative PD patients and 60 age- and sex-matched controls underwent clinical evaluation, 3DT1-weighted and resting-state functional MRI (rs-fMRI) at baseline and, only PD patients, over a 7-year period. Functional connectome for each subject was obtained from rs-fMRI scans as the Pearson’s correlation coefficient between time-series in 83 cortical and subcortical brain regions. Graph analysis and connectomics assessed global and local functional topologic network, and regional functional connectivity using Network Based Statistics (NBS). AND model was developed to predict PD pathology progression as a spreading process from a seed through an averaged functional connectome extracted from healthy controls. Pearson’s correlations were calculated between AND-predicted atrophy and the observed longitudinal atrophy in PD patients over follow-ups of 7-years. Through the use of grid searches and optimization methods, the model was tailored to each individual, maximizing correlation values by refining model parameters, and selecting seed regions. Relative to controls, GBA-positive PD patients showed severe global functional network alterations, while GBA-negative patients showed relatively preserved functional brain architecture. Considering NBS analysis, functional connectivity breakdown in the parietal lobe, temporal, and occipital, sensorimotor, and right frontal differentiated GBA-positive PD patients from controls. AND model produced highly variable results across both groups. Seed region identification showed overlapping patterns across both groups. In GBA-negative patients, the model primarily selected regions in the temporal lobe, with secondary preference for the frontal lobe. Similarly, in GBA-positive patients, regions were mainly located in the temporal lobe, with additional preference for the basal ganglia and sensorimotor lobe. Parameter optimization enabled detailed analysis of simulated atrophy patterns, with the global network "transmissibility" constant, beta, emerging as the key factor. Lower beta values, with a seed region in the temporal lobe, led to early involvement of the basal ganglia, frontal, and temporal lobes at 24 months, followed by progression to the parietal, sensorimotor, and occipital lobes at later stages. In contrast, higher beta values, starting from the same lobe, resulted in early involvement of the basal ganglia, temporal, frontal, parietal, and sensorimotor lobes, with later progression to the occipital lobe. These findings show that GBA mutations drive distinct functional network changes in PD. Patient-specific spreading patterns confirmed the need for individualized model fitting. Integrating predictive modeling with connectivity analysis offers new insights into PD progression.