Simulated brain networks reflecting severity of Parkinson's disease

1718 

Abstract Submission 

Kyesam Jung^1,2, Simon Eickhoff^1,2, Julian Caspers³, Oleksandr Popovych^1,2

¹Institute of Neuroscience and Medicine - Brain and Behaviour (INM-7), Forschungszentrum Jülich, Jülich, Germany, ²Institute of Systems Neuroscience, Medical Faculty, Heinrich Heine University, Düsseldorf, Germany, ³Department of Diagnostic and Interventional Radiology, Medical Faculty, Heinrich Heine University, Düsseldorf, Germany

Kyesam Jung  
Institute of Neuroscience and Medicine - Brain and Behaviour (INM-7), Forschungszentrum Jülich|Institute of Systems Neuroscience, Medical Faculty, Heinrich Heine University
Jülich, Germany|Düsseldorf, Germany

Simon Eickhoff  
Institute of Neuroscience and Medicine - Brain and Behaviour (INM-7), Forschungszentrum Jülich|Institute of Systems Neuroscience, Medical Faculty, Heinrich Heine University
Jülich, Germany|Düsseldorf, Germany

Julian Caspers  
Department of Diagnostic and Interventional Radiology, Medical Faculty, Heinrich Heine University
Düsseldorf, Germany

Oleksandr Popovych  
Institute of Neuroscience and Medicine - Brain and Behaviour (INM-7), Forschungszentrum Jülich|Institute of Systems Neuroscience, Medical Faculty, Heinrich Heine University
Jülich, Germany|Düsseldorf, Germany

Brain connectome of patients with Parkinson's disease (PD) showed different network properties compared to healthy participants (Zuo et al. 2022), which is supposed to develop in time with PD progression and motor impairment (Holden et al. 2018). Empirical investigation of such changes of the brain networks is mostly restricted to longitudinal data, which limits the evaluation of the disease effects on the network properties of the human brains in vivo. Here we suggest an alternative approach and consider the resting-state brain dynamics simulated by whole-brain dynamical models, which allows us to probe network changes using brains in silico. Accordingly, we performed whole-brain simulations and investigated network modularity (segregation) and global efficiency (integration) of the simulated brain networks for behavioral model fitting to clinical measures. As a result, we demonstrate significant enhancements of correlation between simulated network properties and disease severity, which was not observed in empirical cross-sectional neuroimaging PD data used for model derivation.

This study included 51 (30 males) healthy controls and 60 (43 males) PD patients. Empirical functional connectivity (eFC) and empirical structural connectivity (eSC) were calculated using the resting-state blood oxygenation level-dependent (BOLD) signals and streamlines from whole-brain tractography, respectively. We used two parcellation schemes based on functional (Schaefer et al. 2018) and structural (Desikan et al. 2006) brain properties. We performed whole-brain simulations using a network of the Jansen-Rit type models (Jansen and Rit 1995) informed by eSC and the model parameters on a dense 64 coupling × 43 delay parameter grid. The obtained simulated BOLD signals were used to calculate the simulated functional connectivity (sFC). Subsequently, we calculated network properties (modularity and efficiency) of all connectivities (1 eFC, 1 eSC, and 2752 sFCs for each subject and parcellation). We applied a behavioral model fitting (optimizing the whole-brain model to fit to target variables) for group comparison between healthy controls and PD patients and correlation with disease severity based on the unified PD rating scale (UPDRS) (Goetz et al. 2008). We searched for the optimal parameter points corresponding to the largest effect size of the network properties of sFC between healthy controls and PD patients as well as the optimal parameter points corresponding to the strongest correlation between the simulated network properties and the disease severity from 60 patients.

The applied model fitting for the network properties of sFC showed significant group differences between healthy controls and PD patients in network modularity and efficiency (Fig. 1a-b), which was not the case for the empirical data. The largest effect size was 0.19 for eFC and 0.12 for eSC as compared with the largest effect size of 0.38 for sFC. Furthermore, the behavioral network-based model fitting showed that sFC network properties significantly correlate with the disease severity (Fig. 1c-e and Fig. 2e-g), which was again not significant for the empirical connectomes (Fig. 2a-d). Remarkably, efficiency of sFC exhibited two local optima of negative and positive correlations with disease severity (Fig. 1d-e) in contrast to weak and non-significant correlation with the network properties of empirical connectomes (Fig. 2b, d).

Our results indicate that functional segregation (modularity) and integration (efficiency) of simulated brain networks can exhibit an enhanced group difference between healthy subjects and PD patients and significant correlations with severity of motor impairment in patients with PD as compared with empirical data. With this, we suggest using the network-based model fitting of the whole-brain dynamical models to find dynamical regimes closely related to clinical measures of PD for further investigations of disease onset and progression.

Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) ²

Connectivity (eg. functional, effective, structural)

fMRI Connectivity and Network Modeling ¹

Computational Neuroscience

Modeling

Movement Disorder

Other - Whole-brain simulation