Search

Bernstein Conference 2022 abstract. http://bernstein-conference.de

Bernstein Conference 2022 abstract. http://bernstein-conference.de

Large-scale neurophysiological networks are often reconstructed from band-pass filtered time series derived from magnetoencephalography (MEG) data. Common practice is to reconstruct these networks separately for different frequency bands and to treat them independently. Recent evidence suggests that this separation may be inadequate, as there can be significant coupling between frequency bands (interlayer connectivity). A multilayer network approach offers a solution to analyze frequency-specific networks in one framework. We propose to use a recently developed network reconstruction method in conjunction with phase oscillator models to estimate interlayer connectivity that optimally fits the empirical data. This approach determines interlayer connectivity based on observed frequency-specific time series of the phase and a connectome derived from diffusion weighted imaging. The performance of this interlayer reconstruction method was evaluated in-silico. Our reconstruction of the underlying interlayer connectivity agreed to very high degree with the ground truth. Subsequently, we applied our method to empirical resting-state MEG data obtained from healthy subjects and reconstructed two-layered networks consisting of either alpha-to-beta or theta-to-gamma band connectivity. Our analysis revealed that interlayer connectivity is dominated by a multiplex structure, i.e. by one-to-one interactions for both alpha-to-beta band and theta-to-gamma band networks. For theta-gamma band networks, we also found a plenitude of interlayer connections between distant nodes, though weaker connectivity relative to the one-to-one connections. Our work is an stepping stone towards the identification of interdependencies across frequency-specific networks. Our results lay the ground for the use of the promising multilayer framework in this field with more-informed and justified interlayer connections.

International audience; Despite the acceleration of knowledge and data accumulation in neuroscience over the last years, the highly prevalent neurodegenerative disease of AD remains a growing problem. Alzheimer's Disease (AD) is the most common cause of dementia and represents the most prevalent neurodegenerative disease. For AD, disease-modifying treatments are presently lacking, and the understanding of disease mechanisms continues to be incomplete. In the present review, we discuss candidate contributing factors leading to AD, and evaluate novel computational brain simulation methods to further disentangle their potential roles. We first present an overview of existing computational models for AD that aim to provide a mechanistic understanding of the disease. Next, we outline the potential to link molecular aspects of neurodegeneration in AD with large-scale brain network modeling using The Virtual Brain ( www.thevirtualbrain.org ), an open-source, multiscale, whole-brain simulation neuroinformatics platform. Finally, we discuss how this methodological approach may contribute to the understanding, improved diagnostics, and treatment optimization of AD.

Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease and is characterized by upper and lower motor neuron involvement. ALS is a heterogeneous disease in terms of disease onset and progression, making it difficult to determine the cause of the disease. Identifying subgroups within the ALS population and elucidating their corresponding neuroimaging patterns may help as a biomarker to monitor disease progression and possible treatment effects in clinical trials. This thesis combines different perspectives of neuroimaging, network science and machine learning to shine light on disease patterns of gray and white matter involvement in the brain and spinal cord of patients within the ALS population. Patients with ALS, divided into subgroups based on clinical or genetic characteristics, display various degrees of gray and white matter involvement. From these results, four different MRI-detectable brain patterns could be distinguished over time. White matter degeneration seems to occur early in the disease process, followed by gray matter involvement over time. For patients with a C9orf72 gene mutation, the opposite might be true as these patients display thinner cortex throughout the brain at baseline. Not only patients but also asymptomatic family members with this mutation show changes in the brain compared to family members without the mutation. The relative cortical thinning is associated with the gene expression of C9orf72, which suggests a regional vulnerability of brain areas to become affected by ALS. Besides brain alterations, ALS also affects the motor neurons in the spinal cord. Thinning of the upper cervical spinal cord does not occur only in patients with ALS, but throughout the motor neuron disease spectrum. Longitudinal cervical spinal cord measurements were progressive over time for ALS and showed a significant relationship with disease severity. Therefore, neuroimaging of the spinal cord may possibly be an additional measure for diagnosis or disease progression outside the brain. Simulating disease spread can provide insight in potential disease mechanisms. By applying a random walker model on the white matter network, disease spread according to histologically defined neuropathology staging was simulated. Simulated aggregation levels of pTDP-43 proteins correlated with empirical impairment found at follow-up, both for the total group and individual patients. Hence, early-stage white matter alterations define subsequent pathology. Characterization of network metrics has the potential to elucidate affected organization in the connectome (i.e. white matter network), reflecting patterns of neurodegeneration spread across the white matter tracts in the brain. The findings showed that the nodal network metrics based on shortest paths in the network are important network measures in the understanding of disease effects. Survival classes of patients may be predicted by using supervised deep learning techniques on clinical and neuroimaging variables. The combination of modalities improves the prediction accuracy from approximately 60.0% to at least 83.3%, illustrating the added value of neuroimaging to survival prediction. The results of deep learning show its potential in disease prognostication and, especially in the era of big data, pave the way in the development of an automated prognostication tool for the estimation of survival in individual patients.