image: A general disease-oriented BNM pipeline is illustrated in this figure. Specifically, diffusion MRI data is processed to obtain individual large-scale connectome with empirical brain SC (top row). Empirical FC is derived from statistical analyses of time-series data from MEG, EEG, or fMRI (bottom left). The BNM is constructed by coupling neural mass models with SC data and supplies the computational neuronal activity for each brain region (bottom right). Then, simulated signals are compared with empirical functional data to estimate the optimal working points and regimes of a virtual brain (bottom centre).
Credit: Harbin Institute of Technology, Shenzhen
A research team from the International Research Institute for Artificial Intelligence at Harbin Institute of Technology, Shenzhen, has recently published a comprehensive review in the journal Health Data Science on the application of Brain Network Models (BNMs) in the medical field. This study summarizes recent advances and challenges in using BNMs to simulate brain activities, understand neuropathological mechanisms, evaluate therapeutic effects, and predict disease progression.
Brain Network Models (BNMs) are mathematical modeling tools based on neural networks that integrate structural connectivity (SC) and functional connectivity (FC) data to simulate dynamic changes in the brain under various neurological conditions. With advancements in neuroimaging techniques, BNMs have become crucial in studying underlying mechanisms for neurological disorders such as epilepsy, Alzheimer’s disease (AD), and Parkinson’s disease (PD).
Led by Assistant Professor Chenfei Ye from the International Research Institute for Artificial Intelligence at Harbin Institute of Technology, Shenzhen, the team reviewed the current applications of BNMs in medicine. The review highlights how BNMs integrate multimodal neuroimaging data to simulate overall brain dynamics and proposes improvements such as adopting multimodal data fusion strategies to enhance the accuracy of models in representing the brain's complex functional architecture.
The team developed a disease-oriented BNM workflow that demonstrates how to extract an individual’s brain structural connectome (SC) from structural and diffusion-weighted MRI data and derive functional connectivity (FC) through statistical analyses of data from MEG, EEG, or fMRI. Then, by coupling local Neural Mass Models (NMMs) with structural connectivity data, a global BNM is constructed to simulate large-scale brain activities.
The key value of BNMs lies in their ability to quantitatively analyze abnormal network dynamics of the brain under different disease states, offering new possibilities for personalized treatment planning. The study suggests that future BNM development should focus more on individual differences and the integration of multimodal data to achieve more precise disease diagnosis and therapeutic strategies.
The research team indicates that future work will focus on developing new BNM models capable of estimating a broader range of neurodynamic parameters, such as the distribution of presynaptic inputs, frequency-dependent synaptic depression, and intrinsic excitability of postsynaptic neurons. The ultimate goal is to apply these advanced modeling techniques in clinical practice to optimize treatment strategies.
Journal
Health Data Science
Article Title
Recent Progress in Brain Network Models for Medical Applications: A Review
Article Publication Date
8-Jul-2024