By linking MRI anatomy with EEG dynamics, the FEDE model offers a detailed new way to study how brain structure and neural activity interact in autism, while highlighting the need for larger validation studies.
In a recent study published in the journal PLOS Digital Health, researchers developed and validated the high FidElity Digital brain modEl (FEDE) system for creating patient-specific virtual brain models, or digital twins, designed to replicate brain structure and biophysical activity. They used specialized magnetic resonance imaging (MRI) scans to construct the digital twin and then applied it to study brain activity in a young child with an autism spectrum disorder (ASD) diagnosis. The model demonstrated robust performance in replicating brain activity patterns and estimated possible patient-specific alterations in signal transmission through synapses. If validated in larger studies, such models could support future precision medicine approaches to investigate brain disorders and evaluate therapeutic strategies.
The structure of the brain plays a crucial role in shaping signal transmission pathways through different regions. Existing models have not fully replicated the brain’s anatomical and functional characteristics. Scientists are exploring ways to integrate imaging data and computational modeling into a single framework to precisely replicate brain structure and neural activity for research and clinical applications. By enabling large numbers of virtual experiments, such models could help scientists uncover the biophysical and network-level mechanisms underlying complex conditions such as ASD.
About the study
In the present study, researchers used the FEDE approach to develop an interactive digital twin of a young child’s brain with ASD. They used three different types of MRI scans to reconstruct the brain’s anatomical features. The protocol included T1-weighted, T2-weighted, and diffusion-weighted imaging (DWI) sequences. The researchers then simulated brain activity using virtual electrodes placed on the scalp surface. They compared the results with EEG recordings obtained from the ASD patient (age, 2.4 years) and information obtained using standard models to evaluate the reliability and robustness of the findings.
The FEDE pipeline uses the finite-element method (FEM) to combine brain anatomical connections reproduced from medical images with biophysical recordings of brain activity in one integrated framework. The model reconstructed the connection networks of brain cortices, the myelination around nerve fibers, and the conductance properties of different tissues.
The method incorporates parameter-optimization methods to recreate multiscale structural features. The team optimized the parameters directly on the activity of a highly dense cortical mesh (20,484 vertices) for high-resolution reconstruction. They also divided brain regions using a standard atlas (HCPMMP1) to study connections between different parts of the brain. The team additionally measured the lengths of the nerve fiber pathways connecting different regions. They then created detailed maps to determine how quickly nerve signals are likely to travel through different parts of the brain and to identify potential delays in ASD.
The researchers used linear regression models to determine whether the FEDE results matched the EEG recordings obtained from the toddler. They also used a machine learning approach to identify the most important model parameters for reproducing the EEG findings.
Results
The FEDE approach reconstructed brain structure with high spatial resolution and reproduced selected EEG-derived features of brain activity. The simulated findings correlated well with the EEG data. In fact, the model could identify potential alterations in nerve cell transmission inferred by the model, consistent with biological changes observed in ASD. The model suggested possible abnormalities at multiple levels of brain organization in the toddler, including altered communication between brain cells, developmentally relevant features such as myelination, and changes in connections within and between brain regions. Because these findings come from one patient, the ASD-related interpretations should be treated as hypotheses rather than generalizable disease markers.
The FEDE method predicted shorter signal transmission delays than standard models, suggesting that conventional approaches often overestimate the time required for brain signals to travel between regions. This is because standard models do not account for myelination, the insulating covering around nerve fibers that helps electrical signals move more quickly.
How to make hybridoma-based antibody discovery more efficient eBook Sphere Bio showcases solutions to boost hybridoma-based antibody discovery.Download the latest editionThe accuracy of the FEDE model depended mainly on how detailed the brain simulations were and how electrical signals were modeled, rather than the speed of signal transmission. To match the FEDE results with the EEG data, the researchers primarily adjusted background noise levels and the excitatory-to-inhibitory (EI) ratio. The optimal noise level was about 100 times higher than the standard model value, suggesting greater fluctuations in neural activity in ASD. The EI ratio was about three times higher than expected in a healthy brain, indicating an imbalance between neural signals that increase and those that suppress brain activity.
Simultaneous reconstruction of brain anatomy and dynamics from neural data: The FEDE pipeline.
(A): MRI recordings, including (from top to bottom) T1-w, DWI and T2-w, were used to generate a 3D replica of the patient brain. (B): MRI processing steps, from left to right: anatomical constrained tractography analysis, brain regions parcellation and segmentation with reconstruction of cortical surface, reconstruction of biophysical finite element method (FEM) model of the patient’s head, including 3D scalp model with EEG electrodes. (C): Left: the parcellation of brain areas defined connective weights, distances and conduction velocity maps for the whole brain structure, which were integrated in a virtual brain model (top, see Methods). Right: neural activity was computed on the high-density cortical surface, the activity was then projected to the scalp of the patient with an anatomically-accurate lead-field matrix, leveraging the FEM model of the patient’s head and considering anisotropy of brain tissues. This allowed for the precise computation of EEG signal from simulated brain activity. (D): EEG recordings were acquired during resting-state, extracting features such as power spectral distribution and functional connectivity. (E): Parameter optimization through the comparison between experimental and simulated EEG led to the identification of multi-scale structural features underlying patient’s condition.
Conclusions
Based on the findings, the FEDE approach represents a significant advance over conventional methods for modeling the structure and function of the brain within a single framework. If validated in larger studies including more diverse populations of healthy and ASD patients across ages, the FEDE pipeline could be used to create personalized digital twins for various brain diseases and to support research, treatment evaluation, and the development of more individualized therapeutic strategies. Such models could be especially helpful to clarify complex conditions like ASD in toddlers who have rapidly changing brain systems, which may be difficult to image without motion artifacts, and cannot be subjected to invasive procedures due to ethical concerns.
However, the findings should be interpreted cautiously because the study was conducted in a single toddler with ASD, without a control group or additional patients. The results demonstrate the feasibility of creating a high-fidelity digital brain twin and generating plausible, model-based hypotheses about ASD-related neural dynamics, but they do not yet show that FEDE can diagnose ASD, guide treatment, or identify definitive biological abnormalities.
