[codethief]

> To simulate the mechanisms of the human brain, Ritter and her team use digital data from brain scans like magnetic resonance imaging (MRI) as well as mathematical models based on theoretical knowledge about biological processes. This initially results in a “general” human brain model. The scientists then refine this model using data from individual people, thus creating “personalized brain models.” […] “We can reproduce the activity of individual brains very efficiently,” says Ritter. “We found out in the process that these in silico brains behave differently from one another – and in the same way as their biological counterparts. Our virtual avatars match the intellectual performance and reaction times of their biological analogues.”

Huh, I hadn't heard of this before. Does anyone know more about how exactly those "in silico" brains work and how they compare to their real-world counterparts? I mean, the article makes it sound as if the researchers fully understood how the brain works and had managed to create a faithful digital copy, which I find difficult to believe.

EDIT: The original paper says

> To study neuronal processing in silico we created BNMs [brain network models] for the 650 subjects using a tuning algorithm that fits each participant’s simulated FC with their empirical FC (Figs. 2 and 3). The BNMs use coupled neural mass models to simulate the electric, synaptic, firing, and hemodynamic (fMRI) activity of a 379-nodes whole-brain network. Each node consists of one excitatory and one inhibitory population that mutually and recurrently interact. To simulate long-range white matter coupling, the neural masses were connected by each participant’s SC, which were estimated by dwMRI tractography. Importantly, we added feedforward inhibition to increase biological realism

Sounds like they used small neural networks for simulation and adjusted the weights between the neurons to what they saw in participants' MRI measurements.