NeuroImage: brain model reveals the mechanism behind stroke and injury

January 17, 2020 / BIOON / -- a neuroimaging researcher at the University of Buffalo has developed a computer model of the human brain, which more realistically simulates the actual brain damage pattern than the existing methods This advancement represents a combination of two approaches to building a digital analog environment that can serve as a test ground for specific neural injury hypotheses, helping stroke victims and other brain injury patients Christopher mcnogan, assistant professor of psychology at the UB School of art and science, said: "this model is precisely linked to the functional connections of the brain and can prove the real pattern of cognitive impairment Because the model reflects the way the brain is connected, we can manipulate it in a way that provides insight For example, in-depth study of brain regions of patients who may be damaged These findings provide powerful means to identify and understand brain networks and their functions, which may lead to the possibility of discovery and understanding that once could not be realized The results were published in the journal neuroimage To explain mcnorgan's model, we first need to study two basic components of its design: functional connectivity and multi pattern analysis (MVPA) For many years, the traditional brain based model has been dependent on the general linear method This method looks at every part of the brain and how it responds to stimulation This method is used in the traditional study of functional connectivity, which relies on functional magnetic resonance imaging (fMRI) to explore the way of brain connectivity The linear model assumes that there is a direct relationship between two things Although linear models are good at identifying which regions are active under certain conditions, they are usually unable to detect the potential complex relationship between multiple regions MVPA, a "teachable" machine learning technology, can be operated at a more comprehensive level to assess the activity patterns of various regions of the brain MVPA is nonlinear For example, suppose there is a group of neurons dedicated to identifying the meaning of a stop sign When we see red or octagons, these neurons are inactive because there is no one-to-one linear mapping between red and stop symbols, or between octagons and becoming octagons "The nonlinear response ensures that when we see objects that are both red and octagonal, they really light up," mcnorgan explains Therefore, nonlinear methods such as MVPA have always been the core of so-called "deep learning" methods " In itself, the traditional methods of functional connectivity and MVPA have limitations, and integrating the results of each method requires brain researchers to make a lot of efforts and expertise to find out the evidence However, if the two are combined, the limitations will restrict each other Mcnorgan was the first researcher to successfully integrate functional connectivity and MVPA to develop a machine learning model, which is explicitly rooted in the real functional connectivity between brain regions In other words, the result of mutual restriction is a problem of self-assembly "My professional experience has enabled me to use a wide range of different theoretical models This background provides a special set of experiences that makes this combination apparent in hindsight " Sources of information: brain model offers new insights into damage caused by stroke and other interventions: Chris mcnorgan, Gregory J Smith, Erica S Edwards, integrating functional connectivity and MVPA through a multiple constraint network analysis Neuroimage, 2020; 208: 116412 doi: 10.1016/j.neuroimage.2019.116412