The University at Buffalo's Christopher McNorgan has developed a computer model of the human brain that more accurately simulates brain-impairment patterns than previous techniques, by combining functional connectivity and multivariate pattern analyses (MVPA).
Functional connectivity simulates the brain's connections in a linear fashion from magnetic resonance imaging (MRI) scans, while MVPA is a non-linear "teachable" machine learning method that operates on a more holistic level to assess how activity is patterned across brain areas. When combined, these techniques overcome their limitations.
McNorgan collected data on MRI patterns tied to specific brain activities, then digitized the patterns to train pattern-recognition models.
The addition of "virtual lesions" to the models determined that the mutually constrained models exhibited classification errors consistent with lesion sites.
From UB News Center
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA