Network-based Approach for Accurate Brain Modeling
More than a billion people worldwide suffer from some neurological disorders, such as Alzheimer's, Parkinson's, multiple sclerosis, epilepsy, stroke, autism, addiction, and many others. These conditions result in more than 7 million deaths, with associated costs exceeding $1.5 trillion annually. Most neurological diseases cannot be cured, and treatment is limited to medicines that delay or halt disease progression.
A primary challenge with developing new treatment options for brain conditions is the lack of fundamental understanding of how it functions. While neurological data can be captured by many advanced systems, tools for analyzing and interpreting data are severely limited. Most tools consider the neurological system as binary, whereas it’s highly complex and consists of many functional networks.
This technology describes an iteration of Generalized Exponential Random Graph Modeling (GERGM) to analyze and characterize brain networks. It uses a hierarchical Bayesian framework in conjunction with GERM to quantify brain networks over time, thereby testing for temporal dynamics and task-dependent changes. The system uses various imaging data to generate a realistic model of the brain and its function.
This invention provides for neural mapping and diagnostic information for patient specific maladies. It can be used as both a research tool to understand brain function and in the clinical setting to elucidate brain function in different neurological conditions. Further, it can be used as a drug development tool by mapping the effects of compounds on neural pathways and therefore improving drug development.
This invention improves analysis and modeling of neurological data and disease compared to existing binary-based approaches that are inaccurate and do not translate to natural human brain function. This technology only requires a small number of parameters to model the neural network, models it more precisely to capture abnormal patterns, and can differentiate between normal and abnormal cognitive activity. The software can use image data from a variety of sources including fMRI, MRI, EEG, PET infrastructure and takes less than 30 minutes to obtain a diagnosis.