Illustration of a directed brain network.
by Barry Van Veen
One of the scientific frontiers of our time is understanding the human brain. The U.S. National Academy of Engineering declared reverse engineering the brain to be an engineering grand challenge. U.S. President Barak Obama recently inaugurated the BRAIN initiative to "revolutionize our understanding of the human mind." Similar initiatives are being pursued by governments and research institutes around the world.
Signal Processing and Brain Research
The role of signal processing in unraveling the secrets of the human brain is worthy of a long series of articles. In summary, signal processing is foundational for any use of imaging technology - EEG, MEG, MRI, fMRI, PET, etc. Hypotheses about brain function are tested by analysis of data. Signal processing is essential for filtering, artifact rejection, estimation of models, and development of methods for hypotheses testing. The datasets are often very large, involving hundreds of subjects evaluated under multiple conditions and many thousands of voxels and/or time samples.
Brain Network Models and High Throughput Computing
Yesterday (October 17) I participated in a panel discussion on Big Data and the Brain at the Wisconsin Science Festival, held at the Wisconsin Institute for Discovery. The topic was on how high-throughput computing is transforming brain research. My fellow panelists included Barbara Bendlin from the UW Department of Medicine and Mike Koenigs from the UW Psychiatry Department. Prof. Bendlin discussed her use of high throughput computing to study changes in neural connectivity associated with Alzheimer's disease across large populations of subjects. Prof. Koenigs reported on his use of high-throughput computing to study of the functional connectivity in inmates with psychopathy. I shared on my research developing signal-processing algorithms for identifying directed network models of electrical brain activity and how high-throughput computing enables these algorithms to be implemented in practical timeframes. We then entertained questions from the audience. I captured a screencast of my presentation during a practice run to share here, and also included a link to the entire panel presentation. It was a very broad audience, so mine is a high-level talk with little detail about algorithms. I expect to share more about the role of signal processing in brain research in the future.