VLSI systems for neurocomputing and health informatics

Ubiquitous access to computers, cell phones, internet, personal digital devices, cameras and TV can be attributed to advances in the very large scale integration (VLSI) technology and the advances in circuit design to operate circuits at Gigahertz rates. One of the mysteries that we have not been able to unravel is the understanding of how the brain works from different perspectives. Reverse engineering the brain has been identified as one of the grand challenge problems by the National Academies. Advances in sensor technologies and imaging modalities such as electroencephalogram (EEG), intra-cranial electroencephalogram (iEEG), magnetoencephalogram (MEG), and magnetic resonance imaging (MRI) allow us to collect data from hundreds of electrodes from the brain at sample rates ranging from 256 Hz to 15kHz. These data can be key to not only understanding brain functioning and brain connectivity at macro and micro levels in healthy subjects but also in identifying patients with neurological and mental disorder. Extracting the appropriate biomarkers using spectral-temporal-spatial signal processing approaches and classifying states using machine learning approaches can assist clinicians in predicting and detecting seizures in epileptic patients, and in identifying patients with mental disorder such as schizophrenia, depression and personality disorder. The biomarkers can be tracked to design personalized therapy and effectiveness of therapy by closed loop drug delivery or closed loop neuromodulation, i.e., brain stimulation either by invasive or non-invasive means using electrical or magnetic stimulation. High-performance VLSI system design is critical to not-only increasing battery life of VLSI chips for neuromodulation but also for reducing computation time by orders of magnitude in analyzing MRI signals. Another grand challenge problem identified by the National Academies is Advanced Health Informatics. Analysis of health data is key to monitoring biomarkers and delivering drugs as needed. VLSI system design of biomarkers and disease state classification is again critical in improving the health and quality of life of human beings.

In this talk, I will highlight the emerging opportunities in high-performance low-power VLSI system design for neurocomputing and health informatics at various scales. At macroscale, the goal is to design small low-power implantable or wearable devices that can be used to monitor biomarkers and trigger an alarm signal to alert an abnormal state of the brain such as an impending seizure. At microscale, extracting thousands of connections from structural and functional MRI can require many hours or even a day for one subject and one set of parameters using parallel computers. The challenge here is to design parallel multicore computer architectures and compiler tools that can reduce the time for microscale analysis of MRI to an hour or less. I will describe research in my group in use of signal processing and machine learning approaches to identify and track various neurological and mental disorders. I will present some results on VLSI design of feature extractors such as power spectral density (PSD) and classifiers such as support vector machines (SVMs). I will present diabetic retinopathy screening using fundus image analysis and machine learning as an example to illustrate opportunities in design of embedded systems for health informatics. Significant research needs to be pursued in this area. My presentation will hopefully inspire further research in this emerging and important field embedded VLSI system design for neuro, bio and health informatics.