Date: Thursday 12 March 2015
Time: 16:00 - 17:30
Location / Room: Bayard

Organisers:
Luca Daniel, MIT, US
Luis Miguel Silveira, INESC-ID, PT

Chair:
Luca Daniel, MIT, US

Co-Chair:
Luis Miguel Silveira, INESC-ID, PT

Tools and techniques originally developed by the Electronic Design Automation community for parasitic extraction, model reduction, or circuit simulation are having deep impact in alternative and exiting fields outside of the circuit world. In particular, this session shows several applications of such techniques to analyzing the functionality of the brain and of the nervous system, as well aiding the design of biomedical and medical instrumentation and diagnostics.

Time	Label	Presentation Title Authors
16:00	12.6.1	THE OLD, THE NEW, AND THE RECYCLED - EDA ALGORITHMS IN CONNECTOMIC Speaker: Lou Scheffer, Howard Hughes Medical Institute, US Abstract Connectomics seeks to extract detailed wiring diagrams of circuits of the nervous system. This makes it a combination of reverse engineering (as applied to chips), parasitic extraction, and model reduction. As biologists extract and work with the larger neural circuits that can now be extracted, they are running into many of the same problems that EDA faced long ago. This talk compares and contrasts connectomics with the equivalent processes for chips, notes the differences and similarities, and shows where algorithms developed for EDA can help connectomics.
16:30	12.6.2	COMPUTATIONAL MODELING AND SIMULATION OF SYNCHRONIZED FIRING BEHAVIORS OF THE BRAIN Speaker: Peng Li, Texas A&M, US Abstract Computational simulation is a critical enabler for understanding complex functions and neuronal dynamics of mammalian brains. However, several grant challenges, such as retaining biological realism in computer-based models, obtaining and managing a vast amount of biological data, and tackling high computational complexity, exist. Nevertheless, efficient computational techniques, capable of simulating large neural networks with biophysically accurate neuron models, are highly desirable. Such capability will fundamentally enable the test of hypotheses of neurological disorders and development of therapeutic treatments, as well as stimulate new engineering applications. In this talk, we will show how neuronal models of different complexities (behavioral oscillator models vs. Hodgkin-Huxley models) and global connectivity data may be leveraged to reason about the origins of oscillatory behaviors of the brain. The key focus of the talk will be placed on a large-scale biophysically detailed thalamocortical model and parallel numerical techniques that have been developed to efficiently handle widely spread time scales in the network. Our results suggest that computational techniques may shed light on the causes of absence seizures by associating abnormal brain level oscillation with several key cellular level mechanisms.
17:00	12.6.3	ELECTROMAGNETIC POWER DEPOSITION ANALYSIS TOOL FOR HIGH RESOLUTION MAGNETIC RESONANCE IMAGING BRAIN SCANS Speakers: Jorge F. Villena1, Athanasios G. Polimeridis1, Lawrence L. Wald2, Elfar Adalsteinsson1, Jakob K. White1 and Luca Daniel1 1Massachusetts Institute of Technology, US; 2Massachusetts General Hospital, Harvard Medical School, US Abstract MARIE (MAgnetic Resonance Integral Equation suite) is an open domain numerical software platform for fast electromagnetic (EM) analysis and design of Magnetic Resonance Imagine (MRI) scanners. The tool is based on a combination of surface and volume integral equation formulations. It exploits the characteristics of the different parts of an MRI system (coil array, shield and realistic body model), and it applies sophisticated numerical methods to rapidly perform all the required EM simulations to characterize the MRI design: computing the un-tuned coil port parameters; obtaining the current distribution for the tuned coils, and the corresponding electromagnetic field distribution in the inhomogeneous body for each transmit channel. The software runs on MATLAB and is able to solve a complex scattering problem in ~2-3 min. on a standard single GPU-accelerated windows desktop machine. On the same platform it can perform a frequency sweep of a complex coil in ~3-5 min. per frequency point. Furthermore, it can solve the complete inhomogeneous body and coil system in ~5-10 min. per port, depending on the model resolution and error tolerance required. The software could potentially be employed also on more advanced analyses, such as the generation of ultimate intrinsic Signal to Noise Ration (SNR) and Specific Absorption Rare (SAR) on realistic body models, fast coil design and optimization, and generation of patient specific protocols.
17:30		End of session