Martin J. Cole

Subject-specific, multiscale simulation of electrophysiology: a software pipeline for image-based models and application examples

Many simulation studies in biomedicine are based on a similar sequence of processing steps, starting from images and running through geometric model generation, assignment of tissue properties, numerical simulation and visualization of the results—a process known as image-based geometric modelling and simulation. We present an overview of software systems for implementing such a sequence both within highly integrated problem-solving environments and in the form of loosely integrated pipelines. Loose integration in this case indicates that individual programs function largely independently but communicate through files of a common format and support simple scripting, so as to automate multiple executions wherever possible. We then describe three specific applications of such pipelines to translational biomedical research in electrophysiology.

A Note on Data-Driven Contaminant Simulation

In this paper we introduce a numerical procedure for performing dynamic data driven simulations (DDDAS). The main ingredient of our simulation is the multiscale interpolation technique that maps the sensor data into the solution space. We test our method on various synthetic examples. In particular we show that frequent updating of the sensor data in the simulations can significantly improve the prediction results and thus important for applications. The frequency of sensor data updating in the simulations is related to streaming capabilities and addressed within DDDAS framework. A further extension of our approach using local inversion is also discussed.

Dynamic compilation of C++ template code

Generic programming using the C++ template facility has been a successful method for creating high-performance, yet general algorithms for scientific computing and visualization. However, adding template code tends to require more template code in surrounding structures and algorithms to maintain generality. Compiling all possible expansions of these templates can lead to massive template bloat. Furthermore, compile-time binding of templates requires that all possible permutations be known at compile time, limiting the runtime extensibility of the generic code. We present a method for deferring the compilation of these templates until an exact type is needed. This dynamic compilation mechanism will produce the minimum amount of compiled code needed for a particular application, while maintaining the generality and performance that templates innately provide. Through a small amount of supporting code within each templated class, the proper templated code can be generated at runtime without modifying the compiler. We describe the implementation of this goal within the SCIRun dataflow system. SCIRun is freely available online for research purposes.

Evaluation of different meshing algorithms in the computation of defibrillation thresholds in children

In this paper we evaluate different meshing schemes to solve for the bioelectric fields that arise in the human body due to the defibrillation shock generated by an Implantable Cardiac Defibrillator, with particular emphasis on implantation in children. For children, the question of relative performance of different electrode locations remains open. Computational simulation is a critical tool to address this question, and mesh design is a critical component of such simulations. We use the SCIRun software package to address this simulation problem because it combines the powerful numeric tools required with interactive flexibility allowing easy comparison of both algorithms and electrode orientation. We describe a pipeline that starts with segmented CT-images and produces clinically useful parameters. Using this framework we report below that a meshing scheme using regularly spaced hexahedral elements which are locally refined around the electrodes constitute a quick and relatively accurate way of solving this problem.

Dynamic contaminant identification in water

We describe how we plan to convert a traditional data collection sensor and ocean model into a DDDAS enabled system for identifying contaminants and then reacting with different models, simulations, and sensing strategies in a symbiotic manner. The sensor is just as useful in water as it would be on Mars for material identification. A successful terrestrial application of the sensor will lead to many new applications of the device and possible technology transfer to the private sector.

A meshing pipeline for biomedical computing

Biomedical computing applications often require a computational pipeline that integrates data from experimental measurements or from image acquisition into a modeling and visualization environment. The latter process often involves segmentation, mesh generation, and numerical simulations. An important requirement of the numerical approximation and visualization methods is the need to create a discrete decomposition of the model geometry into a ‘mesh’. The meshes produced are used both as input for computational simulation and as the geometric basis for many of the resulting visualizations. Historically, the generation of these meshes has been a significant bottleneck in efforts to efficiently create complex, three-dimensional biomedical models. In this paper, we will outline a pipeline for more efficiently generating meshes suitable for biomedical simulations. Because of the wide array of geometries and phenomena encountered in biomedical computing, this pipeline, SCIRun, will incorporate a flexible suite of tools that will offer some generality to mesh generation of biomedical models. We will discuss several tools that have been successfully used in past problems and how these tools have been incorporated into SCIRun. We will demonstrate mesh generation for example problems along with methods for verifying the quality of the meshes generated. Finally, we will discuss ongoing and future efforts to bring all of these tools into a common environment to dramatically reduce the difficulty of mesh generation for biomedical simulations.

INTERPOLATION AND UPDATE IN DYNAMIC DATA-DRIVEN APPLICATION SIMULATIONS

In this paper we discuss numerical techniques involved in dynamic data driven application simulations (DDDAS). We present an interpolation technique and update procedures. A multiscale interpolation technique is designed to map the sensor data into the solution space. In particular we show that frequent updating of the sensor data in the simulations can significantly improve the prediction results and thus important for applications. The frequency of sensor data updating in the simulations is related to streaming capabilities and addressed within DDDAS framework (Douglas et al., 2003). We discuss the update of permeability and initial data.

Virtual telemetry for dynamic data-driven application simulations

We describe a virtual telemetry system that allows us to devise and augment dynamic data-driven application simulations (DDDAS). Virtual telemetry has the advantage that it is inexpensive to produce from real time simulations and readily transmittable using open source streaming software. Real telemetry is usually expensive to receive (if it is even available long term), tends to be messy, comes in no particular order, and can be incomplete or erroneous due to transmission problems or sensor malfunction. We will generate multiple streams continuously for extended periods (e.g., months or years): clean data, somewhat error prone data, and quite lossy or inaccurate data. By studying all of the streams at once we will be able to devise DDDAS components useful in predictive contaminant modeling.

Dynamically Identifying and Tracking Contaminants in Water Bodies

We present an overview of an ongoing project to build a DDDAS for identifying and tracking chemicals in water. The project involves a new class of intelligent sensor, building a library to optically identify molecules, communication techniques for moving objects, and a problem solving environment. We are developing an innovative environment so that we can create a symbiotic relationship between computational models for contaminant identification and tracking in water bodies and a new instrument, the Solid-State Spectral Imager (SSSI), to gather hydrological and geological data and to perform chemical analyses. The SSSI is both small and light and can scan ranges of up to about 10 meters. It can easily be used with remote sensing applications.

A Software Framework for Solving Bioelectrical Field Problems Based on Finite Elements

Computational modeling and simulation can provide important insights into the electrical and electrophysiological properties of cells, tissues, and organs. Commonly, the modeling is based on Maxwells and Poissons equations for electromagnetic and electric fields, respectively, and numerical techniques are applied for field calculation such as the finite element and finite differences methods. Focus of this work are finite element methods, which are based on an element-wise discretization of the spatial domain. These methods can be classified on the elements geometry, e.g. triangles, tetrahedrons and hexahedrons, and the underlying interpolation functions, e.g. polynomials of various order. Aim of this work is to describe finite element-based approaches and their application to extend the problem-solving environment SCIRun/BioPSE. Finite elements of various types were integrated and methods for interpolation and integration were implemented. General methods for creation of finite element system matrices and boundary conditions were incorporated. The extension provides flexible means for geometric modeling, physical simulation, and visualization with particular application in solving bioelectric field problems