FPGA-Based Acceleration of the Computations Involved in Transcranial Magnetic Stimulation

Abstract

In the last years the interest for magnetic stimulation of the human nervous tissue has increased, because this technique has proved its utility and applicability both as a diagnostic and as a treatment instrument. Research in this domain is aimed at eliminating some disadvantages of the technique: the lack of focalization of the stimulated human body region and the reduced efficiency of the energetic transfer from the stimulating coil to the tissue. Designing better stimulation coils is so far a trial-and-error process, relying on very compute-intensive simulations. In software, such a simulation has a very high running time (several hours for complicated geometries of the coils). This paper proposes and demonstrates an FPGA-based hardware implementation of this simulation, which reduces the computation time by 2-3 orders of magnitude. Thanks to this powerful tool, some significant improvements in the design of the coils have already been obtained.

A flexible communication method for multi-FPGA based designs

“…The Aurora core inserts an overhead (due to the framing protocol) in the streaming that gets smaller as words widths are larger (which would make sense in using the bandwidth more efficiently). The integration of the proposed solution was tested using some existing large applications developed by our team ( [13], [14]) as well as on several open-source designs from the OpenCores database -two of them are presented hereinafter. The first core was designed to implement a FFT-based FIR filter [15] that had already provided a big part of the mechanism necessary for the data handshake.…”

VFloat

“…However, floatingpoint arithmetic, especially floating-point divide and square root, are difficult to design, and often the critical, performance limiting factors. In particularly, applications that require floating-point divide and square root include transcranial magnetic stimulation [Cret et al 2007], Molecular Dynamics (MD) simulations [Govindu et al 2005], Monte Carlo radiative heat transfer simulation · 16: 31 [Gokhale et al 2004], sparse matrix Jacobi solver , QR decomposition [Wang and Leeser 2007a], smoothed particle hydrodynamics method [Lienhart et al 2002], and gravity calculation for N-body simulation [Lienhart et al 2006], radiation dose calculation [Whitton et al 2006], optical flow algorithms for image stabilization [Etiemble et al 2005], and Least Mean Squares (LMS) and Maximum Likelihood (ML) for a space systems [Poznanovic 2004]. As a result, a floating-point library including floating-point divide and square root is very desirable.…”

An FPGA-specific approach to floating-point accumulation and sum-of-products

“…In [3], the FPGA computes a very large integrationseveral hours-and only the final result is relevant. In such Another common case is that a single post-normalisation unit may be shared by several accumulators.…”

“…In the inductance computation of [3], physical expertise tells that the sum will be less than 10 5 (using arbitrary units due to factoring out some physical constants), while profiling showed that the absolute value of a summand was always between 10 −2 and 2.…”