Electrical Tomography Hardware Systems for Real-Time Applications: a Review

Abstract

This paper presents a review of two-dimensional (2D) and three-dimensional (3D) electrical tomography (ET) hardware accelerators for real-time applications. While many recent review papers have discussed various algorithms for image reconstruction or acquisition systems, none of them has considered state-of-the-art hardware implementations of the associated image reconstruction algorithms to achieve real-time performance, especially for 3D ET where the computation requirement is excessively high. A 3D ET is useful in various applications such as robotics, autonomous vehicles, and process control, but it is computationally very expensive with respect to its 2D counterpart. Most implementations are based on single or multi-core CPUs and, to a lesser extent, on either graphics processing units (GPUs) or field programmable gate arrays (FPGAs). However, there is a clear gap between the currently available processors, whose computation power exceeds hundreds of teraflops per second (TOPS) at a reasonable low power consumption, and the ones recently used in ET systems. This gives great potential for nextgeneration ET systems to achieve real-time 2D and 3D ET reconstruction within a small form factor. The paper summarizes the most recent ET hardware systems with respect to their performance in terms of quality and processing frame rate, reconstruction methods, along with optimization and future directions.

A Hybrid FPGA-GPU-based Hardware Accelerator for High Throughput 2D Electrical Impedance Tomography

“…It can be observed that a general-purpose PC is widely used for the reconstruction of the image, with a few implementations utilizing FPGAs or external GPUs. A detailed review of some recent work for image reconstruction can be found in [11]. The reconstruction speed is directly dependent on the number of electrodes and mesh elements.…”

“…The overall throughput of the EIT systems is defined by both the acquisition and the data processing (reconstruction) modules. While there have been multiple solutions to increasing the acquisition speeds, such as multi-frequency or partial measurements or improved hardware, the reconstruction has been mostly limited to utilizing general-purpose PC [11] with some standard library such as EIDORS [12] for carrying out the computation. The computation involves several matrixbased calculations in an iterative or non-iterative way for 2D image reconstruction.…”

CATE: A fast and scalable CUDA implementation to conduct highly parallelized evolutionary tests on large scale genomic data

“…These include dedicated software built around multi-core processing as well as the expansion of statistical tools such as R to enable support for parallel processing. Most of these solutions limit themselves to the CPU, which at best contains tens of cores and hundreds of threads (43, 44). Modern GPUs on the other hand house over thousands to tens of thousands of cores on a single chip.…”

A novel neural network-based data acquisition system targeting high-speed electrical impedance tomography systems

“…In the case of highly conductive medium such as multi-phase flow featuring high water-cut and low gas void fraction (GVF) or in the case of brain/breathe monitoring, the ERT or EIT technique is preferred. An electrical current in the frequency range of a few kHz to up to 10 MHz is applied across two electrodes, and the voltage output across all the other electrodes is recorded and used for 2D or 3D image reconstruction [17]. Synchronous detection (i.e., phase-sensitive detection) is used to eliminate low-frequency drift components of the voltage signal.…”

“…In conclusion, high spatial and temporal resolution ET systems are yet to be developed, despite the fact that they are highly desirable in several applications. The development of powerful edge processors based on GPU, FPGA, and/or multi-core CPUs will enable the rapid execution of iterative and complex tomography forward-inverse algorithms such as iterative Gauss-Newton, iterative Landweber, total variation, and even CNN/DNN algorithms in the near future [17]. This also requires high-speed acquisition solutions.…”