IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 59, NO. 2, FEBRUARY 2012 383 A Four-Shell Diffusion Phantom of the Head for Electrical Impedance Tomography Matteo Sperandio, Marco Guermandi ∗ , and Roberto Guerrieri Abstract—A four-shell head phantom has been built and charac- terized. Its structure is similar to that of nonhomogeneous concen- tric shell domains used by numerical solvers that better approxi- mate current distribution than phantoms currently used to validate electrical impedance tomography systems. Each shell represents a head tissue, namely, skin, skull, cerebrospinal fluid, and brain. A novel technique, which employs a volume conductive impermeable film, has been implemented to prevent ion diffusion between differ- ent agar regions without affecting current distribution inside the phantom. Comparisons between simulations and phantom mea- surements performed over four days are given to prove both the adherence to the model in the frequency range between 10 kHz and 1 MHz and its long-term stability. Index Terms—Electrical impedance tomography (EIT), phan- tom, tissue modeling. I. INTRODUCTION B UILDING phantoms that are able to emulate different properties of human tissues are a widely adopted practice in the biomedical research area. First of all, phantoms are useful in testing medical devices before they are applied to human sub- jects; second, they allow one to mimic normal and pathological conditions of the human body by providing a completely char- acterized system with known properties [1]–[3]. In this study, we present a phantom suitable for the validation of electrical impedance tomography (EIT) brain imaging systems [4]. EIT is a noninvasive imaging technique whose application to the central nervous system is currently at the research stage. EIT estimates the impedance distribution inside a body by applying specific ac current patterns to the surface of the body and measur- ing the corresponding electric potentials on the surface. Injected Manuscript received June 8, 2011; revised August 1, 2011, September 21, 2011, and October 6, 2011; accepted October 8, 2011. Date of publication October 21, 2011; date of current version January 20, 2012. This work has been created in the scope of the HIGH PROFILE project that was supported in part by the ARTEMIS Joint Undertaking under Grant agreement n 269356 and by the national programs/funding authorities of Austria, Belgium, Finland, Italy, and the Netherlands. Asterisk indicates corresponding author. M. Sperandio was with the Advanced Research Center on Electronic Systems (ARCES), University of Bologna, Bologna 40123, Italy. He is now with Aizoon Consulting srl, Torino 52010, Italy (e-mail: matteo.sperandio@aizoon.it). ∗ M. Guermandi is with the Advanced Research Center on Electronic Systems (ARCES), University of Bologna, Bologna 40123, Italy (e-mail: mguermandi@ arces.unibo.it). R. Guerrieri is with the Advanced Research Center on Electronic Systems (ARCES), University of Bologna, Bologna 40123, Italy (e-mail: rguerrieri@ arces.unibo.it). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TBME.2011.2173197 current frequencies vary among different EIT instrumentation capabilities and the goals of clinical studies. However, for prac- tical and safety reasons, currents are usually injected between a few kilohertz and some megahertz. The quality of an EIT phantom lies in its ability to capture the electrical properties of the underlying tissues, emulating as well as possible the current flows inside the body. EIT phantoms are often realized as 2-D domains [5] that can satisfactorily model cylindrical structures like, say, the human thorax, but cannot be applied to the human head. Numerical human head models have evolved from simple uniform spheres [6] to more complex models comprising a number of tissues with different electrical properties [7] since it has been demonstrated that significant differences in current distribution occur inside the domain. In order to have a phantom with known conductivity distribution and high reproducibility, 2-D EIT test phantoms have often been based on meshes composed of discrete resistors [3]. The main advantages of these solutions are their high reproducibility and stability over time. However, they lack the ability to capture the continuous behavior of tissues and would require a huge amount of high-precision components and complex connections to ac- curately reproduce a 3-D domain. On the other hand, gel-based diffusion phantoms (e.g., sodium chloride solutions thickened by an appropriate amount of agar) can accurately model the conductivity of many tissues and are widely used in so-called diffusion phantoms [2]. The main drawback of this solution is the difficulty of emulating structures composed of regions with different electrical properties. Since electrical conductivity is due to ions, different conductivities are achieved by different ion concentrations. The diffusion of the ions between differ- ent agar layers due to gradients in their concentration yields only short-term stability, from a few minutes to a few hours, even when additional materials are added to the recipes with the purpose of increasing phantom lifetime. In this study, we present a 3-D head diffusion phantom com- posed of four hemispheres, each representing a different com- partment in the usual segmentation of the human head that, from the outermost shell, consists of scalp, skull, cerebrospinal fluid (CSF), and brain. Long-term stability is achieved by inserting a thin impermeable conductive polymer sheet between agar layers to prevent ion diffusion. We concentrate on real conductivities though techniques for adding reactive components as in [8] can be similarly implemented. Applications of the developed phan- tom are to test and validate EIT hardware, especially for brain imaging purposes, verifying optimal current patterns and evalu- ating the quality of reconstruction algorithms, in particular those relying on multilayered head models. In contrast to phantoms composed of discrete component meshes, standard electrodes 0018-9294/$26.00 © 2011 IEEE