Osteoporosis is a major health problem for roughly 55% of the US population of 50 years of age or older. It is characterized by low bone mass and structural deterioration which leads to increased fragility and risk of fracture. Fifty percent (50%) of women and 25% of men over age 50 will have an osteoporosis-related bone fracture. The most typical fractures occur in the hip, spine, wrist, and ribs, of which the hip and vertebral fractures can require long-term care and even cause death in as many as 24% of hip fracture cases [1
]. This dynamic aspect of bone physiology may facilitate the use of dielectric interrogation as a means of imaging bone health.
Assessing bone health may be a particularly good opportunity to exploit dielectric properties for screening, diagnosis, and in the overall management of bone treatment. In parallel with previous broadband tissue dielectric property studies [2
], bone has received special attention. For instance, clinicians have used electrical currents to stimulate bone growth for several decades [12
]. With such treatment it is essential to both understand the reaction of the bone to electrical stimulation as well as understand the tissue dielectric properties for guiding the therapy. In addition, the dielectric properties, themselves, may provide clinically useful information with respect to assessing overall bone health as in the case of osteoporosis and monitoring osteogenic response to treatment. These properties have been studied extensively up to 5
]; however, tests beyond this frequency have not proved useful to date [20
]. At lower frequencies, electrical impedance spectroscopy (EIS) techniques, including parallel plate capacitance cells, have been used to retrieve accurate dielectric properties [21
]. In fact, measurements have been made using conventional open-ended coaxial dielectric probes at frequencies as high as 3
GHz, but remain unpublished. The primary reason is that the dielectric probe technique is inherently ill-suited to this type of measurement. While researchers have performed experiments to determine the proximal limits of heterogeneities when testing homogeneous materials and liquids using these probes [23
], first-hand experience suggests that measurements with these probes on inhomogeneous targets are dominated by the tissue in direct contact with the probe. Given the heterogeneous nature of trabecular bone samples and the potential for property and texture variations between the less disturbed internal zones and the bone surfaces because of the extensive manipulation involved in preparing the samples, it is not surprising that these probe measurements have been inconclusive. More recent in vivo
animal studies by Gabriel et al. [8
] indicate that dramatic dielectric property changes occur with age in bone (not seen in other tissue types). This dynamic aspect of bone physiology may facilitate exploitation of dielectric interrogation as a means of bone health imaging.
The baseline studies by Gabriel et al. [5
] and others have proved useful in establishing nominal, frequency-dependent property ranges for different tissue types and have set the stage for further investigation of whether variations in individual tissue dielectric properties can be predictive of various maladies. Studies by Joines et al. [24
] and Lazebnik et al. [25
] have explored whether tumors exhibit dielectric properties distinct from their normal organ. Additionally, at frequencies below 2
GHz where the ionic flow dominates the overall conductivity effect [27
], tissue conductivity has been shown to vary linearly with temperature, and this mechanism has been utilized in investigations of noninvasive temperature monitoring in conjunction with thermal therapy [29
]. Similar work has been performed in other fields to look at tissue property variations based on physiological phenomena other than cancer. For instance, in comparable studies of ultrasound computed tomography, Sehgal et al. [31
] showed that the speed of sound and broadband attenuation varied considerably for liver samples depending on overall fat and tumor content. Given that the only substantive variations in the tissues tested in that study were the fat and water content levels, dielectric mixture laws such as the Maxwell-Fricke relationships would also naturally predict similar variations [3
In the area of ultrasound, researchers have performed numerous studies of the effect of bone density on the speed of sound and broadband attenuation, and several devices have been developed and approved by the FDA for testing bone health. For instance, the Lunar Achilles system produced by GE Healthcare (Waukesha, WI, USA) and the Sahara Clinical Bone Sonometer manufactured by Hologic (Bedford, MA, USA) are both FDA approved and used routinely in clinical situations. In the ultrasound experiments performed by Wu et al. [33
], trabecular bone samples were tested with both dual X-ray absorptiometry (DXA) and ultrasound transmission techniques to assess whether correlations exist between bone mineral density (BMD) and the ultrasound metrics—speed of sound and broadband attenuation. The samples were tested at a baseline with both techniques and at several subsequent times after demineralization through submersion in acidic solvents. The ultrasound measurements were performed with the bone specimens placed in water to assess overall bulk property changes as the hydroxyapatite was progressively removed and artificially replaced by the water. (Note that hydroxyapatite is the main constituent of the mineralized portion of the bone and is generally referred to as bone mineral.)
We have followed a similar path for this dielectric property study except the DXA X-ray tests have been replaced with more exact X-ray micro-CT measurements to determine the bone sample constituent proportions. Furthermore, we have replaced the ultrasound transmission measurements with microwave tomographic images. These tissue samples do not readily lend themselves to standard dielectric probe measurements because the trabecular bone samples are fragile and their surfaces are uneven given their intricate architecture. Instead, we have applied a newly developed microwave tomographic approach utilizing a soft prior regularization that is well suited for testing the bulk properties of small samples of known geometries [34
]. The samples were placed in a test tube filled with saline of known size in a predetermined location within the illumination zone of our microwave tomographic system. Prior knowledge of the sample shape, size, and location was applied in the regularization scheme as part of the standard image reconstruction process [34
] to recover estimates of the specimen dielectric properties. This measurement approach is not unlike the ultrasound CT technique described by Schreiman et al. [36
] for isolated tissue samples where the configuration geometry was also well known. While replacement of the bone marrow with saline does not perfectly mimic the process that normal bone undergoes during aging and/or unnatural bone loss, it does validate the overall notion that tissue bulk dielectric properties are functions of the individual properties and volume fractions of its constituents. In situ
dielectric changes will likely occur in more complicated patterns, but these measurements may still provide important information on tissue pathology and health.
In this paper we describe the process used in these experiments including the tissue preparation, X-ray CT, and microwave imaging approaches. We illustrate the technique of applying the microwave tomographic method with soft prior regularization to recover accurate values of the dielectric properties of the bone tissue samples and the associated tools used to assess the radiographic bone properties from the micro-CT data. Finally, we present results demonstrating the correlation between the bone dielectric properties and the X-ray data metrics.