The microwave imaging research community includes a body of investigators with remarkable intellect, but the group represents a swirling mass of ideas that can be difficult to interpret by an outside observer. The discourse is driven by diverse groups with many individual perspectives. Too often, researchers mostly speak over the heads of others such that the core ideas are rarely shared and consensus is rarely articulated. Important recent developments are beginning to put all of this energy into more focus. For instance, what do the clinical systems at University of Bristol, Dune Medical Systems and Dartmouth College have in common? On the surface, their differences are substantial. The first is a radar detection approach, the second is a coaxial probe used to assess surgical margins during breast conserving surgeries and the latter is a tomographic technique that has been used in both diagnostic and therapy monitoring settings. The main point is that all three are now being used extensively in the clinic and they all exploit the endogenous dielectric property contrast between tumor and benign/normal tissue. How is this possible? And what about the results of Lazebnik et al. Citation[1]? Maybe the emerging clinical data and dielectric probe study can co-exist, even though they seemingly contradict each other. Importantly, these recent dielectric probe results say little about the prevalence or distribution of fibroglandular tissue in the breast. Just as perplexing, the results may mask potentially relevant information related to the bound and free water fractions in tissue that dielectric properties are well-positioned to detect in vivo. Other modalities such as MR and near-infrared optical imaging have demonstrated that free water can be a powerful surrogate for detecting cancer. Pioneering research in the 1970s and 1980s demonstrated that tissue dielectric properties were not equivalent to total water content measures at lower frequencies, and the deviations noted were due to the presence of bound water. Indeed, a closer look at the results from the aforementioned dielectric property study suggests that a single dipole relaxation model over a broad bandwidth may miss important low-frequency features associated with bound water. Several studies Citation[2,3] have noted a distinct break point between 2 and 2.5 GHz under which the bound water effects are dramatic and over which they are minimal. In fact, in a thermal property paper Citation[4], the authors also observed a related property crossover behavior in the same frequency range – albeit as a function of temperature.
Beyond the fundamental question of microwave property contrast in breast tissue in vivo, translation of actual imaging systems into the clinic has, unfortunately, been far too limited. Multipath signals, which confound the data of interest, have remained a significant challenge in most realizations. These waveforms propagate along alternative and unwanted paths relative to the desired ones, that is, the signals that pass through the target tissue. They are well known in radar and communications systems and have been a research topic of substance for many years. With respect to the near-field imaging problems presented by biomedical applications, the opportunity for multipath signals are numerous – poor isolation in multichannel switching networks, signal reflections from the surrounding environment, for example, from chamber walls and surfaces. Surface waves appear in a variety of modes including ones that propagate along the outside of coaxial feed lines and in planar forms along the interfaces of materials. The more successful coping mechanisms have included time gating to eliminate longer propagation times, attenuation through a lossy coupling medium and utilizing a very large tank with only a single transmitter–receiver pair. Time gating is particularly useful in radar and time-domain techniques because the broadband data are already available and requires clever gating to eliminate multipath signal contributions. A lossy coupling medium is seeing broader application by groups using tomographic approaches and has been quite effective. The single transmitter–receiver pair has been used effectively because it also eliminates mutual coupling since no immediately adjacent antennas exist, but examination times can be long because of the mechanical scanning and the tank is also large.
Beyond the hardware challenges, tomographic image reconstruction is particularly enigmatic because so many groups have been studying them for years but with relatively little application of the algorithms to data acquired in the clinic. Most of these efforts can be described as, ‘doomed to success.’ That is, when was the last time a published simulation failed? The real challenge is developing algorithms that are able to process data collected from a living, breathing human and converting it into information (i.e., images) that can inform a clinical decision. Techniques that converge to solutions (local minima) without the benefit of good a priori estimate of the image, which require large amounts of measurement data or take days to create are nonstarters with clinical specialists and external observers. The medical microwave imaging community needs to take a serious look at what has and has not worked. Unfortunately, negative results are rarely published, but these occurrences are often where the real learning occurs. The classic multipath signal corruption noted above might finally get recognition as being the especially challenging problem it is in the setting of producing an image reconstruction algorithm that routinely recovers a useful image from clinical data.
A range of microwave technologies may offer benefits in the clinical management of breast cancer. Although the focus of new imaging modalities has usually been in screening, other roles in diagnosis and monitoring chemotherapy are also important and may present fewer barriers to entry. Once microwave technology is introduced into the clinic in any form, the whole field will benefit. The field should encourage and foster a range of imaging strategies – but the community needs to offer viable paths to implementation. Collaboration and real sharing of ideas – the author recommends students and postdoctoral students spend time at competing institutions – is essential. Tailoring emerging algorithms to existing hardware implementations and fostering collaboration may be a more realistic path to clinical translation than beginning each new effort from ground zero. The funding may simply not exist for multiple new techniques to start from the beginning. The recently funded European Cooperation in Science and Technology action in microwave imaging may be a real start in this regard.
In the end, outside observers will paint us all with the same brush if data beyond simulations are not produced. The microwave imaging community needs to encourage and promote translation to the clinic. The group at large benefits with each success and it will continue down a path of irrelevance, if simulations remain the emphasis.
Financial & competing interests disclosure
PM Meaney is a Professor of the Thayer School of Engineering at Dartmouth College and co-owner of Microwave Imaging System Technologies, Inc., both in Hanover, NH, USA. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.
No writing assistance was utilized in the production of this manuscript.
References
- Lazebnik M, Popovic D, McCartney L et al. A large-scale study of the ultrawideband microwave dielectric properties of normal, benign and malignant breast tissues obtained from cancer surgeries. Phys. Med. Biol. 52, 6093–6115 (2007).
- Foster KR, Schepps JL, Schwan HP. Microwave dielectric relaxation in muscle: a second look. Biophys. J. 29, 271–282 (1980).
- Schepps JL, Foster KR. The UHF and microwave dielectric properties of normal and tumour tissues: variation in dielectric properties with tissue water content. Phys. Med. Biol. 25, 1949–1959 (1980).
- Lazebnik M, Converse MC, Booske JH, Hagness SC. Ultrawideband temperature-dependent dielectric properties of animal liver tissue in the microwave frequency range. Phys. Med. Biol. 51, 1941–1955 (2006).