In contrast, what we’ve developed is a non-ionizing solution in the form of a helmet that is hyper-portable, estimated at around 10–15 kilograms, and at targeted costs in the tens of thousands of dollars. Our system completes acquisitions in a matter of minutes and operates at under three milliwatts of power (far less than the maximum output power of cell phones). Using custom neural networks developed for characterization of the data obtained by the antenna array within the helmet, the system is designed to function as an intelligent, fully diagnostic stand-alone instrument with tomographic capabilities.

HCB News: Why is this the first I’m hearing about the potential diagnostic value of microwaves?
SD and LA: While the use of microwaves as a probe of tissue properties has been considered for years, it has historically proven difficult to recover, or reconstruct, a true anatomic image in the typical sense. This is because the process of transmitting microwaves, allowing their interactions with tissue, and then estimating the signatures of those interactions based on the data received by the antennas introduces many potential paths that cannot conclusively be localized. The difficulty in localizing those effects presents major challenges when attempting to reconstruct an image of a volume or slice of tissue from such data. Historically, this had been performed across relatively narrow ranges of microwave frequencies.
The key insight into our work is the development of ultra-wideband microwave transmission and reflection across far greater ranges than historically used, leveraging advancements in antenna design and technology.

By increasing the range of interrogation across, for instance, 10 GHz or more, an antenna could probe highly varied interactions in a manner not previously performed and could benefit greatly from the biophysical contrasts across those ranges. Taking the additional step of assembling an array of such antennas, for example in a ring around the object, an unprecedented richness of data could be obtained.

HCB News: So modern computational power was necessary to unlock the utility of microwaves?
SD and LA: We realized early on that data of this sort was well-suited to analysis using large artificial neural networks, in what was still the relative infancy of the field approximately seven years ago. Trained, for instance, from contemporaneous brain MR scans, we felt that we could alleviate the challenge of tomographic microwave reconstruction using a subject’s brain MR as a guide in the learning process, and perhaps even develop direct diagnostic capabilities to circumvent the need for expert interpretations, which are unavailable in many places worldwide.

More Magazine Features

Stroke Homepage