MIT researchers are developing a new kind of x-ray imager that uses information that traditional machines ignore. By looking at how tissue refracts the rays, not simply at how it absorbs them, the researchers hope to increase the resolution of mammography, enabling doctors to detect smaller tumors earlier.
The basic physics behind x-ray imaging hasn’t changed in more than 100 years. Most hospitals have gotten rid of film and gone digital, but their images still record the same kind of information: how a part of the patient’s body absorbs the rays.
When radiation hits a material, many other things happen to the waves besides absorption, say Richard Lanza and Antonio Damato, a research scientist and a graduate student, respectively, in nuclear engineering at MIT. The pair is developing a prototype x-ray mammogram system that will record information about absorption and refraction of the radiation as it passes through an object.
Refraction is a change in the direction of a wave as it passes from one medium to another; it’s the same effect that makes a straw in a glass of water look broken.
“Cancers have very similar x-ray absorption to normal tissue,” says Daniel Kopans, director of breast imaging at Massachusetts General Hospital, in Boston. But cancers may have different refractive properties. Kopans says that preliminary studies using a synchrotron accelerator have shown that cancers appear to refract x-rays differently than normal tissue does. If this is so, recording this information during an x-ray scan might indeed help doctors better delineate tumors.
Since the widespread adoption of mammogram screenings in the early 1990s, the breast-cancer death rate has gone down by 25 percent in the United States. But breast cancer is still the second leading cause of cancer death in women. These tumors remain particularly difficult to detect because breast tissue has a variable structure. Five millimeters is the limit of detection with conventional mammography, so doctors know that they are missing tumors, says Richard Moore, research director in the breast-imaging center at Mass General.
Moore says the hallmarks of breast cancer that doctors look for in mammograms reflect disease processes that have been under way for a long time: tiny calcifications that signal cancer-cell death, and healthy surrounding tissue that’s pinched, like fabric. Although they don’t know what they’ll find, Moore says that Lanza’s x-rays might provide new information about breast tumors that enable doctors to see disease sooner, before the calcifications and pinching effects visible in a conventional mammogram even develop.
It’s practically impossible to measure how a sample refracts waves unless the waves all start out going in the same direction at the same frequency. Conventional x-ray tubes–such as those used in mammography–can’t do this, and x-rays can’t be focused with lenses and mirrors. One way to achieve the uniformity, or coherence, necessary for refraction measurements is to run the x-rays through a particle accelerator. But these are too large and expensive for hospital use.
Another way to make coherent radiation is to filter the waves generated by a conventional x-ray tube through a tiny hole in a lead shield. The rays come at the shield every which way, and they exit the hole in uniform circular waves. But because only a small proportion of the waves generated by the tube can squeeze through the hole, using this kind of filter takes a lot of power and generates a very weak image. In such a system, a patient would have to hold still and be scanned for a very long time, says Lanza.
The system Lanza and Damato are developing combines a traditional x-ray tube with a filter that has hundreds of tiny holes for generating coherent beams of x-rays. First, these coherent waves pass through the sample. Next, the waves pass an interferometer, which collects information about how each of the hundreds of sets of waves was refracted by the sample. Then they pass to a traditional absorption detector, which collects information about how each of the hundreds of sets of waves was absorbed by the sample.
Hundreds of interference images are then combined with the hundreds of absorption images into a single picture that a radiologist can read. Berthold Horn, a member of MIT’s computer-science and artificial-intelligence laboratory, is developing algorithms to process the x-ray images. “The computer becomes an integral part of the imaging system,” he says.
Similar filters are used in x-ray astronomy, in which the objects of interest–the stars–are very, very far away. Horn and Lanza were the first to apply this technique, called coded-aperture imaging, to nearby objects. They have already employed it to improve the resolution of a gamma-ray imaging technique used on lab animals, and they believe they will see similar results with x-ray imaging.
If the MIT x-ray system does prove to have a higher resolution, as the gamma-ray system did, it could help doctors find smaller tumors. But it will need to go through randomized control trials to prove that it saves lives. “You want to make sure you don’t make a patient’s life any worse,” says Moore. “They’re coming in for screening before they have any symptoms.” He hopes that more-sensitive mammography will not highlight nonlethal tumors, as a recent controversy suggests that lung CT scans may, but will help eliminate false positives.
For every 1,000 women screened for breast cancer, 80 are called back. “All 80 think they have cancer, and they feel terrible,” says Moore. But only 20 will be called back for biopsies, and only 5 will have cancer. “The funnel is wide at the top, narrow at the bottom,” says Moore. He hopes that better x-rays will help doctors do better.
For now, says Horn, the researchers are working on their prototype and hoping they can “demonstrate that there’s information there that’s not in the absorption images.”
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
Data analytics reveal real business value
Sophisticated analytics tools mine insights from data, optimizing operational processes across the enterprise.
The Biggest Questions: What is death?
New neuroscience is challenging our understanding of the dying process—bringing opportunities for the living.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.