Can technology developed by the Pentagon to search for enemy warheads in space also be used to hunt for potentially cancerous tumors in a woman's breast?
A group of military doctors and engineers have joined forces in an experimental effort to find out. Early results after two months of testing have shown promise, suggesting the approach could provide a means for spotting breast cancers years ahead of current methods, according to officials involved in the experiment.
But they cautioned that the findings have yet to be subjected to rigorous peer review, and that years of further testing remain.
Pentagon authorities cited the effort as a recent example of technologies spawned by the missile defense program and since spun off into beneficial commercial applications.
After spending more than $100 billion over several decades to try to build a nationwide defense against a ballistic missile attack, the Pentagon expects to deploy a very rudimentary system later this year. But technological advances fathered by the program already have found their way into such commercial ventures as cameras that aid eye exams, facial imaging sensors used to identify people for security purposes, and chemical detectors that expose contaminants in water and soil, the officials said.
What distinguishes the breast cancer project is the way it originated. While other spinoffs have tended to come about through formal Pentagon-funded research, the cancer effort grew out of a chance encounter and is being pursued by several Defense Department employees in their spare time.
About a year and a half ago at a party at Fort McNair, Army Col. Craig Shriver, who runs the breast cancer center at the Walter Reed Army Medical Center, struck up a conversation with Jess Granone, director of the Space and Missile Defense Technical Center at the Army's Space and Missile Defense Command in Huntsville, Ala.
Granone mentioned work being done in Huntsville testing Pentagon algorithms -- mathematical formulas -- designed to pick enemy warheads out of fields of decoys and background clutter using a process of pattern recognition. Shriver saw a parallel with work his staff was doing to try to identify patterns in proteins in the bloodstream that may signal the presence of breast cancer.
Currently, detection of breast cancer relies largely on radiologists reading X-rays known as mammograms. But that method is far from perfect.
"How well a mammogram is read is based on a radiologist's experience, whether or not he's tired that day, whether or not he gets distracted by a phone call in the midst of reading it," said Gary Payton, deputy for advanced systems at the Pentagon's Missile Defense Agency. "So there's some variability."
But what if the mammograms were turned into digital images -- that is, put in a format that a computer could read? Could algorithms then be applied to find suspicious lesions and determine whether those lesions are cancerous?
"It would take some of the variability out of what a radiologist might judge as he is looking at this X-ray," said Payton, whose wife was treated for breast cancer two years ago.
Granone, who oversees a staff of more than 400 scientists and engineers, thought the idea was worth a try. He had never seen a mammogram, but he lost a sister-in-law to breast cancer about five years ago and he welcomed the chance to attempt to improve the detection technology.
"Engineers are good at processing lots of data. It's what we do in missile defense," he said.
Granone did some checking and discovered that, during the Clinton administration, Pentagon officials had also tried to apply missile defense algorithms to cancer detection, but without much success. Shriver was familiar with computer-imaging detection programs that were being used to hunt for abnormalities in breast tissue. But these programs did little to characterize any discovered abnormalities as either benign or malignant.
"Most mammograms, when abnormal, turn out to be benign, but you don't know that unless you biopsy," Shriver said. "And that's one of the big complaints that women have about mammography -- it results in a lot of false positives, and we have to go after those surgically. So anything that can cut down on that would be a welcome addition. And that's exactly what we're looking at."
The first challenge, Shriver said, has been turning mammograms into digitized images -- something not usually done. Drawing on past cases, Shriver and his team have digitized nearly 100 mammograms and sent them to Huntsville, where Granone's group has applied algorithms to see whether known abnormalities can be picked up by the computer and correctly identified as benign or malignant.
"The approach seems promising, but we're still working on the data," Shriver said, declining to provide exact statistics for the success rate so far. "I thought when we first began that we wouldn't even get out of the starting gate."
To pay for the project, the researchers applied last year for a grant from the National Institutes of Health but were rejected. They intend to resubmit an application after obtaining more results "to show we're not going down a blind alley," Shriver said.
Once satisfying themselves that the algorithms work reliably in historical cases, the researchers plan to try their approach on new patients, assessing whether a cancer exists, then waiting for biopsies to confirm their findings.
"I don't believe we'll ever replace radiologists with this approach," Granone said. "But perhaps we will be able to give them the tools to be more acute when they're trying to look for something cancerous."