WASHINGTON — Imaging software designed for NASA could soon help medical professionals to better interpret images such as mammograms, X-rays, and MRI’s through its ability to process images on an incredibly detailed scale.
The software is based an algorithm developed by NASA Goddard Space Flight Center engineer James Tilton to analyze digital images for earth science research. With the algorithm, Tilton said he hoped to expand analysis of digital images from one pixel at a time to looking at images in a regional perspective, like the number of lakes in a particular area or a pattern of roofs in an urban development. By looking beyond individual pixels, the algorithm “presents more useful information to the computer for analysis,” Tilton said.
Tilton’s software has been used by Bartron Medical Imaging Inc. to create a medical imaging device called MED-SEG, that Bartron CEO Fitz Walker said he hopes could eventually help doctors identify and potentially diagnose health conditions like breast cancer, more accurately and efficiently. “Our whole mission is to develop a device to make a difference and save lives,” Walker said.
The capability of the algorithm caught the attention of Walker, at a NASA technology conference and the company licensed the technology in 2002. He said the main question when it comes to medical imaging technology is “does it improve [medical professionals’] ability to interpret what they are looking at?” And after completing millions of tests on hundreds of thousands of images, Walker said he believed there was strong evidence that it could.
“It showed great promise in the area of medical analysis, in the use of medical imaging,” he said.
The algorithm, known as HSEG, uses a process called “region growing” to compare each individual pixel in an image to its neighboring pixels. As it finds similar pixels, it groups them together and organizes them into “region classes” based on levels of detail, showing the relationships between the objects. What makes HSEG unique, Tilton said, is that it automatically performs the classification process during the image analysis.
HSEG can even compare and classify objects that aren’t near each other in the image. When describing how it processed a satellite image of Wisconsin lakes, Tilton explained that it grouped the lakes together by depth ranging from light blue to dark blue, and then could also separate out all the lakes from the land into one class.
The MED-SEG device has received FDA clearance this year as a “Picture Archiving and Communication” system, which means it can receive and process images, but cannot yet be used as a screening or diagnostic tool. They system has been installed at the University of Connecticut Medical Center and Walker says Bartron is preparing a product launch this year.
Dr. Molly Brewer, a professor in the Division of Gynecologic Oncology at the University of Connecticut Health Center, has been involved with MED-SEG’s testing process over the past few years said she believes MED-SEG has potential to help improve mammography. Breast density often makes it difficult for a doctor to spot cancer on a mammogram, she said, but MED-SEG’s ability to process images at a more detailed level could provide a clearer picture for physicians.
“You see more nuances – the way to think about it is instead of seeing just white, you see more differences. Really what you’re doing is not changing what you’re looking at; you’re just changing your ability to interpret it, “Brewer said. “What we’re hoping is because mammography depends entirely on density, and because MED-SEG is able to separate out very small differences in density, we think that it may have some promise in enhancing mammography.” Brewer explained at this point they can make no claims, but test results show promising potential for the technology and they are hoping to move forward with clinical trials next year.
Dr. Daniel Kopans, a Harvard Medical School professor and member of the American College of Radiology Breast Imaging Commission, points out that the physical makeup of breast tissue combined with the fundamental limits of conventional mammograms makes it difficult for imaging technologies to really improve mammography. Standard mammograms are three-dimensional images projected as two-dimensional, so “things in front and things behind obscure something that might be buried in the tissue,” Kopans said.
He believes this sort of technology would be more useful when used to analyze three-dimensional images. Kopans has worked on a new three-dimensional mammography technique called Digital Breast Tomosynthesis, which can look at a mammogram in slices so that you can see the breast in layers, getting rid of the superimposition. “I’m not optimistic it’s going to have much use looking at standard two-dimensional mammograms, but it’s entirely possible that it would be useful for looking at Tomosynthesis,” he said.
To learn more about MED-SEG’s potential, Brewer and Walker plan on conducting clinical trials next year.
“It looks very promising, but part of what you have to do is then you have to look at it in a larger series,” Brewer said. “Our goal would be able to say to one woman, you need a biopsy and to say to another woman, no you don’t have cancer, you don’t need a biopsy. That would be optimal down the road” she said.
Brewer and Walker said they hope MED-SEG will eventually receive FDA clearance as a screening and diagnostic tool after clinical trials. In the field of mammography, this could mean more efficient diagnoses and less unnecessary testing for women, they said. They are also working with NASA engineer Tilton on a three-dimensional version of the technology.