University of Rochester

Image-Compression Method Tames Flood of Ultrasound Data

April 29, 1998

By stripping away the interference that ultrasound scanners ordinarily pick up from muscle and skin, engineers at the University of Rochester have devised a method for storing slimmed-down ultrasound scans that saves not only images of the body's inner workings, but also a great deal of space, time, and money.

The new technique addresses a problem that has become increasingly acute for health-care providers in recent years: how to deal with the millions of saved scans that pile up with each year of ultrasound's increasing popularity. A sleek way to digitize, store, and transmit archived scans would make it considerably easier for doctors to access stored ultrasound images; it would also be a boon to the budding field of teleradiology, which allows physicians to confer with distant colleagues by sending them ultrasound scans over telephone lines.

Such advances are made more likely by a novel form of ultrasound compression recently patented (U.S. Patent No. 5,734,754) by the University on behalf of Kevin Parker, professor of electrical engineering, dean of the University's School of Engineering and Applied Sciences, and director of the Rochester Center for Biomedical Ultrasound. By junking the unnecessary echoes from soft tissue and saving data only from the underlying organs, it whittles the size of a digitized ultrasound file image down to one-twentieth of what's now required. It's kind of like conserving toner in your printer or photocopier by erasing a dark background from the web page you're printing or the article you're copying.

"The normal interference patterns from muscle, skin, and other soft tissues add a ton of complexity to ultrasonic scans," says Parker, a well-known ultrasound expert. "Simply removing them could mean substantial savings in both the money and space hospitals devote to saving old scans."

The millions of ultrasound scans taken in the United States every year are used to observe the progress of virtually all babies as they develop in the womb; doctors also examine disorders of the prostate and liver primarily with ultrasonic scans. While an ultrasound scanner can record up to 30 images per second, technicians usually only choose to save several dozen of the scans collected in a half-hour session. These images are stockpiled, but at many facilities, they're saved not in digital form but instead on videotapes or transparency-sized films, which are physically bulky, expensive, hard to access, and subject to deterioration. The saved scans take up rooms and rooms of valuable space at most large hospitals: Stored uncompressed on 3.5-inch diskettes, the 10 million scans taken in the United States in a single year alone would form four stacks the height of Mt. Everest.

For many years now, health care providers have sought more efficient ways to package ultrasound scans. But most of the compression techniques developed in response are just emulations of existing digital imaging technologies, such as JPEG or MPEG -- methods that work best when imaging large, flat expanses, Parker says. Since the inside of the human body is a lot more complex, images stored as MPEG or JPEG files can end up looking like hard-to-interpret cubist renditions of the body's innards. For lack of a better option, though, most doctors tolerate the imprecision of these forms of digital storage, which are still much cheaper and sturdier than storage on video or film.

"By contrast, our technique is built specifically for ultrasound from the ground up, taking into account the special storage needs of ultrasound scans," says Parker. "While JPEG and MPEG work from the assumption that an image has a photographic origin, our technology recreates an image based on the assumption that it's working with data gathered through a pulse-echo system. It's a method of reconstructing images that's in step with the way ultrasound scanners collect data.

"Compression always entails some loss of detail, but the losses in fidelity with this method are almost imperceptible."

Many physicians eagerly await improved ultrasound compression for a different reason: It will slash the time they wait for scans to be electronically transferred across hospitals, states, or even the world. "Compression will be most helpful in rapid data transfer," says Deborah Rubens, associate professor of radiology and director of diagnostic ultrasound at the University's Medical Center. "It will make possible quick real-time image transfer without clogging up data-transfer lines."

Indeed, the physical bulkiness of uncompressed ultrasound scans threatens to undo the benefits of teleradiology in areas not served by high-bandwidth data lines, where it may prove simply impossible to muster enough resources to convey the uncompressed images.

Since they're one-twentieth the size of uncompressed images, Parker's compressed scans should be accessible to physicians via data-transfer lines roughly 20 times faster. Storing scans digitally also should reduce hospital costs, since disk space is generally a cheaper storage medium than the videotapes and X-ray films now used at many facilities. Digital storage also makes it far easier for doctors to perform retrospective medical studies. Researchers can easily search a computerized database of ultrasound scans to find, for example, 60-year-old Hispanic males with a specific pattern of prostate cancer. Such a search is nearly impossible with scans saved on film or video.

Parker was joined in the research by Robert Cramblitt, a former postdoctoral researcher at Rochester now employed by imaging firm SVS Inc. in Albuquerque, N.M. The University is now searching for a corporate partner to commercialize the technology.

The work is sponsored by the Department of Electrical Engineering and the Center for Electronic Imaging Systems, both at the University. CEIS is sponsored by the National Science Foundation, New York State, and several industrial partners.