To peer into the body and watch it work -- with no need for a knife, no convalescence or scars -- is a wondrous boon to doctors and patients alike. Given this capability, Magnetic Resonance Imaging (MRI) has, not surprisingly, become a standard diagnostic technique since its introduction some dozen years ago.

Yet taking an MRI exam involves lying flat on a hospital pallet, being slid into a narrow gleaming tube head first, and "relaxing" while being blasted with a pulsating noise that is the decibel equivalent of a loud rock concert.

"In more cases than they'd like, doctors have to abort the test," says Scott Sommerfeldt, a research associate and assistant professor of acoustics in Penn State's Applied Research Lab. "The impression I get is that the patients go berserk. They decide they're not staying in there."

Sommerfeldt, a noise-control expert, is working with General Electric's corporate research and development center to damp the sound that switches on those anxious fidgets.

The task is tricky for two reasons. First, "it's a complex and time-varying noise," explains Sommerfeldt. The strong magnetic field which, when interrupted by electrical pulses, is used to create the image of the body's interior, incidentally sets up a rhythmic vibrating in the imaging tube's external shell. These vibrations produce a violent noise that varies its character during each imaging pulse. "This is probably the toughest problem I've ever worked on," Sommerfeldt notes, "because the sound field is so loud and so complex."

Second, Sommerfeldt's noise-control mechanisms in the past had all used magnetic components -- sensors, loudspeakers, microphones -- which here would interfere with the imaging process.

For active noise control, somewhat counterintuitively, works on the principle of fighting fire with fire, sound with sound. According to the principles of acoustics, a simple sound can be thought of as a sine wave, a wiggly hump-valley-hump-valley line. If such a sound were perfectly matched with its mirror image (a valley-hump-valley-hump wave) the result would be silence. As Sommerfeldt explains, "If the noises are such that they add up out of phase, they cancel each other out."

The catch is that real-world sounds are not simple sine waves but extraordinarily complex interminglings of peaks and troughs. An active noise-control system thus needs a highly precise sensor to analyze the noise's wave form, a computer to calculate the exact counter noise, and a loudspeaker to blast that counter noise into the system at precisely the right time. The theory has been around since 1933, when a German physicist, Paul Lueg, filed a patent on it. But not until the advent of high-speed digital signal processing in the 1980s and '90s was it feasible. "Vacuum tubes and transistors drifted too much," Sommerfeldt explains. "Active control has to be matched up very precisely in time. It needs very stable, fast electronics.

"What determines how well you can do it," he adds, "is how fast you can get the sound you're cancelling with there. You could get the exact same wave form, but if it's not there soon enough, it's not going to cancel the original sound. It may even make it louder."

The only reason active noise control can work at all, he adds, is "that you can process at the speed of light -- which is faster than the speed of sound."

To quiet an MRI, then, Sommerfeldt and his GE colleagues designed a retrofit involving an array of noise sensors, adaptive signal processing (using a computer algorithm that adjusts its calculations for changes in pressure and temperature, which change the speed of sound), and an innovative, non-magnetic sound source. "It's a relatively compact solution," Sommerfeldt says. "It's designed so that the patient can lie in the chamber, and the system will create a reduced sound-field in the location of his or her ears." He and his collaborators predict a cost of less than $1,000 (compared to the $1 million price tag of the MRI machine).

"To prove success," adds Sommerfeldt, "we had to demonstrate not only that we could bring the noise level down, but that we did not degenerate the images at all."

Testing with the standard dummy ("a jar with liquid and a bunch of things in it") showed no noticeable differences between images taken with and without the active noise-control system.

"Then we ran active control actually with a patient -- one of us -- inside." (Sommerfeldt, on his turn, forgot to take his wallet out of his pocket; the magnetic field erased all his credit cards.) The noise-control system again had no effect on the images. Better, "There was a noticeable difference you could hear."

On a hunch, while testing their laboratory system on themselves, Sommerfeldt and his colleagues also tried its use as an intercom. They found, as they had suspected, "We could pipe in background music to mask out the leftover noise. Or the doctor could speak into a microphone to the patient and the patient could answer back.

"Not only could you get noise improvement, but it's possible that this system could provide a communications channel in a convenient manner," which would go far in humanizing the futuristic MRI machine.

Scott D. Sommerfeldt, Ph.D., is research associate and assistant professor of acoustics in the Applied Research Laboratory, 204 ARL Building, University Park, PA 16802; 814-863-1398. This research was funded by General Electric Medical Systems.