New technologies emerge in medical AI.

New technologies emerge in medical AI

Two decades after medicine and computer science began a starry-eyed courtship, the honeymoon is over, physicians and biomedical engineers says. But scientists designing medical "expert systems" and other forms of clinically useful artificial intelligence (AI) are starting to see a few of their dreams come true. Researchers this week reported significant improvements in several intelligent medical technologies, including computer-generated radiotherapy protocols and polarized video screens that display anatomical images in three dimensions.

New applications of artificial intelligence, especially those aimed at improvements in medical imaging," are beginning to help us realize the goal of being able to reach out and hold that tumor in our hands, roll it around, bounce it off the table," even before scheduling surgery, says Henry A. Swett of the Yale University School of Medicine. This week, he and other biomedical specialists described the state-of-the-art in medical artificial intelligence at the World Congress on Medical Physics and Biomedical Engineering in San Antonio, Tex.

In the past 12 years, computerized tomography (CT scan), ultrasound and magnetic resonance imaging have "almost totally transformed" diagnostic radiology, Swett says. Now, "intelligent radiologic workstations" are starting to help radiologists cope with the "tremendous information overload" that has accompanied these technologies, which produce huge amounts of data but are limited in their ability to process and represent that information in useful ways.

Scientists long have dreamed that artificial intelligence might someday guide them through the complexities of medical problem solving. Rooted in rules rather than numbers, and with its ability to make decisions based on uncertain or incomplete information, artificial intelligence seemed to hold many advantages over the subjective and fallible human brain. However, says Jeffry A. Siegel of Temple University Hospital in Philadelphia, although some newer AI systems are good cancer therapy decisionmakers, "the notion of applied artificial intelligence as automated diagnostician is little more than fantasy." Rather, these systems are proving increasingly useful as "medical decision support systems" that review with their own explicit logic a physician's more intuitive diagnosis or treatment decision.

More important for now, artificial intelligence is spurring a revolution in medical imaging. For example, a new generation of liquid crystal video displays charged with AI-coordinated electrical currents produce rapidly alternating, oppositely polarized images that, when viewed through special polarized glasses, appear completely three-dimensional and are much easier to interpret than conventional "layered," two-dimensional views.

Facing an even more daunting technological challenge, Henry Fuchs and his colleagues at the University of North Carolina in Chapel Hill have developed a hard-wired, high-tech pair of "glasses" that displays for the user a three-dimensional visual field in "real time" through the use of programmed imagery. As the user moves about, the projected, simulated view changes appropriately. With an additional, hand-held sensor, the scene can be manipulated or layers of the image "erased."

Fuchs predicts the system may prove useful for designing more effective radiation therapies by providing simulated, three-dimensional views of a patient's internal anatomy along with a superimposed image of a radiation beam. The user could try various locations and orientations of radiation sources while a dynamic, contour map shows areas of increasing radiation dosage.