Now, Air Force Computers May Home In On Tumors

Radiologists rely on the images produced by CAT scanners, MRI machines, and X-rays to diagnose tumors. But much of the information in the digitized pictures is too dense or too subtle for doctors to make sense of. Computer systems originally designed to seek out enemy tanks could help. Consultant Booz, Allen & Hamilton Inc.'s technology arm in McLean, Va., is collaborating with the U.S. Air Force's research laboratory in Rome, N.Y., and with Cornell University Medical College in New York to adapt military computer systems to detect small malignant tumors at an early stage.

The partners are using advanced computer techniques called neural networks that quickly learn to recognize suspect objects--be they tanks or tumors--given enough examples of the patterns of these images. Over the next four months, the scientists expect to train the military's neural nets on the lung CAT scans of 80 patients. Neural networks could also decipher digitized X-rays and MRI scans. They should help doctors distinguish between malignant and benign tumors with more confidence and reduce the number of biopsies.

    Before it's here, it's on the Bloomberg Terminal. LEARN MORE