Contact: Cynthia C. Rosso, 703-729-3353 x306, [email protected]
Thursday, July 20, 2000, Leesburg, VA--In a study comparing the effectiveness of neural network software and experienced human observers, three Boston radiologists have found that the computer makes an excellent consultant in cases with normal findings on chest radiographs, particularly in cases with large pulmonary emboli. They found also that adding the computer-based diagnostic tools to the interpretation process improved the diagnostic predictions of even the most accurate doctor participating in the study.
The study used computers running artificial neural networks on scan data obtained directly from the anterior and posterior ventilation and perfusion images, pitting the computers against three experienced nuclear medicine practitioners. The computers and the human doctors were given the same cases to diagnose, using digital data from 100 patients with normal findings on chest radiographs who were undergoing both radionuclide ventilation--perfusion scanning and pulmonary angiography.
Drs. James A. Scott, Edwin L. Palmer and Alan J. Fischman of Massachusetts General Hospital's Department of Radiology, Division of Nuclear Medicine, and Harvard Medical School, concluded that in most cases, the machine matched the human in performance. They found, however, that in viewing scans from different perspectives, the humans and computers made different types of errors, but combining the strengths of the two interpreters allows one to correct the errors of the other. In an article in the August issue of the American Journal of Roentgenology, the three investigators conclude that "adjunctive use of artificial neural networks can be helpful...either as a 'second opinion' in the absence of a clinical colleague or for training purposes."
The American Journal of Roentgenology is a monthly publication of the American Roentgen Ray Society (ARRS), the first and oldest radiological society in America. Established in 1900, the Society has been a forum for progress in radiology since shortly after the discovery of the X ray and is dedicated to the goal of the advancement of medicine through the science of radiology and its allied sciences. The ARRS headquarters is located at 44211 Slatestone Ct., Leesburg, VA 20176-5109.
# # #
RSS