NEURAL NETWORK-ASSISTED CERVICAL-CANCER SCREENING

Neural network-assisted (NNA) screening represents a relatively recent development in the application of artificial intelligence computing a nd machine vision as a diagnostic assistant in the evaluation of cytol ogic specimens, Neural network computers have shown promise in address ing highly variable and fuzzy pattern recognition problems. The Papnet Testing System(R) (Neuromedical Systems, Inc., Suffern, New York) emp loys interactive NNA screening to supporting cytologists' analysis of cervical cytologic specimens, or ''Pap smears.'' This report reviews t he clinical effectiveness evidence base describing the performance of NNA cervical smear screening in a variety of modes and as evaluated by an international laboratory base, These data have been applied to cos t-effectiveness models of cervical cancer screening, and the results s upport the economic value of NNA screening. Finally, early results fro m investigators who have evaluated NNA analysis of cytologic specimens from other body sites, such as the bladder, lung, and esophagus, sugg est that the device might be clinically effective in a variety of appl ications.