Lj. Mango, REDUCING FALSE NEGATIVES IN CLINICAL-PRACTICE - THE ROLE OF NEURAL-NETWORK TECHNOLOGY, American journal of obstetrics and gynecology, 175(4), 1996, pp. 1114-1119
The fact that some cervical smears result in false-negative findings i
s an unavoidable and unpredictable consequence of the conventional (ma
nual microscopic) method of screening. Errors in the detection and int
erpretation of abnormality are cited as leading causes of false-negati
ve cytology findings; these are random errors that are not known to co
rrelate with any patient risk factor, which makes the false-negative f
indings a ''silent'' threat that is difficult to prevent. Described by
many as a labor-intensive procedure, the microscopic evaluation of a
cervical smear involves a detailed search among hundreds of thousands
of cells on each smear for a possible few that may indicate abnormalit
y. Investigations into causes of false-negative findings preceding the
discovery of high-grade lesions found that many smears had very few d
iagnostic cells that were often very small in size. These small cells
were initially overlooked or misinterpreted and repeatedly missed on r
escreening. PAPNET testing is designed to supplement conventional scre
ening by detecting abnormal cells that initially may have been missed
by microscopic examination. This interactive system uses neural networ
ks, a type of artificial intelligence well suited for pattern recognit
ion, to automate the arduous search for abnormality. The instrument fo
cuses the review of suspicious cells by a trained cytologist. clinical
studies indicate that PAPNET testing is sensitive to abnormality typi
cally missed by conventional screening and that its use as a supplemen
tal test improves the accuracy of screening.