DETECTION OF MALIGNANCY ASSOCIATED CHANGES IN CERVICAL CELL-NUCLEI USING FEEDFORWARD NEURAL NETWORKS

Normal cells in the presence of a precancerous lesion undergo subtle c hanges of their DNA distribution when observed by visible microscopy. These changes have been termed Malignancy Associated Changes (MACs). U sing statistical models such as neural networks and discriminant funct ions it is possible to design classifiers that can separate these obje cts from truly normal cells. The correct classification rate using fee d-forward neural networks is compared to linear discriminant analysis when applied to detecting MACs. Classifiers were designed using 53 nuc lear features calculated from images for each of 25,360 normal appeari ng cells taken from 344 slides diagnosed as normal or containing sever e dysplasia. A linear discriminant function achieved a correct classif ication rate of 61.6% on the test data while neural networks scored as high as 72.5% on a cell-by-cell basis. The cell classifiers were appl ied to a library of 93,494 cells from 395 slides, and the results were jackknifed using a single slide feature. The discriminant function ac hieved a correct classification rate of 67.6% while the neural network s managed as high as 76.2%.