A COMPARATIVE-STUDY OF 3 VARIATIONS OF THE LEARNING VECTOR QUANTIZER IN THE DISCRIMINATION OF BENIGN FROM MALIGNANT GASTRIC CELLS

A prospective study was undertaken to investigate the potential value of morphometry and artificial neural networks (ANN) for the discrimina tion of benign and malignant gastric lesions. Two thousand five hundre d cells from 23 cases of cancer, 19 cases of gastritis and 58 cases of ulcer were selected as a training set, and an additional 8524 cells f rom an equal number of cases of cancer, gastritis and ulcer were used as a test set. Images of routine processed gastric smears stained by t he Papanicolaou technique were processed by a custom image analysis sy stem. The application of the learning vector quantization (LVQ) classi fier enabled correct classification of > 97% of benign cells and > 95% of malignant cells, obtaining an overall accuracy of > 97%. This stud y presents the capabilities of ANN, and also indicates that ANN and im age morphometry may offer useful information on the potential of malig nancy in gastric cells.