Application of back propagation to the diagnosis of breast lesions by fine needle aspiration

Abstract We investigated the capability of the back propagation (BP) neural network (NN) in the distinction of benign from malignant breast lumps. The study was carried out on Giemsa stained smears from 68 carcinomas and 32 benign lesions. Using a custom image analysis system, 25 parameters describing the size, shape and texture of the cell nucleus were measured. Three thousand nuclei were used as a training set for the NN and the whole data set were used for the test. Additionally, 238 cells from cases without definite cytological diagnosis were evaluated by the system. The application of the BP neural network on the cellular level enabled correct classification of 87% of cells; at the patient level, correct diagnosis was achieved in 98% of cases, by using a hypothesis value of 50%. Our results indicate that the use of neural networks combined with image morphometry and statistical techniques could improve the diagnostic accuracy of fine needle aspiration of breast lesions and reduce the number of unnecessary biopsies.