A HYBRID NEURAL AND STATISTICAL CLASSIFIER SYSTEM FOR HISTOPATHOLOGICGRADING OF PROSTATIC LESIONS

Neural network and statistical classification methods were applied to derive an objective grading for moderately and poorly differentiated l esions of the prostate, based on characteristics of the nuclear placem ent patterns. A partly trained multilayer neural network was used as a feature extractor. A hybrid classifier system using a quadratic Bayes ian classifier applied to these features allowed grade assignment cons ensus with visual diagnosis in 96% of fields from a training set of 50 0 fields and in 77% of 130 fields of a test set.