Cell classification for diagnostic of reactive histocytic hyperplasia using neural networks

The most important step in the medical treatment of patients is getting the correct diagnosis. Diagnostic procedure is based upon many facts gathered through the process of various examinations. Important information can be gathered by visual examination of tissue samples using a microscope. This paper presents one method for automatic analysis of microscopic bone marrow images in order to help physicians in the diagnosis of reactive histiocytosis. Morphological parameters of bone marrow cells are measured by the means of digital image processing and after that measured parameters are analyzed using a neural network to produce helpful diagnostic informations. By just a visual examination a physician can only estimate cell parameters, so the diagnostic procedure is mostly based on his experience. Using the described automatic system, relevant parameters can be precisely measured and a neural network is used to accumulate and keep the physician´s experience and knowledge so that correct conclusions can be made