White Blood Cell Detection Using a Novel Fuzzy Morphological Shared-Weight Neural Network

In most medical diagnostic systems, the numbers of cells, especially white blood cells, can be used to determine some diseases. Due to the complexity of microscopic blood images, the accuracy of white blood cell detection is still an active area of research. In most case, uncertainty is often happened while the images are under such conditions as backgrounds influence, cells reunion and occlusion. By treating these conditions as fuzziness inherent in an image, fuzzy concept can be introduced into white blood cell detection. After that, because of its feasible morphological properties on image processing, a new kind of fuzzy morphological hit/miss operator is presented, then on the basis of which, a kind of fuzzy morphological shared-weight neural network (FMSNN) is developed in detail. During its application on locating of white blood cell, experimental results here show that the FMSNN has the ability to deal with those conditions.