Development of a system for determination of different types of white blood corpuscle (Leucocyte) in dried blood smear

Abstract

The networks which are composed of tightly connected simple processingelements and try to mimic the characteristics of the human brain such as massiveparallelism, fault tolerance and learning from experience, are called artificial neuralnetworks.In this thesis, artificial neural networks were examined and a software forrecognizing white blood corpuscle was developed by using Kohonen's Self OrganizationFeature Map (S.O.F.M.) and perceptron algorithms. The S.O.F.M algorithm imitates theordering of sensory pathways and the high level of organization created during learning inthe human brain.In the present thesis digital white blood corpuscle images were processed by thisalgorithm and a feature vector was obtained. The resulting reduced feature vector isinputted to single-layer perceptron to train it's weights. After the training, the weights ofthe two algorithm was linked serially to make a final classification.